Import Skein 1.3vendor/skein/1.3vendor/skein

Bruce Schneier's hashing algorithm

Used by newer versions of ZFS

69 files changed, 31494 insertions, 0 deletions

diff --git a/Additional_Implementations/Atmel_AVR.c b/Additional_Implementations/Atmel_AVR.c
new file mode 100644
index 000000000000..11cfdd8d74f8
--- /dev/null
+++ b/Additional_Implementations/Atmel_AVR.c
@@ -0,0 +1,77 @@
+#include <stdio.h>
+#include "skein.h"
+
+#define   SKEIN_CODE_SIZE (1)       /* instantiate code size routines */
+#define   SKEIN_LOOP    (111)       /* unroll only 8 rounds */
+#define   SKEIN_USE_ASM (512+1024)  /* what to exclude here */
+#include "skein.c"
+#include "skein_block.c"
+
+/* for code size limitations, make "dummy" versions of unused block functions */
+#if SKEIN_USE_ASM & 256
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd) { }
+#endif
+#if SKEIN_USE_ASM & 512
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd) { }
+#endif
+#if SKEIN_USE_ASM & 1024
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd) { }
+#endif
+
+const u08b_t msg[1] = 
+  {
+  0
+  };
+
+int main(int argc,char *argv[])
+    {
+    u08b_t hash[1024/8];
+	u08b_t i,x;
+    static size_t aBytes,bBytes,uCount;
+
+#if !(SKEIN_USE_ASM & 256)
+    Skein_256_Ctxt_t ctx;
+
+    aBytes = 2*Skein_256_API_CodeSize();
+	bBytes = 2*Skein_256_Process_Block_CodeSize();
+	uCount =   Skein_256_Unroll_Cnt();
+
+    Skein_256_Init  (&ctx,256);
+	Skein_256_Update(&ctx,msg,sizeof(msg));
+	Skein_256_Final (&ctx,hash);
+
+    Skein_256_Process_Block(&ctx,msg,1,256);
+#endif
+
+#if !(SKEIN_USE_ASM & 512)
+    Skein_512_Ctxt_t ctx;
+
+    aBytes = 2*Skein_512_API_CodeSize();
+	bBytes = 2*Skein_512_Process_Block_CodeSize();
+	uCount =   Skein_512_Unroll_Cnt();
+
+    Skein_512_Init  (&ctx,512);
+	Skein_512_Update(&ctx,msg,sizeof(msg));
+	Skein_512_Final (&ctx,hash);
+
+    Skein_512_Process_Block(&ctx,msg,1,512);
+#endif
+
+#if !(SKEIN_USE_ASM & 1024)
+    Skein1024_Ctxt_t ctx;
+
+    aBytes = 2*Skein1024_API_CodeSize();
+	bBytes = 2*Skein1024_Process_Block_CodeSize();
+	uCount =   Skein1024_Unroll_Cnt();
+
+    Skein1024_Init  (&ctx,1024);
+	Skein1024_Update(&ctx,msg,sizeof(msg));
+	Skein1024_Final (&ctx,hash);
+
+    Skein1024_Process_Block(&ctx,msg,1,1024);
+#endif
+    printf("API size = %4d bytes. Block size = %4d bytes. Unroll=%d\n",
+	          aBytes,bBytes,uCount);
+    for (i=x=0;i<5;i++)
+	    printf("hash[%d] = %02X [%02X]\n",i,hash[i],x ^= hash[i]);
+    }
diff --git a/Additional_Implementations/skein_8bit_estimates.xls b/Additional_Implementations/skein_8bit_estimates.xls
new file mode 100644
index 000000000000..ecc66a28f205
--- /dev/null
+++ b/Additional_Implementations/skein_8bit_estimates.xls
diff --git a/Additional_Implementations/skein_MSC_v9_perf.txt b/Additional_Implementations/skein_MSC_v9_perf.txt
new file mode 100644
index 000000000000..9e8f125a45c5
--- /dev/null
+++ b/Additional_Implementations/skein_MSC_v9_perf.txt
@@ -0,0 +1,129 @@
+File STDIN:
+      1_ ||  2802.00  2814.00  |  5952.00  5952.00  | 30606.00 30606.00  | //: 32-bit, MSC_v9.00 [ C =...]
+     10_ ||   278.40   278.40  |   593.40   593.40  |  3063.00  3063.00  | //: 32-bit, MSC_v9.00 [ C =...]
+    100_ ||    65.52    65.58  |    88.02    88.08  |   306.30   306.30  | //: 32-bit, MSC_v9.00 [ C =...]
+   1000_ ||    41.26    41.41  |    47.96    47.96  |   135.28   135.29  | //: 32-bit, MSC_v9.00 [ C =...]
+  10000_ ||    38.86    39.08  |    44.13    44.21  |   119.88   120.11  | //: 32-bit, MSC_v9.00 [ C =...]
+ 100000_ ||    38.85    39.09  |    43.56    43.77  |   105.79   114.18  | //: 32-bit, MSC_v9.00 [ C =...]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =...]
+  Block  ||        10192 bytes |        22960 bytes |        53072 bytes | //: 32-bit, MSC_v9.00 [ C =...]
+      1_ ||   780.00   786.00  |  1110.00  1110.00  |  3288.00  3318.00  | //: 64-bit, MSC_v9.00 [ C =...]
+     10_ ||    78.60    79.80  |   109.80   109.80  |   331.20   331.80  | //: 64-bit, MSC_v9.00 [ C =...]
+    100_ ||    16.74    16.80  |    15.54    15.54  |    33.30    33.30  | //: 64-bit, MSC_v9.00 [ C =...]
+   1000_ ||     9.88    10.67  |     7.38     7.38  |    14.16    14.17  | //: 64-bit, MSC_v9.00 [ C =...]
+  10000_ ||     9.21     9.22  |     6.60     6.60  |    12.27    12.39  | //: 64-bit, MSC_v9.00 [ C =...]
+ 100000_ ||     9.98    10.01  |     7.04     7.08  |    12.36    13.14  | //: 64-bit, MSC_v9.00 [ C =...]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =...]
+  Block  ||         2272 bytes |         4944 bytes |        15264 bytes | //: 64-bit, MSC_v9.00 [ C =...]
+      1_ ||  2484.00  2490.00  |  4830.00  4836.00  | 22182.00 22188.00  | //: 32-bit, MSC_v9.00 [asm=...]
+     10_ ||   250.20   252.00  |   485.40   488.40  |  1936.80  1959.00  | //: 32-bit, MSC_v9.00 [asm=...]
+    100_ ||    58.62    58.68  |    70.74    70.80  |   221.76   221.76  | //: 32-bit, MSC_v9.00 [asm=...]
+   1000_ ||    34.12    34.16  |    35.44    35.44  |    85.27    85.31  | //: 32-bit, MSC_v9.00 [asm=...]
+  10000_ ||    34.78    34.98  |    35.36    35.36  |    86.31    86.35  | //: 32-bit, MSC_v9.00 [asm=...]
+ 100000_ ||    32.96    33.40  |    33.29    33.60  |    75.79    76.81  | //: 32-bit, MSC_v9.00 [asm=...]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=...]
+  Block  ||         7588 bytes |        16636 bytes |        38262 bytes | //: 32-bit, MSC_v9.00 [asm=...]
+      1_ ||   672.00   672.00  |  1068.00  1068.00  |  1920.00  1926.00  | //: 64-bit, MSC_v9.00 [asm=...]
+     10_ ||    64.80    65.40  |   107.40   108.00  |   192.00   192.60  | //: 64-bit, MSC_v9.00 [asm=...]
+    100_ ||    15.54    15.60  |    16.20    16.26  |    21.06    21.06  | //: 64-bit, MSC_v9.00 [asm=...]
+   1000_ ||     8.18     8.18  |     6.97     6.97  |     7.77     7.78  | //: 64-bit, MSC_v9.00 [asm=...]
+  10000_ ||     7.59     7.59  |     6.23     6.23  |     6.69     6.69  | //: 64-bit, MSC_v9.00 [asm=...]
+ 100000_ ||     7.55     7.71  |     6.14     6.38  |     6.56     6.86  | //: 64-bit, MSC_v9.00 [asm=...]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=...]
+  Block  ||         2323 bytes |         4733 bytes |        11817 bytes | //: 64-bit, MSC_v9.00 [asm=...]
+      1_ ||  2952.00  2958.00  |  6030.00  6036.00  | 13668.00 13674.00  | //: 32-bit, MSC_v9.00 [ C =111]
+     10_ ||   295.80   295.80  |   603.00   603.60  |  1366.80  1366.80  | //: 32-bit, MSC_v9.00 [ C =111]
+    100_ ||    69.96    70.02  |    88.98    89.04  |   136.92   137.52  | //: 32-bit, MSC_v9.00 [ C =111]
+   1000_ ||    43.90    43.96  |    48.78    48.85  |    60.08    60.11  | //: 32-bit, MSC_v9.00 [ C =111]
+  10000_ ||    41.53    41.59  |    44.76    44.80  |    53.01    53.01  | //: 32-bit, MSC_v9.00 [ C =111]
+ 100000_ ||    41.32    41.60  |    44.52    44.62  |    51.75    51.92  | //: 32-bit, MSC_v9.00 [ C =111]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =111]
+  Block  ||         1712 bytes |         3664 bytes |         7200 bytes | //: 32-bit, MSC_v9.00 [ C =111]
+      1_ ||   780.00   786.00  |  1422.00  1434.00  |  3810.00  3816.00  | //: 64-bit, MSC_v9.00 [ C =111]
+     10_ ||    75.60    76.20  |   140.40   140.40  |   380.40   381.00  | //: 64-bit, MSC_v9.00 [ C =111]
+    100_ ||    17.16    17.22  |    20.52    21.00  |    38.22    38.28  | //: 64-bit, MSC_v9.00 [ C =111]
+   1000_ ||     9.69     9.69  |    10.42    10.42  |    16.51    16.51  | //: 64-bit, MSC_v9.00 [ C =111]
+  10000_ ||     8.97     8.97  |     9.38     9.38  |    14.38    14.40  | //: 64-bit, MSC_v9.00 [ C =111]
+ 100000_ ||     9.18     9.71  |     9.35     9.49  |    14.79    14.99  | //: 64-bit, MSC_v9.00 [ C =111]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =111]
+  Block  ||          704 bytes |         1456 bytes |         2976 bytes | //: 64-bit, MSC_v9.00 [ C =111]
+      1_ ||  2580.00  2598.00  |  4842.00  4848.00  | 10578.00 10602.00  | //: 32-bit, MSC_v9.00 [asm=111]
+     10_ ||   259.80   259.80  |   484.20   484.20  |  1059.60  1060.20  | //: 32-bit, MSC_v9.00 [asm=111]
+    100_ ||    57.18    57.24  |    66.42    66.48  |    98.40    98.46  | //: 32-bit, MSC_v9.00 [asm=111]
+   1000_ ||    35.56    35.59  |    35.96    35.96  |    42.79    42.80  | //: 32-bit, MSC_v9.00 [asm=111]
+  10000_ ||    33.69    36.50  |    33.29    33.42  |    37.98    41.34  | //: 32-bit, MSC_v9.00 [asm=111]
+ 100000_ ||    33.96    34.57  |    33.93    35.69  |    38.04    38.20  | //: 32-bit, MSC_v9.00 [asm=111]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=111]
+  Block  ||         1276 bytes |         2532 bytes |         4983 bytes | //: 32-bit, MSC_v9.00 [asm=111]
+      1_ ||   678.00   678.00  |  1098.00  1098.00  |  2034.00  2040.00  | //: 64-bit, MSC_v9.00 [asm=111]
+     10_ ||    66.60    66.60  |   109.80   109.80  |   204.00   204.00  | //: 64-bit, MSC_v9.00 [asm=111]
+    100_ ||    15.48    16.68  |    16.98    16.98  |    22.38    22.38  | //: 64-bit, MSC_v9.00 [asm=111]
+   1000_ ||     8.45     8.45  |     7.93     7.93  |     8.39     8.39  | //: 64-bit, MSC_v9.00 [asm=111]
+  10000_ ||     7.81     7.81  |     6.50     6.50  |     7.18     7.18  | //: 64-bit, MSC_v9.00 [asm=111]
+ 100000_ ||     8.08     8.09  |     6.40     6.71  |     6.98     7.21  | //: 64-bit, MSC_v9.00 [asm=111]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=111]
+  Block  ||          664 bytes |         1074 bytes |         2221 bytes | //: 64-bit, MSC_v9.00 [asm=111]
+      1_ ||  2988.00  2994.00  |  6240.00  6246.00  | 13794.00 13800.00  | //: 32-bit, MSC_v9.00 [ C =332]
+     10_ ||   297.60   299.40  |   623.40   624.00  |  1379.40  1380.00  | //: 32-bit, MSC_v9.00 [ C =332]
+    100_ ||    70.26    70.32  |    91.92    91.92  |   138.00   138.06  | //: 32-bit, MSC_v9.00 [ C =332]
+   1000_ ||    44.88    44.89  |    50.20    50.20  |    60.44    60.45  | //: 32-bit, MSC_v9.00 [ C =332]
+  10000_ ||    42.42    42.42  |    46.30    46.31  |    53.29    53.31  | //: 32-bit, MSC_v9.00 [ C =332]
+ 100000_ ||    42.21    42.50  |    43.60    45.77  |    49.55    50.03  | //: 32-bit, MSC_v9.00 [ C =332]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =332]
+  Block  ||         4560 bytes |         9232 bytes |        12560 bytes | //: 32-bit, MSC_v9.00 [ C =332]
+      1_ ||   780.00   798.00  |  1920.00  1920.00  |  3732.00  3732.00  | //: 64-bit, MSC_v9.00 [ C =332]
+     10_ ||    76.80    76.80  |   189.00   191.40  |   402.60   402.60  | //: 64-bit, MSC_v9.00 [ C =332]
+    100_ ||    17.10    17.16  |    27.66    27.90  |    37.62    37.62  | //: 64-bit, MSC_v9.00 [ C =332]
+   1000_ ||     9.98    10.12  |    14.23    14.25  |    16.13    16.13  | //: 64-bit, MSC_v9.00 [ C =332]
+  10000_ ||     9.27     9.28  |    12.89    12.99  |    13.98    13.98  | //: 64-bit, MSC_v9.00 [ C =332]
+ 100000_ ||     9.32     9.56  |    13.12    13.19  |    14.15    14.23  | //: 64-bit, MSC_v9.00 [ C =332]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =332]
+  Block  ||         1200 bytes |         2928 bytes |         5008 bytes | //: 64-bit, MSC_v9.00 [ C =332]
+      1_ ||  2598.00  2604.00  |  4866.00  4878.00  | 10614.00 10632.00  | //: 32-bit, MSC_v9.00 [asm=332]
+     10_ ||   260.40   261.00  |   490.20   490.20  |  1067.40  1067.40  | //: 32-bit, MSC_v9.00 [asm=332]
+    100_ ||    60.78    60.78  |    72.00    72.00  |   106.86   106.92  | //: 32-bit, MSC_v9.00 [asm=332]
+   1000_ ||    38.38    38.42  |    39.17    39.19  |    46.49    46.61  | //: 32-bit, MSC_v9.00 [asm=332]
+  10000_ ||    40.98    47.69  |    35.81    35.86  |    40.96    43.93  | //: 32-bit, MSC_v9.00 [asm=332]
+ 100000_ ||    34.46    36.34  |    34.07    37.16  |    39.60    43.18  | //: 32-bit, MSC_v9.00 [asm=332]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=332]
+  Block  ||         3060 bytes |         6300 bytes |         8835 bytes | //: 32-bit, MSC_v9.00 [asm=332]
+      1_ ||   684.00   690.00  |  1104.00  1104.00  |  2028.00  2034.00  | //: 64-bit, MSC_v9.00 [asm=332]
+     10_ ||    70.80    70.80  |   120.00   120.00  |   219.00   219.00  | //: 64-bit, MSC_v9.00 [asm=332]
+    100_ ||    15.72    15.72  |    16.74    16.74  |    22.20    22.20  | //: 64-bit, MSC_v9.00 [asm=332]
+   1000_ ||     8.42     8.42  |     7.22     7.22  |     8.30     8.30  | //: 64-bit, MSC_v9.00 [asm=332]
+  10000_ ||     7.85     8.51  |     6.58     6.58  |     7.11     7.12  | //: 64-bit, MSC_v9.00 [asm=332]
+ 100000_ ||     7.80     9.43  |     6.90     7.71  |     7.18     8.48  | //: 64-bit, MSC_v9.00 [asm=332]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=332]
+  Block  ||         1288 bytes |         2182 bytes |         3449 bytes | //: 64-bit, MSC_v9.00 [asm=332]
+      1_ ||  2994.00  2994.00  |  6240.00  6240.00  | 14598.00 14604.00  | //: 32-bit, MSC_v9.00 [ C =335]
+     10_ ||   300.60   301.20  |   624.00   624.60  |  1459.20  1461.00  | //: 32-bit, MSC_v9.00 [ C =335]
+    100_ ||    70.62    70.68  |    91.86    91.92  |   146.10   146.16  | //: 32-bit, MSC_v9.00 [ C =335]
+   1000_ ||    44.65    44.65  |    50.20    50.20  |    62.74    62.76  | //: 32-bit, MSC_v9.00 [ C =335]
+  10000_ ||    42.16    42.42  |    46.31    46.73  |    55.11    55.13  | //: 32-bit, MSC_v9.00 [ C =335]
+ 100000_ ||    40.09    40.55  |    45.76    45.97  |    51.00    53.08  | //: 32-bit, MSC_v9.00 [ C =335]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =335]
+  Block  ||         4560 bytes |         9232 bytes |        29280 bytes | //: 32-bit, MSC_v9.00 [ C =335]
+      1_ ||   780.00   798.00  |  1890.00  1920.00  |  3498.00  3498.00  | //: 64-bit, MSC_v9.00 [ C =335]
+     10_ ||    77.40    78.00  |   190.80   195.00  |   350.40   379.20  | //: 64-bit, MSC_v9.00 [ C =335]
+    100_ ||    17.10    17.10  |    27.72    28.08  |    35.28    35.28  | //: 64-bit, MSC_v9.00 [ C =335]
+   1000_ ||     9.95    10.00  |    14.23    14.24  |    15.09    15.10  | //: 64-bit, MSC_v9.00 [ C =335]
+  10000_ ||     9.30    10.06  |    12.94    14.10  |    13.07    14.36  | //: 64-bit, MSC_v9.00 [ C =335]
+ 100000_ ||     9.33     9.58  |    13.94    13.95  |    13.24    13.92  | //: 64-bit, MSC_v9.00 [ C =335]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =335]
+  Block  ||         1200 bytes |         2928 bytes |        10880 bytes | //: 64-bit, MSC_v9.00 [ C =335]
+      1_ ||  2586.00  2592.00  |  4896.00  4902.00  | 10668.00 10668.00  | //: 32-bit, MSC_v9.00 [asm=335]
+     10_ ||   263.40   263.40  |   489.60   489.60  |  1069.20  1069.80  | //: 32-bit, MSC_v9.00 [asm=335]
+    100_ ||    61.08    61.14  |    72.30    72.36  |   107.04   107.10  | //: 32-bit, MSC_v9.00 [asm=335]
+   1000_ ||    35.57    35.57  |    36.11    36.12  |    43.07    43.12  | //: 32-bit, MSC_v9.00 [asm=335]
+  10000_ ||    33.68    34.51  |    33.29    36.32  |    37.91    39.80  | //: 32-bit, MSC_v9.00 [asm=335]
+ 100000_ ||    36.32    36.43  |    35.91    35.98  |    38.02    38.19  | //: 32-bit, MSC_v9.00 [asm=335]
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=335]
+  Block  ||         3060 bytes |         6300 bytes |        20391 bytes | //: 32-bit, MSC_v9.00 [asm=335]
+      1_ ||   684.00   690.00  |  1104.00  1104.00  |  2022.00  2022.00  | //: 64-bit, MSC_v9.00 [asm=335]
+     10_ ||    65.40    65.40  |   109.80   109.80  |   201.60   202.20  | //: 64-bit, MSC_v9.00 [asm=335]
+    100_ ||    15.78    15.78  |    16.80    16.80  |    22.02    22.08  | //: 64-bit, MSC_v9.00 [asm=335]
+   1000_ ||     8.41     8.42  |     7.21     7.22  |     8.24     8.26  | //: 64-bit, MSC_v9.00 [asm=335]
+  10000_ ||     7.84     7.84  |     6.45     6.50  |     7.12     7.12  | //: 64-bit, MSC_v9.00 [asm=335]
+ 100000_ ||     8.11     8.11  |     6.49     6.74  |     6.95     7.26  | //: 64-bit, MSC_v9.00 [asm=335]
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=335]
+  Block  ||         1288 bytes |         2182 bytes |         7133 bytes | //: 64-bit, MSC_v9.00 [asm=335]
diff --git a/Additional_Implementations/skein_block_x64.asm b/Additional_Implementations/skein_block_x64.asm
new file mode 100644
index 000000000000..b5221ae423ad
--- /dev/null
+++ b/Additional_Implementations/skein_block_x64.asm
@@ -0,0 +1,1335 @@
+;

+;----------------------------------------------------------------

+; 64-bit x86 assembler code (Microsoft ML64) for Skein block functions

+;

+; Author: Doug Whiting, Hifn

+;

+; This code is released to the public domain.

+;----------------------------------------------------------------

+;

+    .code

+;

+_MASK_ALL_  equ (256+512+1024)      ;all three algorithm bits

+_MAX_FRAME_ equ 240

+;

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_USE_ASM

+_USE_ASM_        = _MASK_ALL_

+elseif SKEIN_USE_ASM and _MASK_ALL_

+_USE_ASM_        = SKEIN_USE_ASM

+else

+_USE_ASM_        = _MASK_ALL_

+endif

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_LOOP                           ;configure loop unrolling

+_SKEIN_LOOP       = 0                       ;default is all fully unrolled

+else

+_SKEIN_LOOP       = SKEIN_LOOP

+endif

+; the unroll counts (0 --> fully unrolled)

+SKEIN_UNROLL_256  = (_SKEIN_LOOP / 100) mod 10

+SKEIN_UNROLL_512  = (_SKEIN_LOOP /  10) mod 10

+SKEIN_UNROLL_1024 = (_SKEIN_LOOP      ) mod 10

+;

+SKEIN_ASM_UNROLL  = 0

+  irp _NN_,<256,512,1024>

+    if (SKEIN_UNROLL_&_NN_) eq 0

+SKEIN_ASM_UNROLL  = SKEIN_ASM_UNROLL + _NN_

+    endif

+  endm

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_ROUNDS

+ROUNDS_256  =   72

+ROUNDS_512  =   72

+ROUNDS_1024 =   80

+else

+ROUNDS_256  = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)

+ROUNDS_512  = 8*((((SKEIN_ROUNDS /  10) + 5) mod 10) + 5)

+ROUNDS_1024 = 8*((((SKEIN_ROUNDS      ) + 5) mod 10) + 5)

+endif

+;

+irp _NN_,<256,512,1024>

+  if _USE_ASM_ and _NN_

+    irp _RR_,<%(ROUNDS_&_NN_)>

+      if _NN_ eq 1024

+%out  +++ SKEIN_ROUNDS_&_NN_ = _RR_

+      else

+%out  +++ SKEIN_ROUNDS_&_NN_  = _RR_

+      endif

+    endm

+  endif

+endm

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_CODE_SIZE

+ifdef  SKEIN_PERF

+SKEIN_CODE_SIZE equ (1)

+endif

+endif

+;

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_DEBUG

+_SKEIN_DEBUG      = 0

+else

+_SKEIN_DEBUG      = 1

+endif

+;;;;;;;;;;;;;;;;;

+;

+; define offsets of fields in hash context structure

+;

+HASH_BITS   =   0                   ;# bits of hash output

+BCNT        =   8 + HASH_BITS       ;number of bytes in BUFFER[]

+TWEAK       =   8 + BCNT            ;tweak values[0..1]

+X_VARS      =  16 + TWEAK           ;chaining vars

+;

+;(Note: buffer[] in context structure is NOT needed here :-)

+;

+r08     equ     <r8>

+r09     equ     <r9>

+;

+KW_PARITY   =   01BD11BDAA9FC1A22h  ;overall parity of key schedule words

+FIRST_MASK  =   NOT (1 SHL 62)

+;

+; rotation constants for Skein

+;

+RC_256_0_0  = 14

+RC_256_0_1  = 16

+

+RC_256_1_0  = 52

+RC_256_1_1  = 57

+

+RC_256_2_0  = 23

+RC_256_2_1  = 40

+

+RC_256_3_0  =  5

+RC_256_3_1  = 37

+

+RC_256_4_0  = 25

+RC_256_4_1  = 33

+

+RC_256_5_0  = 46

+RC_256_5_1  = 12

+

+RC_256_6_0  = 58

+RC_256_6_1  = 22

+

+RC_256_7_0  = 32

+RC_256_7_1  = 32

+

+RC_512_0_0  = 46

+RC_512_0_1  = 36

+RC_512_0_2  = 19

+RC_512_0_3  = 37

+

+RC_512_1_0  = 33

+RC_512_1_1  = 27

+RC_512_1_2  = 14

+RC_512_1_3  = 42

+

+RC_512_2_0  = 17

+RC_512_2_1  = 49

+RC_512_2_2  = 36

+RC_512_2_3  = 39

+

+RC_512_3_0  = 44

+RC_512_3_1  =  9

+RC_512_3_2  = 54

+RC_512_3_3  = 56

+

+RC_512_4_0  = 39

+RC_512_4_1  = 30

+RC_512_4_2  = 34

+RC_512_4_3  = 24

+

+RC_512_5_0  = 13

+RC_512_5_1  = 50

+RC_512_5_2  = 10

+RC_512_5_3  = 17

+

+RC_512_6_0  = 25

+RC_512_6_1  = 29

+RC_512_6_2  = 39

+RC_512_6_3  = 43

+

+RC_512_7_0  =  8

+RC_512_7_1  = 35

+RC_512_7_2  = 56

+RC_512_7_3  = 22

+

+RC_1024_0_0 = 24

+RC_1024_0_1 = 13

+RC_1024_0_2 =  8

+RC_1024_0_3 = 47

+RC_1024_0_4 =  8

+RC_1024_0_5 = 17

+RC_1024_0_6 = 22

+RC_1024_0_7 = 37

+

+RC_1024_1_0 = 38

+RC_1024_1_1 = 19

+RC_1024_1_2 = 10

+RC_1024_1_3 = 55

+RC_1024_1_4 = 49

+RC_1024_1_5 = 18

+RC_1024_1_6 = 23

+RC_1024_1_7 = 52

+

+RC_1024_2_0 = 33

+RC_1024_2_1 =  4

+RC_1024_2_2 = 51

+RC_1024_2_3 = 13

+RC_1024_2_4 = 34

+RC_1024_2_5 = 41

+RC_1024_2_6 = 59

+RC_1024_2_7 = 17

+

+RC_1024_3_0 =  5

+RC_1024_3_1 = 20

+RC_1024_3_2 = 48

+RC_1024_3_3 = 41

+RC_1024_3_4 = 47

+RC_1024_3_5 = 28

+RC_1024_3_6 = 16

+RC_1024_3_7 = 25

+

+RC_1024_4_0 = 41

+RC_1024_4_1 =  9

+RC_1024_4_2 = 37

+RC_1024_4_3 = 31

+RC_1024_4_4 = 12

+RC_1024_4_5 = 47

+RC_1024_4_6 = 44

+RC_1024_4_7 = 30

+

+RC_1024_5_0 = 16

+RC_1024_5_1 = 34

+RC_1024_5_2 = 56

+RC_1024_5_3 = 51

+RC_1024_5_4 =  4

+RC_1024_5_5 = 53

+RC_1024_5_6 = 42

+RC_1024_5_7 = 41

+

+RC_1024_6_0 = 31

+RC_1024_6_1 = 44

+RC_1024_6_2 = 47

+RC_1024_6_3 = 46

+RC_1024_6_4 = 19

+RC_1024_6_5 = 42

+RC_1024_6_6 = 44

+RC_1024_6_7 = 25

+

+RC_1024_7_0 =  9

+RC_1024_7_1 = 48

+RC_1024_7_2 = 35

+RC_1024_7_3 = 52

+RC_1024_7_4 = 23

+RC_1024_7_5 = 31

+RC_1024_7_6 = 37

+RC_1024_7_7 = 20

+;

+;  Input:  reg

+; Output: <reg> <<< RC_BlkSize_roundNum_mixNum, BlkSize=256/512/1024

+;

+RotL64 macro reg,BLK_SIZE,ROUND_NUM,MIX_NUM

+_RCNT_ = ( RC_&BLK_SIZE&_&ROUND_NUM&_&MIX_NUM AND 63 )

+  if _RCNT_  ;is there anything to do?

+    rol     reg,_RCNT_

+  endif

+endm

+;

+;----------------------------------------------------------------

+;

+; MACROS: define local vars and configure stack

+;

+;----------------------------------------------------------------

+; declare allocated space on the stack

+StackVar    macro localName,localSize

+localName   =   _STK_OFFS_

+_STK_OFFS_  =   _STK_OFFS_+(localSize)

+endm ;StackVar

+;

+;----------------------------------------------------------------

+;

+; MACRO: Configure stack frame, allocate local vars

+;

+Setup_Stack macro BLK_BITS,KS_CNT,NO_FRAME,debugCnt

+    WCNT    =    (BLK_BITS)/64

+;

+_PushCnt_   =   0                   ;save nonvolatile regs on stack

+  irp _reg_,<rbp,rsi,rdi,rbx,r12,r13,r14,r15>

+       push     _reg_

+      .pushreg  _reg_               ;pseudo-op push for exception handling

+_PushCnt_ = _PushCnt_ + 1           ;track count to keep alignment

+  endm

+;

+_STK_OFFS_  =   0                   ;starting offset from rsp

+    ;---- local  variables         ;<-- rsp

+    StackVar    X_stk  ,8*(WCNT)    ;local context vars

+    StackVar    ksTwk  ,8*3         ;key schedule: tweak words

+    StackVar    ksKey  ,8*(WCNT)+8  ;key schedule: key   words

+  if (SKEIN_ASM_UNROLL and (BLK_BITS)) eq 0

+    StackVar    ksRot ,16*(KS_CNT+0);leave space for "rotation" to happen

+  endif

+    StackVar    Wcopy  ,8*(WCNT)    ;copy of input block    

+  if _SKEIN_DEBUG

+  ifnb  <debugCnt>                  ;temp location for debug X[] info

+    StackVar    xDebug_&BLK_BITS ,8*(debugCnt)

+  endif

+  endif

+  if ((8*_PushCnt_ + _STK_OFFS_) and 8) eq 0

+    StackVar    align16,8           ;keep 16-byte aligned (adjust for retAddr?)

+tmpStk_&BLK_BITS = align16          ;use this

+  endif

+LOCAL_SIZE  =   _STK_OFFS_          ;size of local vars

+    ;---- 

+    StackVar    savRegs,8*_PushCnt_ ;saved registers

+    StackVar    retAddr,8           ;return address

+    ;---- caller parameters

+    StackVar    ctxPtr ,8           ;context ptr

+    StackVar    blkPtr ,8           ;pointer to block data

+    StackVar    blkCnt ,8           ;number of full blocks to process

+    StackVar    bitAdd ,8           ;bit count to add to tweak

+    ;---- caller's stack frame

+;

+; set up the stack frame pointer (rbp)

+;

+FRAME_OFFS  =   ksTwk + 128         ;allow short (negative) offset to ksTwk, kwKey

+  if FRAME_OFFS gt _STK_OFFS_       ;keep rbp in the "locals" range

+FRAME_OFFS  =      _STK_OFFS_

+  endif

+  if FRAME_OFFS gt _MAX_FRAME_      ;keep Microsoft .setframe happy

+FRAME_OFFS  =      _MAX_FRAME_

+  endif

+;

+ifdef SKEIN_ASM_INFO

+  if     FRAME_OFFS+128 lt savRegs

+%out +++ SKEIN_&BLK_BITS: Unable to reach all of Wcopy with short offset from rbp.

+  elseif FRAME_OFFS+128 lt Wcopy

+%out +++ SKEIN_&BLK_BITS: Unable to reach end of Wcopy with short offset from rbp.

+  elseif FRAME_OFFS+128 lt _STK_OFFS_

+%out +++ SKEIN_&BLK_BITS: Unable to reach caller parms with short offset from rbp

+  endif

+endif

+  ;put some useful defines in the .lst file (for grep)

+__STK_LCL_SIZE_&BLK_BITS = LOCAL_SIZE

+__STK_TOT_SIZE_&BLK_BITS = _STK_OFFS_

+__STK_FRM_OFFS_&BLK_BITS = FRAME_OFFS

+;

+; Notes on stack frame setup:

+;   * the most frequently used variable is X_stk[], based at [rsp+0]

+;   * the next most used is the key schedule arrays, ksKey and ksTwk

+;       so rbp is "centered" there, allowing short offsets to the key 

+;       schedule even in 1024-bit Skein case

+;   * the Wcopy variables are infrequently accessed, but they have long 

+;       offsets from both rsp and rbp only in the 1024-bit case.

+;   * all other local vars and calling parameters can be accessed 

+;       with short offsets, except in the 1024-bit case

+;

+    sub     rsp,LOCAL_SIZE          ;make room for the locals

+    .allocstack LOCAL_SIZE          ;pseudo op for exception handling

+    lea     rbp,[rsp+FRAME_OFFS]    ;maximize use of short offsets

+  ifb <NO_FRAME>

+    .setframe rbp,   FRAME_OFFS     ;pseudo op for exception handling

+  endif

+    mov         [FP_+ctxPtr],rcx    ;save caller's parameters on the stack

+    mov         [FP_+blkPtr],rdx

+    mov         [FP_+blkCnt],r08

+    mov         [FP_+bitAdd],r09

+    .endprolog                      ;pseudo op to support exception handling

+

+    mov     rdi,[FP_+ctxPtr ]       ;rdi --> context

+;

+endm ;Setup_Stack

+;

+FP_         equ <rbp-FRAME_OFFS>    ;keep as many short offsets as possible

+;

+;----------------------------------------------------------------

+;

+Reset_Stack macro   procStart

+    add     rsp,LOCAL_SIZE          ;get rid of locals (wipe??)

+  irp _reg_,<r15,r14,r13,r12,rbx,rdi,rsi,rbp>

+    pop     _reg_

+_PushCnt_ = _PushCnt_ - 1

+  endm

+  if _PushCnt_

+    .err    "Mismatched push/pops?"

+  endif

+

+    ;display code size in bytes to stdout

+  irp  _BCNT_,<%($+1-procStart)>    ;account for return opcode

+_ProcBytes_ = _BCNT_

+if     _BCNT_ ge 10000

+%out procStart code size = _BCNT_ bytes  

+elseif _BCNT_ ge  1000

+%out procStart code size =  _BCNT_ bytes  

+else

+%out procStart code size =   _BCNT_ bytes  

+endif

+  endm ;irp _BCNT_

+endm ; Reset_Stack

+;

+;----------------------------------------------------------------

+; macros to help debug internals

+;

+if _SKEIN_DEBUG

+    extrn   Skein_Show_Block:proc   ;calls to C routines

+    extrn   Skein_Show_Round:proc

+;

+SKEIN_RND_SPECIAL       =   1000

+SKEIN_RND_KEY_INITIAL   =   SKEIN_RND_SPECIAL+0

+SKEIN_RND_KEY_INJECT    =   SKEIN_RND_SPECIAL+1

+SKEIN_RND_FEED_FWD      =   SKEIN_RND_SPECIAL+2

+;

+Skein_Debug_Block macro BLK_BITS

+;

+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,

+;                     const u08b_t *blkPtr, const u64b_t *wPtr, 

+;                     const u64b_t *ksPtr,const u64b_t *tsPtr);

+;

+  irp _reg_,<rax,rcx,rdx,r08,r09,r10,r11>

+    push    _reg_                   ;save all volatile regs on tack before the call

+  endm

+    ; get and push call parameters

+    lea     rax,[FP_+ksTwk]         ;tweak pointer

+    push    rax

+    lea     rax,[FP_+ksKey]         ;key pointer

+    push    rax

+    lea     rax,[FP_+Wcopy]         ;wPtr

+    push    rax

+    mov     r09,[FP_+blkPtr]        ;blkPtr

+    push    r09                     ;(push register parameters anyway to make room on stack)

+    mov     rdx,[FP_+ctxPtr]        

+    lea     r08,[rdx+X_VARS]        ;X (pointer)

+    push    r08

+    push    rdx                     ;h (pointer)

+    mov     rcx, BLK_BITS           ;bits

+    push    rdx

+    call    Skein_Show_Block        ;call external debug handler

+    add     rsp,7*8                 ;discard parameters on stack

+  irp _reg_,<r11,r10,r09,r08,rdx,rcx,rax>

+    pop     _reg_                   ;restore regs

+  endm

+endm ; Skein_Debug_Block

+;

+;

+; the macro to "call" to debug a round

+;

+Skein_Debug_Round macro BLK_BITS,R,RDI_OFFS,afterOp

+    ; call the appropriate (local) debug function

+    push    r08

+  if (SKEIN_ASM_UNROLL and BLK_BITS) or (R ge SKEIN_RND_SPECIAL)

+    mov     r08, R

+  else                              ;compute round number using edi

+_rOffs_ = RDI_OFFS + 0

+   if BLK_BITS eq 1024

+    mov     r08,[rsp+8+rIdx_offs]   ;get rIdx off the stack (adjust for push r08)

+    lea     r08,[4*r08+1+(((R)-1) and 3)+_rOffs_]

+   else

+    lea     r08,[4*rdi+1+(((R)-1) and 3)+_rOffs_]

+   endif

+  endif

+    call    Skein_Debug_Round_&BLK_BITS

+    pop     r08

+;

+  afterOp

+endm  ;  Skein_Debug_Round

+else  ;------- _SKEIN_DEBUG (dummy macros if debug not enabled)

+Skein_Debug_Block macro BLK_BITS,afterOp

+endm

+;

+Skein_Debug_Round macro BLK_BITS,R,RDI_OFFS,afterOp

+endm

+;

+endif ; _SKEIN_DEBUG

+;

+;----------------------------------------------------------------

+;

+addReg  macro   dstReg,srcReg_A,srcReg_B,useAddOp,immOffs

+  ifnb <immOffs>

+       lea     dstReg,[srcReg_A&&srcReg_B + dstReg + immOffs]

+  elseif ((useAddOp + 0) eq 0)

+    ifndef ASM_NO_LEA

+      ;lea seems to be faster on Core 2 Duo CPUs!

+       lea     dstReg,[srcReg_A&&srcReg_B + dstReg]   

+    else

+       add     dstReg, srcReg_A&&srcReg_B

+    endif

+  else

+       add     dstReg, srcReg_A&&srcReg_B

+  endif

+endm

+;

+;=================================== Skein_256 =============================================

+;

+if _USE_ASM_ and 256

+    public  Skein_256_Process_Block

+;

+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+;

+; code

+;

+Skein_256_Process_Block proc frame

+    Setup_Stack 256,((ROUNDS_256/8)+1)

+    mov     r14,[rdi+TWEAK+8]

+    jmp   short Skein_256_block_loop

+    align   16

+    ; main hash loop for Skein_256

+Skein_256_block_loop:

+    ;

+    ; general register usage:

+    ;   RAX..RDX        = X0..X3    

+    ;   R08..R12        = ks[0..4]

+    ;   R13..R15        = ts[0..2]

+    ;   RSP, RBP        = stack/frame pointers

+    ;   RDI             = round counter or context pointer

+    ;   RSI             = temp

+    ;

+    mov     r13,[rdi+TWEAK+0]

+    add     r13,[FP_+bitAdd]        ;computed updated tweak value T0

+    mov     r15,r14

+    xor     r15,r13                 ;now r13.r15 is set as the tweak 

+

+    mov     r12,KW_PARITY

+    mov     r08,[rdi+X_VARS+ 0]

+    mov     r09,[rdi+X_VARS+ 8]

+    mov     r10,[rdi+X_VARS+16]

+    mov     r11,[rdi+X_VARS+24]

+    mov         [rdi+TWEAK+0],r13   ;save updated tweak value ctx->h.T[0]

+    xor     r12,r08                 ;start accumulating overall parity

+

+    mov     rsi,[FP_+blkPtr ]       ;esi --> input block

+    xor     r12,r09

+    mov     rax,[rsi+ 0]            ;get X[0..3]

+    xor     r12,r10

+    mov     rbx,[rsi+ 8]

+    xor     r12,r11

+    mov     rcx,[rsi+16]

+    mov     rdx,[rsi+24]

+

+    mov         [FP_+Wcopy+ 0],rax  ;save copy of input block

+    mov         [FP_+Wcopy+ 8],rbx

+    mov         [FP_+Wcopy+16],rcx

+    mov         [FP_+Wcopy+24],rdx

+

+    add     rax, r08                ;initial key injection

+    add     rbx, r09

+    add     rcx, r10

+    add     rdx, r11

+    add     rbx, r13

+    add     rcx, r14

+

+if _SKEIN_DEBUG

+    mov         [rdi+TWEAK+ 8],r14  ;save updated tweak T[1] (start bit cleared?)

+    mov         [FP_+ksKey+ 0],r08  ;save key schedule on stack for Skein_Debug_Block

+    mov         [FP_+ksKey+ 8],r09

+    mov         [FP_+ksKey+16],r10

+    mov         [FP_+ksKey+24],r11

+    mov         [FP_+ksKey+32],r12

+

+    mov         [FP_+ksTwk+ 0],r13

+    mov         [FP_+ksTwk+ 8],r14

+    mov         [FP_+ksTwk+16],r15

+

+    mov         [rsp+X_stk + 0],rax ;save X[] on stack for Skein_Debug_Block

+    mov         [rsp+X_stk + 8],rbx

+    mov         [rsp+X_stk +16],rcx

+    mov         [rsp+X_stk +24],rdx

+

+    Skein_Debug_Block 256           ;debug dump

+    Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL

+endif

+;

+if ((SKEIN_ASM_UNROLL and 256) eq 0)

+    mov         [FP_+ksKey+40],r08 ;save key schedule on stack for looping code

+    mov         [FP_+ksKey+ 8],r09

+    mov         [FP_+ksKey+16],r10

+    mov         [FP_+ksKey+24],r11

+    mov         [FP_+ksKey+32],r12

+

+    mov         [FP_+ksTwk+24],r13

+    mov         [FP_+ksTwk+ 8],r14

+    mov         [FP_+ksTwk+16],r15

+endif

+    add     rsi, WCNT*8             ;skip the block

+    mov         [FP_+blkPtr   ],rsi ;update block pointer

+;

+opLoop macro op1,op2

+  if (SKEIN_ASM_UNROLL and 256) eq 0

+    op1

+  else

+    op2

+  endif

+endm

+;

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 256

+_UNROLL_CNT =   ROUNDS_256/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_256

+  if ((ROUNDS_256/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_256"

+  endif

+    xor     rdi,rdi                   ;rdi = iteration count

+Skein_256_round_loop:

+endif

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+    ; all X and ks vars in regs     ; (ops to "rotate" ks vars, via mem, if not unrolled)

+    ; round 4*_RBase_ + 0

+    addReg  rax, rbx

+    RotL64  rbx, 256,%((4*_RBase_+0) and 7),0

+    addReg  rcx, rdx

+                    opLoop  <mov r08,[FP_+ksKey+8*rdi+8*1]>

+    xor     rbx, rax

+    RotL64  rdx, 256,%((4*_RBase_+0) and 7),1

+    xor     rdx, rcx

+ if SKEIN_ASM_UNROLL and 256

+    irp _r0_,<%(08+(_Rbase_+3) mod 5)>

+    irp _r1_,<%(13+(_Rbase_+2) mod 3)>

+      lea   rdi,[r&_r0_+r&_r1_]     ;precompute key injection value for rcx

+    endm

+    endm

+ endif

+                    opLoop  <mov r13,[FP_+ksTwk+8*rdi+8*1]>

+    Skein_Debug_Round 256,%(4*_RBase_+1)

+

+    ; round 4*_RBase_ + 1

+    addReg  rax, rdx

+    RotL64  rdx, 256,%((4*_RBase_+1) and 7),0

+    xor     rdx, rax

+                    opLoop  <mov r09,[FP_+ksKey+8*rdi+8*2]>

+    addReg  rcx, rbx

+    RotL64  rbx, 256,%((4*_RBase_+1) and 7),1

+    xor     rbx, rcx

+                    opLoop  <mov r11,[FP_+ksKey+8*rdi+8*4]>

+    Skein_Debug_Round 256,%(4*_RBase_+2)

+ if SKEIN_ASM_UNROLL and 256

+    irp _r0_,<%(08+(_Rbase_+2) mod 5)>

+    irp _r1_,<%(13+(_Rbase_+1) mod 3)>

+      lea   rsi,[r&_r0_+r&_r1_]     ;precompute key injection value for rbx

+    endm

+    endm

+ endif

+    ; round 4*_RBase_ + 2

+    addReg  rax, rbx

+    RotL64  rbx, 256,%((4*_RBase_+2) and 7),0

+    addReg  rcx, rdx

+                    opLoop  <mov r10,[FP_+ksKey+8*rdi+8*3]>

+    xor     rbx, rax

+    RotL64  rdx, 256,%((4*_RBase_+2) and 7),1

+    xor     rdx, rcx

+                    opLoop  <mov     [FP_+ksKey+8*rdi+8*6],r08> ;"rotate" the key

+                    opLoop  <lea r11,[r11+rdi+1]>   ;precompute key + tweak

+    Skein_Debug_Round 256,%(4*_RBase_+3)

+    ; round 4*_RBase_ + 3

+    addReg  rax, rdx

+    RotL64  rdx, 256,%((4*_RBase_+3) and 7),0

+    addReg  rcx, rbx

+                    opLoop  <add r10,[FP_+ksTwk+8*rdi+8*2]>    ;precompute key + tweak

+                    opLoop  <mov     [FP_+ksTwk+8*rdi+8*4],r13> ;"rotate" the tweak

+    xor     rdx, rax

+    RotL64  rbx, 256,%((4*_RBase_+3) and 7),1

+    xor     rbx, rcx

+    Skein_Debug_Round 256,%(4*_RBase_+4)

+                    opLoop  <addReg r09,r13>    ;precompute key+tweak

+      ;inject key schedule words

+_Rbase_ = _Rbase_+1

+  if SKEIN_ASM_UNROLL and 256

+      addReg    rax,r,%(08+((_Rbase_+0) mod 5))

+      addReg    rbx,rsi

+      addReg    rcx,rdi

+      addReg    rdx,r,%(08+((_Rbase_+3) mod 5)),,_Rbase_

+  else

+      inc       rdi

+      addReg    rax,r08

+      addReg    rcx,r10

+      addReg    rbx,r09

+      addReg    rdx,r11

+  endif

+      Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT

+endm ;rept _UNROLL_CNT

+

+;

+if (SKEIN_ASM_UNROLL and 256) eq 0

+    cmp     rdi,2*(ROUNDS_256/8)

+    jb      Skein_256_round_loop

+endif ; (SKEIN_ASM_UNROLL and 256) eq 0

+    mov     rdi,[FP_+ctxPtr ]           ;restore edi --> context

+

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..3}

+    xor     rax,[FP_+Wcopy + 0]

+    mov     r14,FIRST_MASK

+    xor     rbx,[FP_+Wcopy + 8]

+    xor     rcx,[FP_+Wcopy +16]

+    xor     rdx,[FP_+Wcopy +24]

+    mov         [rdi+X_VARS+ 0],rax     ;store final result

+    and     r14,[rdi+TWEAK + 8]

+    dec     qword ptr [FP_+blkCnt]      ;set zero flag

+    mov         [rdi+X_VARS+ 8],rbx

+    mov         [rdi+X_VARS+16],rcx

+    mov         [rdi+X_VARS+24],rdx

+

+    Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,,<cmp qword ptr [FP_+blkCnt],0>

+

+    ; go back for more blocks, if needed

+    jnz     Skein_256_block_loop

+    mov         [rdi+TWEAK + 8],r14

+    Reset_Stack Skein_256_Process_Block

+    ret

+

+  if _SKEIN_DEBUG

+Skein_Debug_Round_256:

+    mov         [FP_+X_stk+ 0],rax  ;first, save X[] state on stack so debug routines can access it

+    mov         [FP_+X_stk+ 8],rbx  ;(use FP_ since rsp has changed!)

+    mov         [FP_+X_stk+16],rcx

+    mov         [FP_+X_stk+24],rdx

+    push    rdx                     ;save two regs for BLK_BITS-specific parms

+    push    rcx

+    mov     rdx,[FP_+ctxPtr]        ;ctx_hdr_ptr

+    mov     rcx, 256

+    jmp     Skein_Debug_Round_Common

+  endif

+

+Skein_256_Process_Block endp

+;

+ifdef SKEIN_CODE_SIZE

+    public  Skein_256_Process_Block_CodeSize

+Skein_256_Process_Block_CodeSize proc

+    mov     rax,_ProcBytes_

+    ret

+Skein_256_Process_Block_CodeSize endp

+;

+    public  Skein_256_Unroll_Cnt

+Skein_256_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_256/8

+    mov     rax,_UNROLL_CNT

+  else

+    xor     rax,rax

+  endif

+    ret

+Skein_256_Unroll_Cnt endp

+endif

+;

+endif ;_USE_ASM_ and 256

+;

+;=================================== Skein_512 =============================================

+;

+if _USE_ASM_ and 512

+    public  Skein_512_Process_Block

+;

+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+rX_512_0    equ r08         ;register assignments for X[] values during rounds

+rX_512_1    equ r09

+rX_512_2    equ r10

+rX_512_3    equ r11

+rX_512_4    equ r12

+rX_512_5    equ r13

+rX_512_6    equ r14

+rX_512_7    equ r15

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: one round for 512-bit blocks

+;

+R_512_OneRound  macro r0,r1,r2,r3,r4,r5,r6,r7,_Rn_,op1,op2,op3,op4

+;

+    addReg      rX_512_&r0, rX_512_&r1

+    RotL64      rX_512_&r1, 512,%((_Rn_) and 7),0

+    xor         rX_512_&r1, rX_512_&r0

+            op1

+    addReg      rX_512_&r2, rX_512_&r3

+    RotL64      rX_512_&r3, 512,%((_Rn_) and 7),1

+    xor         rX_512_&r3, rX_512_&r2

+            op2

+    addReg      rX_512_&r4, rX_512_&r5

+    RotL64      rX_512_&r5, 512,%((_Rn_) and 7),2

+    xor         rX_512_&r5, rX_512_&r4

+            op3

+    addReg      rX_512_&r6, rX_512_&r7

+    RotL64      rX_512_&r7, 512,%((_Rn_) and 7),3

+    xor         rX_512_&r7, rX_512_&r6

+            op4

+    Skein_Debug_Round 512,%(_Rn_+1),-4

+;

+endm ;R_512_OneRound

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: eight rounds for 512-bit blocks

+;

+R_512_FourRounds macro _RR_    ;RR = base round number (0 mod 8)

+  if SKEIN_ASM_UNROLL and 512

+    ; here for fully unrolled case.

+    _II_ = ((_RR_)/4) + 1       ;key injection counter

+    R_512_OneRound 0,1,2,3,4,5,6,7,%((_RR_)+0),<mov rax,[FP_+ksKey+8*(((_II_)+3) mod 9)]>,,<mov rbx,[FP_+ksKey+8*(((_II_)+4) mod 9)]>

+    R_512_OneRound 2,1,4,7,6,5,0,3,%((_RR_)+1),<mov rcx,[FP_+ksKey+8*(((_II_)+5) mod 9)]>,,<mov rdx,[FP_+ksKey+8*(((_II_)+6) mod 9)]>

+    R_512_OneRound 4,1,6,3,0,5,2,7,%((_RR_)+2),<mov rsi,[FP_+ksKey+8*(((_II_)+7) mod 9)]>,,<add rcx,[FP_+ksTwk+8*(((_II_)+0) mod 3)]>

+    R_512_OneRound 6,1,0,7,2,5,4,3,%((_RR_)+3),<add rdx,[FP_+ksTwk+8*(((_II_)+1) mod 3)]>,

+    ; inject the key schedule

+    add     r08,[FP_+ksKey+8*(((_II_)+0) mod 9)]

+    addReg  r11,rax

+    add     r09,[FP_+ksKey+8*(((_II_)+1) mod 9)]

+    addReg  r12,rbx

+    add     r10,[FP_+ksKey+8*(((_II_)+2) mod 9)]

+    addReg  r13,rcx

+    addReg  r14,rdx

+    addReg  r15,rsi,,,(_II_)

+  else

+    ; here for looping case                                                    ;"rotate" key/tweak schedule (move up on stack)

+    inc     rdi                 ;bump key injection counter

+    R_512_OneRound 0,1,2,3,4,5,6,7,%((_RR_)+0),<mov rdx,[FP_+ksKey+8*rdi+8*6]>,<mov rax,[FP_+ksTwk+8*rdi-8*1]>    ,<mov rsi,[FP_+ksKey+8*rdi-8*1]>

+    R_512_OneRound 2,1,4,7,6,5,0,3,%((_RR_)+1),<mov rcx,[FP_+ksKey+8*rdi+8*5]>,<mov     [FP_+ksTwk+8*rdi+8*2],rax>,<mov     [FP_+ksKey+8*rdi+8*8],rsi>

+    R_512_OneRound 4,1,6,3,0,5,2,7,%((_RR_)+2),<mov rbx,[FP_+ksKey+8*rdi+8*4]>,<add rdx,[FP_+ksTwk+8*rdi+8*1]>    ,<mov rsi,[FP_+ksKey+8*rdi+8*7]>    

+    R_512_OneRound 6,1,0,7,2,5,4,3,%((_RR_)+3),<mov rax,[FP_+ksKey+8*rdi+8*3]>,<add rcx,[FP_+ksTwk+8*rdi+8*0]>

+    ; inject the key schedule

+    add     r08,[FP_+ksKey+8*rdi+8*0]

+    addReg  r11,rax

+    addReg  r12,rbx

+    add     r09,[FP_+ksKey+8*rdi+8*1]

+    addReg  r13,rcx

+    addReg  r14,rdx

+    add     r10,[FP_+ksKey+8*rdi+8*2]

+    addReg  r15,rsi

+    addReg  r15,rdi              ;inject the round number

+  endif

+    ;show the result of the key injection

+    Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT

+endm ;R_512_EightRounds

+;

+;;;;;;;;;;;;;;;;;

+; instantiated code

+;

+Skein_512_Process_Block proc frame

+    Setup_Stack 512,ROUNDS_512/8

+    mov     rbx,[rdi+TWEAK+ 8]

+    jmp   short Skein_512_block_loop

+    align  16

+    ; main hash loop for Skein_512

+Skein_512_block_loop:

+    ; general register usage:

+    ;   RAX..RDX        = temps for key schedule pre-loads

+    ;   R08..R15        = X0..X7

+    ;   RSP, RBP        = stack/frame pointers

+    ;   RDI             = round counter or context pointer

+    ;   RSI             = temp

+    ;

+    mov     rax,[rdi+TWEAK+ 0]

+    add     rax,[FP_+bitAdd]        ;computed updated tweak value T0

+    mov     rcx,rbx

+    xor     rcx,rax                 ;rax/rbx/rcx = tweak schedule

+    mov         [rdi+TWEAK+ 0],rax  ;save updated tweak value ctx->h.T[0]

+    mov         [FP_+ksTwk+ 0],rax

+    mov     rdx,KW_PARITY

+    mov     rsi,[FP_+blkPtr ]       ;rsi --> input block

+    mov         [FP_+ksTwk+ 8],rbx

+    mov         [FP_+ksTwk+16],rcx

+

+    irp _Rn_,<0,1,2,3,4,5,6,7>

+      mov   rX_512_&_Rn_,[rdi+X_VARS+8*(_Rn_)]

+      xor   rdx,rX_512_&_Rn_        ;compute overall parity

+      mov   [FP_+ksKey+8*(_Rn_)],rX_512_&_Rn_

+    endm                            ;load state into r08..r15, compute parity

+      mov   [FP_+ksKey+8*(8)],rdx   ;save key schedule parity

+

+    addReg  rX_512_5,rax            ;precompute key injection for tweak

+    addReg  rX_512_6,rbx

+if _SKEIN_DEBUG

+    mov         [rdi+TWEAK+ 8],rbx  ;save updated tweak value ctx->h.T[1] for Skein_Debug_Block below

+endif

+    mov     rax,[rsi+ 0]            ;load input block

+    mov     rbx,[rsi+ 8]

+    mov     rcx,[rsi+16]

+    mov     rdx,[rsi+24]

+    addReg  r08,rax                 ;do initial key injection

+    addReg  r09,rbx

+    mov         [FP_+Wcopy+ 0],rax  ;keep local copy for feedforward

+    mov         [FP_+Wcopy+ 8],rbx

+    addReg  r10,rcx

+    addReg  r11,rdx

+    mov         [FP_+Wcopy+16],rcx

+    mov         [FP_+Wcopy+24],rdx

+

+    mov     rax,[rsi+32]

+    mov     rbx,[rsi+40]

+    mov     rcx,[rsi+48]

+    mov     rdx,[rsi+56]

+    addReg  r12,rax

+    addReg  r13,rbx

+    addReg  r14,rcx

+    addReg  r15,rdx

+    mov         [FP_+Wcopy+32],rax

+    mov         [FP_+Wcopy+40],rbx

+    mov         [FP_+Wcopy+48],rcx

+    mov         [FP_+Wcopy+56],rdx

+

+if _SKEIN_DEBUG

+    irp _Rn_,<0,1,2,3,4,5,6,7>      ;save values on stack for debug output

+      mov       [rsp+X_stk+8*(_Rn_)],rX_512_&_Rn_

+    endm

+

+    Skein_Debug_Block 512           ;debug dump

+    Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL

+endif

+    add     rsi, 8*WCNT             ;skip the block

+    mov         [FP_+blkPtr   ],rsi ;update block pointer

+    ;

+    ;;;;;;;;;;;;;;;;;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 512

+_UNROLL_CNT =   ROUNDS_512/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_512

+  if ((ROUNDS_512/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_512"

+  endif

+    xor     rdi,rdi                 ;rdi = round counter

+Skein_512_round_loop:

+endif

+;

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+      R_512_FourRounds %(4*_Rbase_+00)

+_Rbase_ = _Rbase_+1

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 512) eq 0

+    cmp     rdi,2*(ROUNDS_512/8)

+    jb      Skein_512_round_loop

+    mov     rdi,[FP_+ctxPtr ]           ;restore rdi --> context

+endif

+    ; end of rounds

+    ;;;;;;;;;;;;;;;;;

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..7}

+    irp _Rn_,<0,1,2,3,4,5,6,7>

+  if (_Rn_ eq 0)

+    mov     rbx,FIRST_MASK

+  endif

+      xor   rX_512_&_Rn_,[FP_+Wcopy+8*(_Rn_)]       ;feedforward XOR

+      mov       [rdi+X_VARS+8*(_Rn_)],rX_512_&_Rn_  ;and store result

+  if (_Rn_ eq 6)

+    and     rbx,[rdi+TWEAK+ 8]

+  endif

+    endm

+    Skein_Debug_Round 512,SKEIN_RND_FEED_FWD

+

+    ; go back for more blocks, if needed

+    dec     qword ptr [FP_+blkCnt]

+    jnz     Skein_512_block_loop

+    mov         [rdi+TWEAK + 8],rbx

+

+    Reset_Stack Skein_512_Process_Block

+    ret

+;

+  if _SKEIN_DEBUG

+; call here with r08 = "round number"

+Skein_Debug_Round_512:

+    push    rdx                     ;save two regs for BLK_BITS-specific parms

+    push    rcx

+    mov     rcx,[rsp+24]            ;get back original r08 (pushed on stack in macro call)

+    mov         [FP_+X_stk],rcx     ;and save it in X_stk

+  irp _Rn_,<1,2,3,4,5,6,7>          ;save rest of X[] state on stack so debug routines can access it

+    mov         [FP_+X_stk+8*(_Rn_)],rX_512_&_Rn_ 

+  endm

+    mov     rdx,[FP_+ctxPtr]        ;ctx_hdr_ptr

+    mov     rcx, 512                ;block size

+    jmp     Skein_Debug_Round_Common

+  endif

+;

+Skein_512_Process_Block endp

+;

+ifdef SKEIN_CODE_SIZE

+    public  Skein_512_Process_Block_CodeSize

+Skein_512_Process_Block_CodeSize proc

+    mov     rax,_ProcBytes_

+    ret

+Skein_512_Process_Block_CodeSize endp

+;

+    public  Skein_512_Unroll_Cnt

+Skein_512_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_512/8

+    mov     rax,_UNROLL_CNT

+  else

+    xor     rax,rax

+  endif

+    ret

+Skein_512_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 512

+;

+;=================================== Skein1024 =============================================

+if _USE_ASM_ and 1024

+    public  Skein1024_Process_Block

+;

+; void Skein1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+; use details of permutation to make register assignments

+;

+r1K_x0    equ rdi 

+r1K_x1    equ rsi

+r1K_x2    equ rbp

+r1K_x3    equ rax

+r1K_x4    equ rcx           ;"shared" with X6, since X4/X6 alternate

+r1K_x5    equ rbx

+r1K_x6    equ rcx

+r1K_x7    equ rdx

+r1K_x8    equ r08

+r1K_x9    equ r09

+r1K_xA    equ r10

+r1K_xB    equ r11

+r1K_xC    equ r12

+r1K_xD    equ r13

+r1K_xE    equ r14

+r1K_xF    equ r15

+;

+rIdx      equ r1K_x0        ;index register for looping versions

+rIdx_offs equ tmpStk_1024

+;

+R1024_Mix  macro w0,w1,_RN0_,_Rn1_,op1

+_w0  = 0&w0&h               ;handle the hex conversion

+_w1  = 0&w1&h

+_II_ = ((_RN0_)/4)+1        ;injection count

+     ;

+    addReg      r1K_x&w0 , r1K_x&w1                     ;perform the MIX

+    RotL64      r1K_x&w1 , 1024,%((_RN0_) and 7),_Rn1_

+    xor         r1K_x&w1 , r1K_x&w0

+ if ((_RN0_) and 3) eq 3                                ;time to do key injection?

+  if _SKEIN_DEBUG

+    mov         [rsp+xDebug_1024+8*_w0],r1K_x&w0        ;save intermediate values for Debug_Round

+    mov         [rsp+xDebug_1024+8*_w1],r1K_x&w1        ; (before inline key injection)

+  endif

+  if SKEIN_ASM_UNROLL and 1024  ;here to do fully unrolled key injection

+    add         r1K_x&w0, [rsp+ksKey+      8*((_II_+_w0) mod 17)]

+    add         r1K_x&w1, [rsp+ksKey+      8*((_II_+_w1) mod 17)]

+   if     _w1 eq 13                                     ;tweak injection

+    add         r1K_x&w1, [rsp+ksTwk+      8*((_II_+0  ) mod  3)]

+   elseif _w0 eq 14

+    add         r1K_x&w0, [rsp+ksTwk+      8*((_II_+1  ) mod  3)]

+   elseif _w1 eq 15

+    add         r1K_x&w1, _II_                          ;(injection counter)

+   endif

+  else                          ;here to do looping  key injection

+   if  (_w0 eq 0)

+    mov                   [rsp+X_stk+8*_w0],r1K_x0      ;if so, store N0 so we can use reg as index

+    mov         rIdx,     [rsp+rIdx_offs]               ;get the injection counter index into rIdx (N0)

+   else

+    add         r1K_x&w0, [rsp+ksKey+8+8*rIdx+8*_w0]    ;even key injection

+   endif

+   if     _w1 eq 13                                     ;tweak injection

+    add         r1K_x&w1, [rsp+ksTwk+8+8*rIdx+8*0  ]

+   elseif _w0 eq 14

+    add         r1K_x&w0, [rsp+ksTwk+8+8*rIdx+8*1  ]

+   elseif _w1 eq 15

+    addReg      r1K_x&w1, rIdx,,,1                      ;(injection counter)

+   endif

+    add         r1K_x&w1, [rsp+ksKey+8+8*rIdx+8*_w1]    ;odd  key injection

+  endif

+ endif

+    ; insert the op provided, if any

+    op1

+endm

+;;;;;;;;;;;;;;;;;

+; MACRO: one round for 1024-bit blocks

+;

+R1024_OneRound  macro x0,x1,x2,x3,x4,x5,x6,x7,x8,x9,xA,xB,xC,xD,xE,xF,_Rn_

+  if (x0 ne 0) or ((x4 ne 4) and (x4 ne 6)) or (x4 ne (x6 xor 2))

+    .err "faulty register assignment!"

+  endif

+    R1024_Mix   x0,x1,_Rn_,0

+    R1024_Mix   x2,x3,_Rn_,1 

+    R1024_Mix   x4,x5,_Rn_,2, <mov        [rsp+X_stk+8*0&x4&h],r1K_x4>  ;save x4  on  stack (x4/x6 alternate)

+    R1024_Mix   x8,x9,_Rn_,4, <mov r1K_x6,[rsp+X_stk+8*0&x6&h]>         ;load x6 from stack 

+    R1024_Mix   xA,xB,_Rn_,5

+    R1024_Mix   xC,xD,_Rn_,6

+    R1024_Mix   x6,x7,_Rn_,3

+    R1024_Mix   xE,xF,_Rn_,7

+  if _SKEIN_DEBUG

+    Skein_Debug_Round 1024,%(_Rn_+1)

+  endif

+endm ;R1024_OneRound

+;;;;;;;;;;;;;;;;;

+; MACRO: four rounds for 1024-bit blocks

+;

+R1024_FourRounds macro _RR_    ;RR = base round number (0 mod 4)

+    ; should be here with r1K_x4 set properly, x6 stored on stack

+    R1024_OneRound 0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,%((_RR_)+0)

+    R1024_OneRound 0,9,2,D,6,B,4,F,A,7,C,3,E,5,8,1,%((_RR_)+1)

+    R1024_Oneround 0,7,2,5,4,3,6,1,C,F,E,D,8,B,A,9,%((_RR_)+2)

+    R1024_Oneround 0,F,2,B,6,D,4,9,E,1,8,5,A,3,C,7,%((_RR_)+3)

+  if (SKEIN_ASM_UNROLL and 1024) eq 0       ;here with r1K_x0 == rIdx, X0 on stack

+    ;rotate the key schedule on the stack

+    mov            [rsp+X_stk+       8* 8],r1K_x8;free up a reg

+    mov     r1K_x8,[rsp+ksKey+8*rIdx+8* 0]          ;get key

+    mov            [rsp+ksKey+8*rIdx+8*17],r1K_x8   ;rotate it (must do key first or tweak clobbers it!)

+    mov     r1K_x8,[rsp+ksTwk+8*rIdx+8* 0]          ;get tweak

+    mov            [rsp+ksTwk+8*rIdx+8* 3],r1K_x8   ;rotate it

+    mov     r1K_x8,[rsp+X_stk+       8* 8]      ;get the reg back

+    inc     rIdx                                ;bump the index

+    mov            [rsp+rIdx_offs],rIdx         ;save it

+    mov     r1K_x0,[rsp+ksKey+8*rIdx]           ;get the key schedule word for X0

+    add     r1K_x0,[rsp+X_stk+8*0]              ;perform the X0 key injection

+  endif

+    ;show the result of the key injection

+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT

+endm ;R1024_FourRounds

+;

+;;;;;;;;;;;;;;;;

+; code

+;

+Skein1024_Process_Block proc frame

+;

+    Setup_Stack 1024,ROUNDS_1024/8,NO_FRAME,<WCNT>

+    mov     r09,[rdi+TWEAK+ 8]

+    jmp   short Skein1024_block_loop

+    align  16

+    ; main hash loop for Skein1024

+Skein1024_block_loop:

+    ; general register usage:

+    ;   RSP             = stack pointer

+    ;   RAX..RDX,RSI,RDI= X1, X3..X7 (state words)

+    ;   R08..R15        = X8..X15    (state words)

+    ;   RBP             = temp (used for X0 and X2)

+    ;

+  if (SKEIN_ASM_UNROLL and 1024) eq 0

+    xor     rax,rax                 ;init loop index on the stack

+    mov     [rsp+rIdx_offs],rax

+  endif

+    mov     r08,[rdi+TWEAK+ 0]

+    add     r08,[FP_+bitAdd]        ;computed updated tweak value T0

+    mov     r10,r09

+    xor     r10,r08                 ;rax/rbx/rcx = tweak schedule

+    mov         [rdi+TWEAK+ 0],r08  ;save updated tweak value ctx->h.T[0]

+    mov         [FP_+ksTwk+ 0],r08

+    mov         [FP_+ksTwk+ 8],r09  ;keep values in r08,r09 for initial tweak injection below

+    mov         [FP_+ksTwk+16],r10

+  if _SKEIN_DEBUG

+    mov         [rdi+TWEAK+ 8],r09  ;save updated tweak value ctx->h.T[1] for Skein_Debug_Block

+  endif

+    mov     rsi ,[FP_+blkPtr ]      ;r1K_x2 --> input block

+    mov     rax , KW_PARITY         ;overall key schedule parity

+

+    ; logic here assumes the set {rdi,rsi,rbp,rax} = r1K_x{0,1,2,3}

+

+    irp _rN_,<0,1,2,3,4,6>            ;process the "initial" words, using r14,r15 as temps

+      mov       r14,[rdi+X_VARS+8*_rN_]                 ;get state word

+      mov       r15,[rsi+       8*_rN_]                 ;get msg   word

+      xor       rax,r14                                 ;update key schedule parity

+      mov           [FP_+ksKey +8*_rN_],r14             ;save key schedule word on stack

+      mov           [FP_+Wcopy +8*_rN_],r15             ;save local msg Wcopy 

+      add       r14,r15                                 ;do the initial key injection

+      mov           [rsp+X_stk +8*_rN_],r14             ;save initial state var on stack

+    endm

+    ; now process the rest, using the "real" registers 

+    ;     (MUST do it in reverse order to inject tweaks r08/r09 first)

+    irp _rN_,<F,E,D,C,B,A,9,8,7,5>

+_rr_ = 0&_rN_&h

+      mov   r1K_x&_rN_,[rdi+X_VARS+8*_rr_]              ;get key schedule word from context

+      mov   r1K_x4    ,[rsi+       8*_rr_]              ;get next input msg word

+      mov              [rsp+ksKey +8*_rr_],r1K_x&_rN_   ;save key schedule on stack

+      xor   rax       , r1K_x&_rN_                      ;accumulate key schedule parity

+      mov              [FP_+Wcopy +8*_rr_],r1K_x4       ;save copy of msg word for feedforward

+      add   r1K_x&_rN_, r1K_x4                          ;do the initial  key  injection

+      if     _rr_ eq 13                                 ;do the initial tweak injection

+        addReg r1K_x&_rN_,r08                           ;          (only in words 13/14)

+      elseif _rr_ eq 14

+        addReg r1K_x&_rN_,r09

+      endif

+    endm

+    mov                [FP_+ksKey+8*WCNT],rax           ;save key schedule parity

+if _SKEIN_DEBUG

+    Skein_Debug_Block 1024           ;debug dump

+endif

+    addReg  rsi,8*WCNT                                  ;bump the msg ptr

+    mov                [FP_+blkPtr],rsi                 ;save bumped msg ptr

+    ; re-load words 0..4 [rbp,rsi,rdi,rax,rbx] from stack, enter the main loop

+    irp _rN_,<0,1,2,3,4>                                ;(no need to re-load x6)

+      mov   r1K_x&_rN_,[rsp+X_stk+8*_rN_]               ;re-load state and get ready to go!

+    endm

+if _SKEIN_DEBUG

+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL        ;show state after initial key injection

+endif

+    ;

+    ;;;;;;;;;;;;;;;;;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 1024

+_UNROLL_CNT =   ROUNDS_1024/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_1024

+  if ((ROUNDS_1024/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL1024"

+  endif

+Skein1024_round_loop:

+endif

+;

+_Rbase_ = 0

+rept _UNROLL_CNT*2                   ;implement the rounds, 4 at a time

+      R1024_FourRounds %(4*_Rbase_+00)

+_Rbase_ = _Rbase_+1

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 1024) eq 0

+    cmp     qword ptr [rsp+tmpStk_1024],2*(ROUNDS_1024/8) ;see if we are done

+    jb      Skein1024_round_loop    

+endif

+    ; end of rounds

+    ;;;;;;;;;;;;;;;;;

+    ;

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..15}

+    mov     [rsp+X_stk+8*7],r1K_x7  ;we need a register. x6 already on stack

+    mov     r1K_x7,[rsp+ctxPtr]

+    

+    irp _rN_,<0,1,2,3,4,5,8,9,A,B,C,D,E,F>              ;do all but x6,x7

+      xor   r1K_x&_rN_,[rsp   +Wcopy +8*(0&_rN_&h)]     ;feedforward XOR

+      mov              [r1K_x7+X_VARS+8*(0&_rN_&h)],r1K_x&_rN_ ;save result into context

+  if (0&_rN_&h eq 9)

+    mov     r09,FIRST_MASK

+  endif

+  if (0&_rN_&h eq 0eh)

+    and     r09,[r1K_x7+TWEAK+ 8]

+  endif

+    endm

+    ; 

+    mov     rax,[rsp+X_stk    +8*6] ;now process x6,x7

+    mov     rbx,[rsp+X_stk    +8*7]

+    xor     rax,[rsp+Wcopy    +8*6]

+    xor     rbx,[rsp+Wcopy    +8*7]

+    mov         [r1K_x7+X_VARS+8*6],rax

+    dec     qword ptr [rsp+blkCnt]  ;set zero flag iff done

+    mov         [r1K_x7+X_VARS+8*7],rbx

+

+    Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD,,<cmp qword ptr [rsp+blkCnt],0>

+    ; go back for more blocks, if needed

+    mov     rdi,[rsp+ctxPtr]        ;don't muck with the flags here!

+    lea     rbp,[rsp+FRAME_OFFS]

+    jnz     Skein1024_block_loop

+    mov         [r1K_x7+TWEAK+ 8],r09

+    Reset_Stack Skein1024_Process_Block

+    ret

+;

+if _SKEIN_DEBUG

+; call here with r08 = "round number"

+Skein_Debug_Round_1024:

+_SP_OFFS_ = 8*2                     ;stack "offset" here: r08, return addr

+ SP_ equ <rsp + _SP_OFFS_>          ;useful shorthand below

+;

+  irp _wN_,<1,2,3,5,7,9,A,B,C,D,E,F> ;save rest of X[] state on stack so debug routines can access it

+    mov         [SP_+X_stk+8*(0&_wN_&h)],r1K_x&_wN_

+  endm

+    ;figure out what to do with x0. On rounds R where R==0 mod 4, it's already on the stack

+    cmp     r08,SKEIN_RND_SPECIAL   ;special rounds always save

+    jae     save_x0

+    test    r08,3

+    jz      save_x0_not

+save_x0:

+    mov     [SP_+X_stk+8*0],r1K_x0

+save_x0_not:

+    ;figure out the x4/x6 swapping state and save the correct one!

+    cmp     r08,SKEIN_RND_SPECIAL   ;special rounds always do x4

+    jae     save_x4

+    test    r08,1                   ;and even ones have r4 as well

+    jz      save_x4

+    mov     [SP_+X_stk+8*6],r1K_x6

+    jmp     short debug_1024_go

+save_x4:

+    mov     [SP_+X_stk+8*4],r1K_x4

+debug_1024_go:

+    ;now all is saved in Xstk[] except for X8

+    push    rdx                     ;save two regs for BLK_BITS-specific parms

+    push    rcx

+_SP_OFFS_ = _SP_OFFS_ + 16          ;adjust stack offset accordingly

+    ; now stack offset is 32 to X_stk

+    mov     rcx,[SP_ - 8]           ;get back original r08 (pushed on stack in macro call)

+    mov         [SP_+X_stk+8*8],rcx ;and save it in its rightful place in X_stk[8]

+    mov     rdx,[SP_+ctxPtr]        ;ctx_hdr_ptr

+    mov     rcx, 1024               ;block size

+    jmp     Skein_Debug_Round_Common

+endif

+;

+Skein1024_Process_Block endp

+;

+ifdef SKEIN_CODE_SIZE

+    public  Skein1024_Process_Block_CodeSize

+Skein1024_Process_Block_CodeSize proc

+    mov     rax,_ProcBytes_

+    ret

+Skein1024_Process_Block_CodeSize endp

+;

+    public  Skein1024_Unroll_Cnt

+Skein1024_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_1024/8

+    mov     rax,_UNROLL_CNT

+  else

+    xor     rax,rax

+  endif

+    ret

+Skein1024_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 1024

+;

+if _SKEIN_DEBUG

+;----------------------------------------------------------------

+;local debug routine to set up for calls to:

+;  void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);

+;

+; here with r08 = round number

+;           rdx = ctx_hdr_ptr

+;           rcx = block size (256/512/1024)

+;

+Skein_Debug_Round_Common:

+_SP_OFFS_ = 32                      ;current stack "offset": r08, retAddr, rcx, rdx

+    irp _rr_,<rax,rbx,rsi,rdi,rbp,r09,r10,r11,r12,r13,r14,r15>  ;save the rest of the regs

+      push  _rr_

+_SP_OFFS_ = _SP_OFFS_+8

+    endm

+ if (_SP_OFFS_ and 0Fh)             ; make sure stack is still 16-byte aligned here

+    .err    "Debug_Round_Common: stack alignment"

+ endif

+    ; compute r09 = ptr to the X[] array on the stack

+    lea     r09,[SP_+X_stk]         ;adjust for reg pushes, return address

+    cmp     r08,SKEIN_RND_FEED_FWD  ;special handling for feedforward "round"?

+    jnz     _got_r09a

+    lea     r09,[rdx+X_VARS]

+_got_r09a:

+  if _USE_ASM_ and 1024

+    ; special handling for 1024-bit case

+    ;    (for rounds right before with key injection: 

+    ;        use xDebug_1024[] instead of X_stk[])

+    cmp     r08,SKEIN_RND_SPECIAL

+    jae     _got_r09b               ;must be a normal round

+    or      r08,r08

+    jz      _got_r09b               ;just before key injection

+    test    r08,3

+    jne     _got_r09b

+    cmp     rcx,1024                ;only 1024-bit(s) for now

+    jne     _got_r09b

+    lea     r09,[SP_+xDebug_1024]

+_got_r09b:

+  endif

+    sub     rsp, 8*4                ;make room for parms on stack

+    call    Skein_Show_Round        ;call external debug handler

+    add     rsp, 8*4                ;discard parm space on the stack

+

+    irp _rr_,<r15,r14,r13,r12,r11,r10,r09,rbp,rdi,rsi,rbx,rax>  ;restore regs

+      pop   _rr_

+_SP_OFFS_ = _SP_OFFS_-8

+    endm

+ if _SP_OFFS_ - 32

+    .err    "Debug_Round_Common: push/pop misalignment!"

+ endif    

+    pop     rcx

+    pop     rdx

+    ret

+endif

+;----------------------------------------------------------------

+    end

diff --git a/Additional_Implementations/skein_block_x64.s b/Additional_Implementations/skein_block_x64.s
new file mode 100644
index 000000000000..b2d0a83acbe9
--- /dev/null
+++ b/Additional_Implementations/skein_block_x64.s
@@ -0,0 +1,1328 @@
+#
+#----------------------------------------------------------------
+# 64-bit x86 assembler code (gnu as) for Skein block functions
+#
+# Author: Doug Whiting, Hifn/Exar
+#
+# This code is released to the public domain.
+#----------------------------------------------------------------
+#
+    .text
+    .altmacro
+    .psize 0,128                            #list file has no page boundaries
+#
+_MASK_ALL_  =  (256+512+1024)               #all three algorithm bits
+_MAX_FRAME_ =  240
+#
+#################
+.ifndef SKEIN_USE_ASM
+_USE_ASM_         = _MASK_ALL_
+.else
+_USE_ASM_         = SKEIN_USE_ASM
+.endif
+#################
+.ifndef SKEIN_LOOP                          #configure loop unrolling
+_SKEIN_LOOP       =   2                     #default is fully unrolled for 256/512, twice for 1024
+.else
+_SKEIN_LOOP       = SKEIN_LOOP
+  .irp _NN_,%_SKEIN_LOOP                #only display loop unrolling if default changed on command line
+.print  "+++ SKEIN_LOOP = \_NN_"
+  .endr
+.endif
+# the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256  = (_SKEIN_LOOP / 100) % 10
+SKEIN_UNROLL_512  = (_SKEIN_LOOP /  10) % 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP      ) % 10
+#
+SKEIN_ASM_UNROLL  = 0
+  .irp _NN_,256,512,1024
+    .if (SKEIN_UNROLL_\_NN_) == 0
+SKEIN_ASM_UNROLL  = SKEIN_ASM_UNROLL + \_NN_
+    .endif
+  .endr
+#################
+#
+.ifndef SKEIN_ROUNDS
+ROUNDS_256  =   72
+ROUNDS_512  =   72
+ROUNDS_1024 =   80
+.else
+ROUNDS_256  = 8*((((SKEIN_ROUNDS / 100) + 5) % 10) + 5)
+ROUNDS_512  = 8*((((SKEIN_ROUNDS /  10) + 5) % 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS      ) + 5) % 10) + 5)
+# only display rounds if default size is changed on command line
+.irp _NN_,256,512,1024
+  .if _USE_ASM_ && \_NN_
+    .irp _RR_,%(ROUNDS_\_NN_)
+      .if _NN_ < 1024
+.print  "+++ SKEIN_ROUNDS_\_NN_  = \_RR_"
+      .else
+.print  "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+      .endif
+    .endr
+  .endif
+.endr
+.endif
+#################
+#
+.ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE = (1)
+.else
+.ifdef  SKEIN_PERF                           #use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE = (1)
+.else
+_SKEIN_CODE_SIZE = (0)
+.endif
+.endif
+#
+#################
+#
+.ifndef SKEIN_DEBUG
+_SKEIN_DEBUG      = 0
+.else
+_SKEIN_DEBUG      = 1
+.endif
+#################
+#
+# define offsets of fields in hash context structure
+#
+HASH_BITS   =   0                   #bits of hash output
+BCNT        =   8 + HASH_BITS       #number of bytes in BUFFER[]
+TWEAK       =   8 + BCNT            #tweak values[0..1]
+X_VARS      =  16 + TWEAK           #chaining vars
+#
+#(Note: buffer[] in context structure is NOT needed here :-)
+#
+KW_PARITY   =   0x1BD11BDAA9FC1A22  #overall parity of key schedule words
+FIRST_MASK  =   ~ (1 <<  6)
+FIRST_MASK64=   ~ (1 << 62)
+#
+# rotation constants for Skein
+#
+RC_256_0_0  = 14
+RC_256_0_1  = 16
+
+RC_256_1_0  = 52
+RC_256_1_1  = 57
+
+RC_256_2_0  = 23
+RC_256_2_1  = 40
+
+RC_256_3_0  =  5
+RC_256_3_1  = 37
+
+RC_256_4_0  = 25
+RC_256_4_1  = 33
+
+RC_256_5_0  = 46
+RC_256_5_1  = 12
+
+RC_256_6_0  = 58
+RC_256_6_1  = 22
+
+RC_256_7_0  = 32
+RC_256_7_1  = 32
+
+RC_512_0_0  = 46
+RC_512_0_1  = 36
+RC_512_0_2  = 19
+RC_512_0_3  = 37
+
+RC_512_1_0  = 33
+RC_512_1_1  = 27
+RC_512_1_2  = 14
+RC_512_1_3  = 42
+
+RC_512_2_0  = 17
+RC_512_2_1  = 49
+RC_512_2_2  = 36
+RC_512_2_3  = 39
+
+RC_512_3_0  = 44
+RC_512_3_1  =  9
+RC_512_3_2  = 54
+RC_512_3_3  = 56
+
+RC_512_4_0  = 39
+RC_512_4_1  = 30
+RC_512_4_2  = 34
+RC_512_4_3  = 24
+
+RC_512_5_0  = 13
+RC_512_5_1  = 50
+RC_512_5_2  = 10
+RC_512_5_3  = 17
+
+RC_512_6_0  = 25
+RC_512_6_1  = 29
+RC_512_6_2  = 39
+RC_512_6_3  = 43
+
+RC_512_7_0  =  8
+RC_512_7_1  = 35
+RC_512_7_2  = 56
+RC_512_7_3  = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 =  8
+RC_1024_0_3 = 47
+RC_1024_0_4 =  8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 =  4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 =  5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 =  9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 =  4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 =  9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+#
+#  Input:  reg
+# Output: <reg> <<< RC_BlkSize_roundNum_mixNum, BlkSize=256/512/1024
+#
+.macro RotL64   reg,BLK_SIZE,ROUND_NUM,MIX_NUM
+_RCNT_ = RC_\BLK_SIZE&_\ROUND_NUM&_\MIX_NUM
+  .if _RCNT_  #is there anything to do?
+    rolq    $_RCNT_,%\reg
+  .endif
+.endm
+#
+#----------------------------------------------------------------
+#
+# MACROS: define local vars and configure stack
+#
+#----------------------------------------------------------------
+# declare allocated space on the stack
+.macro StackVar localName,localSize
+\localName  =   _STK_OFFS_
+_STK_OFFS_  =   _STK_OFFS_+(\localSize)
+.endm #StackVar
+#
+#----------------------------------------------------------------
+#
+# MACRO: Configure stack frame, allocate local vars
+#
+.macro Setup_Stack BLK_BITS,KS_CNT,debugCnt
+    WCNT    =    (\BLK_BITS)/64
+#
+_PushCnt_   =   0                   #save nonvolatile regs on stack
+  .irp _reg_,rbp,rbx,r12,r13,r14,r15
+       pushq    %\_reg_
+_PushCnt_ = _PushCnt_ + 1           #track count to keep alignment
+  .endr
+#
+_STK_OFFS_  =   0                   #starting offset from rsp
+    #---- local  variables         #<-- rsp
+    StackVar    X_stk  ,8*(WCNT)    #local context vars
+    StackVar    ksTwk  ,8*3         #key schedule: tweak words
+    StackVar    ksKey  ,8*(WCNT)+8  #key schedule: key   words
+  .if (SKEIN_ASM_UNROLL && (\BLK_BITS)) == 0
+    StackVar    ksRot ,16*(\KS_CNT) #leave space for "rotation" to happen
+  .endif
+    StackVar    Wcopy  ,8*(WCNT)    #copy of input block    
+  .if _SKEIN_DEBUG
+  .if \debugCnt + 0                 #temp location for debug X[] info
+    StackVar    xDebug_\BLK_BITS ,8*(\debugCnt)
+  .endif
+  .endif
+  .if ((8*_PushCnt_ + _STK_OFFS_) % 8) == 0
+    StackVar    align16,8           #keep 16-byte aligned (adjust for retAddr?)
+tmpStk_\BLK_BITS = align16          #use this
+  .endif
+    #---- saved caller parameters (from regs rdi, rsi, rdx, rcx)
+    StackVar    ctxPtr ,8           #context ptr
+    StackVar    blkPtr ,8           #pointer to block data
+    StackVar    blkCnt ,8           #number of full blocks to process
+    StackVar    bitAdd ,8           #bit count to add to tweak
+LOCAL_SIZE  =   _STK_OFFS_          #size of "local" vars
+    #---- 
+    StackVar    savRegs,8*_PushCnt_ #saved registers
+    StackVar    retAddr,8           #return address
+    #---- caller's stack frame (aligned mod 16)
+#
+# set up the stack frame pointer (rbp)
+#
+FRAME_OFFS  =   ksTwk + 128         #allow short (negative) offset to ksTwk, kwKey
+  .if FRAME_OFFS > _STK_OFFS_       #keep rbp in the "locals" range
+FRAME_OFFS  =      _STK_OFFS_
+  .endif
+F_O         =   -FRAME_OFFS
+#
+  #put some useful defines in the .lst file (for grep)
+__STK_LCL_SIZE_\BLK_BITS = LOCAL_SIZE
+__STK_TOT_SIZE_\BLK_BITS = _STK_OFFS_
+__STK_FRM_OFFS_\BLK_BITS = FRAME_OFFS
+#
+# Notes on stack frame setup:
+#   * the most frequently used variable is X_stk[], based at [rsp+0]
+#   * the next most used is the key schedule arrays, ksKey and ksTwk
+#       so rbp is "centered" there, allowing short offsets to the key 
+#       schedule even in 1024-bit Skein case
+#   * the Wcopy variables are infrequently accessed, but they have long 
+#       offsets from both rsp and rbp only in the 1024-bit case.
+#   * all other local vars and calling parameters can be accessed 
+#       with short offsets, except in the 1024-bit case
+#
+    subq    $LOCAL_SIZE,%rsp        #make room for the locals
+    leaq    FRAME_OFFS(%rsp),%rbp   #maximize use of short offsets
+    movq    %rdi, ctxPtr+F_O(%rbp)  #save caller's parameters on the stack
+    movq    %rsi, blkPtr+F_O(%rbp)
+    movq    %rdx, blkCnt+F_O(%rbp)
+    movq    %rcx, bitAdd+F_O(%rbp)
+#
+.endm #Setup_Stack
+#
+#----------------------------------------------------------------
+#
+.macro Reset_Stack
+    addq    $LOCAL_SIZE,%rsp        #get rid of locals (wipe??)
+  .irp _reg_,r15,r14,r13,r12,rbx,rbp
+    popq    %\_reg_                 #restore caller's regs
+_PushCnt_ = _PushCnt_ - 1
+  .endr
+  .if _PushCnt_
+    .error  "Mismatched push/pops?"
+  .endif
+.endm # Reset_Stack
+#
+#----------------------------------------------------------------
+# macros to help debug internals
+#
+.if _SKEIN_DEBUG
+    .extern  Skein_Show_Block     #calls to C routines
+    .extern  Skein_Show_Round
+#
+SKEIN_RND_SPECIAL       =   1000
+SKEIN_RND_KEY_INITIAL   =   SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT    =   SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD      =   SKEIN_RND_SPECIAL+2
+#
+.macro Skein_Debug_Block BLK_BITS
+#
+#void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+#                     const u08b_t *blkPtr, const u64b_t *wPtr, 
+#                     const u64b_t *ksPtr,const u64b_t *tsPtr)
+#
+_NN_ = 0
+  .irp _reg_,rax,rcx,rdx,rsi,rdi,r8,r9,r10,r11
+    pushq   %\_reg_                 #save all volatile regs on tack before the call
+_NN_ = _NN_ + 1
+  .endr
+    # get and push call parameters
+    movq    $\BLK_BITS      ,%rdi   #bits
+    movq    ctxPtr+F_O(%rbp),%rsi   #h (pointer)
+    leaq    X_VARS    (%rsi),%rdx   #X (pointer)
+    movq    blkPtr+F_O(%rbp),%rcx   #blkPtr
+    leaq    Wcopy +F_O(%rbp),%r8    #wPtr
+    leaq    ksKey +F_O(%rbp),%r9    #key pointer
+    leaq    ksTwk +F_O(%rbp),%rax   #tweak pointer
+    pushq   %rax                    #   (pass on the stack)
+    call    Skein_Show_Block        #call external debug handler
+    addq    $8*1,%rsp               #discard parameters on stack
+  .if (_NN_ % 2 ) == 0              #check stack alignment
+    .error "Stack misalignment problem in Skein_Debug_Block_\_BLK_BITS"
+  .endif
+  .irp _reg_,r11,r10,r9,r8,rdi,rsi,rdx,rcx,rax
+    popq    %\_reg_                 #restore regs
+_NN_ = _NN_ - 1
+  .endr
+  .if _NN_
+    .error "Push/pop mismatch problem in Skein_Debug_Block_\_BLK_BITS"
+  .endif
+.endm # Skein_Debug_Block
+#
+# the macro to "call" to debug a round
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+    # call the appropriate (local) debug "function"
+    pushq   %rdx                    #save rdx, so we can use it for round "number"
+  .if (SKEIN_ASM_UNROLL && \BLK_BITS) || (\R >= SKEIN_RND_SPECIAL)
+    movq    $\R,%rdx
+  .else                             #compute round number using edi
+_rOffs_ = \RDI_OFFS + 0
+   .if \BLK_BITS == 1024
+    movq    rIdx_offs+8(%rsp),%rdx  #get rIdx off the stack (adjust for pushq rdx above)
+    leaq    1+(((\R)-1) && 3)+_rOffs_(,%rdx,4),%rdx
+   .else
+    leaq    1+(((\R)-1) && 3)+_rOffs_(,%rdi,4),%rdx
+   .endif
+  .endif
+    call    Skein_Debug_Round_\BLK_BITS
+    popq    %rdx                    #restore origianl rdx value
+#
+    afterOp
+.endm  #  Skein_Debug_Round
+.else  #------- _SKEIN_DEBUG (dummy macros if debug not enabled)
+.macro Skein_Debug_Block BLK_BITS
+.endm
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+.endm
+#
+.endif # _SKEIN_DEBUG
+#
+#----------------------------------------------------------------
+#
+.macro  addReg dstReg,srcReg_A,srcReg_B,useAddOp,immOffs
+  .if \immOffs + 0
+       leaq    \immOffs(%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+  .elseif ((\useAddOp + 0) == 0)
+    .ifndef ASM_NO_LEA  #lea seems to be faster on Core 2 Duo CPUs!
+       leaq   (%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+    .else
+       addq    %\srcReg_A\srcReg_B,%\dstReg
+    .endif
+  .else
+       addq    %\srcReg_A\srcReg_B,%\dstReg
+  .endif
+.endm
+
+# keep Intel-style ordering here, to match addReg
+.macro  xorReg dstReg,srcReg_A,srcReg_B
+        xorq   %\srcReg_A\srcReg_B,%\dstReg
+.endm
+#
+#----------------------------------------------------------------
+#
+.macro C_label lName
+ \lName:        #use both "genders" to work across linkage conventions
+_\lName:
+    .global  \lName
+    .global _\lName
+.endm
+#
+#=================================== Skein_256 =============================================
+#
+.if _USE_ASM_ & 256
+#
+# void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+#
+# code
+#
+C_label Skein_256_Process_Block
+    Setup_Stack 256,((ROUNDS_256/8)+1)
+    movq    TWEAK+8(%rdi),%r14
+    jmp     Skein_256_block_loop
+    .p2align 4
+    # main hash loop for Skein_256
+Skein_256_block_loop:
+    #
+    # general register usage:
+    #   RAX..RDX        = X0..X3    
+    #   R08..R12        = ks[0..4]
+    #   R13..R15        = ts[0..2]
+    #   RSP, RBP        = stack/frame pointers
+    #   RDI             = round counter or context pointer
+    #   RSI             = temp
+    #
+    movq    TWEAK+0(%rdi)     ,%r13
+    addq    bitAdd+F_O(%rbp)  ,%r13  #computed updated tweak value T0
+    movq    %r14              ,%r15
+    xorq    %r13              ,%r15  #now %r13.%r15 is set as the tweak 
+
+    movq    $KW_PARITY        ,%r12
+    movq       X_VARS+ 0(%rdi),%r8
+    movq       X_VARS+ 8(%rdi),%r9 
+    movq       X_VARS+16(%rdi),%r10
+    movq       X_VARS+24(%rdi),%r11
+    movq    %r13,TWEAK+0(%rdi)       #save updated tweak value ctx->h.T[0]
+    xorq    %r8               ,%r12  #start accumulating overall parity
+
+    movq    blkPtr +F_O(%rbp) ,%rsi  #esi --> input block
+    xorq    %r9               ,%r12
+    movq     0(%rsi)          ,%rax  #get X[0..3]
+    xorq    %r10              ,%r12
+    movq     8(%rsi)          ,%rbx
+    xorq    %r11              ,%r12
+    movq    16(%rsi)          ,%rcx
+    movq    24(%rsi)          ,%rdx
+
+    movq    %rax,Wcopy+ 0+F_O(%rbp)  #save copy of input block
+    movq    %rbx,Wcopy+ 8+F_O(%rbp)    
+    movq    %rcx,Wcopy+16+F_O(%rbp)    
+    movq    %rdx,Wcopy+24+F_O(%rbp)    
+
+    addq    %r8 ,%rax                #initial key injection
+    addq    %r9 ,%rbx 
+    addq    %r10,%rcx
+    addq    %r11,%rdx
+    addq    %r13,%rbx
+    addq    %r14,%rcx
+
+.if _SKEIN_DEBUG
+    movq    %r14,TWEAK+ 8(%rdi)      #save updated tweak T[1] (start bit cleared?)
+    movq    %r8 ,ksKey+ 0+F_O(%rbp)  #save key schedule on stack for Skein_Debug_Block
+    movq    %r9 ,ksKey+ 8+F_O(%rbp)    
+    movq    %r10,ksKey+16+F_O(%rbp)    
+    movq    %r11,ksKey+24+F_O(%rbp)    
+    movq    %r12,ksKey+32+F_O(%rbp)    
+                                       
+    movq    %r13,ksTwk+ 0+F_O(%rbp)    
+    movq    %r14,ksTwk+ 8+F_O(%rbp)    
+    movq    %r15,ksTwk+16+F_O(%rbp)    
+                                       
+    movq    %rax,X_stk + 0(%rsp)     #save X[] on stack for Skein_Debug_Block
+    movq    %rbx,X_stk + 8(%rsp)       
+    movq    %rcx,X_stk +16(%rsp)       
+    movq    %rdx,X_stk +24(%rsp)       
+
+    Skein_Debug_Block 256            #debug dump
+    Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL
+.endif
+#
+.if ((SKEIN_ASM_UNROLL & 256) == 0)
+    movq    %r8 ,ksKey+40+F_O(%rbp)  #save key schedule on stack for looping code
+    movq    %r9 ,ksKey+ 8+F_O(%rbp)    
+    movq    %r10,ksKey+16+F_O(%rbp)    
+    movq    %r11,ksKey+24+F_O(%rbp)    
+    movq    %r12,ksKey+32+F_O(%rbp)    
+                                       
+    movq    %r13,ksTwk+24+F_O(%rbp)    
+    movq    %r14,ksTwk+ 8+F_O(%rbp)    
+    movq    %r15,ksTwk+16+F_O(%rbp)    
+.endif
+    addq    $WCNT*8,%rsi             #skip the block
+    movq    %rsi,blkPtr  +F_O(%rbp)  #update block pointer
+    #
+    # now the key schedule is computed. Start the rounds
+    #
+.if SKEIN_ASM_UNROLL & 256
+_UNROLL_CNT =   ROUNDS_256/8
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_256
+  .if ((ROUNDS_256/8) % _UNROLL_CNT)
+    .error "Invalid SKEIN_UNROLL_256"
+  .endif
+    xorq    %rdi,%rdi                #rdi = iteration count
+Skein_256_round_loop:
+.endif
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+    # all X and ks vars in regs      # (ops to "rotate" ks vars, via mem, if not unrolled)
+    # round 4*_RBase_ + 0
+    addReg  rax, rbx
+    RotL64  rbx, 256,%((4*_Rbase_+0) % 8),0
+    addReg  rcx, rdx
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq ksKey+8*1+F_O(%rbp,%rdi,8),%r8
+                .endif
+    xorReg  rbx, rax
+    RotL64  rdx, 256,%((4*_Rbase_+0) % 8),1
+    xorReg  rdx, rcx
+  .if SKEIN_ASM_UNROLL & 256
+    .irp _r0_,%( 8+(_Rbase_+3) % 5)
+    .irp _r1_,%(13+(_Rbase_+2) % 3)
+      leaq   (%r\_r0_,%r\_r1_),%rdi    #precompute key injection value for %rcx
+    .endr
+    .endr
+  .endif
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq ksTwk+8*1+F_O(%rbp,%rdi,8),%r13
+                .endif
+    Skein_Debug_Round 256,%(4*_Rbase_+1)
+
+    # round 4*_Rbase_ + 1
+    addReg  rax, rdx
+    RotL64  rdx, 256,%((4*_Rbase_+1) % 8),0
+    xorReg  rdx, rax
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq ksKey+8*2+F_O(%rbp,%rdi,8),%r9
+                .endif
+    addReg  rcx, rbx
+    RotL64  rbx, 256,%((4*_Rbase_+1) % 8),1
+    xorReg  rbx, rcx
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq ksKey+8*4+F_O(%rbp,%rdi,8),%r11
+                .endif
+    Skein_Debug_Round 256,%(4*_Rbase_+2)
+ .if SKEIN_ASM_UNROLL & 256
+    .irp _r0_,%( 8+(_Rbase_+2) % 5)
+    .irp _r1_,%(13+(_Rbase_+1) % 3)
+      leaq   (%r\_r0_,%r\_r1_),%rsi     #precompute key injection value for %rbx
+    .endr
+    .endr
+ .endif
+    # round 4*_Rbase_ + 2
+    addReg  rax, rbx
+    RotL64  rbx, 256,%((4*_Rbase_+2) % 8),0
+    addReg  rcx, rdx
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq ksKey+8*3+F_O(%rbp,%rdi,8),%r10
+                .endif
+    xorReg  rbx, rax
+    RotL64  rdx, 256,%((4*_Rbase_+2) % 8),1
+    xorReg  rdx, rcx
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    movq %r8,ksKey+8*6+F_O(%rbp,%rdi,8)  #"rotate" the key
+                    leaq 1(%r11,%rdi),%r11               #precompute key + tweak
+                .endif
+    Skein_Debug_Round 256,%(4*_Rbase_+3)
+    # round 4*_Rbase_ + 3
+    addReg  rax, rdx
+    RotL64  rdx, 256,%((4*_Rbase_+3) % 8),0
+    addReg  rcx, rbx
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    addq      ksTwk+8*2+F_O(%rbp,%rdi,8),%r10  #precompute key + tweak
+                    movq %r13,ksTwk+8*4+F_O(%rbp,%rdi,8)       #"rotate" the tweak
+                .endif
+    xorReg  rdx, rax
+    RotL64  rbx, 256,%((4*_Rbase_+3) % 8),1
+    xorReg  rbx, rcx
+    Skein_Debug_Round 256,%(4*_Rbase_+4)
+                .if (SKEIN_ASM_UNROLL & 256) == 0
+                    addReg r9 ,r13           #precompute key+tweak
+                .endif
+      #inject key schedule words
+_Rbase_ = _Rbase_+1
+  .if SKEIN_ASM_UNROLL & 256
+    addReg    rax,r,%(8+((_Rbase_+0) % 5))
+    addReg    rbx,rsi
+    addReg    rcx,rdi
+    addReg    rdx,r,%(8+((_Rbase_+3) % 5)),,_Rbase_
+  .else
+    incq      %rdi
+    addReg    rax,r8 
+    addReg    rcx,r10
+    addReg    rbx,r9 
+    addReg    rdx,r11
+  .endif
+    Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 256) == 0
+    cmpq    $2*(ROUNDS_256/8),%rdi
+    jb      Skein_256_round_loop
+.endif # (SKEIN_ASM_UNROLL & 256) == 0
+    movq    ctxPtr +F_O(%rbp),%rdi           #restore rdi --> context
+
+    #----------------------------
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..3}
+    movq    $FIRST_MASK64 ,%r14
+    xorq    Wcopy + 0+F_O (%rbp),%rax
+    xorq    Wcopy + 8+F_O (%rbp),%rbx
+    xorq    Wcopy +16+F_O (%rbp),%rcx
+    xorq    Wcopy +24+F_O (%rbp),%rdx
+    andq    TWEAK + 8     (%rdi),%r14
+    movq    %rax,X_VARS+ 0(%rdi)             #store final result
+    movq    %rbx,X_VARS+ 8(%rdi)        
+    movq    %rcx,X_VARS+16(%rdi)        
+    movq    %rdx,X_VARS+24(%rdi)        
+
+    Skein_Debug_Round 256,SKEIN_RND_FEED_FWD
+
+    # go back for more blocks, if needed
+    decq    blkCnt+F_O(%rbp)
+    jnz     Skein_256_block_loop
+    movq    %r14,TWEAK + 8(%rdi)
+    Reset_Stack
+    ret
+Skein_256_Process_Block_End:
+
+  .if _SKEIN_DEBUG
+Skein_Debug_Round_256:               #here with rdx == round "number" from macro
+    pushq   %rsi                     #save two regs for BLK_BITS-specific parms
+    pushq   %rdi
+    movq    24(%rsp),%rdi            #get back original rdx (pushed on stack in macro call) to rdi
+    movq    %rax,X_stk+ 0+F_O(%rbp)  #save X[] state on stack so debug routines can access it
+    movq    %rbx,X_stk+ 8+F_O(%rbp)  #(use FP_ since rsp has changed!)
+    movq    %rcx,X_stk+16+F_O(%rbp)
+    movq    %rdi,X_stk+24+F_O(%rbp)
+
+    movq    ctxPtr+F_O(%rbp),%rsi    #ctx_hdr_ptr
+    movq    $256,%rdi                #now <rdi,rsi,rdx> are set for the call
+    jmp     Skein_Debug_Round_Common
+  .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label  Skein_256_Process_Block_CodeSize
+    movq    $(Skein_256_Process_Block_End-Skein_256_Process_Block),%rax
+    ret
+#
+C_label Skein_256_Unroll_Cnt
+  .if _UNROLL_CNT <> ROUNDS_256/8
+    movq    $_UNROLL_CNT,%rax
+  .else
+    xorq    %rax,%rax
+  .endif
+    ret
+.endif
+#
+.endif #_USE_ASM_ & 256
+#
+#=================================== Skein_512 =============================================
+#
+.if _USE_ASM_ & 512
+#
+# void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)
+#
+# X[i] == %r[8+i]          #register assignments for X[] values during rounds (i=0..7)
+#
+#################
+# MACRO: one round for 512-bit blocks
+#
+.macro R_512_OneRound rn0,rn1,rn2,rn3,rn4,rn5,rn6,rn7,_Rn_,op1,op2,op3,op4
+#
+    addReg      r\rn0, r\rn1
+    RotL64      r\rn1, 512,%((_Rn_) % 8),0
+    xorReg      r\rn1, r\rn0
+            op1
+    addReg      r\rn2, r\rn3
+    RotL64      r\rn3, 512,%((_Rn_) % 8),1
+    xorReg      r\rn3, r\rn2
+            op2
+    addReg      r\rn4, r\rn5
+    RotL64      r\rn5, 512,%((_Rn_) % 8),2
+    xorReg      r\rn5, r\rn4
+            op3
+    addReg      r\rn6, r\rn7
+    RotL64      r\rn7, 512,%((_Rn_) % 8),3
+    xorReg      r\rn7, r\rn6
+            op4
+    Skein_Debug_Round 512,%(_Rn_+1),-4
+#
+.endm #R_512_OneRound
+#
+#################
+# MACRO: eight rounds for 512-bit blocks
+#
+.macro R_512_FourRounds _RR_    #RR = base round number (0 % 8)
+  .if (SKEIN_ASM_UNROLL && 512)
+    # here for fully unrolled case.
+    _II_ = ((_RR_)/4) + 1       #key injection counter
+    R_512_OneRound  8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*(((_II_)+3) % 9)+F_O(%rbp),%rax>,,<movq ksKey+8*(((_II_)+4) % 9)+F_O(%rbp),%rbx>
+    R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*(((_II_)+5) % 9)+F_O(%rbp),%rcx>,,<movq ksKey+8*(((_II_)+6) % 9)+F_O(%rbp),%rdx>
+    R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*(((_II_)+7) % 9)+F_O(%rbp),%rsi>,,<addq ksTwk+8*(((_II_)+0) % 3)+F_O(%rbp),%rcx>
+    R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<addq ksTwk+8*(((_II_)+1) % 3)+F_O(%rbp),%rdx>,
+    # inject the key schedule
+    addq    ksKey+8*(((_II_)+0)%9)+F_O(%rbp),%r8
+    addReg   r11, rax
+    addq    ksKey+8*(((_II_)+1)%9)+F_O(%rbp),%r9
+    addReg   r12, rbx
+    addq    ksKey+8*(((_II_)+2)%9)+F_O(%rbp),%r10
+    addReg   r13, rcx
+    addReg   r14, rdx
+    addReg   r15, rsi,,,(_II_)
+  .else
+    # here for looping case                                                    #"rotate" key/tweak schedule (move up on stack)
+    incq    %rdi                 #bump key injection counter
+    R_512_OneRound  8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*6+F_O(%rbp,%rdi,8),%rdx>,<movq      ksTwk-8*1+F_O(%rbp,%rdi,8),%rax>,<movq      ksKey-8*1+F_O(%rbp,%rdi,8),%rsi>
+    R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*5+F_O(%rbp,%rdi,8),%rcx>,<movq %rax,ksTwk+8*2+F_O(%rbp,%rdi,8)     >,<movq %rsi,ksKey+8*8+F_O(%rbp,%rdi,8)>
+    R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*4+F_O(%rbp,%rdi,8),%rbx>,<addq      ksTwk+8*1+F_O(%rbp,%rdi,8),%rdx>,<movq      ksKey+8*7+F_O(%rbp,%rdi,8),%rsi>    
+    R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<movq ksKey+8*3+F_O(%rbp,%rdi,8),%rax>,<addq      ksTwk+8*0+F_O(%rbp,%rdi,8),%rcx>
+    # inject the key schedule
+    addq    ksKey+8*0+F_O(%rbp,%rdi,8),%r8
+    addReg   r11, rax
+    addReg   r12, rbx
+    addq    ksKey+8*1+F_O(%rbp,%rdi,8),%r9
+    addReg   r13, rcx
+    addReg   r14, rdx
+    addq    ksKey+8*2+F_O(%rbp,%rdi,8),%r10
+    addReg   r15, rsi
+    addReg   r15, rdi              #inject the round number
+  .endif
+
+    #show the result of the key injection
+    Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT
+.endm #R_512_EightRounds
+#
+#################
+# instantiated code
+#
+C_label Skein_512_Process_Block
+    Setup_Stack 512,ROUNDS_512/8
+    movq    TWEAK+ 8(%rdi),%rbx
+    jmp     Skein_512_block_loop
+    .p2align 4
+    # main hash loop for Skein_512
+Skein_512_block_loop:
+    # general register usage:
+    #   RAX..RDX       = temps for key schedule pre-loads
+    #   R8 ..R15       = X0..X7
+    #   RSP, RBP       = stack/frame pointers
+    #   RDI            = round counter or context pointer
+    #   RSI            = temp
+    #
+    movq    TWEAK +  0(%rdi),%rax
+    addq    bitAdd+F_O(%rbp),%rax     #computed updated tweak value T0
+    movq    %rbx,%rcx
+    xorq    %rax,%rcx                 #%rax/%rbx/%rcx = tweak schedule
+    movq    %rax,TWEAK+ 0    (%rdi)   #save updated tweak value ctx->h.T[0]
+    movq    %rax,ksTwk+ 0+F_O(%rbp)
+    movq    $KW_PARITY,%rdx
+    movq    blkPtr +F_O(%rbp),%rsi    #%rsi --> input block
+    movq    %rbx,ksTwk+ 8+F_O(%rbp)
+    movq    %rcx,ksTwk+16+F_O(%rbp)
+    .irp _Rn_,8,9,10,11,12,13,14,15
+      movq  X_VARS+8*(_Rn_-8)(%rdi),%r\_Rn_
+      xorq  %r\_Rn_,%rdx              #compute overall parity
+      movq  %r\_Rn_,ksKey+8*(_Rn_-8)+F_O(%rbp)
+    .endr                             #load state into %r8 ..%r15, compute parity
+      movq  %rdx,ksKey+8*(8)+F_O(%rbp)#save key schedule parity
+
+    addReg   r13,rax                  #precompute key injection for tweak
+    addReg   r14, rbx
+.if _SKEIN_DEBUG
+    movq    %rbx,TWEAK+ 8(%rdi)       #save updated tweak value ctx->h.T[1] for Skein_Debug_Block below
+.endif
+    movq     0(%rsi),%rax             #load input block
+    movq     8(%rsi),%rbx 
+    movq    16(%rsi),%rcx 
+    movq    24(%rsi),%rdx 
+    addReg   r8 , rax                 #do initial key injection
+    addReg   r9 , rbx
+    movq    %rax,Wcopy+ 0+F_O(%rbp)   #keep local copy for feedforward
+    movq    %rbx,Wcopy+ 8+F_O(%rbp)
+    addReg   r10, rcx
+    addReg   r11, rdx
+    movq    %rcx,Wcopy+16+F_O(%rbp)
+    movq    %rdx,Wcopy+24+F_O(%rbp)
+
+    movq    32(%rsi),%rax
+    movq    40(%rsi),%rbx 
+    movq    48(%rsi),%rcx 
+    movq    56(%rsi),%rdx
+    addReg   r12, rax
+    addReg   r13, rbx
+    addReg   r14, rcx
+    addReg   r15, rdx
+    movq    %rax,Wcopy+32+F_O(%rbp)    
+    movq    %rbx,Wcopy+40+F_O(%rbp)    
+    movq    %rcx,Wcopy+48+F_O(%rbp)    
+    movq    %rdx,Wcopy+56+F_O(%rbp)    
+
+.if _SKEIN_DEBUG
+    .irp _Rn_,8,9,10,11,12,13,14,15   #save values on stack for debug output
+      movq  %r\_Rn_,X_stk+8*(_Rn_-8)(%rsp)
+    .endr
+
+    Skein_Debug_Block 512             #debug dump
+    Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL
+.endif
+    addq    $8*WCNT,%rsi              #skip the block
+    movq    %rsi,blkPtr+F_O(%rbp)     #update block pointer
+    #
+    #################
+    # now the key schedule is computed. Start the rounds
+    #
+.if SKEIN_ASM_UNROLL & 512
+_UNROLL_CNT =   ROUNDS_512/8
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_512
+  .if ((ROUNDS_512/8) % _UNROLL_CNT)
+    .err "Invalid SKEIN_UNROLL_512"
+  .endif
+    xorq    %rdi,%rdi                 #rdi = round counter
+Skein_512_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+      R_512_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 512) == 0
+    cmpq    $2*(ROUNDS_512/8),%rdi
+    jb      Skein_512_round_loop
+    movq    ctxPtr +F_O(%rbp),%rdi           #restore rdi --> context
+.endif
+    # end of rounds
+    #################
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..7}
+    .irp _Rn_,8,9,10,11,12,13,14,15
+  .if (_Rn_ == 8)
+    movq    $FIRST_MASK64,%rbx
+  .endif
+      xorq  Wcopy+8*(_Rn_-8)+F_O(%rbp),%r\_Rn_  #feedforward XOR
+      movq  %r\_Rn_,X_VARS+8*(_Rn_-8)(%rdi)     #and store result
+  .if (_Rn_ == 14)
+    andq    TWEAK+ 8(%rdi),%rbx
+  .endif
+    .endr
+    Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+
+    # go back for more blocks, if needed
+    decq    blkCnt+F_O(%rbp)
+    jnz     Skein_512_block_loop
+    movq    %rbx,TWEAK + 8(%rdi)
+
+    Reset_Stack
+    ret
+Skein_512_Process_Block_End:
+#
+  .if _SKEIN_DEBUG
+# call here with rdx  = "round number"
+Skein_Debug_Round_512:
+    pushq   %rsi                     #save two regs for BLK_BITS-specific parms
+    pushq   %rdi
+  .irp _Rn_,8,9,10,11,12,13,14,15    #save X[] state on stack so debug routines can access it
+    movq    %r\_Rn_,X_stk+8*(_Rn_-8)+F_O(%rbp)
+  .endr
+    movq    ctxPtr+F_O(%rbp),%rsi    #ctx_hdr_ptr
+    movq    $512,%rdi                #now <rdi,rsi,rdx> are set for the call
+    jmp     Skein_Debug_Round_Common
+  .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein_512_Process_Block_CodeSize
+    movq    $(Skein_512_Process_Block_End-Skein_512_Process_Block),%rax
+    ret
+#
+C_label Skein_512_Unroll_Cnt
+  .if _UNROLL_CNT <> (ROUNDS_512/8)
+    movq    $_UNROLL_CNT,%rax
+  .else
+    xorq    %rax,%rax
+  .endif
+    ret
+.endif
+#
+.endif # _USE_ASM_ & 512
+#
+#=================================== Skein1024 =============================================
+.if _USE_ASM_ & 1024
+#
+# void Skein1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+# use details of permutation to make register assignments
+# 
+o1K_rdi =  0        #offsets in X[] associated with each register
+o1K_rsi =  1 
+o1K_rbp =  2 
+o1K_rax =  3 
+o1K_rcx =  4        #rcx is "shared" with X6, since X4/X6 alternate
+o1K_rbx =  5 
+o1K_rdx =  7 
+o1K_r8  =  8  
+o1K_r9  =  9  
+o1K_r10 = 10
+o1K_r11 = 11
+o1K_r12 = 12
+o1K_r13 = 13
+o1K_r14 = 14
+o1K_r15 = 15
+#
+rIdx_offs = tmpStk_1024
+#
+.macro r1024_Mix w0,w1,reg0,reg1,_RN0_,_Rn1_,op1
+    addReg      \reg0 , \reg1                      #perform the MIX
+    RotL64      \reg1 , 1024,%((_RN0_) % 8),_Rn1_
+    xorReg      \reg1 , \reg0
+.if ((_RN0_) && 3) == 3         #time to do key injection?
+ .if _SKEIN_DEBUG
+    movq       %\reg0 , xDebug_1024+8*w0(%rsp)     #save intermediate values for Debug_Round
+    movq       %\reg1 , xDebug_1024+8*w1(%rsp)     # (before inline key injection)
+ .endif
+_II_ = ((_RN0_)/4)+1            #injection count
+ .if SKEIN_ASM_UNROLL && 1024   #here to do fully unrolled key injection
+    addq        ksKey+ 8*((_II_+w0) % 17)(%rsp),%\reg0
+    addq        ksKey+ 8*((_II_+w1) % 17)(%rsp),%\reg1
+  .if     w1 == 13                                 #tweak injection
+    addq        ksTwk+ 8*((_II_+ 0) %  3)(%rsp),%\reg1
+  .elseif w0 == 14
+    addq        ksTwk+ 8*((_II_+ 1) %  3)(%rsp),%\reg0
+  .elseif w1 == 15
+    addq        $_II_, %\reg1                      #(injection counter)
+  .endif
+ .else                          #here to do looping  key injection
+  .if  (w0 == 0)
+    movq        %rdi, X_stk+8*w0(%rsp)             #if so, store N0 so we can use reg as index
+    movq         rIdx_offs(%rsp),%rdi              #get the injection counter index into rdi
+  .else
+    addq         ksKey+8+8*w0(%rsp,%rdi,8),%\reg0  #even key injection
+  .endif
+  .if     w1 == 13                                 #tweak injection
+    addq         ksTwk+8+8* 0(%rsp,%rdi,8),%\reg1  
+  .elseif w0 == 14
+    addq         ksTwk+8+8* 1(%rsp,%rdi,8),%\reg0  
+  .elseif w1 == 15
+    addReg      \reg1,rdi,,,1                      #(injection counter)
+  .endif
+    addq         ksKey+8+8*w1(%rsp,%rdi,8),%\reg1  #odd key injection
+ .endif
+.endif
+    # insert the op provided, .if any
+    op1
+.endm
+#################
+# MACRO: four rounds for 1024-bit blocks
+#
+.macro r1024_FourRounds _RR_    #RR = base round number (0 mod 4)
+    # should be here with X4 set properly, X6 stored on stack
+_Rn_ = (_RR_) + 0
+        r1024_Mix  0, 1,rdi,rsi,_Rn_,0
+        r1024_Mix  2, 3,rbp,rax,_Rn_,1
+        r1024_Mix  4, 5,rcx,rbx,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)>       #save X4  on  stack (x4/x6 alternate)
+        r1024_Mix  8, 9,r8 ,r9 ,_Rn_,4,<movq      X_stk+8*6(%rsp),%rcx>  #load X6 from stack 
+        r1024_Mix 10,11,r10,r11,_Rn_,5
+        r1024_Mix 12,13,r12,r13,_Rn_,6
+        r1024_Mix  6, 7,rcx,rdx,_Rn_,3
+        r1024_Mix 14,15,r14,r15,_Rn_,7
+    .if _SKEIN_DEBUG
+      Skein_Debug_Round 1024,%(_Rn_+1)
+    .endif
+_Rn_ = (_RR_) + 1
+        r1024_Mix  0, 9,rdi,r9 ,_Rn_,0
+        r1024_Mix  2,13,rbp,r13,_Rn_,1
+        r1024_Mix  6,11,rcx,r11,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)>       #save X6  on  stack (x4/x6 alternate)
+        r1024_Mix 10, 7,r10,rdx,_Rn_,4,<movq      X_stk+8*4(%rsp),%rcx>  #load X4 from stack 
+        r1024_Mix 12, 3,r12,rax,_Rn_,5
+        r1024_Mix 14, 5,r14,rbx,_Rn_,6
+        r1024_Mix  4,15,rcx,r15,_Rn_,3
+        r1024_Mix  8, 1,r8 ,rsi,_Rn_,7
+    .if _SKEIN_DEBUG
+      Skein_Debug_Round 1024,%(_Rn_+1)
+    .endif
+_Rn_ = (_RR_) + 2
+        r1024_Mix  0, 7,rdi,rdx,_Rn_,0
+        r1024_Mix  2, 5,rbp,rbx,_Rn_,1
+        r1024_Mix  4, 3,rcx,rax,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)>       #save X4  on  stack (x4/x6 alternate)
+        r1024_Mix 12,15,r12,r15,_Rn_,4,<movq      X_stk+8*6(%rsp),%rcx>  #load X6 from stack 
+        r1024_Mix 14,13,r14,r13,_Rn_,5
+        r1024_Mix  8,11,r8 ,r11,_Rn_,6
+        r1024_Mix  6, 1,rcx,rsi,_Rn_,3
+        r1024_Mix 10, 9,r10,r9 ,_Rn_,7
+    .if _SKEIN_DEBUG
+      Skein_Debug_Round 1024,%(_Rn_+1)
+    .endif
+_Rn_ = (_RR_) + 3
+        r1024_Mix  0,15,rdi,r15,_Rn_,0
+        r1024_Mix  2,11,rbp,r11,_Rn_,1
+        r1024_Mix  6,13,rcx,r13,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)>       #save X6  on  stack (x4/x6 alternate)
+        r1024_Mix 14, 1,r14,rsi,_Rn_,4,<movq      X_stk+8*4(%rsp),%rcx>  #load X4 from stack 
+        r1024_Mix  8, 5,r8 ,rbx,_Rn_,5
+        r1024_Mix 10, 3,r10,rax,_Rn_,6
+        r1024_Mix  4, 9,rcx,r9 ,_Rn_,3
+        r1024_Mix 12, 7,r12,rdx,_Rn_,7
+    .if _SKEIN_DEBUG
+      Skein_Debug_Round 1024,%(_Rn_+1)
+    .endif
+
+  .if (SKEIN_ASM_UNROLL && 1024) == 0           #here with rdi == rIdx, X0 on stack
+    #"rotate" the key schedule on the stack
+i8 = o1K_r8
+i0 = o1K_rdi
+    movq    %r8 , X_stk+8*i8(%rsp)              #free up a register (save it on the stack)
+    movq          ksKey+8* 0(%rsp,%rdi,8),%r8   #get  key  word
+    movq    %r8 , ksKey+8*17(%rsp,%rdi,8)       #rotate key (must do key first or tweak clobbers it!)
+    movq          ksTwk+8* 0(%rsp,%rdi,8),%r8   #get tweak word
+    movq    %r8 , ksTwk+8* 3(%rsp,%rdi,8)       #rotate tweak (onto the stack)
+    movq          X_stk+8*i8(%rsp)       ,%r8   #get the reg back
+    incq    %rdi                                #bump the index
+    movq    %rdi, rIdx_offs (%rsp)              #save rdi again
+    movq          ksKey+8*i0(%rsp,%rdi,8),%rdi  #get the key schedule word for X0 back
+    addq          X_stk+8*i0(%rsp)       ,%rdi  #perform the X0 key injection
+  .endif
+    #show the result of the key injection
+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT
+.endm #r1024_FourRounds
+#
+################
+# code
+#
+C_label Skein1024_Process_Block
+#
+    Setup_Stack 1024,ROUNDS_1024/8,WCNT
+    movq    TWEAK+ 8(%rdi),%r9
+    jmp     Skein1024_block_loop
+    # main hash loop for Skein1024
+    .p2align 4
+Skein1024_block_loop:
+    # general register usage:
+    #   RSP              = stack pointer
+    #   RAX..RDX,RSI,RDI = X1, X3..X7 (state words)
+    #   R8 ..R15         = X8..X15    (state words)
+    #   RBP              = temp (used for X0 and X2)
+    #
+  .if (SKEIN_ASM_UNROLL & 1024) == 0
+    xorq    %rax,%rax                      #init loop index on the stack
+    movq    %rax,rIdx_offs(%rsp)
+  .endif
+    movq         TWEAK+     0(%rdi),%r8
+    addq         bitAdd+  F_O(%rbp),%r8    #computed updated tweak value T0
+    movq    %r9 ,%r10 
+    xorq    %r8 ,%r10                      #%rax/%rbx/%rcx = tweak schedule
+    movq    %r8 ,TWEAK+     0(%rdi)        #save updated tweak value ctx->h.T[0]
+    movq    %r8 ,ksTwk+ 0+F_O(%rbp)
+    movq    %r9 ,ksTwk+ 8+F_O(%rbp)        #keep values in %r8 ,%r9  for initial tweak injection below
+    movq    %r10,ksTwk+16+F_O(%rbp)
+  .if _SKEIN_DEBUG
+    movq    %r9 ,TWEAK+     8(%rdi)        #save updated tweak value ctx->h.T[1] for Skein_Debug_Block
+  .endif
+    movq         blkPtr +F_O(%rbp),%rsi    # rsi --> input block
+    movq        $KW_PARITY        ,%rax    #overall key schedule parity
+
+    # the logic here assumes the set {rdi,rsi,rbp,rax} = X[0,1,2,3]
+    .irp _rN_,0,1,2,3,4,6                  #process the "initial" words, using r14/r15 as temps
+      movq       X_VARS+8*_rN_(%rdi),%r14  #get state word
+      movq              8*_rN_(%rsi),%r15  #get msg   word
+      xorq  %r14,%rax                      #update key schedule overall parity
+      movq  %r14,ksKey +8*_rN_+F_O(%rbp)   #save key schedule word on stack
+      movq  %r15,Wcopy +8*_rN_+F_O(%rbp)   #save local msg Wcopy 
+      addq  %r15,%r14                      #do the initial key injection
+      movq  %r14,X_stk +8*_rN_    (%rsp)   #save initial state var on stack
+    .endr
+    # now process the rest, using the "real" registers 
+    #     (MUST do it in reverse order to inject tweaks r8/r9 first)
+    .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rdx,rbx
+_oo_ = o1K_\_rr_                           #offset assocated with the register
+      movq  X_VARS+8*_oo_(%rdi),%\_rr_     #get key schedule word from context
+      movq         8*_oo_(%rsi),%rcx       #get next input msg word
+      movq  %\_rr_, ksKey +8*_oo_(%rsp)    #save key schedule on stack
+      xorq  %\_rr_, %rax                   #accumulate key schedule parity
+      movq  %rcx,Wcopy+8*_oo_+F_O(%rbp)    #save copy of msg word for feedforward
+      addq  %rcx,%\_rr_                    #do the initial  key  injection
+      .if    _oo_ == 13                    #do the initial tweak injection
+        addReg _rr_,r8                     #          (only in words 13/14)
+      .elseif _oo_ == 14
+        addReg _rr_,r9 
+      .endif
+    .endr
+    movq    %rax,ksKey+8*WCNT+F_O(%rbp)    #save key schedule parity
+.if _SKEIN_DEBUG
+    Skein_Debug_Block 1024                 #initial debug dump
+.endif
+    addq     $8*WCNT,%rsi                  #bump the msg ptr
+    movq     %rsi,blkPtr+F_O(%rbp)         #save bumped msg ptr
+    # re-load words 0..4 from stack, enter the main loop
+    .irp _rr_,rdi,rsi,rbp,rax,rcx          #(no need to re-load x6, already on stack)
+      movq  X_stk+8*o1K_\_rr_(%rsp),%\_rr_ #re-load state and get ready to go!
+    .endr
+.if _SKEIN_DEBUG
+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL        #show state after initial key injection
+.endif
+    #
+    #################
+    # now the key schedule is computed. Start the rounds
+    #
+.if SKEIN_ASM_UNROLL & 1024
+_UNROLL_CNT =   ROUNDS_1024/8
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_1024
+  .if ((ROUNDS_1024/8) % _UNROLL_CNT)
+    .error "Invalid SKEIN_UNROLL_1024"
+  .endif
+Skein1024_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2                        #implement the rounds, 4 at a time
+      r1024_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 1024) == 0
+    cmpq    $2*(ROUNDS_1024/8),tmpStk_1024(%rsp) #see .if we are done
+    jb      Skein1024_round_loop    
+.endif
+    # end of rounds
+    #################
+    #
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..15}
+    movq    %rdx,X_stk+8*o1K_rdx(%rsp) #we need a register. x6 already on stack
+    movq       ctxPtr(%rsp),%rdx
+    
+    .irp _rr_,rdi,rsi,rbp,rax,rcx,rbx,r8,r9,r10,r11,r12,r13,r14,r15   #do all but x6,x7
+_oo_ = o1K_\_rr_
+      xorq  Wcopy +8*_oo_(%rsp),%\_rr_ #feedforward XOR
+      movq  %\_rr_,X_VARS+8*_oo_(%rdx) #save result into context
+      .if (_oo_ ==  9)
+        movq   $FIRST_MASK64 ,%r9
+      .endif
+      .if (_oo_ == 14)
+        andq   TWEAK+ 8(%rdx),%r9
+      .endif
+    .endr
+    # 
+    movq         X_stk +8*6(%rsp),%rax #now process x6,x7 (skipped in .irp above)
+    movq         X_stk +8*7(%rsp),%rbx
+    xorq         Wcopy +8*6(%rsp),%rax
+    xorq         Wcopy +8*7(%rsp),%rbx
+    movq    %rax,X_VARS+8*6(%rdx)
+    decq             blkCnt(%rsp)      #set zero flag iff done
+    movq    %rbx,X_VARS+8*7(%rdx)
+
+    Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD,,<cmpq $0,blkCnt(%rsp)>
+    # go back for more blocks, if needed
+    movq             ctxPtr(%rsp),%rdi #don't muck with the flags here!
+    lea          FRAME_OFFS(%rsp),%rbp
+    jnz     Skein1024_block_loop
+    movq    %r9 ,TWEAK+   8(%rdx)
+    Reset_Stack
+    ret
+#
+Skein1024_Process_Block_End:
+#
+.if _SKEIN_DEBUG
+Skein_Debug_Round_1024:
+    # call here with rdx  = "round number",
+_SP_OFFS_ = 8*2                     #stack "offset" here: rdx, return addr
+    #
+  #save rest of X[] state on stack so debug routines can access it
+  .irp _rr_,rsi,rbp,rax,rbx,r8,r9,r10,r11,r12,r13,r14,r15
+    movq    %\_rr_,X_stk+8*o1K_\_rr_+_SP_OFFS_(%rsp)
+  .endr
+    # Figure out what to do with x0 (rdi).  When rdx == 0 mod 4, it's already on stack
+    cmpq    $SKEIN_RND_SPECIAL,%rdx #special rounds always save
+    jae     save_x0
+    testq   $3,%rdx                 #otherwise only if rdx != 0 mod 4
+    jz      save_x0_not
+save_x0:
+    movq    %rdi,X_stk+8*o1K_rdi+_SP_OFFS_(%rsp)
+save_x0_not:
+    #figure out the x4/x6 swapping state and save the correct one!
+    cmpq    $SKEIN_RND_SPECIAL,%rdx #special rounds always do x4
+    jae     save_x4
+    testq   $1,%rdx                  #and even ones have r4 as well
+    jz      save_x4
+    movq    %rcx,X_stk+8*6+_SP_OFFS_(%rsp)
+    jmp     debug_1024_go
+save_x4:
+    movq    %rcx,X_stk+8*4+_SP_OFFS_(%rsp)
+debug_1024_go:
+    #now all is saved in Xstk[] except for rdx
+    push    %rsi                    #save two regs for BLK_BITS-specific parms
+    push    %rdi
+_SP_OFFS_ = _SP_OFFS_ + 16          #adjust stack offset accordingly (now 32)
+
+    movq    _SP_OFFS_-8(%rsp),%rsi  #get back original %rdx (pushed on stack in macro call)
+    movq    %rsi,X_stk+8*o1K_rdx+_SP_OFFS_(%rsp) #and save it in its rightful place in X_stk[]
+
+    movq    ctxPtr+_SP_OFFS_(%rsp),%rsi  #rsi = ctx_hdr_ptr
+    movq    $1024,%rdi                   #rdi = block size
+    jmp     Skein_Debug_Round_Common
+.endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein1024_Process_Block_CodeSize
+    movq    $(Skein1024_Process_Block_End-Skein1024_Process_Block),%rax
+    ret
+#
+C_label Skein1024_Unroll_Cnt
+  .if _UNROLL_CNT <> (ROUNDS_1024/8)
+    movq    $_UNROLL_CNT,%rax
+  .else
+    xorq    %rax,%rax
+  .endif
+    ret
+.endif
+#
+.endif # _USE_ASM_ and 1024
+#
+.if _SKEIN_DEBUG
+#----------------------------------------------------------------
+#local debug routine to set up for calls to:
+#  void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X)
+#                       [       rdi                        rsi   rdx              rcx]
+#
+# here with %rdx = round number
+#           %rsi = ctx_hdr_ptr
+#           %rdi = block size (256/512/1024)
+# on stack: saved rdi, saved rsi, retAddr, saved rdx  
+#
+Skein_Debug_Round_Common:
+_SP_OFFS_ = 32                        #account for four words on stack already
+  .irp _rr_,rax,rbx,rcx,rbp,r8,r9,r10,r11,r12,r13,r14,r15  #save the rest of the regs
+    pushq %\_rr_
+_SP_OFFS_ = _SP_OFFS_+8
+  .endr
+  .if (_SP_OFFS_ % 16)                # make sure stack is still 16-byte aligned here
+    .error  "Debug_Round_Common: stack alignment"
+  .endif
+    # compute %rcx  = ptr to the X[] array on the stack (final parameter to call)
+    leaq    X_stk+_SP_OFFS_(%rsp),%rcx #adjust for reg pushes, return address
+    cmpq    $SKEIN_RND_FEED_FWD,%rdx   #special handling for feedforward "round"?
+    jnz     _got_rcxA
+    leaq    X_VARS(%rsi),%rcx
+_got_rcxA:
+  .if _USE_ASM_ & 1024
+    # special handling for 1024-bit case
+    #    (for rounds right before with key injection: 
+    #        use xDebug_1024[] instead of X_stk[])
+    cmpq    $SKEIN_RND_SPECIAL,%rdx
+    jae     _got_rcxB               #must be a normal round
+    orq     %rdx,%rdx
+    jz      _got_rcxB               #just before key injection
+    test    $3,%rdx
+    jne     _got_rcxB
+    cmp     $1024,%rdi              #only 1024-bit(s) for now
+    jne     _got_rcxB
+    leaq    xDebug_1024+_SP_OFFS_(%rsp),%rcx
+_got_rcxB:
+  .endif
+    call    Skein_Show_Round        #call external debug handler
+
+  .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rbp,rcx,rbx,rax  #restore regs
+    popq  %\_rr_
+_SP_OFFS_ = _SP_OFFS_-8
+  .endr
+  .if _SP_OFFS_ - 32
+    .error   "Debug_Round_Common: push/pop misalignment!"
+  .endif    
+    popq    %rdi
+    popq    %rsi
+    ret
+.endif
+#----------------------------------------------------------------
+    .end
diff --git a/Additional_Implementations/skein_block_x86.asm b/Additional_Implementations/skein_block_x86.asm
new file mode 100644
index 000000000000..4679e991fe04
--- /dev/null
+++ b/Additional_Implementations/skein_block_x86.asm
@@ -0,0 +1,1180 @@
+;

+;----------------------------------------------------------------

+; 32-bit x86 assembler code for Skein block functions

+;

+; Author: Doug Whiting, Hifn

+;

+; This code is released to the public domain.

+;----------------------------------------------------------------

+;

+    .386p

+    .model flat

+    .code

+;

+_MASK_ALL_  equ (256+512+1024)              ;all three algorithm bits

+;

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_USE_ASM

+_USE_ASM_        = _MASK_ALL_

+elseif SKEIN_USE_ASM and _MASK_ALL_

+_USE_ASM_        = SKEIN_USE_ASM

+else

+_USE_ASM_        = _MASK_ALL_

+endif

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_LOOP  

+_SKEIN_LOOP       = 0                       ;default is all fully unrolled

+else

+_SKEIN_LOOP       = SKEIN_LOOP

+endif

+; the unroll counts (0 --> fully unrolled)

+SKEIN_UNROLL_256  = (_SKEIN_LOOP / 100) mod 10

+SKEIN_UNROLL_512  = (_SKEIN_LOOP /  10) mod 10

+SKEIN_UNROLL_1024 = (_SKEIN_LOOP      ) mod 10

+;

+SKEIN_ASM_UNROLL  = 0

+  irp _NN_,<256,512,1024>

+    if (SKEIN_UNROLL_&_NN_) eq 0

+SKEIN_ASM_UNROLL  = SKEIN_ASM_UNROLL + _NN_

+    endif

+  endm

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_ROUNDS

+ROUNDS_256  =   72

+ROUNDS_512  =   72

+ROUNDS_1024 =   80

+else

+ROUNDS_256  = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)

+ROUNDS_512  = 8*((((SKEIN_ROUNDS /  10) + 5) mod 10) + 5)

+ROUNDS_1024 = 8*((((SKEIN_ROUNDS      ) + 5) mod 10) + 5)

+endif

+irp _NN_,<256,512,1024>

+  if _USE_ASM_ and _NN_

+    irp _RR_,<%(ROUNDS_&_NN_)>

+      if _NN_ eq 1024

+%out  +++ SKEIN_ROUNDS_&_NN_ = _RR_

+      else

+%out  +++ SKEIN_ROUNDS_&_NN_  = _RR_

+      endif

+    endm

+  endif

+endm

+;;;;;;;;;;;;;;;;;

+;

+ifdef SKEIN_CODE_SIZE

+_SKEIN_CODE_SIZE equ (1)

+else

+ifdef  SKEIN_PERF                           ;use code size if SKEIN_PERF is defined

+_SKEIN_CODE_SIZE equ (1)

+endif

+endif

+;

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_DEBUG

+_SKEIN_DEBUG      = 0

+else

+_SKEIN_DEBUG      = 1

+endif

+;;;;;;;;;;;;;;;;;

+;

+; define offsets of fields in hash context structure

+;

+HASH_BITS   =   0                           ;# bits of hash output

+BCNT        =   4 + HASH_BITS               ;number of bytes in BUFFER[]

+TWEAK       =   4 + BCNT                    ;tweak values[0..1]

+X_VARS      =  16 + TWEAK                   ;chaining vars

+;

+;(Note: buffer[] in context structure is NOT needed here :-)

+;

+KW_PARITY_LO=   0A9FC1A22h                  ;overall parity of key schedule words (hi32/lo32)

+KW_PARITY_HI=   01BD11BDAh                  ;overall parity of key schedule words (hi32/lo32)

+FIRST_MASK  =   NOT (1 SHL 30)              ;FIRST block flag bit

+;

+; rotation constants for Skein

+;

+RC_256_0_0  = 14

+RC_256_0_1  = 16

+

+RC_256_1_0  = 52

+RC_256_1_1  = 57

+

+RC_256_2_0  = 23

+RC_256_2_1  = 40

+

+RC_256_3_0  =  5

+RC_256_3_1  = 37

+

+RC_256_4_0  = 25

+RC_256_4_1  = 33

+

+RC_256_5_0  = 46

+RC_256_5_1  = 12

+

+RC_256_6_0  = 58

+RC_256_6_1  = 22

+

+RC_256_7_0  = 32

+RC_256_7_1  = 32

+

+RC_512_0_0  = 46

+RC_512_0_1  = 36

+RC_512_0_2  = 19

+RC_512_0_3  = 37

+

+RC_512_1_0  = 33

+RC_512_1_1  = 27

+RC_512_1_2  = 14

+RC_512_1_3  = 42

+

+RC_512_2_0  = 17

+RC_512_2_1  = 49

+RC_512_2_2  = 36

+RC_512_2_3  = 39

+

+RC_512_3_0  = 44

+RC_512_3_1  =  9

+RC_512_3_2  = 54

+RC_512_3_3  = 56

+

+RC_512_4_0  = 39

+RC_512_4_1  = 30

+RC_512_4_2  = 34

+RC_512_4_3  = 24

+

+RC_512_5_0  = 13

+RC_512_5_1  = 50

+RC_512_5_2  = 10

+RC_512_5_3  = 17

+

+RC_512_6_0  = 25

+RC_512_6_1  = 29

+RC_512_6_2  = 39

+RC_512_6_3  = 43

+

+RC_512_7_0  =  8

+RC_512_7_1  = 35

+RC_512_7_2  = 56

+RC_512_7_3  = 22

+

+RC_1024_0_0 = 24

+RC_1024_0_1 = 13

+RC_1024_0_2 =  8

+RC_1024_0_3 = 47

+RC_1024_0_4 =  8

+RC_1024_0_5 = 17

+RC_1024_0_6 = 22

+RC_1024_0_7 = 37

+

+RC_1024_1_0 = 38

+RC_1024_1_1 = 19

+RC_1024_1_2 = 10

+RC_1024_1_3 = 55

+RC_1024_1_4 = 49

+RC_1024_1_5 = 18

+RC_1024_1_6 = 23

+RC_1024_1_7 = 52

+

+RC_1024_2_0 = 33

+RC_1024_2_1 =  4

+RC_1024_2_2 = 51

+RC_1024_2_3 = 13

+RC_1024_2_4 = 34

+RC_1024_2_5 = 41

+RC_1024_2_6 = 59

+RC_1024_2_7 = 17

+

+RC_1024_3_0 =  5

+RC_1024_3_1 = 20

+RC_1024_3_2 = 48

+RC_1024_3_3 = 41

+RC_1024_3_4 = 47

+RC_1024_3_5 = 28

+RC_1024_3_6 = 16

+RC_1024_3_7 = 25

+

+RC_1024_4_0 = 41

+RC_1024_4_1 =  9

+RC_1024_4_2 = 37

+RC_1024_4_3 = 31

+RC_1024_4_4 = 12

+RC_1024_4_5 = 47

+RC_1024_4_6 = 44

+RC_1024_4_7 = 30

+

+RC_1024_5_0 = 16

+RC_1024_5_1 = 34

+RC_1024_5_2 = 56

+RC_1024_5_3 = 51

+RC_1024_5_4 =  4

+RC_1024_5_5 = 53

+RC_1024_5_6 = 42

+RC_1024_5_7 = 41

+

+RC_1024_6_0 = 31

+RC_1024_6_1 = 44

+RC_1024_6_2 = 47

+RC_1024_6_3 = 46

+RC_1024_6_4 = 19

+RC_1024_6_5 = 42

+RC_1024_6_6 = 44

+RC_1024_6_7 = 25

+

+RC_1024_7_0 =  9

+RC_1024_7_1 = 48

+RC_1024_7_2 = 35

+RC_1024_7_3 = 52

+RC_1024_7_4 = 23

+RC_1024_7_5 = 31

+RC_1024_7_6 = 37

+RC_1024_7_7 = 20

+;

+;  Input:  rHi,rLo

+; Output: <rHi,rLo> <<< _RCNT_

+Rol64 macro rHi,rLo,tmp,_RCNT_

+  if _RCNT_  ;is there anything to do?

+    if _RCNT_ lt 32

+      mov   tmp,rLo

+      shld  rLo,rHi,_RCNT_

+      shld  rHi,tmp,_RCNT_

+    elseif _RCNT_ gt 32

+      mov   tmp,rLo

+      shrd  rLo,rHi,((64-_RCNT_) AND 63)

+      shrd  rHi,tmp,((64-_RCNT_) AND 63)

+    else    

+      xchg  rHi,rLo ;special case for _RCNT_ == 32

+    endif

+  endif

+endm

+;

+;  Input:  rHi,rLo

+; Output: <rHi,rLo> <<< rName&&rNum, and tmp trashed;

+RotL64 macro rHi,rLo,tmp,BLK_SIZE,ROUND_NUM,MIX_NUM

+_RCNT_ = ( RC_&BLK_SIZE&_&ROUND_NUM&_&MIX_NUM AND 63 )

+    Rol64 rHi,rLo,tmp,_RCNT_

+endm

+;

+;----------------------------------------------------------------

+; declare allocated space on the stack

+StackVar    macro localName,localSize

+localName   =   _STK_OFFS_

+_STK_OFFS_  =   _STK_OFFS_+(localSize)

+endm ;StackVar

+;

+;----------------------------------------------------------------

+;

+; MACRO: Configure stack frame, allocate local vars

+;

+Setup_Stack macro WCNT,KS_CNT

+_STK_OFFS_  =   0                   ;starting offset from esp 

+    ;----- local  variables         ;<-- esp

+    StackVar    X_stk  ,8*(WCNT)    ;local context vars

+    StackVar    Wcopy  ,8*(WCNT)    ;copy of input block    

+    StackVar    ksTwk  ,8*3         ;key schedule: tweak words

+    StackVar    ksKey  ,8*(WCNT)+8  ;key schedule: key   words

+  if WCNT le 8

+FRAME_OFFS  =   _STK_OFFS_          ;<-- ebp

+  else

+FRAME_OFFS  =   _STK_OFFS_-8*4      ;<-- ebp

+  endif

+  if (SKEIN_ASM_UNROLL and (WCNT*64)) eq 0

+    StackVar    ksRot ,16*(KS_CNT+0);leave space for "rotation" to happen

+  endif

+LOCAL_SIZE  =   _STK_OFFS_          ;size of local vars

+    ;----- 

+    StackVar    savRegs,8*4         ;pushad data

+    StackVar    retAddr,4           ;return address

+    ;----- caller parameters

+    StackVar    ctxPtr ,4           ;context ptr

+    StackVar    blkPtr ,4           ;pointer to block data

+    StackVar    blkCnt ,4           ;number of full blocks to process

+    StackVar    bitAdd ,4           ;bit count to add to tweak

+    ;----- caller's stack frame

+;

+; Notes on stack frame setup:

+;   * the most frequently used variable is X_stk[], based at [esp+0]

+;   * the next most used is the key schedule words

+;       so ebp is "centered" there, allowing short offsets to the key/tweak

+;       schedule even in 1024-bit Skein case

+;   * the Wcopy variables are infrequently accessed, but they have long 

+;       offsets from both esp and ebp only in the 1024-bit case.

+;   * all other local vars and calling parameters can be accessed 

+;       with short offsets, except in the 1024-bit case

+;

+    pushad                          ;save all regs

+    sub     esp,LOCAL_SIZE          ;make room for the locals

+    lea     ebp,[esp+FRAME_OFFS]    ;maximize use of short offsets

+    mov     edi,[FP_+ctxPtr ]       ;edi --> context

+;

+endm ;Setup_Stack

+;

+FP_         equ <ebp-FRAME_OFFS>    ;keep as many short offsets as possible

+;

+;----------------------------------------------------------------

+;

+Reset_Stack macro   procStart

+    add     esp,LOCAL_SIZE          ;get rid of locals (wipe??)

+    popad                           ;restore all regs

+

+    ;display code size in bytes to stdout

+  irp  _BCNT_,<%($+1-procStart)>    ;account for return opcode

+if     _BCNT_ ge 10000              ;(align it all pretty)

+%out procStart code size = _BCNT_ bytes  

+elseif _BCNT_ ge  1000

+%out procStart code size =  _BCNT_ bytes  

+else

+%out procStart code size =   _BCNT_ bytes  

+endif

+  endm ;irp _BCNT_

+

+endm ; Reset_Stack

+;

+;----------------------------------------------------------------

+; macros to help debug internals

+;

+if _SKEIN_DEBUG

+    extrn   _Skein_Show_Block:near   ;calls to C routines

+    extrn   _Skein_Show_Round:near

+;

+SKEIN_RND_SPECIAL       =   1000

+SKEIN_RND_KEY_INITIAL   =   SKEIN_RND_SPECIAL+0

+SKEIN_RND_KEY_INJECT    =   SKEIN_RND_SPECIAL+1

+SKEIN_RND_FEED_FWD      =   SKEIN_RND_SPECIAL+2

+;

+Skein_Debug_Block macro BLK_BITS

+;

+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,

+;                     const u08b_t *blkPtr, const u64b_t *wPtr, 

+;                     const u64b_t *ksPtr,const u64b_t *tsPtr);

+;

+    pushad                          ;save all regs

+    lea     eax,[FP_+ksTwk]

+    lea     ebx,[FP_+ksKey]

+    lea     ecx,[esp+32+Wcopy]

+    mov     edx,[FP_+ctxPtr]        ;ctx_hdr_ptr

+    lea     edx,[edx+X_VARS]        ;edx ==> cxt->X[]

+    push    eax                     ;tsPtr

+    push    ebx                     ;ksPtr

+    push    ecx                     ;wPtr

+    push    dword ptr [FP_+blkPtr]  ;blkPtr

+    push    edx                     ;ctx->Xptr

+    push    dword ptr [FP_+ctxPtr]  ;ctx_hdr_ptr

+    mov     eax,BLK_BITS

+    push    eax                     ;bits

+  ifdef _MINGW_

+    call    _Skein_Show_Block-4     ;strange linkage??

+  else

+    call    _Skein_Show_Block

+  endif

+    add     esp,7*4                 ;discard parameter space on stack

+    popad                           ;restore regs

+endm ;Skein_Debug_Block

+

+;

+Skein_Debug_Round macro BLK_SIZE,R,saveRegs

+;

+;void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);

+;

+  ifnb <saveRegs>

+    mov         [esp+X_stk+ 0],eax  ;save internal vars for debug dump

+    mov         [esp+X_stk+ 4],ebx

+    mov         [esp+X_stk+ 8],ecx

+    mov         [esp+X_stk+12],edx

+  endif

+    pushad                          ;save all regs

+  if R ne SKEIN_RND_FEED_FWD

+    lea     eax,[esp+32+X_stk]

+  else

+    mov     eax,[FP_+ctxPtr]

+    add     eax,X_VARS

+  endif

+    push    eax                     ;Xptr

+  if (SKEIN_ASM_UNROLL and BLK_SIZE) or (R ge SKEIN_RND_SPECIAL)

+    mov     eax,R

+  else

+    lea     eax,[4*edi+1+(((R)-1) and 3)] ;compute round number using edi

+  endif

+    push    eax                     ;round number

+    push    dword ptr [FP_+ctxPtr]  ;ctx_hdr_ptr

+    mov     eax,BLK_SIZE

+    push    eax                     ;bits

+  ifdef _MINGW_

+    call    _Skein_Show_Round-4     ;strange linkage??

+  else

+    call    _Skein_Show_Round

+  endif

+    add     esp,4*4                 ;discard parameter space on stack

+    popad                           ;restore regs

+endm  ;Skein_Debug_Round

+endif ;ifdef SKEIN_DEBUG

+;

+;----------------------------------------------------------------

+;

+; MACRO: a mix step

+;

+MixStep     macro   BLK_SIZE,ld_A,ld_C,st_A,st_C,RotNum0,RotNum1,_debug_

+  ifnb <ld_A>

+    mov     eax,[esp+X_stk+8*(ld_A)+0]

+    mov     ebx,[esp+X_stk+8*(ld_A)+4]

+  endif

+  ifnb <ld_C>

+    mov     ecx,[esp+X_stk+8*(ld_C)+0]

+    mov     edx,[esp+X_stk+8*(ld_C)+4]

+  endif

+    add     eax, ecx                ;X[A] += X[C]

+    adc     ebx, edx

+  ifnb <st_A>

+    mov         [esp+X_stk+8*(st_A)+0],eax

+    mov         [esp+X_stk+8*(st_A)+4],ebx

+  endif

+__rNum0 = (RotNum0) AND 7

+    RotL64  ecx, edx, esi,%(BLK_SIZE),%(__rNum0),%(RotNum1) ;X[C] <<<= RC_<BLK_BITS,RotNum0,RotNum1>

+    xor     ecx, eax                ;X[C] ^= X[A]

+    xor     edx, ebx

+  if _SKEIN_DEBUG or  (0 eq (_debug_ + 0))

+   ifb <st_C>

+    mov         [esp+X_stk+8*(ld_C)+0],ecx

+    mov         [esp+X_stk+8*(ld_C)+4],edx

+   else

+    mov         [esp+X_stk+8*(st_C)+0],ecx

+    mov         [esp+X_stk+8*(st_C)+4],edx

+   endif

+  endif

+  if _SKEIN_DEBUG and (0 ne (_debug_ + 0))

+    Skein_Debug_Round BLK_SIZE,%(RotNum0+1)

+  endif

+endm ;MixStep

+;

+;;;;;;;;;;;;;;;;;

+;

+; MACRO: key schedule injection

+;

+ks_Inject macro BLK_SIZE,X_load,X_stor,rLo,rHi,rndBase,keyIdx,twkIdx,ROUND_ADD

+    ;are rLo,rHi values already loaded? if not, load them now

+  ifnb <X_load> 

+    mov     rLo,[esp+X_stk +8*(X_load)  ]

+    mov     rHi,[esp+X_stk +8*(X_load)+4]

+  endif

+

+  ;inject the 64-bit key schedule value (and maybe the tweak as well)

+if SKEIN_ASM_UNROLL and BLK_SIZE

+_kOffs_ = ((rndBase)+(keyIdx)) mod ((BLK_SIZE/64)+1)

+    add     rLo,[FP_+ksKey+8*_kOffs_+ 0]

+    adc     rHi,[FP_+ksKey+8*_kOffs_+ 4]

+  ifnb <twkIdx>

+_tOffs_ = ((rndBase)+(twkIdx)) mod 3

+    add     rLo,[FP_+ksTwk+8*_tOffs_+ 0]

+    adc     rHi,[FP_+ksTwk+8*_tOffs_+ 4]

+  endif

+  ifnb <ROUND_ADD>

+    add     rLo,(ROUND_ADD)

+    adc     rHi,0

+  endif

+else

+    add     rLo,[FP_+ksKey+8*(keyIdx)+8*edi  ]

+    adc     rHi,[FP_+ksKey+8*(keyIdx)+8*edi+4]

+  ifnb <twkIdx>

+    add     rLo,[FP_+ksTwk+8*(twkIdx)+8*edi  ]

+    adc     rHi,[FP_+ksTwk+8*(twkIdx)+8*edi+4]

+  endif

+  ifnb <ROUND_ADD>

+    add     rLo,edi                     ;edi is the round number 

+    adc     rHi,0

+  endif

+endif

+

+  ;do we need to store updated rLo,rHi values? if so, do it now

+  ifnb <X_stor>

+    mov         [esp+X_stk +8*(X_stor)  ],rLo

+    mov         [esp+X_stk +8*(X_stor)+4],rHi

+  endif

+endm ;ks_Inject

+;

+;----------------------------------------------------------------

+; MACRO: key schedule rotation

+;

+ks_Rotate macro rLo,rHi,WCNT

+    mov   rLo,[FP_+ksKey+8*edi+ 0]       ;"rotate" the key schedule in memory

+    mov   rHi,[FP_+ksKey+8*edi+ 4]

+    mov       [FP_+ksKey+8*edi+8*(WCNT+1)+ 0],rLo

+    mov       [FP_+ksKey+8*edi+8*(WCNT+1)+ 4],rHi

+    mov   rLo,[FP_+ksTwk+8*edi+ 0]

+    mov   rHi,[FP_+ksTwk+8*edi+ 4]

+    mov       [FP_+ksTwk+8*edi+8*3+ 0],rLo

+    mov       [FP_+ksTwk+8*edi+8*3+ 4],rHi

+endm

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 256

+    public      _Skein_256_Process_Block

+;

+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+;

+; MACRO: two rounds

+;

+R_256_TwoRounds macro _RR_,ld_0

+    ; here with edx:ecx = X[1]

+    ;--------- round _RR_

+    MixStep 256,ld_0, ,0,1,((_RR_)+0),0

+    MixStep 256,   2,3,2,3,((_RR_)+0),1,1

+

+    ; here with edx:ecx = X[3]

+    ;--------- round _RR_ + 1

+    MixStep 256,   0, ,0,3,((_RR_)+1),0

+    MixStep 256,   2,1,2,1,((_RR_)+1),1,1

+

+    ; here with edx:ecx = X[1]

+endm ;R_256_TwoRounds

+;

+;;;;;;;;;;;;;;;;;

+;

+; code

+;

+_Skein_256_Process_Block proc near

+    WCNT    =   4                   ;WCNT=4 for Skein-256

+    Setup_Stack WCNT,(ROUNDS_256/8)

+

+    ; main hash loop for Skein_256

+Skein_256_block_loop:

+    mov     eax,[edi+TWEAK+ 0]      ;ebx:eax = tweak word T0

+    mov     ebx,[edi+TWEAK+ 4]

+    mov     ecx,[edi+TWEAK+ 8]      ;edx:ecx = tweak word T1

+    mov     edx,[edi+TWEAK+12]

+

+    add     eax,[FP_+bitAdd  ]      ;bump T0 by the bitAdd parameter

+    adc     ebx, 0

+    mov         [edi+TWEAK   ],eax  ;save updated tweak value T0

+    mov         [edi+TWEAK+ 4],ebx

+

+    mov         [FP_+ksTwk   ],eax  ;build the tweak schedule on the stack

+    mov         [FP_+ksTwk+ 4],ebx

+    xor     eax,ecx                 ;ebx:eax = T0 ^ T1

+    xor     ebx,edx

+    mov         [FP_+ksTwk+ 8],ecx

+    mov         [FP_+ksTwk+12],edx

+    mov         [FP_+ksTwk+16],eax

+    mov         [FP_+ksTwk+20],ebx

+

+    mov     eax,KW_PARITY_LO        ;init parity accumulator

+    mov     ebx,KW_PARITY_HI

+;

+_NN_ = 0

+  rept WCNT                         ;copy in the chaining vars

+    mov     ecx,[edi+X_VARS+_NN_   ]

+    mov     edx,[edi+X_VARS+_NN_+ 4]

+    xor     eax,ecx                 ;compute overall parity along the way

+    xor     ebx,edx

+    mov         [FP_+ksKey +_NN_   ],ecx

+    mov         [FP_+ksKey +_NN_+ 4],edx

+_NN_ = _NN_+8

+  endm

+;

+    mov         [FP_+ksKey +_NN_   ],eax ;save overall parity at the end of the array

+    mov         [FP_+ksKey +_NN_+ 4],ebx

+

+    mov     esi,[FP_+blkPtr ]       ;esi --> input block

+;

+_NN_ = WCNT*8-16                    ;work down from the end

+  rept WCNT/2                       ;perform initial key injection

+    mov     eax,[esi+_NN_       + 0]

+    mov     ebx,[esi+_NN_       + 4]

+    mov     ecx,[esi+_NN_       + 8]

+    mov     edx,[esi+_NN_       +12]

+    mov         [esp+_NN_+Wcopy + 0],eax

+    mov         [esp+_NN_+Wcopy + 4],ebx

+    mov         [esp+_NN_+Wcopy + 8],ecx

+    mov         [esp+_NN_+Wcopy +12],edx

+    add     eax,[FP_+_NN_+ksKey + 0]

+    adc     ebx,[FP_+_NN_+ksKey + 4]

+    add     ecx,[FP_+_NN_+ksKey + 8]

+    adc     edx,[FP_+_NN_+ksKey +12]

+   if     _NN_ eq (WCNT*8-16)       ;inject the tweak words

+    add     eax,[FP_+     ksTwk + 8];   (at the appropriate points)

+    adc     ebx,[FP_+     ksTwk +12]

+   elseif _NN_ eq (WCNT*8-32)

+    add     ecx,[FP_+     ksTwk + 0]

+    adc     edx,[FP_+     ksTwk + 4]

+   endif

+   if _NN_ or _SKEIN_DEBUG

+    mov         [esp+_NN_+X_stk + 0],eax

+    mov         [esp+_NN_+X_stk + 4],ebx

+    mov         [esp+_NN_+X_stk + 8],ecx

+    mov         [esp+_NN_+X_stk +12],edx

+   endif

+_NN_ = _NN_ - 16                    ;end at X[0], so regs are already loaded for first MIX!

+  endm

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block WCNT*64 

+    Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL

+endif

+    add     esi, WCNT*8             ;skip the block

+    mov         [FP_+blkPtr   ],esi ;update block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 256

+_UNROLL_CNT =   ROUNDS_256/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_256    ;unroll count

+  if ((ROUNDS_256/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_256"

+  endif

+    xor     edi,edi                 ;edi = iteration count

+Skein_256_round_loop:

+endif

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+      ; here with X[0], X[1] already loaded into eax..edx

+      R_256_TwoRounds %(4*_Rbase_+00),

+      R_256_TwoRounds %(4*_Rbase_+02),0

+

+      ;inject key schedule

+  if _UNROLL_CNT ne (ROUNDS_256/8)

+      ks_Rotate eax,ebx,WCNT

+      inc   edi                     ;edi = round number

+  endif

+_Rbase_ = _Rbase_+1

+      ks_Inject 256,3,3,eax,ebx,_Rbase_,3, ,_Rbase_

+      ks_Inject 256,2,2,eax,ebx,_Rbase_,2,1

+      ks_Inject 256, , ,ecx,edx,_Rbase_,1,0

+      ks_Inject 256,0, ,eax,ebx,_Rbase_,0

+  if _SKEIN_DEBUG

+      Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,saveRegs

+  endif

+endm ;rept _UNROLL_CNT

+;

+  if _UNROLL_CNT ne (ROUNDS_256/8)

+    cmp     edi,2*(ROUNDS_256/8)

+    jb      Skein_256_round_loop

+    mov     edi,[FP_+ctxPtr ]           ;restore edi --> context

+  endif

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..3}

+_NN_ = 0

+ rept WCNT/2

+   if _NN_  ;eax..edx already loaded the first time

+    mov     eax,[esp+X_stk + _NN_ + 0]

+    mov     ebx,[esp+X_stk + _NN_ + 4]

+    mov     ecx,[esp+X_stk + _NN_ + 8]

+    mov     edx,[esp+X_stk + _NN_ +12]

+   endif

+   if _NN_ eq 0

+    and     dword ptr [edi +TWEAK +12],FIRST_MASK

+   endif

+    xor     eax,[esp+Wcopy + _NN_ + 0]

+    xor     ebx,[esp+Wcopy + _NN_ + 4]

+    xor     ecx,[esp+Wcopy + _NN_ + 8]

+    xor     edx,[esp+Wcopy + _NN_ +12]

+    mov         [edi+X_VARS+ _NN_ + 0],eax

+    mov         [edi+X_VARS+ _NN_ + 4],ebx

+    mov         [edi+X_VARS+ _NN_ + 8],ecx

+    mov         [edi+X_VARS+ _NN_ +12],edx

+_NN_ = _NN_+16

+  endm

+if _SKEIN_DEBUG

+    Skein_Debug_Round 256,SKEIN_RND_FEED_FWD

+endif

+    ; go back for more blocks, if needed

+    dec     dword ptr [FP_+blkCnt]

+    jnz     Skein_256_block_loop

+    

+    Reset_Stack _Skein_256_Process_Block

+    ret

+_Skein_256_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein_256_Process_Block_CodeSize

+_Skein_256_Process_Block_CodeSize proc

+    mov     eax,_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block

+    ret

+_Skein_256_Process_Block_CodeSize endp

+;

+    public  _Skein_256_Unroll_Cnt

+_Skein_256_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_256/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein_256_Unroll_Cnt endp

+endif

+endif ;_USE_ASM_ and 256

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 512

+    public      _Skein_512_Process_Block

+;

+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: four rounds

+;

+R_512_FourRounds macro _RR_,ld_0

+    ; here with edx:ecx = X[1]

+    ;--------- round _RR_

+    ; R512(0,1,2,3,4,5,6,7,R_0, 1);

+    MixStep 512, ld_0, ,0,1,((_RR_)+0),0

+    MixStep 512,    2,3,2,3,((_RR_)+0),1

+    MixStep 512,    4,5,4,5,((_RR_)+0),2

+    MixStep 512,    6,7,6, ,((_RR_)+0),3,1

+

+    ; here with edx:ecx = X[7]

+    ; R512(2,1,4,7,6,5,0,3,R_1, 2);

+    MixStep 512,    4, ,4,7,((_RR_)+1),1

+    MixStep 512,    6,5,6,5,((_RR_)+1),2

+    MixStep 512,    0,3,0,3,((_RR_)+1),3

+    MixStep 512,    2,1,2, ,((_RR_)+1),0,1

+

+    ; here with edx:ecx = X[1]

+    ; R512(4,1,6,3,0,5,2,7,R_2, 3);

+    MixStep 512,    4, ,4,1,((_RR_)+2),0

+    MixStep 512,    6,3,6,3,((_RR_)+2),1

+    MixStep 512,    0,5,0,5,((_RR_)+2),2

+    MixStep 512,    2,7,2, ,((_RR_)+2),3,1

+

+    ; here with edx:ecx = X[7]

+    ; R512(6,1,0,7,2,5,4,3,R_3, 4);

+    MixStep 512,    0, ,0,7,((_RR_)+3),1

+    MixStep 512,    2,5,2,5,((_RR_)+3),2

+    MixStep 512,    4,3,4,3,((_RR_)+3),3

+    MixStep 512,    6,1,6, ,((_RR_)+3),0,1

+

+endm ;R_512_FourRounds

+;

+;;;;;;;;;;;;;;;;;

+; code

+;

+_Skein_512_Process_Block proc near

+    WCNT    =   8                   ;WCNT=8 for Skein-512

+    Setup_Stack WCNT,(ROUNDS_512/8)

+

+    ; main hash loop for Skein_512

+Skein_512_block_loop:

+    mov     eax,[edi+TWEAK+ 0]      ;ebx:eax = tweak word T0

+    mov     ebx,[edi+TWEAK+ 4]

+    mov     ecx,[edi+TWEAK+ 8]      ;edx:ecx = tweak word T1

+    mov     edx,[edi+TWEAK+12]

+

+    add     eax,[FP_+bitAdd  ]      ;bump T0 by the bitAdd parameter

+    adc     ebx, 0

+    mov         [edi+TWEAK   ],eax  ;save updated tweak value T0

+    mov         [edi+TWEAK+ 4],ebx

+

+    mov         [FP_+ksTwk   ],eax  ;build the tweak schedule on the stack

+    mov         [FP_+ksTwk+ 4],ebx

+    xor     eax,ecx                 ;ebx:eax = T0 ^ T1

+    xor     ebx,edx

+    mov         [FP_+ksTwk+ 8],ecx

+    mov         [FP_+ksTwk+12],edx

+    mov         [FP_+ksTwk+16],eax

+    mov         [FP_+ksTwk+20],ebx

+

+    mov     eax,KW_PARITY_LO        ;init parity accumulator

+    mov     ebx,KW_PARITY_HI

+;

+_NN_ = 0

+  rept WCNT                         ;copy in the chaining vars

+    mov     ecx,[edi+X_VARS+_NN_   ]

+    mov     edx,[edi+X_VARS+_NN_+ 4]

+    xor     eax,ecx                 ;compute overall parity along the way

+    xor     ebx,edx

+    mov         [FP_+ksKey +_NN_   ],ecx

+    mov         [FP_+ksKey +_NN_+ 4],edx

+_NN_ = _NN_+8

+  endm

+;

+    mov         [FP_+ksKey +_NN_   ],eax ;save overall parity at the end of the array

+    mov         [FP_+ksKey +_NN_+ 4],ebx

+

+    mov     esi,[FP_+blkPtr ]       ;esi --> input block

+;

+_NN_ = WCNT*8-16                    ;work down from the end

+  rept WCNT/2                       ;perform initial key injection

+    mov     eax,[esi+_NN_       + 0]

+    mov     ebx,[esi+_NN_       + 4]

+    mov     ecx,[esi+_NN_       + 8]

+    mov     edx,[esi+_NN_       +12]

+    mov         [esp+_NN_+Wcopy + 0],eax

+    mov         [esp+_NN_+Wcopy + 4],ebx

+    mov         [esp+_NN_+Wcopy + 8],ecx

+    mov         [esp+_NN_+Wcopy +12],edx

+    add     eax,[FP_+_NN_+ksKey + 0]

+    adc     ebx,[FP_+_NN_+ksKey + 4]

+    add     ecx,[FP_+_NN_+ksKey + 8]

+    adc     edx,[FP_+_NN_+ksKey +12]

+   if     _NN_ eq (WCNT*8-16)       ;inject the tweak words

+    add     eax,[FP_+     ksTwk + 8];   (at the appropriate points)

+    adc     ebx,[FP_+     ksTwk +12]

+   elseif _NN_ eq (WCNT*8-32)

+    add     ecx,[FP_+     ksTwk + 0]

+    adc     edx,[FP_+     ksTwk + 4]

+   endif

+   if _NN_ or _SKEIN_DEBUG

+    mov         [esp+_NN_+X_stk + 0],eax

+    mov         [esp+_NN_+X_stk + 4],ebx

+    mov         [esp+_NN_+X_stk + 8],ecx

+    mov         [esp+_NN_+X_stk +12],edx

+   endif

+_NN_ = _NN_ - 16                    ;end at X[0], so regs are already loaded for first MIX!

+  endm

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block WCNT*64 

+    Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL

+endif

+    add     esi, WCNT*8             ;skip the block

+    mov         [FP_+blkPtr   ],esi ;update block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 512

+_UNROLL_CNT =   ROUNDS_512/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_512

+  if ((ROUNDS_512/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_512"

+  endif

+    xor     edi,edi                 ;edi = round counter

+Skein_512_round_loop:

+endif

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+      ; here with X[0], X[1] already loaded into eax..edx

+      R_512_FourRounds %(4*_Rbase_+00),

+

+      ;inject odd  key schedule words

+  if _UNROLL_CNT ne (ROUNDS_512/8)

+      ks_Rotate eax,ebx,WCNT

+      inc   edi                     ;edi = round number

+  endif

+_Rbase_ = _Rbase_+1

+      ks_Inject 512,7,7,eax,ebx,_Rbase_,7, ,_Rbase_

+      ks_Inject 512,6,6,eax,ebx,_Rbase_,6,1

+      ks_Inject 512,5,5,eax,ebx,_Rbase_,5,0

+      ks_Inject 512,4,4,eax,ebx,_Rbase_,4

+      ks_Inject 512,3,3,eax,ebx,_Rbase_,3

+      ks_Inject 512,2,2,eax,ebx,_Rbase_,2

+      ks_Inject 512, , ,ecx,edx,_Rbase_,1

+      ks_Inject 512,0, ,eax,ebx,_Rbase_,0

+  if _SKEIN_DEBUG

+      Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT ,saveRegs

+  endif

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 512) eq 0

+    cmp     edi,2*(ROUNDS_512/8)

+    jb      Skein_512_round_loop

+    mov     edi,[FP_+ctxPtr ]           ;restore edi --> context

+endif

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..7}

+_NN_ = 0

+ rept WCNT/2

+   if _NN_  ;eax..edx already loaded the first time

+    mov     eax,[esp+X_stk + _NN_ + 0]

+    mov     ebx,[esp+X_stk + _NN_ + 4]

+    mov     ecx,[esp+X_stk + _NN_ + 8]

+    mov     edx,[esp+X_stk + _NN_ +12]

+   endif

+   if _NN_ eq 0

+    and     dword ptr [edi + TWEAK+12],FIRST_MASK

+   endif

+    xor     eax,[esp+Wcopy + _NN_ + 0]

+    xor     ebx,[esp+Wcopy + _NN_ + 4]

+    xor     ecx,[esp+Wcopy + _NN_ + 8]

+    xor     edx,[esp+Wcopy + _NN_ +12]

+    mov         [edi+X_VARS+ _NN_ + 0],eax

+    mov         [edi+X_VARS+ _NN_ + 4],ebx

+    mov         [edi+X_VARS+ _NN_ + 8],ecx

+    mov         [edi+X_VARS+ _NN_ +12],edx

+_NN_ = _NN_+16

+  endm

+if _SKEIN_DEBUG

+    Skein_Debug_Round 512,SKEIN_RND_FEED_FWD

+endif

+    ; go back for more blocks, if needed

+    dec     dword ptr [FP_+blkCnt]

+    jnz     Skein_512_block_loop

+

+    Reset_Stack _Skein_512_Process_Block

+    ret

+_Skein_512_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein_512_Process_Block_CodeSize

+_Skein_512_Process_Block_CodeSize proc

+    mov     eax,_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block

+    ret

+_Skein_512_Process_Block_CodeSize endp

+;

+    public  _Skein_512_Unroll_Cnt

+_Skein_512_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_512/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein_512_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 512

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 1024

+    public      _Skein1024_Process_Block

+;

+; void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: four rounds

+;

+R_1024_FourRounds macro _RR_,ld_0

+    ; here with edx:ecx = X[1]

+

+    ;--------- round _RR_

+    MixStep 1024, ld_0,  , 0, 1,((_RR_)+0),0

+    MixStep 1024,    2, 3, 2, 3,((_RR_)+0),1

+    MixStep 1024,    4, 5, 4, 5,((_RR_)+0),2

+    MixStep 1024,    6, 7, 6, 7,((_RR_)+0),3

+    MixStep 1024,    8, 9, 8, 9,((_RR_)+0),4

+    MixStep 1024,   10,11,10,11,((_RR_)+0),5

+    MixStep 1024,   12,13,12,13,((_RR_)+0),6

+    MixStep 1024,   14,15,14,  ,((_RR_)+0),7,1

+    ; here with edx:ecx = X[15]

+

+    ;--------- round _RR_+1

+    MixStep 1024,    4,  , 4,15,((_RR_)+1),3

+    MixStep 1024,    0, 9, 0, 9,((_RR_)+1),0

+    MixStep 1024,    2,13, 2,13,((_RR_)+1),1

+    MixStep 1024,    6,11, 6,11,((_RR_)+1),2

+    MixStep 1024,   10, 7,10, 7,((_RR_)+1),4

+    MixStep 1024,   12, 3,12, 3,((_RR_)+1),5

+    MixStep 1024,   14, 5,14, 5,((_RR_)+1),6

+    MixStep 1024,    8, 1, 8,  ,((_RR_)+1),7,1

+    ; here with edx:ecx = X[1]

+

+    ;--------- round _RR_+2

+    MixStep 1024,    6,  , 6, 1,((_RR_)+2),3    

+    MixStep 1024,    0, 7, 0, 7,((_RR_)+2),0    

+    MixStep 1024,    2, 5, 2, 5,((_RR_)+2),1

+    MixStep 1024,    4, 3, 4, 3,((_RR_)+2),2    

+    MixStep 1024,   12,15,12,15,((_RR_)+2),4

+    MixStep 1024,   14,13,14,13,((_RR_)+2),5    

+    MixStep 1024,    8,11, 8,11,((_RR_)+2),6    

+    MixStep 1024,   10, 9,10,  ,((_RR_)+2),7,1

+    ; here with edx:ecx = X[9]

+

+    ;--------- round _RR_+3

+    MixStep 1024,    4,  , 4, 9,((_RR_)+3),3

+    MixStep 1024,    0,15, 0,15,((_RR_)+3),0

+    MixStep 1024,    2,11, 2,11,((_RR_)+3),1

+    MixStep 1024,    6,13, 6,13,((_RR_)+3),2

+    MixStep 1024,    8, 5, 8, 5,((_RR_)+3),5

+    MixStep 1024,   10, 3,10, 3,((_RR_)+3),6

+    MixStep 1024,   12, 7,12, 7,((_RR_)+3),7

+    MixStep 1024,   14, 1,14,  ,((_RR_)+3),4,1

+

+    ; here with edx:ecx = X[1]

+endm ;R_1024_FourRounds

+;

+;;;;;;;;;;;;;;;;;

+; code

+;

+_Skein1024_Process_Block proc near

+;

+    WCNT    =   16                   ;WCNT=16 for Skein-1024

+    Setup_Stack WCNT,(ROUNDS_1024/8)

+

+    ; main hash loop for Skein1024

+Skein1024_block_loop:

+    mov     eax,[edi+TWEAK+ 0]      ;ebx:eax = tweak word T0

+    mov     ebx,[edi+TWEAK+ 4]

+    mov     ecx,[edi+TWEAK+ 8]      ;edx:ecx = tweak word T1

+    mov     edx,[edi+TWEAK+12]

+

+    add     eax,[FP_+bitAdd  ]      ;bump T0 by the bitAdd parameter

+    adc     ebx, 0

+    mov         [edi+TWEAK   ],eax  ;save updated tweak value T0

+    mov         [edi+TWEAK+ 4],ebx

+

+    mov         [FP_+ksTwk   ],eax  ;build the tweak schedule on the stack

+    mov         [FP_+ksTwk+ 4],ebx

+    xor     eax,ecx                 ;ebx:eax = T0 ^ T1

+    xor     ebx,edx

+    mov         [FP_+ksTwk+ 8],ecx

+    mov         [FP_+ksTwk+12],edx

+    mov         [FP_+ksTwk+16],eax

+    mov         [FP_+ksTwk+20],ebx

+

+    mov     eax,KW_PARITY_LO        ;init parity accumulator

+    mov     ebx,KW_PARITY_HI

+EDI_BIAS    equ 70h                 ;bias the edi offsets to make them short!

+    add     edi, EDI_BIAS

+CT_ equ     <edi-EDI_BIAS>

+;

+_NN_ = 0

+  rept WCNT                         ;copy in the chaining vars

+    mov     ecx,[CT_+X_VARS+_NN_   ]

+    mov     edx,[CT_+X_VARS+_NN_+ 4]

+    xor     eax,ecx                 ;compute overall parity along the way

+    xor     ebx,edx

+    mov         [FP_+ksKey +_NN_   ],ecx

+    mov         [FP_+ksKey +_NN_+ 4],edx

+_NN_ = _NN_+8

+  endm

+;

+    mov         [FP_+ksKey +_NN_   ],eax ;save overall parity at the end of the array

+    mov         [FP_+ksKey +_NN_+ 4],ebx

+

+    mov     esi,[FP_+blkPtr ]       ;esi --> input block

+    lea     edi,[esp+Wcopy]

+;

+_NN_ = WCNT*8-16                    ;work down from the end

+  rept WCNT/2                       ;perform initial key injection

+    mov     eax,[esi+_NN_       + 0]

+    mov     ebx,[esi+_NN_       + 4]

+    mov     ecx,[esi+_NN_       + 8]

+    mov     edx,[esi+_NN_       +12]

+    mov         [edi+_NN_+      + 0],eax

+    mov         [edi+_NN_+      + 4],ebx

+    mov         [edi+_NN_+      + 8],ecx

+    mov         [edi+_NN_+      +12],edx

+    add     eax,[FP_+_NN_+ksKey + 0]

+    adc     ebx,[FP_+_NN_+ksKey + 4]

+    add     ecx,[FP_+_NN_+ksKey + 8]

+    adc     edx,[FP_+_NN_+ksKey +12]

+   if     _NN_ eq (WCNT*8-16)       ;inject the tweak words

+    add     eax,[FP_+     ksTwk + 8];   (at the appropriate points)

+    adc     ebx,[FP_+     ksTwk +12]

+   elseif _NN_ eq (WCNT*8-32)

+    add     ecx,[FP_+     ksTwk + 0]

+    adc     edx,[FP_+     ksTwk + 4]

+   endif

+   if _NN_ or _SKEIN_DEBUG

+    mov         [esp+_NN_+X_stk + 0],eax

+    mov         [esp+_NN_+X_stk + 4],ebx

+    mov         [esp+_NN_+X_stk + 8],ecx

+    mov         [esp+_NN_+X_stk +12],edx

+   endif

+_NN_ = _NN_ - 16                    ;end at X[0], so regs are already loaded for first MIX!

+  endm

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block WCNT*64 

+    Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL

+endif

+    sub     esi,-WCNT*8             ;skip the block (short immediate)

+    mov         [FP_+blkPtr   ],esi ;update block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+if SKEIN_ASM_UNROLL and 1024

+_UNROLL_CNT =   ROUNDS_1024/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_1024

+  if ((ROUNDS_1024/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_1024"

+  endif

+    xor     edi,edi                 ;edi = round counter

+Skein_1024_round_loop:

+endif

+

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+      ; here with X[0], X[1] already loaded into eax..edx

+      R_1024_FourRounds %(4*_Rbase_+00),

+

+      ;inject odd  key schedule words

+      ;inject odd  key schedule words

+  if _UNROLL_CNT ne (ROUNDS_1024/8)

+      ks_Rotate eax,ebx,WCNT

+      inc   edi                     ;edi = round number

+  endif

+_Rbase_ = _Rbase_+1

+      ks_Inject 1024,15,15,eax,ebx,_Rbase_,15, ,_Rbase_

+      ks_Inject 1024,14,14,eax,ebx,_Rbase_,14,1

+      ks_Inject 1024,13,13,eax,ebx,_Rbase_,13,0

+  irp _w,<12,11,10,9,8,7,6,5,4,3,2>

+      ks_Inject 1024,_w,_w,eax,ebx,_Rbase_,_w

+  endm

+      ks_Inject 1024,  ,  ,ecx,edx,_Rbase_,1

+      ks_Inject 1024, 0,  ,eax,ebx,_Rbase_,0

+

+  if _SKEIN_DEBUG

+      Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,saveRegs

+  endif

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 1024) eq 0

+    cmp     edi,2*(ROUNDS_1024/8)

+    jb      Skein_1024_round_loop

+endif

+    mov     edi,[FP_+ctxPtr ]           ;restore edi --> context

+    add     edi,EDI_BIAS                ;and bias it for short offsets below

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..15}

+    lea     esi,[esp+Wcopy]             ;use short offsets below

+_NN_ = 0

+ rept WCNT/2

+   if _NN_  ;eax..edx already loaded the first time

+    mov     eax,[esp+X_stk + _NN_ + 0]

+    mov     ebx,[esp+X_stk + _NN_ + 4]

+    mov     ecx,[esp+X_stk + _NN_ + 8]

+    mov     edx,[esp+X_stk + _NN_ +12]

+   endif

+   if _NN_ eq 0

+    and     dword ptr [CT_ + TWEAK+12],FIRST_MASK

+   endif

+    xor     eax,[esi       + _NN_ + 0]

+    xor     ebx,[esi       + _NN_ + 4]

+    xor     ecx,[esi       + _NN_ + 8]

+    xor     edx,[esi       + _NN_ +12]

+    mov         [CT_+X_VARS+ _NN_ + 0],eax

+    mov         [CT_+X_VARS+ _NN_ + 4],ebx

+    mov         [CT_+X_VARS+ _NN_ + 8],ecx

+    mov         [CT_+X_VARS+ _NN_ +12],edx

+_NN_ = _NN_+16

+  endm

+    sub     edi,EDI_BIAS                ;undo the bias for return

+

+if _SKEIN_DEBUG

+    Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD

+endif

+    ; go back for more blocks, if needed

+    dec     dword ptr [FP_+blkCnt]

+    jnz     Skein1024_block_loop

+

+    Reset_Stack _Skein1024_Process_Block

+    ret

+_Skein1024_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein1024_Process_Block_CodeSize

+_Skein1024_Process_Block_CodeSize proc

+    mov     eax,_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block

+    ret

+_Skein1024_Process_Block_CodeSize endp

+;

+    public  _Skein1024_Unroll_Cnt

+_Skein1024_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_1024/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein1024_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 1024

+;----------------------------------------------------------------

+    end

diff --git a/Additional_Implementations/skein_block_xmm32.asm b/Additional_Implementations/skein_block_xmm32.asm
new file mode 100644
index 000000000000..96ef121cd49a
--- /dev/null
+++ b/Additional_Implementations/skein_block_xmm32.asm
@@ -0,0 +1,1167 @@
+;

+;----------------------------------------------------------------

+; 32-bit x86 assembler code for Skein block functions using XMM registers

+;

+; Author: Doug Whiting, Hifn

+;

+; This code is released to the public domain.

+;----------------------------------------------------------------

+;

+    .386p

+    .model flat

+    .code

+    .xmm                                    ;enable XMM instructions

+;

+_MASK_ALL_  equ (256+512+1024)              ;all three algorithm bits

+;

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_USE_ASM

+_USE_ASM_        = _MASK_ALL_

+elseif SKEIN_USE_ASM and _MASK_ALL_

+_USE_ASM_        = SKEIN_USE_ASM

+else

+_USE_ASM_        = _MASK_ALL_

+endif

+;

+;;;;;;;;;;;;;;;;;

+ifndef SKEIN_LOOP  

+_SKEIN_LOOP       = 0                       ;default is all fully unrolled

+else

+_SKEIN_LOOP       = SKEIN_LOOP

+endif

+;--------------

+; the unroll counts (0 --> fully unrolled)

+SKEIN_UNROLL_256  = (_SKEIN_LOOP / 100) mod 10

+SKEIN_UNROLL_512  = (_SKEIN_LOOP /  10) mod 10

+SKEIN_UNROLL_1024 = (_SKEIN_LOOP      ) mod 10

+;

+SKEIN_ASM_UNROLL  = 0

+  irp _NN_,<256,512,1024>

+    if (SKEIN_UNROLL_&_NN_) eq 0

+SKEIN_ASM_UNROLL  = SKEIN_ASM_UNROLL + _NN_

+    endif

+  endm

+;

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_ROUNDS

+ROUNDS_256  =   72

+ROUNDS_512  =   72

+ROUNDS_1024 =   80

+else

+ROUNDS_256  = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)

+ROUNDS_512  = 8*((((SKEIN_ROUNDS /  10) + 5) mod 10) + 5)

+ROUNDS_1024 = 8*((((SKEIN_ROUNDS      ) + 5) mod 10) + 5)

+endif

+irp _NN_,<256,512,1024>

+  if _USE_ASM_ and _NN_

+    irp _RR_,<%(ROUNDS_&_NN_)>

+      if _NN_ eq 1024

+%out  +++ SKEIN_ROUNDS_&_NN_ = _RR_

+      else

+%out  +++ SKEIN_ROUNDS_&_NN_  = _RR_

+      endif

+    endm

+  endif

+endm

+;;;;;;;;;;;;;;;;;

+;

+ifdef SKEIN_CODE_SIZE

+_SKEIN_CODE_SIZE equ (1)

+else

+ifdef  SKEIN_PERF                           ;use code size if SKEIN_PERF is defined

+_SKEIN_CODE_SIZE equ (1)

+endif

+endif

+;

+;;;;;;;;;;;;;;;;;

+;

+ifndef SKEIN_DEBUG

+_SKEIN_DEBUG      = 0

+else

+_SKEIN_DEBUG      = 1

+endif

+;;;;;;;;;;;;;;;;;

+;

+; define offsets of fields in hash context structure

+;

+HASH_BITS   =   0                           ;# bits of hash output

+BCNT        =   4 + HASH_BITS               ;number of bytes in BUFFER[]

+TWEAK       =   4 + BCNT                    ;tweak values[0..1]

+X_VARS      =  16 + TWEAK                   ;chaining vars

+;

+;(Note: buffer[] in context structure is NOT needed here :-)

+;

+KW_PARITY_LO=   0A9FC1A22h                  ;overall parity of key schedule words (hi32/lo32)

+KW_PARITY_HI=   01BD11BDAh

+FIRST_MASK8 =   NOT (1 SHL 6)               ;FIRST block flag bit

+;

+; rotation constants for Skein

+;

+RC_256_0_0  = 14

+RC_256_0_1  = 16

+

+RC_256_1_0  = 52

+RC_256_1_1  = 57

+

+RC_256_2_0  = 23

+RC_256_2_1  = 40

+

+RC_256_3_0  =  5

+RC_256_3_1  = 37

+

+RC_256_4_0  = 25

+RC_256_4_1  = 33

+

+RC_256_5_0  = 46

+RC_256_5_1  = 12

+

+RC_256_6_0  = 58

+RC_256_6_1  = 22

+

+RC_256_7_0  = 32

+RC_256_7_1  = 32

+

+RC_512_0_0  = 46

+RC_512_0_1  = 36

+RC_512_0_2  = 19

+RC_512_0_3  = 37

+

+RC_512_1_0  = 33

+RC_512_1_1  = 27

+RC_512_1_2  = 14

+RC_512_1_3  = 42

+

+RC_512_2_0  = 17

+RC_512_2_1  = 49

+RC_512_2_2  = 36

+RC_512_2_3  = 39

+

+RC_512_3_0  = 44

+RC_512_3_1  =  9

+RC_512_3_2  = 54

+RC_512_3_3  = 56

+

+RC_512_4_0  = 39

+RC_512_4_1  = 30

+RC_512_4_2  = 34

+RC_512_4_3  = 24

+

+RC_512_5_0  = 13

+RC_512_5_1  = 50

+RC_512_5_2  = 10

+RC_512_5_3  = 17

+

+RC_512_6_0  = 25

+RC_512_6_1  = 29

+RC_512_6_2  = 39

+RC_512_6_3  = 43

+

+RC_512_7_0  =  8

+RC_512_7_1  = 35

+RC_512_7_2  = 56

+RC_512_7_3  = 22

+

+RC_1024_0_0 = 24

+RC_1024_0_1 = 13

+RC_1024_0_2 =  8

+RC_1024_0_3 = 47

+RC_1024_0_4 =  8

+RC_1024_0_5 = 17

+RC_1024_0_6 = 22

+RC_1024_0_7 = 37

+

+RC_1024_1_0 = 38

+RC_1024_1_1 = 19

+RC_1024_1_2 = 10

+RC_1024_1_3 = 55

+RC_1024_1_4 = 49

+RC_1024_1_5 = 18

+RC_1024_1_6 = 23

+RC_1024_1_7 = 52

+

+RC_1024_2_0 = 33

+RC_1024_2_1 =  4

+RC_1024_2_2 = 51

+RC_1024_2_3 = 13

+RC_1024_2_4 = 34

+RC_1024_2_5 = 41

+RC_1024_2_6 = 59

+RC_1024_2_7 = 17

+

+RC_1024_3_0 =  5

+RC_1024_3_1 = 20

+RC_1024_3_2 = 48

+RC_1024_3_3 = 41

+RC_1024_3_4 = 47

+RC_1024_3_5 = 28

+RC_1024_3_6 = 16

+RC_1024_3_7 = 25

+

+RC_1024_4_0 = 41

+RC_1024_4_1 =  9

+RC_1024_4_2 = 37

+RC_1024_4_3 = 31

+RC_1024_4_4 = 12

+RC_1024_4_5 = 47

+RC_1024_4_6 = 44

+RC_1024_4_7 = 30

+

+RC_1024_5_0 = 16

+RC_1024_5_1 = 34

+RC_1024_5_2 = 56

+RC_1024_5_3 = 51

+RC_1024_5_4 =  4

+RC_1024_5_5 = 53

+RC_1024_5_6 = 42

+RC_1024_5_7 = 41

+

+RC_1024_6_0 = 31

+RC_1024_6_1 = 44

+RC_1024_6_2 = 47

+RC_1024_6_3 = 46

+RC_1024_6_4 = 19

+RC_1024_6_5 = 42

+RC_1024_6_6 = 44

+RC_1024_6_7 = 25

+

+RC_1024_7_0 =  9

+RC_1024_7_1 = 48

+RC_1024_7_2 = 35

+RC_1024_7_3 = 52

+RC_1024_7_4 = 23

+RC_1024_7_5 = 31

+RC_1024_7_6 = 37

+RC_1024_7_7 = 20

+;

+mov64 macro x0,x1

+    movq    x0,x1

+endm

+;

+;----------------------------------------------------------------

+; declare allocated space on the stack

+StackVar    macro localName,localSize

+localName   =   _STK_OFFS_

+_STK_OFFS_  =   _STK_OFFS_+(localSize)

+endm ;StackVar

+;

+;----------------------------------------------------------------

+;

+; MACRO: Configure stack frame, allocate local vars

+;

+Setup_Stack macro WCNT,RND_CNT

+_STK_OFFS_  =   0                   ;starting offset from esp, forced on 16-byte alignment

+    ;----- local  variables         ;<-- esp

+    StackVar    X_stk  , 8*(WCNT)   ;local context vars

+    StackVar    Wcopy  , 8*(WCNT)   ;copy of input block    

+    StackVar    ksTwk  ,16*3        ;key schedule: tweak words

+    StackVar    ksKey  ,16*(WCNT)+16;key schedule: key   words

+FRAME_OFFS  =   ksTwk+128           ;<-- ebp

+  if (SKEIN_ASM_UNROLL and (WCNT*64)) eq 0

+    StackVar    ksRot,16*(RND_CNT/4);leave space for ks "rotation" to happen

+  endif

+LOCAL_SIZE  =   _STK_OFFS_          ;size of local vars

+    ;

+    ;"restart" the stack defns, because we relocate esp to guarantee alignment

+    ;    (i.e., these vars are NOT at fixed offsets from esp)

+_STK_OFFS_  =   0

+    ;----- 

+    StackVar    savRegs,8*4         ;pushad data

+    StackVar    retAddr,4           ;return address

+    ;----- caller parameters

+    StackVar    ctxPtr ,4           ;context ptr

+    StackVar    blkPtr ,4           ;pointer to block data

+    StackVar    blkCnt ,4           ;number of full blocks to process

+    StackVar    bitAdd ,4           ;bit count to add to tweak

+    ;----- caller's stack frame

+;

+; Notes on stack frame setup:

+;   * the most used variable (except for Skein-256) is X_stk[], based at [esp+0]

+;   * the next most used is the key schedule words

+;       so ebp is "centered" there, allowing short offsets to the key/tweak

+;       schedule in 256/512-bit Skein cases, but not posible for Skein-1024 :-(

+;   * the Wcopy variables are infrequently accessed, and they have long 

+;       offsets from both esp and ebp only in the 1024-bit case.

+;   * all other local vars and calling parameters can be accessed 

+;       with short offsets, except in the 1024-bit case

+;

+    pushad                          ;save all regs

+    mov     ebx,esp                 ;keep ebx as pointer to caller parms

+    sub     esp,LOCAL_SIZE          ;make room for the locals

+    and     esp,not 15              ;force alignment

+    mov     edi,[ebx+ctxPtr ]       ;edi --> Skein context

+    lea     ebp,[esp+FRAME_OFFS]    ;maximize use of short offsets from ebp

+    mov     ecx,ptr32 [ebx+blkCnt]  ;keep block cnt in ecx

+;

+endm ;Setup_Stack

+;

+FP_         equ <ebp-FRAME_OFFS>    ;keep as many short offsets as possible

+SI_         equ <esi-FRAME_OFFS>    ;keep as many short offsets as possible

+ptr64       equ <qword ptr>         ;useful abbreviations

+ptr32       equ <dword ptr>

+ptr08       equ <byte  ptr>

+;

+;----------------------------------------------------------------

+;

+Reset_Stack macro   procStart

+    mov     esp,ebx                 ;get rid of locals (wipe??)

+    popad                           ;restore all regs

+

+    ;display code size in bytes to stdout

+  irp  _BCNT_,<%($+1-procStart)>    ;account for return opcode

+if     _BCNT_ ge 10000              ;(align it all pretty)

+%out procStart code size = _BCNT_ bytes  

+elseif _BCNT_ ge  1000

+%out procStart code size =  _BCNT_ bytes  

+else

+%out procStart code size =   _BCNT_ bytes  

+endif

+  endm ;irp _BCNT_

+

+endm ; Reset_Stack

+;

+;----------------------------------------------------------------

+; macros to help debug internals

+;

+if _SKEIN_DEBUG

+    extrn   _Skein_Show_Block:near   ;calls to C routines

+    extrn   _Skein_Show_Round:near

+;

+SKEIN_RND_SPECIAL       =   1000

+SKEIN_RND_KEY_INITIAL   =   SKEIN_RND_SPECIAL+0

+SKEIN_RND_KEY_INJECT    =   SKEIN_RND_SPECIAL+1

+SKEIN_RND_FEED_FWD      =   SKEIN_RND_SPECIAL+2

+;

+Skein_Debug_Block macro BLK_BITS

+;

+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,

+;                     const u08b_t *blkPtr, const u64b_t *wPtr, 

+;                     const u64b_t *ksPtr,const u64b_t *tsPtr);

+;

+    Put_XMM_&BLK_BITS

+    pushad                          ;save all regs

+    lea     eax,[FP_+ksTwk+1]       ;+1 = flag: "stride" size = 2 qwords

+    lea     esi,[FP_+ksKey+1]

+    lea     ecx,[esp+32+Wcopy]      ;adjust offset by 32 for pushad

+    mov     edx,[ebx+ctxPtr]        ;ctx_hdr_ptr

+    lea     edx,[edx+X_VARS]        ;edx ==> cxt->X[]

+    push    eax                     ;tsPtr

+    push    esi                     ;ksPtr

+    push    ecx                     ;wPtr

+    push    ptr32 [ebx+blkPtr]      ;blkPtr

+    push    edx                     ;ctx->Xptr

+    push    ptr32 [ebx+ctxPtr]      ;ctx_hdr_ptr

+    mov     eax,BLK_BITS

+    push    eax                     ;bits

+  ifdef _MINGW_

+    call    _Skein_Show_Block-4     ;strange linkage??

+  else

+    call    _Skein_Show_Block

+  endif

+    add     esp,7*4                 ;discard parameter space on stack

+    popad                           ;restore regs

+;

+    Get_XMM_&BLK_BITS

+endm ;Skein_Debug_Block

+

+;

+Skein_Debug_Round macro BLK_BITS,R,saveRegs

+;

+;void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);

+;

+  ifnb <saveRegs>

+      Put_XMM_&BLK_BITS

+  endif

+    pushad                          ;save all regs

+  if R ne SKEIN_RND_FEED_FWD

+    lea     eax,[esp+32+X_stk]      ;adjust offset by 32 for pushad

+  else

+    mov     eax,[ebx+ctxPtr]

+    add     eax,X_VARS

+  endif

+    push    eax                     ;Xptr

+  if (SKEIN_ASM_UNROLL and BLK_BITS) or (R ge SKEIN_RND_SPECIAL)

+    mov     eax,R

+  else

+    lea     eax,[4*edx+1+(((R)-1) and 3)] ;compute round number using edx

+  endif

+    push    eax                     ;round number

+    push    ptr32 [ebx+ctxPtr]      ;ctx_hdr_ptr

+    mov     eax,BLK_BITS

+    push    eax                     ;bits

+  ifdef _MINGW_

+    call    _Skein_Show_Round-4     ;strange linkage??

+  else

+    call    _Skein_Show_Round

+  endif

+    add     esp,4*4                 ;discard parameter space on stack

+    popad                           ;restore regs

+

+  ifnb <saveRegs>

+      Get_XMM_&BLK_BITS       ;save internal vars for debug dump

+  endif

+endm  ;Skein_Debug_Round

+endif ;ifdef SKEIN_DEBUG

+;

+;----------------------------------------------------------------

+; useful macros

+_ldX macro xn

+  ifnb <xn>

+    mov64 xmm&xn,ptr64 [esp+X_stk+8*xn]

+  endif

+endm

+

+_stX macro xn

+  ifnb <xn>

+    mov64        ptr64 [esp+X_stk+8*xn],xmm&xn

+  endif

+endm

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 256

+    public      _Skein_256_Process_Block

+;

+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+;

+; Skein-256 round macros

+;

+R_256_OneRound macro _RR_,x0,x1,x2,x3,t0,t1

+  irp _qq_,<%((_RR_) and 7)>        ;figure out which rotation constants to use

+    if x0 eq 0

+_RC0_ =   RC_256_&_qq_&_0

+_RC1_ =   RC_256_&_qq_&_1

+    else

+_RC0_ =   RC_256_&_qq_&_1

+_RC1_ =   RC_256_&_qq_&_0

+    endif

+  endm

+;

+    paddq   xmm&x0,xmm&x1

+    mov64   xmm&t0,xmm&x1

+    psllq   xmm&x1,   _RC0_

+    psrlq   xmm&t0,64-_RC0_

+    xorpd   xmm&x1,xmm&x0

+    xorpd   xmm&x1,xmm&t0

+;

+    paddq   xmm&x2,xmm&x3

+    mov64   xmm&t1,xmm&x3

+    psllq   xmm&x3,   _RC1_

+    psrlq   xmm&t1,64-_RC1_

+    xorpd   xmm&x3,xmm&x2

+    xorpd   xmm&x3,xmm&t1

+  if _SKEIN_DEBUG

+    Skein_Debug_Round 256,%(_RR_+1),saveRegs

+  endif

+endm ;R_256_OneRound

+;

+R_256_FourRounds macro _RN_

+    R_256_OneRound (_RN_+0),0,1,2,3,4,5

+    R_256_OneRound (_RN_+1),2,1,0,3,4,5

+

+    R_256_OneRound (_RN_+2),0,1,2,3,4,5

+    R_256_OneRound (_RN_+3),2,1,0,3,4,5

+

+    ;inject key schedule

+    inc   edx                     ;bump round number

+    movd  xmm4,edx

+  if _UNROLL_CNT eq (ROUNDS_256/8)

+    ;fully unrolled version

+_RK_ = ((_RN_)/4)                 ;key injection counter

+    paddq xmm0,[FP_+ksKey+16*((_RK_+1) mod 5)]

+    paddq xmm1,[FP_+ksKey+16*((_RK_+2) mod 5)]

+    paddq xmm2,[FP_+ksKey+16*((_RK_+3) mod 5)]

+    paddq xmm3,[FP_+ksKey+16*((_RK_+4) mod 5)]

+    paddq xmm1,[FP_+ksTwk+16*((_RK_+1) mod 3)]

+    paddq xmm2,[FP_+ksTwk+16*((_RK_+2) mod 3)]

+    paddq xmm3,xmm4

+  else ;looping version

+    paddq xmm0,[SI_+ksKey+16*1]

+    paddq xmm1,[SI_+ksKey+16*2]

+    paddq xmm2,[SI_+ksKey+16*3]

+    paddq xmm3,[SI_+ksKey+16*4]

+    paddq xmm1,[SI_+ksTwk+16*1]

+    paddq xmm2,[SI_+ksTwk+16*2]

+    paddq xmm3,xmm4

+;   

+    mov64 xmm4,<ptr64 [SI_+ksKey]>;first, "rotate" key schedule on the stack

+    mov64 xmm5,<ptr64 [SI_+ksTwk]>;    (for next time through)

+    mov64      <ptr64 [SI_+ksKey+16*(WCNT+1)]>,xmm4

+    mov64      <ptr64 [SI_+ksTwk+16*3]>,xmm5

+    add   esi,16                  ;bump rolling pointer

+  endif

+  if _SKEIN_DEBUG

+      Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,saveRegs

+  endif

+endm ;R256_FourRounds

+;

+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines

+Put_XMM_256 equ <call _Put_XMM_256>

+Get_XMM_256 equ <call _Get_XMM_256>

+

+_Put_XMM_256:

+  irp _NN_,<0,1,2,3>

+    mov64           ptr64 [esp+X_stk+4+_NN_*8],xmm&_NN_

+  endm

+    ret

+;

+_Get_XMM_256:

+  irp _NN_,<0,1,2,3>

+    mov64  xmm&_NN_,ptr64 [esp+X_stk+4+_NN_*8]

+  endm

+    ret

+endif

+;

+;;;;;;;;;;;;;;;;;

+;

+; code

+;

+_Skein_256_Process_Block proc near

+    WCNT    =   4                   ;WCNT=4 for Skein-256

+    Setup_Stack WCNT,ROUNDS_256

+    ; main hash loop for Skein_256

+Skein_256_block_loop:

+    movd    xmm4,ptr32 [ebx+bitAdd]

+    mov64   xmm5,ptr64 [edi+TWEAK+0]

+    mov64   xmm6,ptr64 [edi+TWEAK+8]

+    paddq   xmm5,xmm4               ;bump T0 by the bitAdd parameter

+    mov64   ptr64 [edi+TWEAK],xmm5  ;save updated tweak value T0 (for next time)

+    movapd  xmm7,xmm6

+    xorpd   xmm7,xmm5               ;compute overall tweak parity

+    movdqa  [FP_+ksTwk   ],xmm5     ;save the expanded tweak schedule on the stack

+    movdqa  [FP_+ksTwk+16],xmm6

+    movdqa  [FP_+ksTwk+32],xmm7

+

+    mov     esi,[ebx+blkPtr]        ;esi --> input block

+    mov     eax,KW_PARITY_LO        ;init key schedule parity accumulator

+    mov     edx,KW_PARITY_HI

+    movd    xmm4,eax

+    movd    xmm0,edx

+    unpcklps xmm4,xmm0              ;pack two 32-bit words into xmm4

+;

+  irp _NN_,<0,1,2,3>                ;copy in the chaining vars

+    mov64   xmm&_NN_,ptr64 [edi+X_VARS+8*_NN_]

+    xorpd   xmm4,xmm&_NN_           ;update overall parity

+    movdqa  [FP_+ksKey+16*_NN_],xmm&_NN_

+  endm

+    movdqa  [FP_+ksKey+16*WCNT],xmm4;save overall parity at the end of the array

+;

+    paddq   xmm1,xmm5               ;inject the initial tweak words

+    paddq   xmm2,xmm6

+;

+  irp _NN_,<0,1,2,3>                ;perform the initial key injection

+    mov64   xmm4,ptr64 [esi+8*_NN_] ;and save a copy of the input block on stack

+    mov64        ptr64 [esp+8*_NN_+Wcopy],xmm4

+    paddq   xmm&_NN_,xmm4

+  endm

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block 256

+    Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL,saveRegs

+endif

+    add     esi, WCNT*8             ;skip to the next block

+    mov         [ebx+blkPtr   ],esi ;save the updated block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+    xor     edx,edx                 ;edx = iteration count

+if SKEIN_ASM_UNROLL and 256

+_UNROLL_CNT =   ROUNDS_256/8        ;fully unrolled

+else

+_UNROLL_CNT =   SKEIN_UNROLL_256    ;partial unroll count

+  if ((ROUNDS_256/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_256" ;sanity check

+  endif

+    mov     esi,ebp                 ;use this as "rolling" pointer into ksTwk/ksKey

+Skein_256_round_loop:               ;   (since there's no 16* scaled address mode)

+endif

+;

+_Rbase_ = 0

+rept _UNROLL_CNT*2                  ; here with X[0..3] in XMM0..XMM3

+      R_256_FourRounds _Rbase_

+_Rbase_ = _Rbase_+4

+endm ;rept _UNROLL_CNT*2

+;

+  if _UNROLL_CNT ne (ROUNDS_256/8)

+    cmp     edx,2*(ROUNDS_256/8)

+    jb      Skein_256_round_loop

+  endif

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..3}

+    irp _NN_,<0,1,2,3>

+        mov64   xmm4,ptr64 [esp+Wcopy+8*_NN_]

+        xorpd   xmm&_NN_,xmm4

+        mov64        ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_

+    endm

+    and     ptr08 [edi +TWEAK +15],FIRST_MASK8

+if _SKEIN_DEBUG

+    Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,saveRegs

+endif

+    ; go back for more blocks, if needed

+    dec     ecx

+    jnz     Skein_256_block_loop

+    

+    Reset_Stack _Skein_256_Process_Block

+    ret

+;

+_Skein_256_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein_256_Process_Block_CodeSize

+_Skein_256_Process_Block_CodeSize proc

+    mov     eax,_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block

+    ret

+_Skein_256_Process_Block_CodeSize endp

+;

+    public  _Skein_256_Unroll_Cnt

+_Skein_256_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_256/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein_256_Unroll_Cnt endp

+endif

+endif ;_USE_ASM_ and 256

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 512

+    public      _Skein_512_Process_Block

+;

+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: one round

+;

+R_512_Round macro _RR_, a0,a1,Ra, b0,b1,Rb, c0,c1,Rc, d0,d1,Rd

+irp _nr_,<%((_RR_) and 7)>

+_Ra_ = RC_512_&_nr_&_&Ra

+_Rb_ = RC_512_&_nr_&_&Rb

+_Rc_ = RC_512_&_nr_&_&Rc

+_Rd_ = RC_512_&_nr_&_&Rd

+endm

+    paddq   xmm&a0,xmm&a1

+                            _stX c0

+    mov64   xmm&c0,xmm&a1

+    psllq   xmm&a1,   _Ra_

+    psrlq   xmm&c0,64-_Ra_

+    xorpd   xmm&a1,xmm&c0

+    xorpd   xmm&a1,xmm&a0

+

+    paddq   xmm&b0,xmm&b1

+                            _stX a0

+    mov64   xmm&a0,xmm&b1

+    psllq   xmm&b1,   _Rb_

+    psrlq   xmm&a0,64-_Rb_

+    xorpd   xmm&b1,xmm&b0

+                            _ldX c0

+    xorpd   xmm&b1,xmm&a0

+                             

+    paddq   xmm&c0,xmm&c1

+    mov64   xmm&a0,xmm&c1

+    psllq   xmm&c1,   _Rc_

+    psrlq   xmm&a0,64-_Rc_

+    xorpd   xmm&c1,xmm&c0

+    xorpd   xmm&c1,xmm&a0

+                             

+    paddq   xmm&d0,xmm&d1

+    mov64   xmm&a0,xmm&d1           

+    psllq   xmm&d1,   _Rd_

+    psrlq   xmm&a0,64-_Rd_

+    xorpd   xmm&d1,xmm&a0

+                            _ldX a0

+    xorpd   xmm&d1,xmm&d0

+  if _SKEIN_DEBUG

+    Skein_Debug_Round 512,%(_RR_+1),saveRegs

+  endif

+endm

+;

+; MACRO: four rounds

+R_512_FourRounds macro _RN_

+    R_512_Round (_RN_)  , 0,1,0, 2,3,1, 4,5,2, 6,7,3

+    R_512_Round (_RN_)+1, 2,1,0, 4,7,1, 6,5,2, 0,3,3

+    R_512_Round (_RN_)+2, 4,1,0, 6,3,1, 0,5,2, 2,7,3

+    R_512_Round (_RN_)+3, 6,1,0, 0,7,1, 2,5,2, 4,3,3

+

+    ;inject key schedule

+  irp _NN_,<0,1,2,3,4,5,6,7>

+   if _UNROLL_CNT eq (ROUNDS_512/8)

+    paddq xmm&_NN_,[FP_+ksKey+16*((((_RN_)/4)+(_NN_)+1) mod 9)]

+    else

+    paddq xmm&_NN_,[SI_+ksKey+16*((_NN_)+1)]

+    endif

+  endm

+    _stX  0                       ;free up a register

+    inc   edx                     ;bump round counter

+    movd  xmm0,edx                ;inject the tweak

+  if _UNROLL_CNT eq (ROUNDS_512/8)

+    paddq xmm5,[FP_+ksTwk+16*(((_RN_)+1) mod 3)]

+    paddq xmm6,[FP_+ksTwk+16*(((_RN_)+2) mod 3)]

+    paddq xmm7,xmm0

+  else ;looping version

+    paddq xmm5,[SI_+ksTwk+16*1]

+    paddq xmm6,[SI_+ksTwk+16*2]

+    paddq xmm7,xmm0

+;   

+    mov64 xmm0,<ptr64 [SI_+ksKey]>;first, "rotate" key schedule on the stack

+    mov64      <ptr64 [SI_+ksKey+16*(WCNT+1)]>,xmm0

+    mov64 xmm0,<ptr64 [SI_+ksTwk]>;    (for next time through)

+    mov64      <ptr64 [SI_+ksTwk+16*3]>,xmm0

+    add   esi,16                  ;bump rolling pointer

+  endif

+    _ldX  0                       ;restore X0

+  if _SKEIN_DEBUG

+      Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT,saveRegs

+  endif

+endm ;R_512_FourRounds

+;;;;;;;;;;;;;;;;;

+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines

+Put_XMM_512 equ <call _Put_XMM_512>

+Get_XMM_512 equ <call _Get_XMM_512>

+

+_Put_XMM_512:

+  irp _NN_,<0,1,2,3,4,5,6,7>

+    mov64          ptr64 [esp+X_stk+4+_NN_*8],xmm&_NN_

+  endm

+    ret

+;

+_Get_XMM_512:

+  irp _NN_,<0,1,2,3,4,5,6,7>

+    mov64  xmm&_NN_,ptr64 [esp+X_stk+4+_NN_*8]

+  endm

+    ret

+endif

+;

+;;;;;;;;;;;;;;;;;

+; code

+;

+_Skein_512_Process_Block proc near

+    WCNT    =   8                   ;WCNT=8 for Skein-512

+    Setup_Stack WCNT,ROUNDS_512

+    ; main hash loop for Skein_512

+Skein_512_block_loop:

+    movd    xmm0,ptr32 [ebx+bitAdd]

+    mov64   xmm1,ptr64 [edi+TWEAK+0]

+    mov64   xmm2,ptr64 [edi+TWEAK+8]

+    paddq   xmm1,xmm0               ;bump T0 by the bitAdd parameter

+    mov64   ptr64 [edi+TWEAK],xmm1  ;save updated tweak value T0 (for next time)

+    mov64   xmm0,xmm2

+    xorpd   xmm0,xmm1               ;compute overall tweak parity

+    movdqa  [FP_+ksTwk     ],xmm1   ;save the expanded tweak schedule on the stack

+    movdqa  [FP_+ksTwk+16*1],xmm2

+    movdqa  [FP_+ksTwk+16*2],xmm0

+

+    mov     esi,[ebx+blkPtr]        ;esi --> input block

+    mov     eax,KW_PARITY_LO        ;init key schedule parity accumulator

+    mov     edx,KW_PARITY_HI

+    movd    xmm0,eax

+    movd    xmm7,edx

+    unpcklps xmm0,xmm7              ;pack two 32-bit words into xmm0

+;

+  irp _NN_,<7,6,5,4,3,2,1>          ;copy in the chaining vars (skip #0 for now)

+    mov64   xmm&_NN_,ptr64 [edi+X_VARS+8*_NN_]

+    xorpd   xmm0,xmm&_NN_           ;update overall parity

+    movdqa  [FP_+ksKey+16*_NN_],xmm&_NN_

+   if _NN_ eq 5

+    paddq   xmm5,xmm1               ;inject the initial tweak words

+    paddq   xmm6,xmm2               ;  (before they get trashed in xmm1/2)

+   endif

+  endm

+    mov64   xmm4,ptr64 [edi+X_VARS] ;handle #0 now

+    xorpd   xmm0,xmm4               ;update overall parity

+    movdqa  [FP_+ksKey+16* 0  ],xmm4;save the key value in slot #0

+    movdqa  [FP_+ksKey+16*WCNT],xmm0;save overall parity at the end of the array

+;

+    mov64   xmm0,xmm4

+  irp _NN_,<7,6,5,  4,3,2,1,0>      ;perform the initial key injection (except #4)

+    mov64   xmm4,ptr64 [esi+ 8*_NN_];and save a copy of the input block on stack

+    mov64        ptr64 [esp+ 8*_NN_+Wcopy],xmm4

+    paddq   xmm&_NN_,xmm4

+  endm

+    mov64   xmm4,ptr64 [esi+ 8*4]   ;get input block word #4

+    mov64        ptr64 [esp+ 8*4+Wcopy],xmm4

+    paddq   xmm4,[FP_+ksKey+16*4]   ;inject the initial key

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block 512

+    Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL,saveRegs

+endif

+    add     esi, WCNT*8             ;skip to the next block

+    mov         [ebx+blkPtr],esi    ;save the updated block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+    xor     edx,edx                 ;edx = round counter

+if SKEIN_ASM_UNROLL and 512

+_UNROLL_CNT =   ROUNDS_512/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_512

+  if ((ROUNDS_512/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_512"

+  endif

+    mov     esi,ebp                 ;use this as "rolling" pointer into ksTwk/ksKey

+Skein_512_round_loop:               ;   (since there's no 16* scaled address mode)

+endif

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+      R_512_FourRounds _Rbase_

+_Rbase_ = _Rbase_+4

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 512) eq 0

+    cmp     edx,2*(ROUNDS_512/8)

+    jb      Skein_512_round_loop

+endif

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..7}

+    and     ptr08 [edi +TWEAK +15],FIRST_MASK8

+irp _NN_,<0,2,4,6>                  ;do the aligned ones first

+    xorpd   xmm&_NN_,[esp+Wcopy+8*_NN_]

+    mov64   ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_

+endm

+irp _NN_,<1,3,5,7>                  ;now we have some register space available

+    mov64   xmm0,ptr64 [esp+Wcopy+8*_NN_]

+    xorpd   xmm&_NN_,xmm0

+    mov64   ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_

+endm

+if _SKEIN_DEBUG

+    Skein_Debug_Round 512,SKEIN_RND_FEED_FWD

+endif

+    ; go back for more blocks, if needed

+    dec     ecx

+    jnz     Skein_512_block_loop

+

+    Reset_Stack _Skein_512_Process_Block

+    ret

+_Skein_512_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein_512_Process_Block_CodeSize

+_Skein_512_Process_Block_CodeSize proc

+    mov     eax,_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block

+    ret

+_Skein_512_Process_Block_CodeSize endp

+;

+    public  _Skein_512_Unroll_Cnt

+_Skein_512_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_512/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein_512_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 512

+;

+;----------------------------------------------------------------

+;

+if _USE_ASM_ and 1024

+    public      _Skein1024_Process_Block

+;

+; void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);

+;

+R_1024_REGS equ     (5)     ;keep this many block variables in registers

+;

+;;;;;;;;;;;;;;;;

+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines

+Put_XMM_1024 equ <call _Put_XMM_1024>

+Get_XMM_1024 equ <call _Get_XMM_1024>

+

+_Put_XMM_1024:

+_NN_ = 0

+  rept R_1024_REGS

+   irp _rr_,<%(_NN_)>

+    mov64           ptr64 [esp+X_stk+4+8*_NN_],xmm&_rr_

+   endm

+_NN_ = _NN_+1

+  endm

+    ret

+;

+_Get_XMM_1024:

+_NN_ = 0

+  rept R_1024_REGS

+   irp _rr_,<%(_NN_)>

+    mov64  xmm&_rr_,ptr64 [esp+X_stk+4+8*_NN_]

+   endm

+_NN_ = _NN_+1

+  endm

+    ret

+endif

+;

+;;;;;;;;;;;;;;;;;

+; MACRO: one mix step

+MixStep_1024 macro  x0,x1,rotIdx0,rotIdx1,_debug_

+_r0_ =  x0      ;default, if already loaded

+_r1_ =  x1

+  ; load the regs (if necessary)

+  if (x0 ge R_1024_REGS)

+_r0_ =       5

+  mov64   xmm5,ptr64 [esp+X_stk+8*(x0)]

+  endif

+  if (x1 ge R_1024_REGS)

+_r1_ =       6     

+    mov64 xmm6,ptr64 [esp+X_stk+8*(x1)]

+  endif

+  ; do the mix

+  irp _rx_,<%((rotIdx0) and 7)>

+_Rc_ = RC_1024_&_rx_&_&rotIdx1  ;rotation constant

+  endm

+  irp _x0_,<%_r0_>

+  irp _x1_,<%_r1_>

+    paddq   xmm&_x0_,xmm&_x1_

+    mov64   xmm7    ,xmm&_x1_

+    psllq   xmm&_x1_,   _Rc_

+    psrlq   xmm7    ,64-_Rc_

+    xorpd   xmm&_x1_,xmm&_x0_

+    xorpd   xmm&_x1_,xmm7

+  endm

+  endm

+  ; save the regs (if necessary)

+  if (x0 ge R_1024_REGS)

+    mov64   ptr64 [esp+X_stk+8*(x0)],xmm5

+  endif

+  if (x1 ge R_1024_REGS)

+    mov64   ptr64 [esp+X_stk+8*(x1)],xmm6

+  endif

+  ; debug output

+  if _SKEIN_DEBUG and (0 ne (_debug_ + 0))

+    Skein_Debug_Round 1024,%((RotIdx0)+1),saveRegs

+  endif

+endm

+;;;;;;;;;;;;;;;;;

+; MACRO: four rounds

+;

+R_1024_FourRounds macro _RR_

+    ;--------- round _RR_

+    MixStep_1024     0, 1,%((_RR_)+0),0

+    MixStep_1024     2, 3,%((_RR_)+0),1

+    MixStep_1024     4, 5,%((_RR_)+0),2

+    MixStep_1024     6, 7,%((_RR_)+0),3

+    MixStep_1024     8, 9,%((_RR_)+0),4

+    MixStep_1024    10,11,%((_RR_)+0),5

+    MixStep_1024    12,13,%((_RR_)+0),6

+    MixStep_1024    14,15,%((_RR_)+0),7,1

+    ;--------- round _RR_+1

+    MixStep_1024     0, 9,%((_RR_)+1),0

+    MixStep_1024     2,13,%((_RR_)+1),1

+    MixStep_1024     6,11,%((_RR_)+1),2

+    MixStep_1024     4,15,%((_RR_)+1),3

+    MixStep_1024    10, 7,%((_RR_)+1),4

+    MixStep_1024    12, 3,%((_RR_)+1),5

+    MixStep_1024    14, 5,%((_RR_)+1),6

+    MixStep_1024     8, 1,%((_RR_)+1),7,1

+    ;--------- round _RR_+2

+    MixStep_1024     0, 7,%((_RR_)+2),0    

+    MixStep_1024     2, 5,%((_RR_)+2),1

+    MixStep_1024     4, 3,%((_RR_)+2),2    

+    MixStep_1024     6, 1,%((_RR_)+2),3    

+    MixStep_1024    12,15,%((_RR_)+2),4

+    MixStep_1024    14,13,%((_RR_)+2),5    

+    MixStep_1024     8,11,%((_RR_)+2),6    

+    MixStep_1024    10, 9,%((_RR_)+2),7,1

+    ;--------- round _RR_+3

+    MixStep_1024     0,15,%((_RR_)+3),0

+    MixStep_1024     2,11,%((_RR_)+3),1

+    MixStep_1024     6,13,%((_RR_)+3),2

+    MixStep_1024     4, 9,%((_RR_)+3),3

+    MixStep_1024    14, 1,%((_RR_)+3),4

+    MixStep_1024     8, 5,%((_RR_)+3),5

+    MixStep_1024    10, 3,%((_RR_)+3),6

+    MixStep_1024    12, 7,%((_RR_)+3),7,1

+

+    inc   edx                     ;edx = round number

+    movd  xmm7,edx

+    ;inject the key

+irp _NN_,<15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0>

+  if _UNROLL_CNT ne (ROUNDS_1024/8)

+    if _NN_ lt R_1024_REGS

+      paddq xmm&_NN_,ptr64 [SI_+ksKey+16*_NN_+16]

+    else

+      mov64 xmm6    ,ptr64 [esp+X_stk+ 8*_NN_]

+     if     _NN_ eq 15

+      paddq xmm6,xmm7

+     elseif _NN_ eq 14

+      paddq xmm6,ptr64 [SI_+ksTwk+16*2]

+     elseif _NN_ eq 13

+      paddq xmm6,ptr64 [SI_+ksTwk+16*1]

+     endif

+      paddq xmm6    ,ptr64 [SI_+ksKey+16*_NN_+16]

+      mov64          ptr64 [esp+X_stk+ 8*_NN_],xmm6

+    endif

+  else

+    if _NN_ lt R_1024_REGS

+      paddq xmm&_NN_,ptr64 [FP_+ksKey+16*(((_Rbase_/4)+(_NN_)+1) mod 17)]

+    else

+      mov64 xmm6,ptr64 [esp+X_stk+ 8*_NN_]

+      paddq xmm6,ptr64 [FP_+ksKey+16*(((_Rbase_/4)+(_NN_)+1) mod 17)]

+     if     _NN_ eq 15

+      paddq xmm6,xmm7

+     elseif _NN_ eq 14

+      paddq xmm6,ptr64 [FP_+ksTwk+16*(((_Rbase_/4)+2) mod  3)]

+     elseif _NN_ eq 13

+      paddq xmm6,ptr64 [FP_+ksTwk+16*(((_Rbase_/4)+1) mod  3)]

+     endif

+      mov64      ptr64 [esp+X_stk+ 8*_NN_],xmm6

+    endif

+  endif

+endm

+if _UNROLL_CNT ne (ROUNDS_1024/8) ;rotate the key schedule on the stack

+    mov64 xmm6,ptr64 [SI_+ksKey]

+    mov64 xmm7,ptr64 [SI_+ksTwk]

+    mov64      ptr64 [SI_+ksKey+16*(WCNT+1)],xmm6

+    mov64      ptr64 [SI_+ksTwk+16* 3      ],xmm7

+    add   esi,16                  ;bump rolling pointer

+endif

+if _SKEIN_DEBUG

+      Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,saveRegs

+endif

+endm ;R_1024_FourRounds

+;;;;;;;;;;;;;;;;

+; code

+;

+_Skein1024_Process_Block proc near

+;

+    WCNT    =   16                  ;WCNT=16 for Skein-1024

+    Setup_Stack WCNT,ROUNDS_1024

+    add     edi,80h                 ;bias the edi ctxt offsets to keep them all short

+ctx equ    <edi-80h>                ;offset alias

+    ; main hash loop for Skein1024

+Skein1024_block_loop:

+    movd    xmm0,ptr32 [ebx+bitAdd]

+    mov64   xmm1,ptr64 [ctx+TWEAK+0]

+    mov64   xmm2,ptr64 [ctx+TWEAK+8]

+    paddq   xmm1,xmm0               ;bump T0 by the bitAdd parameter

+    mov64   ptr64 [ctx+TWEAK],xmm1  ;save updated tweak value T0 (for next time)

+    mov64   xmm0,xmm2

+    xorpd   xmm0,xmm1               ;compute overall tweak parity

+    movdqa  [FP_+ksTwk   ],xmm1     ;save the expanded tweak schedule on the stack

+    movdqa  [FP_+ksTwk+16],xmm2

+    movdqa  [FP_+ksTwk+32],xmm0

+

+    mov     esi,[ebx+blkPtr]        ;esi --> input block

+    mov     eax,KW_PARITY_LO        ;init key schedule parity accumulator

+    mov     edx,KW_PARITY_HI

+    movd    xmm7,eax

+    movd    xmm6,edx

+    unpcklps xmm7,xmm6              ;pack two 32-bit words into xmm7

+;

+    lea     eax,[esp+80h]           ;use short offsets for Wcopy, X_stk writes below

+SP_ equ    <eax-80h>                ;[eax+OFFS] mode is one byte shorter than [esp+OFFS]

+irp _NN_,<15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0>

+    mov64   xmm6,ptr64 [ctx+X_VARS+8*_NN_]

+    xorpd   xmm7,xmm6               ;update overall parity

+    movdqa  [FP_+ksKey+16*_NN_],xmm6;save the key schedule on the stack

+  if _NN_ lt R_1024_REGS

+     _rr_  =  _NN_

+  else

+     _rr_  =   R_1024_REGS

+  endif

+  irp _rn_,<%(_rr_)>

+    mov64   xmm&_rn_,ptr64 [esi+         8*_NN_];save copy of the input block on stack

+    mov64            ptr64 [SP_+ Wcopy + 8*_NN_],xmm&_rn_   ;(for feedforward later)

+    paddq   xmm&_rn_,xmm6               ;inject the key into the block

+   if _NN_ eq 13

+    paddq   xmm&_rn_,xmm1               ;inject the initial tweak words

+   elseif _NN_ eq 14

+    paddq   xmm&_rn_,xmm2

+   endif

+   if _NN_ ge R_1024_REGS               ;only save X[5..15] on stack, leave X[0..4] in regs

+    mov64   ptr64 [SP_+X_stk+8*_NN_],xmm&_rn_ 

+   endif

+  endm

+endm

+    movdqa  [FP_+ksKey+16*WCNT],xmm7;save overall key parity at the end of the array

+;

+if _SKEIN_DEBUG                     ;debug dump of state at this point

+    Skein_Debug_Block 1024

+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL,saveRegs

+endif

+    add     esi, WCNT*8             ;skip to the next block

+    mov         [ebx+blkPtr],esi    ;save the updated block pointer

+    ;

+    ; now the key schedule is computed. Start the rounds

+    ;

+    xor     edx,edx                 ;edx = round counter

+if SKEIN_ASM_UNROLL and 1024

+_UNROLL_CNT =   ROUNDS_1024/8

+else

+_UNROLL_CNT =   SKEIN_UNROLL_1024

+  if ((ROUNDS_1024/8) mod _UNROLL_CNT)

+    .err "Invalid SKEIN_UNROLL_1024"

+  endif

+    mov     esi,ebp                 ;use this as "rolling" pointer into ksTwk/ksKey

+Skein_1024_round_loop:

+endif

+;

+_Rbase_ = 0

+rept _UNROLL_CNT*2

+    R_1024_FourRounds %_Rbase_

+_Rbase_ = _Rbase_+4

+endm ;rept _UNROLL_CNT

+;

+if (SKEIN_ASM_UNROLL and 1024) eq 0

+    cmp     edx,2*(ROUNDS_1024/8)

+    jb      Skein_1024_round_loop

+endif

+    and     ptr08 [ctx +TWEAK +15],FIRST_MASK8      ;clear tweak bit for next time thru

+    ;----------------------------

+    ; feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..15}

+    lea     eax,[esp+80h]                           ;allow short offsets to X_stk and Wcopy

+irp _NN_,<0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15>

+  if _NN_ lt R_1024_REGS

+    if _NN_ and 1                                   ;already in regs: no load needed

+      mov64 xmm7  ,ptr64 [SP_+ Wcopy + 8*_NN_]      ;unaligned

+      xorpd xmm&_NN_,xmm7

+    else

+      xorpd xmm&_NN_,    [SP_+ Wcopy + 8*_NN_]      ;aligned

+    endif

+      mov64        ptr64 [ctx+ X_vars+ 8*_NN_],xmm&_NN_

+  else

+      mov64   xmm7,ptr64 [SP_+ X_stk + 8*_NN_]      ;load X value from stack

+    if _NN_ and 1

+      mov64   xmm6,ptr64 [SP_+ Wcopy + 8*_NN_]      ;unaligned

+      xorpd   xmm7,xmm6

+    else

+      xorpd   xmm7,      [SP_+ Wcopy + 8*_NN_]      ;aligned

+    endif

+      mov64        ptr64 [ctx+ X_vars+ 8*_NN_],xmm7

+ endif

+endm

+if _SKEIN_DEBUG

+    Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD   ;no need to save regs on stack here

+endif

+    ; go back for more blocks, if needed

+    dec     ecx

+    jnz     Skein1024_block_loop

+

+    Reset_Stack _Skein1024_Process_Block

+    ret

+_Skein1024_Process_Block endp

+;

+ifdef _SKEIN_CODE_SIZE

+    public  _Skein1024_Process_Block_CodeSize

+_Skein1024_Process_Block_CodeSize proc

+    mov     eax,_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block

+    ret

+_Skein1024_Process_Block_CodeSize endp

+;

+    public  _Skein1024_Unroll_Cnt

+_Skein1024_Unroll_Cnt proc

+  if _UNROLL_CNT ne ROUNDS_1024/8

+    mov     eax,_UNROLL_CNT

+  else

+    xor     eax,eax

+  endif

+    ret

+_Skein1024_Unroll_Cnt endp

+endif

+;

+endif ; _USE_ASM_ and 1024

+;----------------------------------------------------------------

+    end

diff --git a/Additional_Implementations/skein_block_xmm32.s b/Additional_Implementations/skein_block_xmm32.s
new file mode 100644
index 000000000000..fa10bd2b98c1
--- /dev/null
+++ b/Additional_Implementations/skein_block_xmm32.s
@@ -0,0 +1,1110 @@
+#
+#----------------------------------------------------------------
+# 32-bit x86 assembler code for Skein block functions using XMM registers
+#
+# Author: Doug Whiting, Hifn/Exar
+#
+# This code is released to the public domain.
+#----------------------------------------------------------------
+#
+    .text
+    .altmacro                               #use advanced macro features
+    .psize 0,128                            #list file has no page boundaries
+#
+_MASK_ALL_  =   (256+512+1024)              #all three algorithm bits
+SAVE_REGS   =   1
+#
+#################
+.ifndef SKEIN_USE_ASM
+_USE_ASM_        = _MASK_ALL_
+.elseif SKEIN_USE_ASM & _MASK_ALL_
+_USE_ASM_        = SKEIN_USE_ASM
+.else
+_USE_ASM_        = _MASK_ALL_
+.endif
+#
+#################
+.ifndef SKEIN_LOOP  
+_SKEIN_LOOP       = 002                     #default is all fully unrolled, except Skein1024
+.else
+_SKEIN_LOOP       = SKEIN_LOOP
+.endif
+#--------------
+# the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256  = (_SKEIN_LOOP / 100) % 10
+SKEIN_UNROLL_512  = (_SKEIN_LOOP /  10) % 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP      ) % 10
+#
+SKEIN_ASM_UNROLL  = 0
+  .irp _NN_,256,512,1024
+    .if (SKEIN_UNROLL_\_NN_) == 0
+SKEIN_ASM_UNROLL  = SKEIN_ASM_UNROLL + \_NN_
+    .endif
+  .endr
+#
+#################
+#
+.ifndef SKEIN_ROUNDS
+ROUNDS_256  =   72
+ROUNDS_512  =   72
+ROUNDS_1024 =   80
+.else
+ROUNDS_256  = 8*((((SKEIN_ROUNDS / 100) + 5) % 10) + 5)
+ROUNDS_512  = 8*((((SKEIN_ROUNDS /  10) + 5) % 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS      ) + 5) % 10) + 5)
+.irp _NN_,256,512,1024
+  .if _USE_ASM_ && \_NN_
+    .irp _RR_,%(ROUNDS_\_NN_)
+      .if \_NN_ < 1024
+.print  "+++ SKEIN_ROUNDS_\_NN_  = \_RR_"
+      .else
+.print  "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+      .endif
+    .endr
+  .endif
+.endr
+.endif
+#################
+#
+.ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE = (1)
+.else
+.ifdef  SKEIN_PERF                           #use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE = (1)
+.endif
+.endif
+#
+#################
+#
+.ifndef SKEIN_DEBUG
+_SKEIN_DEBUG      = 0
+.else
+_SKEIN_DEBUG      = 1
+.endif
+#################
+#
+# define offsets of fields in hash context structure
+#
+HASH_BITS   =   0                           ## bits of hash output
+BCNT        =   4 + HASH_BITS               #number of bytes in BUFFER[]
+TWEAK       =   4 + BCNT                    #tweak values[0..1]
+X_VARS      =  16 + TWEAK                   #chaining vars
+#
+#(Note: buffer[] in context structure is NOT needed here :-)
+#
+KW_PARITY_LO=   0xA9FC1A22                  #overall parity of key schedule words (hi32/lo32)
+KW_PARITY_HI=   0x1BD11BDA
+FIRST_MASK8 =   ~ (1 << 6)                  #FIRST block flag bit
+#
+# rotation constants for Skein
+#
+RC_256_0_0  = 14
+RC_256_0_1  = 16
+
+RC_256_1_0  = 52
+RC_256_1_1  = 57
+
+RC_256_2_0  = 23
+RC_256_2_1  = 40
+
+RC_256_3_0  =  5
+RC_256_3_1  = 37
+
+RC_256_4_0  = 25
+RC_256_4_1  = 33
+
+RC_256_5_0  = 46
+RC_256_5_1  = 12
+
+RC_256_6_0  = 58
+RC_256_6_1  = 22
+
+RC_256_7_0  = 32
+RC_256_7_1  = 32
+
+RC_512_0_0  = 46
+RC_512_0_1  = 36
+RC_512_0_2  = 19
+RC_512_0_3  = 37
+
+RC_512_1_0  = 33
+RC_512_1_1  = 27
+RC_512_1_2  = 14
+RC_512_1_3  = 42
+
+RC_512_2_0  = 17
+RC_512_2_1  = 49
+RC_512_2_2  = 36
+RC_512_2_3  = 39
+
+RC_512_3_0  = 44
+RC_512_3_1  =  9
+RC_512_3_2  = 54
+RC_512_3_3  = 56
+
+RC_512_4_0  = 39
+RC_512_4_1  = 30
+RC_512_4_2  = 34
+RC_512_4_3  = 24
+
+RC_512_5_0  = 13
+RC_512_5_1  = 50
+RC_512_5_2  = 10
+RC_512_5_3  = 17
+
+RC_512_6_0  = 25
+RC_512_6_1  = 29
+RC_512_6_2  = 39
+RC_512_6_3  = 43
+
+RC_512_7_0  =  8
+RC_512_7_1  = 35
+RC_512_7_2  = 56
+RC_512_7_3  = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 =  8
+RC_1024_0_3 = 47
+RC_1024_0_4 =  8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 =  4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 =  5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 =  9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 =  4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 =  9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+#
+#----------------------------------------------------------------
+# declare allocated space on the stack
+.macro StackVar  localName,localSize
+\localName  =   _STK_OFFS_
+_STK_OFFS_  =   _STK_OFFS_+(\localSize)
+.endm #StackVar
+#
+#----------------------------------------------------------------
+#
+# MACRO: Configure stack frame, allocate local vars
+#
+.macro Setup_Stack WCNT,RND_CNT
+_STK_OFFS_  =   0                   #starting offset from esp, forced on 16-byte alignment
+    #----- local  variables         #<-- esp
+    StackVar    X_stk  , 8*(WCNT)   #local context vars
+    StackVar    Wcopy  , 8*(WCNT)   #copy of input block    
+    StackVar    ksTwk  ,16*3        #key schedule: tweak words
+    StackVar    ksKey  ,16*(WCNT)+16#key schedule: key   words
+FRAME_OFFS  =   ksTwk+128           #<-- ebp
+F_O         =   FRAME_OFFS          #syntactic shorthand
+  .if (SKEIN_ASM_UNROLL && (WCNT*64)) == 0
+    StackVar    ksRot,16*(RND_CNT/4)#leave space for ks "rotation" to happen
+  .endif
+LOCAL_SIZE  =   _STK_OFFS_          #size of local vars
+    #
+    #"restart" the stack defns, because we relocate esp to guarantee alignment
+    #    (i.e., these vars are NOT at fixed offsets from esp)
+_STK_OFFS_  =   0
+    #----- 
+    StackVar    savRegs,8*4         #pushad data
+    StackVar    retAddr,4           #return address
+    #----- caller parameters
+    StackVar    ctxPtr ,4           #context ptr
+    StackVar    blkPtr ,4           #pointer to block data
+    StackVar    blkCnt ,4           #number of full blocks to process
+    StackVar    bitAdd ,4           #bit count to add to tweak
+    #----- caller's stack frame
+#
+# Notes on stack frame setup:
+#   * the most used variable (except for Skein-256) is X_stk[], based at [esp+0]
+#   * the next most used is the key schedule words
+#       so ebp is "centered" there, allowing short offsets to the key/tweak
+#       schedule in 256/512-bit Skein cases, but not posible for Skein-1024 :-(
+#   * the Wcopy variables are infrequently accessed, and they have long 
+#       offsets from both esp and ebp only in the 1024-bit case.
+#   * all other local vars and calling parameters can be accessed 
+#       with short offsets, except in the 1024-bit case
+#
+    pushal                          #save all regs
+    movl    %esp,%ebx               #keep ebx as pointer to caller parms
+    subl    $LOCAL_SIZE,%esp        #make room for the locals
+    andl    $~15,%esp               #force alignment
+    movl    ctxPtr(%ebx),%edi       #edi --> Skein context
+    leal    FRAME_OFFS(%esp),%ebp   #maximize use of short offsets from ebp
+    movl    blkCnt(%ebx),%ecx       #keep block cnt in ecx
+.endm #Setup_Stack
+#
+#----------------------------------------------------------------
+#
+.macro Reset_Stack,procStart
+    movl     %ebx,%esp              #get rid of locals (wipe??)
+    popal                           #restore all regs
+.endm # Reset_Stack
+#
+#----------------------------------------------------------------
+# macros to help debug internals
+#
+.if _SKEIN_DEBUG
+    .extern   _Skein_Show_Block   #calls to C routines
+    .extern   _Skein_Show_Round
+#
+SKEIN_RND_SPECIAL       =   1000
+SKEIN_RND_KEY_INITIAL   =   SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT    =   SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD      =   SKEIN_RND_SPECIAL+2
+#
+.macro Skein_Debug_Block BLK_BITS
+#
+#void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+#                     const u08b_t *blkPtr, const u64b_t *wPtr, 
+#                     const u64b_t *ksPtr,const u64b_t *tsPtr)#
+#
+    call    _Put_XMM_\BLK_BITS
+    pushal                          #save all regs
+    leal    ksTwk+1-F_O(%ebp),%eax  #+1 = flag: "stride" size = 2 qwords
+    leal    ksKey+1-F_O(%ebp),%esi
+    leal    Wcopy+32(%esp),%ecx     #adjust offset by 32 for pushad
+    movl    ctxPtr(%ebx)  ,%edx     #ctx_hdr_ptr
+    leal    X_VARS(%edx)  ,%edx     #edx ==> cxt->X[]
+    pushl   %eax                    #tsPtr
+    pushl   %esi                    #ksPtr
+    pushl   %ecx                    #wPtr
+    pushl   blkPtr(%ebx)            #blkPtr
+    pushl   %edx                    #ctx->Xptr
+    pushl   ctxPtr(%ebx)            #ctx_hdr_ptr
+    movl    $\BLK_BITS,%eax
+    pushl   %eax                    #bits
+    call    _Skein_Show_Block
+    addl    $7*4,%esp               #discard parameter space on stack
+    popal                           #restore regs
+#
+    call    _Get_XMM_\BLK_BITS
+.endm #Skein_Debug_Block
+
+#
+.macro Skein_Debug_Round BLK_BITS,R,saveRegs=0
+#
+#void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X)#
+#
+  .if \saveRegs
+    call    _Put_XMM_\BLK_BITS
+  .endif
+    pushal                          #save all regs
+  .if R <> SKEIN_RND_FEED_FWD
+    leal    32+X_stk(%esp),%eax     #adjust offset by 32 for pushal
+  .else
+    movl    ctxPtr(%ebx),%eax
+    addl    $X_VARS,%eax
+  .endif
+    pushl   %eax                    #Xptr
+  .if (SKEIN_ASM_UNROLL && \BLK_BITS) || (\R >= SKEIN_RND_SPECIAL)
+    movl    $\R,%eax
+  .else     #compute round number from edx, R
+    leal    1+(((\R)-1) && 3)(,%edx,4),%eax
+  .endif
+    pushl   %eax                    #round number
+    pushl   ctxPtr(%ebx)            #ctx_hdr_ptr
+    movl    $\BLK_BITS,%eax
+    pushl   %eax                    #bits
+    call    _Skein_Show_Round
+    addl    $4*4,%esp               #discard parameter space on stack
+    popal                           #restore regs
+  .if \saveRegs
+    call  _Get_XMM_\BLK_BITS        #save internal vars for debug dump
+  .endif
+.endm  #Skein_Debug_Round
+.endif #ifdef SKEIN_DEBUG
+#
+#----------------------------------------------------------------
+# useful macros
+.macro _ldX xn
+    movq          X_stk+8*(\xn)(%esp),%xmm\xn
+.endm
+
+.macro _stX xn
+    movq  %xmm\xn,X_stk+8*(\xn)(%esp)
+.endm
+#
+#----------------------------------------------------------------
+#
+.macro C_label lName
+ \lName:        #use both "genders" to work across linkage conventions
+_\lName:
+    .global  \lName
+    .global _\lName
+.endm
+#
+
+.if _USE_ASM_ & 256
+#
+# void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+#
+# Skein-256 round macros
+#
+.macro R_256_OneRound _RR_,x0,x1,x2,x3,t0,t1
+  .irp _qq_,%((\_RR_) && 7)        #figure out which rotation constants to use
+    .if \x0 == 0
+_RC0_ =   RC_256_\_qq_&&_0
+_RC1_ =   RC_256_\_qq_&&_1
+    .else
+_RC0_ =   RC_256_\_qq_&&_1
+_RC1_ =   RC_256_\_qq_&&_0
+    .endif
+  .endr
+#
+    paddq    %xmm\x1,%xmm\x0
+    movq     %xmm\x1,%xmm\t0
+    psllq  $   _RC0_,%xmm\x1
+    psrlq  $64-_RC0_,%xmm\t0
+    xorpd    %xmm\x0,%xmm\x1
+    xorpd    %xmm\t0,%xmm\x1
+#                         
+    paddq    %xmm\x3,%xmm\x2
+    movq     %xmm\x3,%xmm\t1
+    psllq  $   _RC1_,%xmm\x3
+    psrlq  $64-_RC1_,%xmm\t1
+    xorpd    %xmm\x2,%xmm\x3
+    xorpd    %xmm\t1,%xmm\x3
+  .if _SKEIN_DEBUG
+    Skein_Debug_Round 256,%(\_RR_+1),SAVE_REGS
+  .endif
+.endm #R_256_OneRound
+#
+.macro R_256_FourRounds _RN_
+    R_256_OneRound %(_RN_+0),0,1,2,3,4,5
+    R_256_OneRound (_RN_+1),2,1,0,3,4,5
+
+    R_256_OneRound (_RN_+2),0,1,2,3,4,5
+    R_256_OneRound (_RN_+3),2,1,0,3,4,5
+
+    #inject key schedule
+    incl  %edx                     #bump round number
+    movd  %edx,%xmm4
+  .if _UNROLL_CNT == (ROUNDS_256/8)
+    #fully unrolled version
+_RK_ = ((_RN_)/4)                 #key injection counter
+    paddq ksKey+16*((_RK_+1) % 5)-F_O(%ebp),%xmm0
+    paddq ksKey+16*((_RK_+2) % 5)-F_O(%ebp),%xmm1
+    paddq ksKey+16*((_RK_+3) % 5)-F_O(%ebp),%xmm2
+    paddq ksKey+16*((_RK_+4) % 5)-F_O(%ebp),%xmm3
+    paddq ksTwk+16*((_RK_+1) % 3)-F_O(%ebp),%xmm1
+    paddq ksTwk+16*((_RK_+2) % 3)-F_O(%ebp),%xmm2
+    paddq %xmm4,%xmm3
+  .else #looping version
+    paddq ksKey+16*1-F_O(%esi),%xmm0
+    paddq ksKey+16*2-F_O(%esi),%xmm1
+    paddq ksKey+16*3-F_O(%esi),%xmm2
+    paddq ksKey+16*4-F_O(%esi),%xmm3
+    paddq ksTwk+16*1-F_O(%esi),%xmm1
+    paddq ksTwk+16*2-F_O(%esi),%xmm2
+    paddq %xmm4,%xmm3
+#   
+    movq        ksKey-F_O(%esi),%xmm4   #first, "rotate" key schedule on the stack
+    movq        ksTwk-F_O(%esi),%xmm5   #    (for next time through)
+    movq  %xmm4,ksKey+16*(WCNT+1)-F_O(%esi)
+    movq  %xmm5,ksTwk+16*3-F_O(%esi)
+    addl  $16,%esi                     #bump rolling pointer
+  .endif
+  .if _SKEIN_DEBUG
+      Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,SAVE_REGS
+  .endif
+.endm #R256_FourRounds
+#
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_256:
+  .irp _NN_,0,1,2,3
+    movq  %xmm\_NN_,X_stk+4+\_NN_*8(%esp)
+  .endr
+    ret
+#
+_Get_XMM_256:
+  .irp _NN_,0,1,2,3
+    movq            X_stk+4+_NN_*8(%esp),%xmm\_NN_
+  .endr
+    ret
+.endif
+#
+#################
+#
+# code
+#
+C_label Skein_256_Process_Block
+    WCNT    =   4                   #WCNT=4 for Skein-256
+    Setup_Stack WCNT,ROUNDS_256
+    # main hash loop for Skein_256
+Skein_256_block_loop:
+    movd    bitAdd (%ebx),%xmm4
+    movq    TWEAK+0(%edi),%xmm5
+    movq    TWEAK+8(%edi),%xmm6
+    paddq   %xmm4        ,%xmm5     #bump T0 by the bitAdd parameter
+    movq    %xmm5,TWEAK(%edi)       #save updated tweak value T0 (for next time)
+    movapd  %xmm6,%xmm7
+    xorpd   %xmm5,%xmm7             #compute overall tweak parity
+    movdqa  %xmm5,ksTwk   -F_O(%ebp)#save the expanded tweak schedule on the stack
+    movdqa  %xmm6,ksTwk+16-F_O(%ebp)        
+    movdqa  %xmm7,ksTwk+32-F_O(%ebp)        
+
+    movl    blkPtr(%ebx),%esi       #esi --> input block
+    movl    $KW_PARITY_LO,%eax      #init key schedule parity accumulator
+    movl    $KW_PARITY_HI,%edx 
+    movd    %eax ,%xmm4
+    movd    %edx ,%xmm0
+    unpcklps %xmm0,%xmm4            #replicate parity dword to 64 bits
+#
+  .irp _NN_,0,1,2,3                 #copy in the chaining vars
+    movq    X_VARS+8*\_NN_(%edi),%xmm\_NN_
+    xorpd   %xmm\_NN_,%xmm4         #update overall parity
+    movdqa  %xmm\_NN_,ksKey+16*_NN_-F_O(%ebp)
+  .endr
+    movdqa  %xmm4,ksKey+16*WCNT-F_O(%ebp)#save overall parity at the end of the array
+#
+    paddq   %xmm5,%xmm1             #inject the initial tweak words
+    paddq   %xmm6,%xmm2
+#
+  .irp _NN_,0,1,2,3                 #perform the initial key injection
+    movq          8*\_NN_(%esi),%xmm4#and save a copy of the input block on stack
+    movq    %xmm4,8*\_NN_+Wcopy(%esp)
+    paddq   %xmm4,%xmm\_NN_         #inject the key word
+  .endr
+#
+.if _SKEIN_DEBUG                    #debug dump of state at this point
+    Skein_Debug_Block 256
+    Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+    addl    $WCNT*8,%esi            #skip to the next block
+    movl    %esi,blkPtr(%ebx)       #save the updated block pointer
+    #
+    # now the key schedule is computed. Start the rounds
+    #
+    xorl    %edx,%edx               #edx = iteration count
+.if SKEIN_ASM_UNROLL & 256
+_UNROLL_CNT =   ROUNDS_256/8        #fully unrolled
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_256    #partial unroll count
+  .if ((ROUNDS_256/8) % _UNROLL_CNT)
+    .error "Invalid SKEIN_UNROLL_256" #sanity check
+  .endif
+    movl    %ebp,%esi               #use this as "rolling" pointer into ksTwk/ksKey
+Skein_256_round_loop:               #   (since there's no 16* scaled address mode)
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2                  # here with X[0..3] in XMM0..XMM3
+      R_256_FourRounds _Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT*2
+#
+  .if _UNROLL_CNT <> (ROUNDS_256/8)
+    cmpl    $2*(ROUNDS_256/8),%edx
+    jb      Skein_256_round_loop
+  .endif
+    #----------------------------
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..3}
+  .irp _NN_,0,1,2,3
+    movq    Wcopy+8*\_NN_(%esp),%xmm4
+    xorpd   %xmm4,%xmm\_NN_
+    movq    %xmm\_NN_,X_VARS+8*\_NN_(%edi)
+  .endr
+    andb    $FIRST_MASK8,TWEAK +15(%edi)
+.if _SKEIN_DEBUG
+    Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,SAVE_REGS
+.endif
+    # go back for more blocks, if needed
+    decl    %ecx
+    jnz     Skein_256_block_loop
+    Reset_Stack _Skein_256_Process_Block
+    ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label  Skein_256_Process_Block_CodeSize
+    movl    $_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block,%eax
+    ret
+#
+C_label  Skein_256_Unroll_Cnt
+  .if _UNROLL_CNT <> ROUNDS_256/8
+    movl    $_UNROLL_CNT,%eax
+  .else
+    xorl    %eax,%eax
+  .endif
+    ret
+.endif
+.endif #_USE_ASM_ & 256
+#
+#----------------------------------------------------------------
+#
+.if _USE_ASM_ & 512
+#
+# void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+# MACRO: one round
+#
+.macro R_512_Round _RR_, a0,a1,Ra, b0,b1,Rb, c0,c1,Rc, d0,d1,Rd
+  .irp _qq_,%((\_RR_) && 7)
+_Ra_ = RC_512_\_qq_&&_\Ra
+_Rb_ = RC_512_\_qq_&&_\Rb
+_Rc_ = RC_512_\_qq_&&_\Rc
+_Rd_ = RC_512_\_qq_&&_\Rd
+  .endr
+    paddq   %xmm\a1 , %xmm\a0 
+                              _stX c0
+    movq    %xmm\a1 , %xmm\c0 
+    psllq  $   _Ra_ , %xmm\a1 
+    psrlq  $64-_Ra_ , %xmm\c0 
+    xorpd   %xmm\c0 , %xmm\a1 
+    xorpd   %xmm\a0 , %xmm\a1 
+                                    
+    paddq   %xmm\b1 , %xmm\b0 
+                              _stX a0
+    movq    %xmm\b1 , %xmm\a0 
+    psllq  $   _Rb_ , %xmm\b1 
+    psrlq  $64-_Rb_ , %xmm\a0 
+    xorpd   %xmm\b0 , %xmm\b1 
+                              _ldX c0
+    xorpd   %xmm\a0 , %xmm\b1 
+                               
+    paddq   %xmm\c1 , %xmm\c0 
+    movq    %xmm\c1 , %xmm\a0 
+    psllq  $   _Rc_ , %xmm\c1 
+    psrlq  $64-_Rc_ , %xmm\a0 
+    xorpd   %xmm\c0 , %xmm\c1 
+    xorpd   %xmm\a0 , %xmm\c1 
+                               
+    paddq   %xmm\d1 , %xmm\d0 
+    movq    %xmm\d1 , %xmm\a0           
+    psllq  $   _Rd_ , %xmm\d1 
+    psrlq  $64-_Rd_ , %xmm\a0 
+    xorpd   %xmm\a0 , %xmm\d1 
+                              _ldX a0
+    xorpd   %xmm\d0 , %xmm\d1 
+  .if _SKEIN_DEBUG
+    Skein_Debug_Round 512,%(_RR_+1),SAVE_REGS
+  .endif
+.endm
+#
+# MACRO: four rounds
+.macro R_512_FourRounds _RN_
+    R_512_Round %((_RN_)  ), 0,1,0, 2,3,1, 4,5,2, 6,7,3
+    R_512_Round %((_RN_)+1), 2,1,0, 4,7,1, 6,5,2, 0,3,3
+    R_512_Round %((_RN_)+2), 4,1,0, 6,3,1, 0,5,2, 2,7,3
+    R_512_Round %((_RN_)+3), 6,1,0, 0,7,1, 2,5,2, 4,3,3
+
+    #inject key schedule
+.irp _NN_,0,1,2,3,4,5,6,7
+  .if _UNROLL_CNT == (ROUNDS_512/8)
+    paddq ksKey+16*((((\_RN_)/4)+(\_NN_)+1)%9)-F_O(%ebp),%xmm\_NN_
+  .else
+    paddq ksKey+16*((\_NN_)+1)-F_O(%esi),%xmm\_NN_
+  .endif
+.endr
+    _stX  0                       #free up a register
+    incl  %edx                    #bump round counter
+    movd  %edx,%xmm0              #inject the tweak
+  .if _UNROLL_CNT == (ROUNDS_512/8)
+    paddq ksTwk+16*(((_RN_)+1) % 3)-F_O(%ebp),%xmm5
+    paddq ksTwk+16*(((_RN_)+2) % 3)-F_O(%ebp),%xmm6
+    paddq %xmm0                              ,%xmm7
+  .else #looping version
+    paddq ksTwk+16*1-F_O(%esi),%xmm5
+    paddq ksTwk+16*2-F_O(%esi),%xmm6
+    paddq %xmm0               ,%xmm7
+    # "rotate" key schedule on the stack (for next time through)
+    movq        ksKey            -F_O(%esi),%xmm0
+    movq  %xmm0,ksKey+16*(WCNT+1)-F_O(%esi)
+    movq        ksTwk            -F_O(%esi),%xmm0
+    movq  %xmm0,ksTwk+16*3       -F_O(%esi)
+    addl  $16,%esi                #bump rolling pointer
+  .endif
+    _ldX  0                       #restore X0
+  .if _SKEIN_DEBUG
+    Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT,SAVE_REGS
+  .endif
+.endm #R_512_FourRounds
+#################
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_512:
+  .irp _NN_,0,1,2,3,4,5,6,7
+    movq  %xmm\_NN_,X_stk+4+\_NN_*8(%esp)
+  .endr
+    ret
+#
+_Get_XMM_512:
+  .irp _NN_,0,1,2,3,4,5,6,7
+    movq            X_stk+4+\_NN_*8(%esp),%xmm\_NN_
+  .endr
+    ret
+.endif
+#
+#################
+#
+C_label Skein_512_Process_Block
+    WCNT    =   8                   #WCNT=8 for Skein-512
+    Setup_Stack WCNT,ROUNDS_512
+    # main hash loop for Skein_512
+Skein_512_block_loop:
+    movd    bitAdd(%ebx) ,%xmm0
+    movq    TWEAK+0(%edi),%xmm1
+    movq    TWEAK+8(%edi),%xmm2
+    paddq   %xmm0,%xmm1               #bump T0 by the bitAdd parameter
+    movq    %xmm1,TWEAK(%edi)         #save updated tweak value T0 (for next time)
+    movq    %xmm2,%xmm0
+    xorpd   %xmm1,%xmm0               #compute overall tweak parity
+    movdqa  %xmm1,ksTwk     -F_O(%ebp)#save the expanded tweak schedule on the stack
+    movdqa  %xmm2,ksTwk+16*1-F_O(%ebp)    
+    movdqa  %xmm0,ksTwk+16*2-F_O(%ebp)    
+
+    movl    blkPtr(%ebx),%esi         #esi --> input block
+    movl    $KW_PARITY_LO,%eax        #init key schedule parity accumulator
+    movl    $KW_PARITY_HI,%edx 
+    movd    %eax ,%xmm0
+    movd    %edx ,%xmm7
+    unpcklps %xmm7,%xmm0              #replicate parity dword to 64 bits
+#
+  .irp _NN_,7,6,5,4,3,2,1             #copy in the chaining vars (skip #0 for now)
+    movq    X_VARS+8*\_NN_(%edi),%xmm\_NN_
+    xorpd   %xmm\_NN_,%xmm0           #update overall parity
+    movdqa  %xmm\_NN_,ksKey+16*\_NN_-F_O(%ebp)
+   .if \_NN_ == 5
+    paddq   %xmm1,%xmm5               #inject the initial tweak words
+    paddq   %xmm2,%xmm6               #  (before they get trashed in %xmm1/2)
+   .endif
+  .endr
+    movq    X_VARS(%edi),%xmm4        #handle #0 now
+    xorpd   %xmm4,%xmm0               #update overall parity
+    movdqa  %xmm4,ksKey+16* 0  -F_O(%ebp) #save the key value in slot #0
+    movdqa  %xmm0,ksKey+16*WCNT-F_O(%ebp) #save overall parity at the end of the array
+#
+    movq    %xmm4,%xmm0
+  .irp _NN_,7,6,5,  4,3,2,1,0         #perform the initial key injection (except #4)
+    movq    8*\_NN_(%esi),%xmm4       #and save a copy of the input block on stack
+    movq    %xmm4,8*\_NN_+Wcopy(%esp)
+    paddq   %xmm4,%xmm\_NN_
+  .endr
+    movq    8*4(%esi),%xmm4           #get input block word #4
+    movq    %xmm4,8*4+Wcopy(%esp)
+    paddq   ksKey+16*4-F_O(%ebp),%xmm4#inject the initial key
+#
+.if _SKEIN_DEBUG                      #debug dump of state at this point
+    Skein_Debug_Block 512
+    Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+    addl    $WCNT*8,%esi              #skip to the next block
+    movl    %esi,blkPtr(%ebx)         #save the updated block pointer
+    #
+    # now the key schedule is computed. Start the rounds
+    #
+    xorl    %edx,%edx                 #edx = round counter
+.if SKEIN_ASM_UNROLL & 512
+_UNROLL_CNT =   ROUNDS_512/8
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_512
+  .if ((ROUNDS_512/8) % _UNROLL_CNT)
+    .error "Invalid SKEIN_UNROLL_512"
+  .endif
+    movl    %ebp,%esi                 #use this as "rolling" pointer into ksTwk/ksKey
+Skein_512_round_loop:                 #   (since there's no 16* scaled address mode)
+.endif
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+      R_512_FourRounds %_Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 512) == 0
+    cmpl    $2*(ROUNDS_512/8),%edx
+    jb      Skein_512_round_loop
+.endif
+    #----------------------------
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..7}
+    andb    $FIRST_MASK8,TWEAK +15(%edi)
+.irp _NN_,0,2,4,6                   #do the aligned ones first
+    xorpd   Wcopy+8*\_NN_(%esp),%xmm\_NN_
+    movq    %xmm\_NN_,X_VARS+8*_NN_(%edi)
+.endr
+.irp _NN_,1,3,5,7                   #now we have some register space available
+    movq    Wcopy+8*\_NN_(%esp),%xmm0
+    xorpd   %xmm0,%xmm&\_NN_
+    movq    %xmm&\_NN_,X_VARS+8*\_NN_(%edi)
+.endr
+.if _SKEIN_DEBUG
+    Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+.endif
+    # go back for more blocks, if needed
+    decl    %ecx
+    jnz     Skein_512_block_loop
+
+    Reset_Stack _Skein_512_Process_Block
+    ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label Skein_512_Process_Block_CodeSize
+    movl    $(_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block),%eax
+    ret
+#
+C_label Skein_512_Unroll_Cnt
+  .if _UNROLL_CNT <> ROUNDS_512/8
+    movl    $_UNROLL_CNT,%eax
+  .else
+    xorl    %eax,%eax
+  .endif
+    ret
+.endif
+#
+.endif # _USE_ASM_ & 512
+#
+#----------------------------------------------------------------
+#
+.if _USE_ASM_ & 1024
+    .global      _Skein1024_Process_Block
+#
+# void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+R_1024_REGS =     (5)     #keep this many block variables in registers
+#
+################
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_1024:
+_NN_ = 0
+  .rept R_1024_REGS
+   .irp _rr_,%(_NN_)
+    movq   %xmm\_rr_,X_stk+4+8*_NN_(%esp)
+   .endr
+_NN_ = _NN_+1
+  .endr
+    ret
+#
+_Get_XMM_1024:
+_NN_ = 0
+  .rept R_1024_REGS
+   .irp _rr_,%(_NN_)
+    movq             X_stk+4+8*_NN_(%esp),%xmm\_rr_
+   .endr
+_NN_ = _NN_+1
+  .endr
+    ret
+.endif
+#
+#################
+# MACRO: one mix step
+.macro MixStep_1024  x0,x1,rotIdx0,rotIdx1,_debug_=0
+_r0_ =  \x0      #default, if already loaded
+_r1_ =  \x1
+  # load the regs (if necessary)
+  .if (\x0 >= R_1024_REGS)
+_r0_ =       5
+    movq    X_stk+8*(\x0)(%esp),%xmm5
+  .endif
+  .if (\x1 >= R_1024_REGS)
+_r1_ =       6     
+    movq  X_stk+8*(\x1)(%esp),%xmm6
+  .endif
+  # do the mix
+  .irp _rx_,%((rotIdx0) && 7)
+_Rc_ = RC_1024_\_rx_&&_\rotIdx1  #rotation constant
+  .endr
+  .irp _x0_,%_r0_
+  .irp _x1_,%_r1_
+    paddq   %xmm\_x1_,%xmm\_x0_
+    movq    %xmm\_x1_,%xmm7    
+    psllq  $   _Rc_  ,%xmm\_x1_
+    psrlq  $64-_Rc_  ,%xmm7    
+    xorpd   %xmm\_x0_,%xmm\_x1_
+    xorpd   %xmm7    ,%xmm\_x1_
+  .endr
+  .endr
+  # save the regs (if necessary)
+  .if (\x0 >= R_1024_REGS)
+    movq    %xmm5,X_stk+8*(\x0)(%esp)
+  .endif
+  .if (\x1 >= R_1024_REGS)
+    movq    %xmm6,X_stk+8*(\x1)(%esp)
+  .endif
+  # debug output
+  .if _SKEIN_DEBUG && (\_debug_)
+    Skein_Debug_Round 1024,%((\RotIdx0)+1),SAVE_REGS
+  .endif
+.endm
+#################
+# MACRO: four rounds
+#
+.macro R_1024_FourRounds _RR_
+    #--------- round _RR_
+    MixStep_1024     0, 1,%((\_RR_)+0),0
+    MixStep_1024     2, 3,%((\_RR_)+0),1
+    MixStep_1024     4, 5,%((\_RR_)+0),2
+    MixStep_1024     6, 7,%((\_RR_)+0),3
+    MixStep_1024     8, 9,%((\_RR_)+0),4
+    MixStep_1024    10,11,%((\_RR_)+0),5
+    MixStep_1024    12,13,%((\_RR_)+0),6
+    MixStep_1024    14,15,%((\_RR_)+0),7,1
+    #--------- round _RR_+1
+    MixStep_1024     0, 9,%((\_RR_)+1),0
+    MixStep_1024     2,13,%((\_RR_)+1),1
+    MixStep_1024     6,11,%((\_RR_)+1),2
+    MixStep_1024     4,15,%((\_RR_)+1),3
+    MixStep_1024    10, 7,%((\_RR_)+1),4
+    MixStep_1024    12, 3,%((\_RR_)+1),5
+    MixStep_1024    14, 5,%((\_RR_)+1),6
+    MixStep_1024     8, 1,%((\_RR_)+1),7,1
+    #--------- round _RR_+2
+    MixStep_1024     0, 7,%((\_RR_)+2),0    
+    MixStep_1024     2, 5,%((\_RR_)+2),1
+    MixStep_1024     4, 3,%((\_RR_)+2),2    
+    MixStep_1024     6, 1,%((\_RR_)+2),3    
+    MixStep_1024    12,15,%((\_RR_)+2),4
+    MixStep_1024    14,13,%((\_RR_)+2),5    
+    MixStep_1024     8,11,%((\_RR_)+2),6    
+    MixStep_1024    10, 9,%((\_RR_)+2),7,1
+    #--------- round _RR_+3
+    MixStep_1024     0,15,%((\_RR_)+3),0
+    MixStep_1024     2,11,%((\_RR_)+3),1
+    MixStep_1024     6,13,%((\_RR_)+3),2
+    MixStep_1024     4, 9,%((\_RR_)+3),3
+    MixStep_1024    14, 1,%((\_RR_)+3),4
+    MixStep_1024     8, 5,%((\_RR_)+3),5
+    MixStep_1024    10, 3,%((\_RR_)+3),6
+    MixStep_1024    12, 7,%((\_RR_)+3),7,1
+
+    incl  %edx                     #edx = round number
+    movd  %edx,%xmm7
+
+    #inject the key
+.irp _NN_,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+  .if _UNROLL_CNT <> (ROUNDS_1024/8)
+    .if \_NN_ < R_1024_REGS
+      paddq ksKey+16*\_NN_+16-F_O(%esi),%xmm&\_NN_
+    .else
+      movq  X_stk+ 8*\_NN_(%esp),%xmm6
+     .if     \_NN_ == 15
+      paddq %xmm7,%xmm6
+     .elseif \_NN_ == 14
+      paddq ksTwk+16*2-F_O(%esi),%xmm6
+     .elseif \_NN_ == 13
+      paddq ksTwk+16*1-F_O(%esi),%xmm6
+     .endif
+      paddq       ksKey+16*\_NN_+16-F_O(%esi),%xmm6
+      movq  %xmm6,X_stk+ 8*\_NN_(%esp)
+    .endif
+  .else
+    .if \_NN_ < R_1024_REGS
+      paddq ksKey+16*(((_Rbase_/4)+(\_NN_)+1) % 17)-F_O(%ebp),%xmm&\_NN_
+    .else
+      movq  X_stk+ 8*\_NN_(%esp), %xmm6
+      paddq ksKey+16*(((_Rbase_/4)+(\_NN_)+1) % 17)-F_O(%ebp),%xmm6
+     .if     \_NN_ == 15
+      paddq %xmm7,%xmm6
+     .elseif \_NN_ == 14
+      paddq ksTwk+16*(((_Rbase_/4)+2) %  3)-F_O(%ebp),%xmm6
+     .elseif \_NN_ == 13
+      paddq ksTwk+16*(((_Rbase_/4)+1) %  3)-F_O(%ebp),%xmm6
+     .endif
+      movq %xmm6,X_stk+ 8*\_NN_(%esp)
+    .endif
+  .endif
+.endr
+  .if _UNROLL_CNT <> (ROUNDS_1024/8) #rotate the key schedule on the stack
+    movq ksKey-F_O(%esi), %xmm6
+    movq ksTwk-F_O(%esi), %xmm7
+    movq %xmm6,ksKey+16*(WCNT+1)-F_O(%esi)
+    movq %xmm7,ksTwk+16* 3      -F_O(%esi)
+    addl $16,%esi                   #bump rolling pointer
+  .endif
+  .if _SKEIN_DEBUG
+      Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,SAVE_REGS
+  .endif
+.endm #R_1024_FourRounds
+#
+################
+#
+C_label Skein1024_Process_Block
+#
+    WCNT    =   16                  #WCNT=16 for Skein-1024
+    Setup_Stack WCNT,ROUNDS_1024
+    addl    $0x80,%edi              #bias the edi ctxt offsets to keep them all short
+    # main hash loop for Skein1024
+Skein1024_block_loop:
+    movd    bitAdd(%ebx)      ,%xmm0
+    movq    TWEAK+0-0x80(%edi),%xmm1
+    movq    TWEAK+8-0x80(%edi),%xmm2
+    paddq   %xmm0,%xmm1             #bump T0 by the bitAdd parameter
+    movq    %xmm1,TWEAK-0x80(%edi)  #save updated tweak value T0 (for next time)
+    movq    %xmm2,%xmm0
+    xorpd   %xmm1,%xmm0             #compute overall tweak parity
+    movdqa  %xmm1,ksTwk   -F_O(%ebp)#save the expanded tweak schedule on the stack
+    movdqa  %xmm2,ksTwk+16-F_O(%ebp)
+    movdqa  %xmm0,ksTwk+32-F_O(%ebp)
+
+    movl    blkPtr(%ebx),%esi       #esi --> input block
+    movl    $KW_PARITY_LO,%eax      #init key schedule parity accumulator
+    movl    $KW_PARITY_HI,%edx 
+    movd    %eax ,%xmm7
+    movd    %edx ,%xmm6
+    unpcklps %xmm6,%xmm7            #replicate parity dword to 64 bits
+#
+    leal    0x80(%esp),%eax         #use short offsets for Wcopy, X_stk writes below
+.irp _NN_,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+    movq    X_VARS+8*\_NN_-0x80(%edi),%xmm6
+    xorpd   %xmm6,%xmm7             #update overall parity
+    movdqa  %xmm6,ksKey+16*\_NN_-F_O(%ebp) #save the key schedule on the stack
+  .if \_NN_ < R_1024_REGS
+    _rr_  =  \_NN_
+  .else
+    _rr_  =   R_1024_REGS
+  .endif
+  .irp _rn_,%(_rr_)
+    movq    8*\_NN_(%esi),%xmm\_rn_ #save copy of the input block on stack
+    movq    %xmm\_rn_,Wcopy+8*\_NN_-0x80(%eax)  #(for feedforward later)
+    paddq   %xmm6,%xmm\_rn_         #inject the key into the block
+   .if \_NN_ == 13
+    paddq   %xmm1,%xmm\_rn_         #inject the initial tweak words
+   .elseif \_NN_ == 14
+    paddq   %xmm2,%xmm\_rn_
+   .endif
+   .if \_NN_ >= R_1024_REGS         #only save X[5..15] on stack, leave X[0..4] in regs
+    movq    %xmm\_rn_,X_stk+8*\_NN_-0x80(%eax)
+   .endif
+  .endr
+.endr
+    movdqa  %xmm7,ksKey+16*WCNT-F_O(%ebp) #save overall key parity at the end of the array
+#
+.if _SKEIN_DEBUG                    #debug dump of state at this point
+    Skein_Debug_Block 1024
+    Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+    addl    $WCNT*8,%esi            #skip to the next block
+    movl    %esi,blkPtr(%ebx)       #save the updated block pointer
+    #
+    # now the key schedule is computed. Start the rounds
+    #
+    xorl    %edx,%edx               #edx = round counter
+.if SKEIN_ASM_UNROLL & 1024
+_UNROLL_CNT =   ROUNDS_1024/8
+.else
+_UNROLL_CNT =   SKEIN_UNROLL_1024
+  .if ((ROUNDS_1024/8) % _UNROLL_CNT)
+    .error "Invalid SKEIN_UNROLL_1024"
+  .endif
+    movl    %ebp,%esi               #use this as "rolling" pointer into ksTwk/ksKey
+Skein_1024_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+    R_1024_FourRounds %_Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 1024) == 0
+    cmp     $2*(ROUNDS_1024/8),%edx
+    jb      Skein_1024_round_loop
+.endif
+    andb    $FIRST_MASK8,TWEAK +15-0x80(%edi)      #clear tweak bit for next time thru
+    #----------------------------
+    # feedforward:   ctx->X[i] = X[i] ^ w[i], {i=0..15}
+    leal    0x80(%esp),%eax                        #allow short offsets to X_stk and Wcopy
+.irp _NN_,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+  .if \_NN_ < R_1024_REGS
+    .if \_NN_ && 1                                 #already in regs: no load needed
+      movq  Wcopy+ 8*\_NN_-0x80(%eax),%xmm7        #unaligned
+      xorpd %xmm7,%xmm\_NN_
+    .else
+      xorpd Wcopy+ 8*\_NN_-0x80(%eax),%xmm\_NN_    #aligned
+    .endif
+      movq  %xmm\_NN_,X_VARS+8*\_NN_-0x80(%edi)
+  .else
+      movq    X_stk+8*\_NN_-0x80(%eax),%xmm7       #load X value from stack
+    .if \_NN_ && 1
+      movq    Wcopy+8*\_NN_-0x80(%eax),%xmm6       #unaligned
+      xorpd   %xmm6,%xmm7
+    .else
+      xorpd   Wcopy+8*\_NN_-0x80(%eax),%xmm7       #aligned
+    .endif
+      movq    %xmm7,X_VARS+8*\_NN_-0x80(%edi)
+ .endif
+.endr
+.if _SKEIN_DEBUG
+    Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD   #no need to save regs on stack here
+.endif
+    # go back for more blocks, if needed
+    decl    %ecx
+    jnz     Skein1024_block_loop
+
+    Reset_Stack _Skein1024_Process_Block
+    ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label Skein1024_Process_Block_CodeSize
+    movl    $(_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block),%eax
+    ret
+#
+C_label Skein1024_Unroll_Cnt
+  .if _UNROLL_CNT <> ROUNDS_1024/8
+    movl    $_UNROLL_CNT,%eax
+  .else
+    xorl    %eax,%eax
+  .endif
+    ret
+.endif
+#
+.endif # _USE_ASM_ & 1024
+#----------------------------------------------------------------
+    .end
diff --git a/Additional_Implementations/skein_perf_core2.txt b/Additional_Implementations/skein_perf_core2.txt
new file mode 100644
index 000000000000..d8b795675c3d
--- /dev/null
+++ b/Additional_Implementations/skein_perf_core2.txt
@@ -0,0 +1,1440 @@
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:27:59,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  3450.00  3450.00  |  8718.00  8718.00  | 41700.00 41730.00  | //: 32-bit, GCC_v3.42 [ C =...]
+      2_ ||  1719.00  1725.00  |  4323.00  4326.00  | 20835.00 20850.00  | //: 32-bit, GCC_v3.42 [ C =...]
+      4_ ||   861.00   861.00  |  2149.50  2151.00  | 10408.50 11277.00  | //: 32-bit, GCC_v3.42 [ C =...]
+      8_ ||   429.00   429.75  |  1074.75  1074.75  |  5204.25  5205.00  | //: 32-bit, GCC_v3.42 [ C =...]
+     10_ ||   343.80   344.40  |   865.80   866.40  |  4167.00  4167.60  | //: 32-bit, GCC_v3.42 [ C =...]
+     16_ ||   214.88   214.88  |   538.50   538.50  |  2603.25  2603.63  | //: 32-bit, GCC_v3.42 [ C =...]
+     32_ ||   107.06   115.88  |   269.25   269.25  |  1301.25  1301.25  | //: 32-bit, GCC_v3.42 [ C =...]
+     64_ ||    85.31    85.41  |   132.66   132.75  |   650.53   650.63  | //: 32-bit, GCC_v3.42 [ C =...]
+    100_ ||    82.20    88.86  |   126.78   126.78  |   416.46   416.46  | //: 32-bit, GCC_v3.42 [ C =...]
+    128_ ||    69.42    69.56  |    97.83    97.83  |   324.98   325.03  | //: 32-bit, GCC_v3.42 [ C =...]
+    256_ ||    56.70    56.74  |    76.34    76.34  |   242.95   242.98  | //: 32-bit, GCC_v3.42 [ C =...]
+    512_ ||    53.06    53.12  |    65.50    65.53  |   200.66   200.67  | //: 32-bit, GCC_v3.42 [ C =...]
+   1000_ ||    52.33    52.42  |    61.66    61.69  |   183.89   183.92  | //: 32-bit, GCC_v3.42 [ C =...]
+   1024_ ||    51.15    51.23  |    60.07    60.08  |   179.52   179.55  | //: 32-bit, GCC_v3.42 [ C =...]
+   2048_ ||    50.20    50.30  |    57.36    57.42  |   168.86   168.97  | //: 32-bit, GCC_v3.42 [ C =...]
+   4096_ ||    49.71    49.77  |    56.00    56.01  |   163.65   166.96  | //: 32-bit, GCC_v3.42 [ C =...]
+   8192_ ||    49.48    50.94  |    55.33    57.07  |   169.60   184.62  | //: 32-bit, GCC_v3.42 [ C =...]
+  10000_ ||    53.64    53.70  |    60.89    60.99  |   186.13   186.98  | //: 32-bit, GCC_v3.42 [ C =...]
+  16384_ ||    53.48    53.80  |    60.35    60.70  |   164.26   167.23  | //: 32-bit, GCC_v3.42 [ C =...]
+  32768_ ||    53.47    53.50  |    60.22    60.37  |   164.15   182.33  | //: 32-bit, GCC_v3.42 [ C =...]
+ 100000_ ||    53.85   100.93  |    60.43    61.36  |   164.25   169.78  | //: 32-bit, GCC_v3.42 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [ C =...]
+  Block  ||        14464 bytes |        32544 bytes |        83024 bytes | //: 32-bit, GCC_v3.42 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:28:12,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2802.00  2814.00  |  5952.00  5952.00  | 30606.00 30606.00  | //: 32-bit, MSC_v9.00 [ C =...]
+      2_ ||  1392.00  1395.00  |  2976.00  2979.00  | 15309.00 15309.00  | //: 32-bit, MSC_v9.00 [ C =...]
+      4_ ||   696.00   697.50  |  1486.50  1486.50  |  7653.00  7654.50  | //: 32-bit, MSC_v9.00 [ C =...]
+      8_ ||   347.25   348.00  |   741.75   742.50  |  3825.75  3827.25  | //: 32-bit, MSC_v9.00 [ C =...]
+     10_ ||   278.40   278.40  |   593.40   593.40  |  3063.00  3063.00  | //: 32-bit, MSC_v9.00 [ C =...]
+     16_ ||   174.38   174.38  |   370.50   370.50  |  1913.25  1913.25  | //: 32-bit, MSC_v9.00 [ C =...]
+     32_ ||    86.25    86.25  |   186.00   186.75  |   957.00   957.19  | //: 32-bit, MSC_v9.00 [ C =...]
+     64_ ||    62.91    62.91  |    92.91    92.91  |   478.50   478.50  | //: 32-bit, MSC_v9.00 [ C =...]
+    100_ ||    65.52    65.58  |    88.02    88.08  |   306.30   306.30  | //: 32-bit, MSC_v9.00 [ C =...]
+    128_ ||    50.72    50.72  |    68.53    68.58  |   238.64   238.88  | //: 32-bit, MSC_v9.00 [ C =...]
+    256_ ||    44.88    45.05  |    56.11    56.13  |   178.17   178.24  | //: 32-bit, MSC_v9.00 [ C =...]
+    512_ ||    41.79    41.86  |    49.79    49.91  |   147.39   147.47  | //: 32-bit, MSC_v9.00 [ C =...]
+   1000_ ||    41.26    41.41  |    47.96    47.96  |   135.28   135.29  | //: 32-bit, MSC_v9.00 [ C =...]
+   1024_ ||    40.40    40.40  |    46.79    46.81  |   132.05   132.08  | //: 32-bit, MSC_v9.00 [ C =...]
+   2048_ ||    39.62    39.62  |    45.23    45.23  |   124.39   124.40  | //: 32-bit, MSC_v9.00 [ C =...]
+   4096_ ||    38.98    38.99  |    44.34    44.44  |   120.58   120.60  | //: 32-bit, MSC_v9.00 [ C =...]
+   8192_ ||    38.83    38.87  |    44.06    47.57  |   118.65   119.36  | //: 32-bit, MSC_v9.00 [ C =...]
+  10000_ ||    38.86    39.08  |    44.13    44.21  |   119.88   120.11  | //: 32-bit, MSC_v9.00 [ C =...]
+  16384_ ||    38.74    39.03  |    43.76    44.01  |   108.36   117.94  | //: 32-bit, MSC_v9.00 [ C =...]
+  32768_ ||    36.77    38.19  |    41.28    41.57  |   105.50   114.79  | //: 32-bit, MSC_v9.00 [ C =...]
+ 100000_ ||    38.85    39.09  |    43.56    43.77  |   105.79   114.18  | //: 32-bit, MSC_v9.00 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =...]
+  Block  ||        10192 bytes |        22960 bytes |        53072 bytes | //: 32-bit, MSC_v9.00 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:28:29,Oct  7 2008  by  'MSC_v6.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  8688.00  8712.00  | 26466.00 26472.00  | 61638.00 61680.00  | //: 32-bit, MSC_v6.00 [ C =...]
+      2_ ||  4347.00  4362.00  | 13293.00 13302.00  | 30036.00 30372.00  | //: 32-bit, MSC_v6.00 [ C =...]
+      4_ ||  2184.00  2199.00  |  6457.50  6508.50  | 15267.00 15285.00  | //: 32-bit, MSC_v6.00 [ C =...]
+      8_ ||  1093.50  1098.75  |  3227.25  3227.25  |  7398.75  7467.75  | //: 32-bit, MSC_v6.00 [ C =...]
+     10_ ||   873.60   878.40  |  2405.40  2574.00  |  5661.00  5668.20  | //: 32-bit, MSC_v6.00 [ C =...]
+     16_ ||   522.00   524.25  |  1455.00  1455.38  |  3459.38  3489.38  | //: 32-bit, MSC_v6.00 [ C =...]
+     32_ ||   260.06   261.00  |   727.69   732.56  |  1727.44  1728.00  | //: 32-bit, MSC_v6.00 [ C =...]
+     64_ ||   186.66   186.94  |   362.16   362.25  |   848.25   856.97  | //: 32-bit, MSC_v6.00 [ C =...]
+    100_ ||   194.10   194.10  |   344.52   344.58  |   542.22   545.28  | //: 32-bit, MSC_v6.00 [ C =...]
+    128_ ||   151.27   151.36  |   266.06   268.59  |   426.23   426.33  | //: 32-bit, MSC_v6.00 [ C =...]
+    256_ ||   137.67   137.91  |   219.66   219.68  |   314.74   317.74  | //: 32-bit, MSC_v6.00 [ C =...]
+    512_ ||   130.21   130.22  |   195.96   218.33  |   263.45   266.96  | //: 32-bit, MSC_v6.00 [ C =...]
+   1000_ ||   129.17   129.60  |   183.96   185.04  |   237.88   240.37  | //: 32-bit, MSC_v6.00 [ C =...]
+   1024_ ||   126.35   126.53  |   178.10   178.44  |   251.47   256.21  | //: 32-bit, MSC_v6.00 [ C =...]
+   2048_ ||   133.77   133.81  |   190.95   191.16  |   240.21   242.87  | //: 32-bit, MSC_v6.00 [ C =...]
+   4096_ ||   116.36   124.81  |   169.39   178.51  |   220.31   222.41  | //: 32-bit, MSC_v6.00 [ C =...]
+   8192_ ||   123.60   125.60  |   171.36   174.54  |   215.79   233.44  | //: 32-bit, MSC_v6.00 [ C =...]
+  10000_ ||   124.42   127.19  |   183.83   188.08  |   236.37   238.04  | //: 32-bit, MSC_v6.00 [ C =...]
+  16384_ ||   133.15   133.46  |   172.38   183.65  |   222.17   232.39  | //: 32-bit, MSC_v6.00 [ C =...]
+  32768_ ||   126.67   128.66  |   180.67   186.04  |   225.01   227.24  | //: 32-bit, MSC_v6.00 [ C =...]
+ 100000_ ||   123.83   125.06  |   172.26   186.33  |   205.87   224.72  | //: 32-bit, MSC_v6.00 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1486 bytes |         1348 bytes |         1445 bytes | //: 32-bit, MSC_v6.00 [ C =...]
+  Block  ||        14094 bytes |        35580 bytes |        69258 bytes | //: 32-bit, MSC_v6.00 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:28:44,Oct  7 2008  by  'MSC_v4.20', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5028.00  5058.00  |  9948.00 10044.00  | 35466.00 35520.00  | //: 32-bit, MSC_v4.20 [ C =...]
+      2_ ||  2508.00  2511.00  |  5070.00  5076.00  | 18090.00 18132.00  | //: 32-bit, MSC_v4.20 [ C =...]
+      4_ ||  1255.50  1255.50  |  2523.00  2523.00  |  9063.00  9063.00  | //: 32-bit, MSC_v4.20 [ C =...]
+      8_ ||   627.75   627.75  |  1261.50  1261.50  |  4536.00  4536.00  | //: 32-bit, MSC_v4.20 [ C =...]
+     10_ ||   502.20   502.20  |  1013.40  1014.60  |  3685.80  3685.80  | //: 32-bit, MSC_v4.20 [ C =...]
+     16_ ||   313.88   313.88  |   624.75   632.63  |  2284.88  2287.88  | //: 32-bit, MSC_v4.20 [ C =...]
+     32_ ||   155.25   155.25  |   312.38   312.38  |  1143.75  1143.75  | //: 32-bit, MSC_v4.20 [ C =...]
+     64_ ||   114.56   114.66  |   155.72   155.72  |   569.91   569.91  | //: 32-bit, MSC_v4.20 [ C =...]
+    100_ ||   120.66   120.78  |   148.92   148.98  |   363.60   363.66  | //: 32-bit, MSC_v4.20 [ C =...]
+    128_ ||    93.84    93.89  |   116.58   116.63  |   284.58   284.58  | //: 32-bit, MSC_v4.20 [ C =...]
+    256_ ||    83.46    83.48  |    95.20    95.20  |   213.77   213.77  | //: 32-bit, MSC_v4.20 [ C =...]
+    512_ ||    78.18    78.19  |    85.08    85.09  |   177.38   177.38  | //: 32-bit, MSC_v4.20 [ C =...]
+   1000_ ||    77.42    77.42  |    81.88    81.88  |   161.92   161.92  | //: 32-bit, MSC_v4.20 [ C =...]
+   1024_ ||    75.54    75.55  |    79.53    79.53  |   158.23   158.23  | //: 32-bit, MSC_v4.20 [ C =...]
+   2048_ ||    74.22    74.23  |    77.37    77.38  |   148.59   149.28  | //: 32-bit, MSC_v4.20 [ C =...]
+   4096_ ||    73.56    73.57  |    76.51    76.51  |   142.82   145.02  | //: 32-bit, MSC_v4.20 [ C =...]
+   8192_ ||    73.23    73.24  |    73.69    74.43  |   143.02   143.30  | //: 32-bit, MSC_v4.20 [ C =...]
+  10000_ ||    73.30    73.36  |    73.75    75.53  |   144.04   144.80  | //: 32-bit, MSC_v4.20 [ C =...]
+  16384_ ||    69.35    73.11  |    71.79    73.01  |   131.31   140.35  | //: 32-bit, MSC_v4.20 [ C =...]
+  32768_ ||    73.05    73.36  |    72.43    74.54  |   126.01   139.95  | //: 32-bit, MSC_v4.20 [ C =...]
+ 100000_ ||    69.28    70.12  |    66.33    67.35  |   129.68   136.97  | //: 32-bit, MSC_v4.20 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1152 bytes |         1024 bytes |         1088 bytes | //: 32-bit, MSC_v4.20 [ C =...]
+  Block  ||        11968 bytes |        23776 bytes |        55360 bytes | //: 32-bit, MSC_v4.20 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:28:57,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   780.00   786.00  |  1110.00  1110.00  |  3288.00  3318.00  | //: 64-bit, MSC_v9.00 [ C =...]
+      2_ ||   402.00   402.00  |   549.00   552.00  |  1659.00  1659.00  | //: 64-bit, MSC_v9.00 [ C =...]
+      4_ ||   199.50   201.00  |   274.50   276.00  |   829.50   829.50  | //: 64-bit, MSC_v9.00 [ C =...]
+      8_ ||    96.75    97.50  |   134.25   135.00  |   414.75   414.75  | //: 64-bit, MSC_v9.00 [ C =...]
+     10_ ||    78.60    79.80  |   109.80   109.80  |   331.20   331.80  | //: 64-bit, MSC_v9.00 [ C =...]
+     16_ ||    48.38    48.38  |    67.13    67.13  |   224.25   224.63  | //: 64-bit, MSC_v9.00 [ C =...]
+     32_ ||    26.63    26.81  |    36.38    36.38  |   112.31   112.31  | //: 64-bit, MSC_v9.00 [ C =...]
+     64_ ||    17.06    17.06  |    16.78    16.78  |    51.66    51.75  | //: 64-bit, MSC_v9.00 [ C =...]
+    100_ ||    16.74    16.80  |    15.54    15.54  |    33.30    33.30  | //: 64-bit, MSC_v9.00 [ C =...]
+    128_ ||    12.98    13.08  |    11.95    12.00  |    25.78    25.83  | //: 64-bit, MSC_v9.00 [ C =...]
+    256_ ||    10.99    10.99  |     9.19     9.21  |    19.03    19.03  | //: 64-bit, MSC_v9.00 [ C =...]
+    512_ ||    10.14    10.18  |     7.84     7.85  |    15.60    15.60  | //: 64-bit, MSC_v9.00 [ C =...]
+   1000_ ||     9.88    10.67  |     7.38     7.38  |    14.16    14.17  | //: 64-bit, MSC_v9.00 [ C =...]
+   1024_ ||     9.60     9.64  |     7.18     7.18  |    13.74    13.74  | //: 64-bit, MSC_v9.00 [ C =...]
+   2048_ ||     9.35     9.38  |     6.83     6.83  |    12.84    12.84  | //: 64-bit, MSC_v9.00 [ C =...]
+   4096_ ||     9.28     9.28  |     6.69     6.70  |    12.40    12.40  | //: 64-bit, MSC_v9.00 [ C =...]
+   8192_ ||     9.18     9.21  |     6.58     6.59  |    12.28    12.28  | //: 64-bit, MSC_v9.00 [ C =...]
+  10000_ ||     9.21     9.22  |     6.60     6.60  |    12.27    12.39  | //: 64-bit, MSC_v9.00 [ C =...]
+  16384_ ||     9.19     9.20  |     6.53     6.55  |    12.12    12.12  | //: 64-bit, MSC_v9.00 [ C =...]
+  32768_ ||     9.16     9.17  |     6.51     6.55  |    12.08    12.53  | //: 64-bit, MSC_v9.00 [ C =...]
+ 100000_ ||     9.98    10.01  |     7.04     7.08  |    12.36    13.14  | //: 64-bit, MSC_v9.00 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =...]
+  Block  ||         2272 bytes |         4944 bytes |        15264 bytes | //: 64-bit, MSC_v9.00 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:28:59,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  6204.00  6252.00  | 11058.00 11124.00  | 25662.00 25788.00  | //: 32-bit, BCC_v5.51 [ C =...]
+      2_ ||  3048.00  3060.00  |  5469.00  5481.00  | 12576.00 12672.00  | //: 32-bit, BCC_v5.51 [ C =...]
+      4_ ||  1515.00  1521.00  |  2731.50  2733.00  |  6303.00  6348.00  | //: 32-bit, BCC_v5.51 [ C =...]
+      8_ ||   756.75   760.50  |  1364.25  1367.25  |  3147.75  3162.75  | //: 32-bit, BCC_v5.51 [ C =...]
+     10_ ||   605.40   607.80  |  1092.60  1095.00  |  2541.60  2545.80  | //: 32-bit, BCC_v5.51 [ C =...]
+     16_ ||   379.50   380.62  |   682.88   683.25  |  1584.38  1590.00  | //: 32-bit, BCC_v5.51 [ C =...]
+     32_ ||   187.88   188.62  |   340.69   341.06  |   794.81   797.62  | //: 32-bit, BCC_v5.51 [ C =...]
+     64_ ||   138.19   138.28  |   169.69   169.78  |   420.75   420.84  | //: 32-bit, BCC_v5.51 [ C =...]
+    100_ ||   145.02   145.08  |   160.80   160.86  |   269.16   269.52  | //: 32-bit, BCC_v5.51 [ C =...]
+    128_ ||   112.92   112.92  |   125.39   125.44  |   210.00   210.23  | //: 32-bit, BCC_v5.51 [ C =...]
+    256_ ||   100.27   100.29  |   103.08   103.08  |   156.33   156.42  | //: 32-bit, BCC_v5.51 [ C =...]
+    512_ ||    93.98    94.00  |    91.90    91.91  |   129.40   129.41  | //: 32-bit, BCC_v5.51 [ C =...]
+   1000_ ||    93.02    93.02  |    88.44    88.44  |   118.61   118.69  | //: 32-bit, BCC_v5.51 [ C =...]
+   1024_ ||    90.80    90.81  |    86.34    86.34  |   115.78   115.86  | //: 32-bit, BCC_v5.51 [ C =...]
+   2048_ ||    89.22    89.22  |    77.11    83.54  |    95.12    95.16  | //: 32-bit, BCC_v5.51 [ C =...]
+   4096_ ||    81.62    81.62  |    75.83    75.83  |    92.17   100.45  | //: 32-bit, BCC_v5.51 [ C =...]
+   8192_ ||    81.80    88.11  |    75.18    78.19  |    90.69    92.09  | //: 32-bit, BCC_v5.51 [ C =...]
+  10000_ ||    81.32    84.70  |    76.65    78.80  |    92.85    94.82  | //: 32-bit, BCC_v5.51 [ C =...]
+  16384_ ||    83.13    83.59  |    76.92    77.00  |    92.05    93.27  | //: 32-bit, BCC_v5.51 [ C =...]
+  32768_ ||    83.07    84.01  |    76.76    77.91  |    92.12    94.08  | //: 32-bit, BCC_v5.51 [ C =...]
+ 100000_ ||    83.48    84.08  |    77.08    78.59  |    93.38   102.21  | //: 32-bit, BCC_v5.51 [ C =...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [ C =...]
+  Block  ||        10732 bytes |        20964 bytes |        45988 bytes | //: 32-bit, BCC_v5.51 [ C =...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:07,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2592.00  2604.00  |  4848.00  4854.00  | 22278.00 22284.00  | //: 32-bit, BCC_v5.51 [asm=...]
+      2_ ||  1287.00  1293.00  |  2430.00  2430.00  | 11139.00 11139.00  | //: 32-bit, BCC_v5.51 [asm=...]
+      4_ ||   637.50   639.00  |  1213.50  1213.50  |  5565.00  5566.50  | //: 32-bit, BCC_v5.51 [asm=...]
+      8_ ||   318.75   319.50  |   606.75   606.75  |  2782.50  2783.25  | //: 32-bit, BCC_v5.51 [asm=...]
+     10_ ||   255.60   255.60  |   486.00   486.60  |  2228.40  2228.40  | //: 32-bit, BCC_v5.51 [asm=...]
+     16_ ||   159.75   159.75  |   301.88   302.25  |  1391.25  1391.62  | //: 32-bit, BCC_v5.51 [asm=...]
+     32_ ||    78.75    78.75  |   151.31   151.31  |   695.44   695.62  | //: 32-bit, BCC_v5.51 [asm=...]
+     64_ ||    55.69    57.28  |    74.81    74.91  |   347.81   347.81  | //: 32-bit, BCC_v5.51 [asm=...]
+    100_ ||    57.42    57.48  |    69.84    69.90  |   222.60   222.60  | //: 32-bit, BCC_v5.51 [asm=...]
+    128_ ||    44.53    44.58  |    54.38    54.38  |   173.67   173.67  | //: 32-bit, BCC_v5.51 [asm=...]
+    256_ ||    38.55    38.55  |    43.99    44.02  |   129.05   129.05  | //: 32-bit, BCC_v5.51 [asm=...]
+    512_ ||    35.60    35.60  |    38.66    38.67  |   106.62   106.62  | //: 32-bit, BCC_v5.51 [asm=...]
+   1000_ ||    34.89    34.89  |    37.18    37.18  |    97.72    97.72  | //: 32-bit, BCC_v5.51 [asm=...]
+   1024_ ||    34.23    34.98  |    35.85    35.86  |    95.40    95.40  | //: 32-bit, BCC_v5.51 [asm=...]
+   2048_ ||    33.86    33.86  |    34.66    34.66  |    89.79    89.80  | //: 32-bit, BCC_v5.51 [asm=...]
+   4096_ ||    33.22    33.59  |    33.92    34.26  |    86.99    86.99  | //: 32-bit, BCC_v5.51 [asm=...]
+   8192_ ||    33.11    33.11  |    33.80    33.92  |    74.64    77.45  | //: 32-bit, BCC_v5.51 [asm=...]
+  10000_ ||    31.46    33.40  |    31.80    32.58  |    78.61    78.66  | //: 32-bit, BCC_v5.51 [asm=...]
+  16384_ ||    31.49    32.78  |    32.16    33.51  |    76.07    76.13  | //: 32-bit, BCC_v5.51 [asm=...]
+  32768_ ||    32.01    32.58  |    32.74    33.18  |    75.73    76.02  | //: 32-bit, BCC_v5.51 [asm=...]
+ 100000_ ||    32.23    32.45  |    33.33    61.75  |    84.30    85.34  | //: 32-bit, BCC_v5.51 [asm=...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [asm=...]
+  Block  ||         7588 bytes |        16636 bytes |        38262 bytes | //: 32-bit, BCC_v5.51 [asm=...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:12,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2484.00  2490.00  |  4830.00  4836.00  | 22182.00 22188.00  | //: 32-bit, MSC_v9.00 [asm=...]
+      2_ ||  1254.00  1254.00  |  2415.00  2415.00  | 11091.00 11091.00  | //: 32-bit, MSC_v9.00 [asm=...]
+      4_ ||   627.00   627.00  |  1207.50  1207.50  |  5545.50  5545.50  | //: 32-bit, MSC_v9.00 [asm=...]
+      8_ ||   313.50   313.50  |   603.00   603.75  |  2390.25  2478.00  | //: 32-bit, MSC_v9.00 [asm=...]
+     10_ ||   250.20   252.00  |   485.40   488.40  |  1936.80  1959.00  | //: 32-bit, MSC_v9.00 [asm=...]
+     16_ ||   156.00   156.75  |   301.50   301.50  |  1386.00  1386.00  | //: 32-bit, MSC_v9.00 [asm=...]
+     32_ ||    77.81    77.81  |   150.94   151.31  |   692.81   692.81  | //: 32-bit, MSC_v9.00 [asm=...]
+     64_ ||    56.34    56.34  |    74.81    74.81  |   343.78   346.41  | //: 32-bit, MSC_v9.00 [asm=...]
+    100_ ||    58.62    58.68  |    70.74    70.80  |   221.76   221.76  | //: 32-bit, MSC_v9.00 [asm=...]
+    128_ ||    45.47    45.47  |    55.08    55.08  |   168.94   173.02  | //: 32-bit, MSC_v9.00 [asm=...]
+    256_ ||    40.10    40.10  |    44.95    44.95  |   128.88   128.88  | //: 32-bit, MSC_v9.00 [asm=...]
+    512_ ||    37.49    37.55  |    39.94    39.94  |    92.99    92.99  | //: 32-bit, MSC_v9.00 [asm=...]
+   1000_ ||    34.12    34.16  |    35.44    35.44  |    85.27    85.31  | //: 32-bit, MSC_v9.00 [asm=...]
+   1024_ ||    33.30    33.30  |    34.58    34.59  |    83.24    83.25  | //: 32-bit, MSC_v9.00 [asm=...]
+   2048_ ||    32.70    32.70  |    36.20    36.20  |    89.82    89.82  | //: 32-bit, MSC_v9.00 [asm=...]
+   4096_ ||    35.09    35.09  |    35.50    35.57  |    87.04    87.05  | //: 32-bit, MSC_v9.00 [asm=...]
+   8192_ ||    34.83    35.38  |    35.12    35.64  |    76.07    84.71  | //: 32-bit, MSC_v9.00 [asm=...]
+  10000_ ||    34.78    34.98  |    35.36    35.36  |    86.31    86.35  | //: 32-bit, MSC_v9.00 [asm=...]
+  16384_ ||    34.76    34.80  |    35.07    35.36  |    80.55    85.21  | //: 32-bit, MSC_v9.00 [asm=...]
+  32768_ ||    32.88    33.17  |    33.06    33.37  |    75.87    76.15  | //: 32-bit, MSC_v9.00 [asm=...]
+ 100000_ ||    32.96    33.40  |    33.29    33.60  |    75.79    76.81  | //: 32-bit, MSC_v9.00 [asm=...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=...]
+  Block  ||         7588 bytes |        16636 bytes |        38262 bytes | //: 32-bit, MSC_v9.00 [asm=...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:17,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2490.00  2496.00  |  4824.00  4836.00  | 22332.00 22356.00  | //: 32-bit, GCC_v3.42 [asm=...]
+      2_ ||  1251.00  1260.00  |  2412.00  2415.00  | 11157.00 11166.00  | //: 32-bit, GCC_v3.42 [asm=...]
+      4_ ||   621.00   622.50  |  1204.50  1204.50  |  5571.00  5572.50  | //: 32-bit, GCC_v3.42 [asm=...]
+      8_ ||   310.50   311.25  |   602.25   602.25  |  2785.50  2786.25  | //: 32-bit, GCC_v3.42 [asm=...]
+     10_ ||   249.00   249.60  |   482.40   482.40  |  2233.20  2233.80  | //: 32-bit, GCC_v3.42 [asm=...]
+     16_ ||   155.25   155.63  |   300.75   301.50  |  1393.88  1393.88  | //: 32-bit, GCC_v3.42 [asm=...]
+     32_ ||    76.50    77.06  |   151.31   151.31  |   696.38   696.56  | //: 32-bit, GCC_v3.42 [asm=...]
+     64_ ||    55.78    56.06  |    75.19    75.19  |   348.19   348.19  | //: 32-bit, GCC_v3.42 [asm=...]
+    100_ ||    58.32    58.44  |    70.80    70.80  |   222.96   222.96  | //: 32-bit, GCC_v3.42 [asm=...]
+    128_ ||    45.14    45.52  |    55.08    55.13  |   173.72   173.77  | //: 32-bit, GCC_v3.42 [asm=...]
+    256_ ||    40.03    40.13  |    44.91    44.93  |   129.33   129.33  | //: 32-bit, GCC_v3.42 [asm=...]
+    512_ ||    37.38    37.50  |    39.77    39.79  |   106.58   106.66  | //: 32-bit, GCC_v3.42 [asm=...]
+   1000_ ||    36.94    37.03  |    38.19    38.19  |    97.66    97.69  | //: 32-bit, GCC_v3.42 [asm=...]
+   1024_ ||    35.75    36.13  |    37.24    37.24  |    95.29    95.32  | //: 32-bit, GCC_v3.42 [asm=...]
+   2048_ ||    35.36    35.44  |    35.94    35.94  |    88.77    89.67  | //: 32-bit, GCC_v3.42 [asm=...]
+   4096_ ||    35.02    35.02  |    35.31    35.38  |    77.07    86.35  | //: 32-bit, GCC_v3.42 [asm=...]
+   8192_ ||    32.18    32.20  |    32.30    32.31  |    74.72    77.04  | //: 32-bit, GCC_v3.42 [asm=...]
+  10000_ ||    32.28    32.34  |    32.41    32.43  |    78.36    78.77  | //: 32-bit, GCC_v3.42 [asm=...]
+  16384_ ||    32.16    33.29  |    32.20    33.16  |    76.16    78.39  | //: 32-bit, GCC_v3.42 [asm=...]
+  32768_ ||    33.20    33.70  |    33.22    33.33  |    75.84    76.10  | //: 32-bit, GCC_v3.42 [asm=...]
+ 100000_ ||    33.26    33.96  |    33.11    33.41  |    75.75    76.29  | //: 32-bit, GCC_v3.42 [asm=...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [asm=...]
+  Block  ||         7588 bytes |        16636 bytes |        38262 bytes | //: 32-bit, GCC_v3.42 [asm=...]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:29:22,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   672.00   672.00  |  1068.00  1068.00  |  1920.00  1926.00  | //: 64-bit, MSC_v9.00 [asm=...]
+      2_ ||   336.00   336.00  |   534.00   534.00  |   963.00   963.00  | //: 64-bit, MSC_v9.00 [asm=...]
+      4_ ||   166.50   168.00  |   267.00   267.00  |   481.50   483.00  | //: 64-bit, MSC_v9.00 [asm=...]
+      8_ ||    81.00    81.00  |   130.50   131.25  |   240.00   240.75  | //: 64-bit, MSC_v9.00 [asm=...]
+     10_ ||    64.80    65.40  |   107.40   108.00  |   192.00   192.60  | //: 64-bit, MSC_v9.00 [asm=...]
+     16_ ||    40.13    40.13  |    65.63    65.63  |   120.00   120.00  | //: 64-bit, MSC_v9.00 [asm=...]
+     32_ ||    20.06    20.06  |    32.81    32.81  |    59.63    59.81  | //: 64-bit, MSC_v9.00 [asm=...]
+     64_ ||    14.25    14.34  |    16.31    16.31  |    32.44    32.44  | //: 64-bit, MSC_v9.00 [asm=...]
+    100_ ||    15.54    15.60  |    16.20    16.26  |    21.06    21.06  | //: 64-bit, MSC_v9.00 [asm=...]
+    128_ ||    11.81    11.86  |    11.44    11.48  |    14.86    14.86  | //: 64-bit, MSC_v9.00 [asm=...]
+    256_ ||     9.28     9.28  |     8.81     8.81  |    10.83    10.83  | //: 64-bit, MSC_v9.00 [asm=...]
+    512_ ||     8.43     8.43  |     7.46     7.46  |     8.66     8.66  | //: 64-bit, MSC_v9.00 [asm=...]
+   1000_ ||     8.18     8.18  |     6.97     6.97  |     7.77     7.78  | //: 64-bit, MSC_v9.00 [asm=...]
+   1024_ ||     7.98     8.50  |     6.81     7.38  |     7.58     7.58  | //: 64-bit, MSC_v9.00 [asm=...]
+   2048_ ||     7.75     7.75  |     6.47     6.47  |     7.05     7.05  | //: 64-bit, MSC_v9.00 [asm=...]
+   4096_ ||     7.65     7.65  |     6.30     6.30  |     6.78     6.78  | //: 64-bit, MSC_v9.00 [asm=...]
+   8192_ ||     7.59     7.59  |     6.21     6.21  |     6.64     6.64  | //: 64-bit, MSC_v9.00 [asm=...]
+  10000_ ||     7.59     7.59  |     6.23     6.23  |     6.69     6.69  | //: 64-bit, MSC_v9.00 [asm=...]
+  16384_ ||     7.57     7.57  |     6.17     6.17  |     6.57     6.57  | //: 64-bit, MSC_v9.00 [asm=...]
+  32768_ ||     7.55     7.56  |     6.15     6.15  |     6.53     6.55  | //: 64-bit, MSC_v9.00 [asm=...]
+ 100000_ ||     7.55     7.71  |     6.14     6.38  |     6.56     6.86  | //: 64-bit, MSC_v9.00 [asm=...]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=...]
+  Block  ||         2323 bytes |         4733 bytes |        11817 bytes | //: 64-bit, MSC_v9.00 [asm=...]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:24,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  4728.00  4728.00  |  8352.00  8352.00  | 20034.00 20040.00  | //: 32-bit, GCC_v3.42 [ C =111]
+      2_ ||  2370.00  2370.00  |  4179.00  4179.00  |  9261.00  9264.00  | //: 32-bit, GCC_v3.42 [ C =111]
+      4_ ||  1092.00  1096.50  |  1924.50  1926.00  |  4624.50  4624.50  | //: 32-bit, GCC_v3.42 [ C =111]
+      8_ ||   544.50   545.25  |  1040.25  1047.75  |  2312.25  2313.00  | //: 32-bit, GCC_v3.42 [ C =111]
+     10_ ||   436.20   436.80  |   768.60   769.20  |  1852.20  1852.20  | //: 32-bit, GCC_v3.42 [ C =111]
+     16_ ||   272.63   273.00  |   480.38   519.38  |  1156.88  1157.25  | //: 32-bit, GCC_v3.42 [ C =111]
+     32_ ||   135.94   135.94  |   240.56   240.75  |   579.00   579.00  | //: 32-bit, GCC_v3.42 [ C =111]
+     64_ ||   100.88   101.53  |   129.75   129.84  |   289.59   289.69  | //: 32-bit, GCC_v3.42 [ C =111]
+    100_ ||   106.44   106.44  |   113.94   114.18  |   185.46   200.94  | //: 32-bit, GCC_v3.42 [ C =111]
+    128_ ||    83.06    83.06  |    89.11    89.11  |   144.61   144.61  | //: 32-bit, GCC_v3.42 [ C =111]
+    256_ ||    73.83    79.99  |    73.34    79.45  |   107.55   107.55  | //: 32-bit, GCC_v3.42 [ C =111]
+    512_ ||    69.16    69.18  |    65.32    65.39  |    88.89    88.92  | //: 32-bit, GCC_v3.42 [ C =111]
+   1000_ ||    68.45    68.45  |    62.84    62.92  |    81.36    81.38  | //: 32-bit, GCC_v3.42 [ C =111]
+   1024_ ||    66.83    66.86  |    61.34    61.34  |    79.42    79.43  | //: 32-bit, GCC_v3.42 [ C =111]
+   2048_ ||    65.67    65.73  |    59.33    59.33  |    74.70    74.71  | //: 32-bit, GCC_v3.42 [ C =111]
+   4096_ ||    65.08    65.15  |    58.33    58.33  |    72.33    72.34  | //: 32-bit, GCC_v3.42 [ C =111]
+   8192_ ||    65.76    70.08  |    62.66    62.66  |    77.08    77.15  | //: 32-bit, GCC_v3.42 [ C =111]
+  10000_ ||    70.01    70.33  |    62.84    62.92  |    77.70    77.70  | //: 32-bit, GCC_v3.42 [ C =111]
+  16384_ ||    69.93    70.32  |    62.63    62.71  |    72.64    72.73  | //: 32-bit, GCC_v3.42 [ C =111]
+  32768_ ||    69.31    69.90  |    58.90    59.54  |    73.37    76.24  | //: 32-bit, GCC_v3.42 [ C =111]
+ 100000_ ||    67.54    70.40  |    59.09    59.39  |    72.65    73.26  | //: 32-bit, GCC_v3.42 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [ C =111]
+  Block  ||         2928 bytes |         5568 bytes |        11712 bytes | //: 32-bit, GCC_v3.42 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:31,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2952.00  2958.00  |  6030.00  6036.00  | 13668.00 13674.00  | //: 32-bit, MSC_v9.00 [ C =111]
+      2_ ||  1476.00  1476.00  |  3015.00  3015.00  |  6831.00  6834.00  | //: 32-bit, MSC_v9.00 [ C =111]
+      4_ ||   738.00   739.50  |  1507.50  1507.50  |  3415.50  3415.50  | //: 32-bit, MSC_v9.00 [ C =111]
+      8_ ||   369.00   369.75  |   751.50   751.50  |  1707.00  1707.00  | //: 32-bit, MSC_v9.00 [ C =111]
+     10_ ||   295.80   295.80  |   603.00   603.60  |  1366.80  1366.80  | //: 32-bit, MSC_v9.00 [ C =111]
+     16_ ||   184.88   185.25  |   376.50   376.50  |   855.38   855.38  | //: 32-bit, MSC_v9.00 [ C =111]
+     32_ ||    91.31    91.50  |   188.44   188.63  |   427.50   427.50  | //: 32-bit, MSC_v9.00 [ C =111]
+     64_ ||    66.56    66.66  |    93.84    93.84  |   213.56   213.66  | //: 32-bit, MSC_v9.00 [ C =111]
+    100_ ||    69.96    70.02  |    88.98    89.04  |   136.92   137.52  | //: 32-bit, MSC_v9.00 [ C =111]
+    128_ ||    54.14    54.23  |    69.52    69.75  |   106.69   106.88  | //: 32-bit, MSC_v9.00 [ C =111]
+    256_ ||    47.70    47.77  |    57.12    57.19  |    79.24    79.29  | //: 32-bit, MSC_v9.00 [ C =111]
+    512_ ||    44.46    44.54  |    50.75    50.81  |    65.52    65.55  | //: 32-bit, MSC_v9.00 [ C =111]
+   1000_ ||    43.90    43.96  |    48.78    48.85  |    60.08    60.11  | //: 32-bit, MSC_v9.00 [ C =111]
+   1024_ ||    42.83    42.87  |    47.44    47.65  |    58.49    58.51  | //: 32-bit, MSC_v9.00 [ C =111]
+   2048_ ||    42.17    42.17  |    45.83    45.83  |    55.01    55.16  | //: 32-bit, MSC_v9.00 [ C =111]
+   4096_ ||    41.76    41.76  |    45.02    45.03  |    53.27    53.44  | //: 32-bit, MSC_v9.00 [ C =111]
+   8192_ ||    38.35    41.55  |    41.20    41.26  |    48.54    51.37  | //: 32-bit, MSC_v9.00 [ C =111]
+  10000_ ||    41.53    41.59  |    44.76    44.80  |    53.01    53.01  | //: 32-bit, MSC_v9.00 [ C =111]
+  16384_ ||    41.38    41.69  |    44.43    44.47  |    52.00    52.07  | //: 32-bit, MSC_v9.00 [ C =111]
+  32768_ ||    41.36    41.38  |    44.43    44.44  |    51.86    52.07  | //: 32-bit, MSC_v9.00 [ C =111]
+ 100000_ ||    41.32    41.60  |    44.52    44.62  |    51.75    51.92  | //: 32-bit, MSC_v9.00 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =111]
+  Block  ||         1712 bytes |         3664 bytes |         7200 bytes | //: 32-bit, MSC_v9.00 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:36,Oct  7 2008  by  'MSC_v6.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  9150.00  9156.00  | 16794.00 16836.00  | 36456.00 37386.00  | //: 32-bit, MSC_v6.00 [ C =111]
+      2_ ||  4335.00  4335.00  |  8244.00  8352.00  | 18156.00 18246.00  | //: 32-bit, MSC_v6.00 [ C =111]
+      4_ ||  2167.50  2167.50  |  4117.50  4201.50  |  9031.50  9060.00  | //: 32-bit, MSC_v6.00 [ C =111]
+      8_ ||  1083.00  1083.75  |  2122.50  2125.50  |  4515.00  4611.00  | //: 32-bit, MSC_v6.00 [ C =111]
+     10_ ||   874.80   874.80  |  1683.60  1695.60  |  3621.00  3705.00  | //: 32-bit, MSC_v6.00 [ C =111]
+     16_ ||   541.50   541.88  |  1041.38  1042.50  |  2268.38  2274.00  | //: 32-bit, MSC_v6.00 [ C =111]
+     32_ ||   271.88   272.25  |   515.63   526.13  |  1133.81  1139.06  | //: 32-bit, MSC_v6.00 [ C =111]
+     64_ ||   201.00   201.09  |   259.59   263.72  |   567.47   569.34  | //: 32-bit, MSC_v6.00 [ C =111]
+    100_ ||   211.92   211.98  |   250.32   251.10  |   363.06   363.60  | //: 32-bit, MSC_v6.00 [ C =111]
+    128_ ||   166.78   167.11  |   196.73   198.28  |   283.45   284.20  | //: 32-bit, MSC_v6.00 [ C =111]
+    256_ ||   147.94   147.94  |   160.57   160.71  |   212.18   212.72  | //: 32-bit, MSC_v6.00 [ C =111]
+    512_ ||   139.32   139.37  |   143.68   143.70  |   175.95   176.36  | //: 32-bit, MSC_v6.00 [ C =111]
+   1000_ ||   138.17   138.18  |   140.23   140.80  |   168.46   168.46  | //: 32-bit, MSC_v6.00 [ C =111]
+   1024_ ||   134.92   134.92  |   135.90   136.72  |   164.48   164.48  | //: 32-bit, MSC_v6.00 [ C =111]
+   2048_ ||   132.76   132.76  |   132.19   132.25  |   154.34   155.67  | //: 32-bit, MSC_v6.00 [ C =111]
+   4096_ ||   131.66   131.74  |   132.76   133.34  |   149.64   150.49  | //: 32-bit, MSC_v6.00 [ C =111]
+   8192_ ||   131.21   135.88  |   120.29   124.46  |   142.16   147.73  | //: 32-bit, MSC_v6.00 [ C =111]
+  10000_ ||   124.18   125.11  |   123.38   125.46  |   139.12   140.88  | //: 32-bit, MSC_v6.00 [ C =111]
+  16384_ ||   124.27   130.94  |   122.04   127.55  |   137.91   146.32  | //: 32-bit, MSC_v6.00 [ C =111]
+  32768_ ||   123.57   128.76  |   120.72   121.97  |   138.10   140.89  | //: 32-bit, MSC_v6.00 [ C =111]
+ 100000_ ||   123.30   129.21  |   123.83   125.61  |   145.19   145.40  | //: 32-bit, MSC_v6.00 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1486 bytes |         1348 bytes |         1445 bytes | //: 32-bit, MSC_v6.00 [ C =111]
+  Block  ||         2435 bytes |         5119 bytes |         8894 bytes | //: 32-bit, MSC_v6.00 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:48,Oct  7 2008  by  'MSC_v4.20', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5418.00  5430.00  |  9450.00  9462.00  | 20436.00 20472.00  | //: 32-bit, MSC_v4.20 [ C =111]
+      2_ ||  2709.00  2721.00  |  4725.00  4731.00  | 10212.00 10245.00  | //: 32-bit, MSC_v4.20 [ C =111]
+      4_ ||  1351.50  1354.50  |  2359.50  2361.00  |  5097.00  5107.50  | //: 32-bit, MSC_v4.20 [ C =111]
+      8_ ||   675.00   678.75  |  1179.75  1179.75  |  2549.25  2552.25  | //: 32-bit, MSC_v4.20 [ C =111]
+     10_ ||   540.60   546.60  |   943.20   944.40  |  2041.20  2041.80  | //: 32-bit, MSC_v4.20 [ C =111]
+     16_ ||   337.88   338.25  |   589.50   589.50  |  1273.88  1275.38  | //: 32-bit, MSC_v4.20 [ C =111]
+     32_ ||   167.81   167.81  |   294.94   295.13  |   636.75   637.13  | //: 32-bit, MSC_v4.20 [ C =111]
+     64_ ||   124.41   124.41  |   147.19   147.84  |   318.28   318.47  | //: 32-bit, MSC_v4.20 [ C =111]
+    100_ ||   131.46   131.52  |   140.10   140.28  |   203.76   203.94  | //: 32-bit, MSC_v4.20 [ C =111]
+    128_ ||   102.42   102.47  |   109.22   109.41  |   159.05   159.38  | //: 32-bit, MSC_v4.20 [ C =111]
+    256_ ||    91.10    91.27  |    90.59    90.59  |   118.73   118.78  | //: 32-bit, MSC_v4.20 [ C =111]
+    512_ ||    85.43    85.43  |    80.78    80.79  |    98.43    98.48  | //: 32-bit, MSC_v4.20 [ C =111]
+   1000_ ||    84.56    84.56  |    77.74    77.75  |    90.24    90.28  | //: 32-bit, MSC_v4.20 [ C =111]
+   1024_ ||    82.55    82.55  |    75.83    75.83  |    88.15    88.19  | //: 32-bit, MSC_v4.20 [ C =111]
+   2048_ ||    81.07    81.07  |    73.35    73.36  |    83.00    83.02  | //: 32-bit, MSC_v4.20 [ C =111]
+   4096_ ||    80.34    80.36  |    72.12    72.13  |    80.42    80.44  | //: 32-bit, MSC_v4.20 [ C =111]
+   8192_ ||    79.97    80.54  |    71.56    71.64  |    79.11    79.62  | //: 32-bit, MSC_v4.20 [ C =111]
+  10000_ ||    75.11    80.03  |    66.25    69.37  |    73.59    74.99  | //: 32-bit, MSC_v4.20 [ C =111]
+  16384_ ||    75.57    80.04  |    67.66    71.51  |    74.32    74.42  | //: 32-bit, MSC_v4.20 [ C =111]
+  32768_ ||    75.61    80.15  |    67.03    67.84  |    74.04    78.41  | //: 32-bit, MSC_v4.20 [ C =111]
+ 100000_ ||    77.96    80.31  |    67.58    67.84  |    74.31    74.73  | //: 32-bit, MSC_v4.20 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1152 bytes |         1024 bytes |         1088 bytes | //: 32-bit, MSC_v4.20 [ C =111]
+  Block  ||         2064 bytes |         3840 bytes |         7616 bytes | //: 32-bit, MSC_v4.20 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:29:54,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   780.00   786.00  |  1422.00  1434.00  |  3810.00  3816.00  | //: 64-bit, MSC_v9.00 [ C =111]
+      2_ ||   384.00   390.00  |   705.00   708.00  |  1902.00  1902.00  | //: 64-bit, MSC_v9.00 [ C =111]
+      4_ ||   193.50   193.50  |   355.50   355.50  |   951.00   952.50  | //: 64-bit, MSC_v9.00 [ C =111]
+      8_ ||    93.75    93.75  |   171.00   171.75  |   474.75   475.50  | //: 64-bit, MSC_v9.00 [ C =111]
+     10_ ||    75.60    76.20  |   140.40   140.40  |   380.40   381.00  | //: 64-bit, MSC_v9.00 [ C =111]
+     16_ ||    51.38    51.38  |    93.00    93.00  |   257.25   257.63  | //: 64-bit, MSC_v9.00 [ C =111]
+     32_ ||    25.31    25.31  |    46.50    46.50  |   118.69   118.69  | //: 64-bit, MSC_v9.00 [ C =111]
+     64_ ||    16.69    16.69  |    21.38    21.38  |    59.53    59.53  | //: 64-bit, MSC_v9.00 [ C =111]
+    100_ ||    17.16    17.22  |    20.52    21.00  |    38.22    38.28  | //: 64-bit, MSC_v9.00 [ C =111]
+    128_ ||    13.27    13.27  |    15.80    15.80  |    29.63    29.67  | //: 64-bit, MSC_v9.00 [ C =111]
+    256_ ||    11.16    11.18  |    12.61    12.73  |    22.10    22.10  | //: 64-bit, MSC_v9.00 [ C =111]
+    512_ ||    10.05    10.07  |    11.00    11.07  |    18.18    19.68  | //: 64-bit, MSC_v9.00 [ C =111]
+   1000_ ||     9.69     9.69  |    10.42    10.42  |    16.51    16.51  | //: 64-bit, MSC_v9.00 [ C =111]
+   1024_ ||     9.44     9.44  |    10.18    10.18  |    16.11    16.12  | //: 64-bit, MSC_v9.00 [ C =111]
+   2048_ ||     9.21     9.21  |     9.62     9.62  |    15.06    15.06  | //: 64-bit, MSC_v9.00 [ C =111]
+   4096_ ||     9.10     9.10  |     9.36     9.37  |    14.55    14.55  | //: 64-bit, MSC_v9.00 [ C =111]
+   8192_ ||     8.97     8.97  |     9.20     9.21  |    14.48    14.66  | //: 64-bit, MSC_v9.00 [ C =111]
+  10000_ ||     8.97     8.97  |     9.38     9.38  |    14.38    14.40  | //: 64-bit, MSC_v9.00 [ C =111]
+  16384_ ||     8.95     9.01  |     9.26     9.26  |    14.16    14.29  | //: 64-bit, MSC_v9.00 [ C =111]
+  32768_ ||     8.90     9.24  |     9.18     9.18  |    14.46    14.75  | //: 64-bit, MSC_v9.00 [ C =111]
+ 100000_ ||     9.18     9.71  |     9.35     9.49  |    14.79    14.99  | //: 64-bit, MSC_v9.00 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =111]
+  Block  ||          704 bytes |         1456 bytes |         2976 bytes | //: 64-bit, MSC_v9.00 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:29:57,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  6420.00  6420.00  | 11040.00 11040.00  | 23358.00 23364.00  | //: 32-bit, BCC_v5.51 [ C =111]
+      2_ ||  3210.00  3210.00  |  5517.00  5520.00  | 11679.00 11682.00  | //: 32-bit, BCC_v5.51 [ C =111]
+      4_ ||  1605.00  1605.00  |  2758.50  2758.50  |  5832.00  5833.50  | //: 32-bit, BCC_v5.51 [ C =111]
+      8_ ||   802.50   802.50  |  1379.25  1379.25  |  2916.00  2916.75  | //: 32-bit, BCC_v5.51 [ C =111]
+     10_ ||   642.00   642.00  |  1103.40  1103.40  |  2335.80  2335.80  | //: 32-bit, BCC_v5.51 [ C =111]
+     16_ ||   400.88   401.25  |   689.25   689.62  |  1458.00  1458.00  | //: 32-bit, BCC_v5.51 [ C =111]
+     32_ ||   199.50   199.50  |   344.44   344.44  |   729.00   729.00  | //: 32-bit, BCC_v5.51 [ C =111]
+     64_ ||   146.06   146.25  |   171.66   172.50  |   364.41   364.50  | //: 32-bit, BCC_v5.51 [ C =111]
+    100_ ||   152.28   152.28  |   162.78   162.78  |   233.16   233.16  | //: 32-bit, BCC_v5.51 [ C =111]
+    128_ ||   118.69   118.69  |   126.89   126.89  |   181.88   181.88  | //: 32-bit, BCC_v5.51 [ C =111]
+    256_ ||   104.62   104.62  |   104.48   104.48  |   135.30   135.33  | //: 32-bit, BCC_v5.51 [ C =111]
+    512_ ||    97.50    97.50  |    93.13    93.14  |   112.00   112.00  | //: 32-bit, BCC_v5.51 [ C =111]
+   1000_ ||    96.26    96.26  |    89.53    89.54  |   102.70   102.71  | //: 32-bit, BCC_v5.51 [ C =111]
+   1024_ ||    93.91    93.91  |    87.40    87.40  |   100.27   100.27  | //: 32-bit, BCC_v5.51 [ C =111]
+   2048_ ||    92.14    92.14  |    84.56    84.56  |    94.38    94.39  | //: 32-bit, BCC_v5.51 [ C =111]
+   4096_ ||    91.28    91.28  |    76.72    83.12  |    84.42    86.14  | //: 32-bit, BCC_v5.51 [ C =111]
+   8192_ ||    83.85    86.88  |    76.06    80.17  |    83.06    87.27  | //: 32-bit, BCC_v5.51 [ C =111]
+  10000_ ||    83.92    87.25  |    76.30    83.56  |    86.42    87.19  | //: 32-bit, BCC_v5.51 [ C =111]
+  16384_ ||    85.71    87.12  |    77.78    77.82  |    84.43    84.51  | //: 32-bit, BCC_v5.51 [ C =111]
+  32768_ ||    85.60    86.59  |    77.64    78.17  |    84.32    84.94  | //: 32-bit, BCC_v5.51 [ C =111]
+ 100000_ ||    86.18    87.75  |    78.03    79.63  |    84.77    88.79  | //: 32-bit, BCC_v5.51 [ C =111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [ C =111]
+  Block  ||         1888 bytes |         3028 bytes |         5864 bytes | //: 32-bit, BCC_v5.51 [ C =111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:04,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2664.00  2664.00  |  4998.00  4998.00  | 10704.00 10704.00  | //: 32-bit, BCC_v5.51 [asm=111]
+      2_ ||  1338.00  1338.00  |  2505.00  2508.00  |  5352.00  5352.00  | //: 32-bit, BCC_v5.51 [asm=111]
+      4_ ||   669.00   669.00  |  1246.50  1246.50  |  2668.50  2670.00  | //: 32-bit, BCC_v5.51 [asm=111]
+      8_ ||   334.50   334.50  |   623.25   623.25  |  1334.25  1334.25  | //: 32-bit, BCC_v5.51 [asm=111]
+     10_ ||   266.40   266.40  |   501.00   501.00  |  1058.40  1058.40  | //: 32-bit, BCC_v5.51 [asm=111]
+     16_ ||   166.50   166.50  |   312.75   321.00  |   628.50   629.25  | //: 32-bit, BCC_v5.51 [asm=111]
+     32_ ||    79.88    79.88  |   147.75   147.75  |   312.19   312.38  | //: 32-bit, BCC_v5.51 [asm=111]
+     64_ ||    56.53    56.53  |    73.22    73.22  |   156.09   156.09  | //: 32-bit, BCC_v5.51 [asm=111]
+    100_ ||    58.08    58.08  |    68.52    74.10  |    99.36   107.52  | //: 32-bit, BCC_v5.51 [asm=111]
+    128_ ||    45.19    45.23  |    53.20    53.20  |    77.81    77.81  | //: 32-bit, BCC_v5.51 [asm=111]
+    256_ ||    39.26    39.28  |    43.24    43.24  |    57.52    62.32  | //: 32-bit, BCC_v5.51 [asm=111]
+    512_ ||    36.13    36.13  |    37.76    37.77  |    47.17    47.24  | //: 32-bit, BCC_v5.51 [asm=111]
+   1000_ ||    35.51    35.71  |    36.22    36.23  |    42.92    43.04  | //: 32-bit, BCC_v5.51 [asm=111]
+   1024_ ||    34.51    34.51  |    34.78    35.12  |    42.05    42.05  | //: 32-bit, BCC_v5.51 [asm=111]
+   2048_ ||    33.69    33.70  |    33.82    33.83  |    38.84    39.04  | //: 32-bit, BCC_v5.51 [asm=111]
+   4096_ ||    32.01    33.99  |    33.64    33.64  |    37.82    37.97  | //: 32-bit, BCC_v5.51 [asm=111]
+   8192_ ||    31.77    32.58  |    32.80    33.00  |    36.98    37.59  | //: 32-bit, BCC_v5.51 [asm=111]
+  10000_ ||    33.75    33.75  |    33.13    33.25  |    37.32    37.86  | //: 32-bit, BCC_v5.51 [asm=111]
+  16384_ ||    31.90    36.52  |    35.86    35.90  |    37.26    40.33  | //: 32-bit, BCC_v5.51 [asm=111]
+  32768_ ||    34.29    34.47  |    33.87    34.03  |    37.77    38.04  | //: 32-bit, BCC_v5.51 [asm=111]
+ 100000_ ||    33.20    34.48  |    33.75    33.91  |    37.98    38.23  | //: 32-bit, BCC_v5.51 [asm=111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [asm=111]
+  Block  ||         1276 bytes |         2532 bytes |         4983 bytes | //: 32-bit, BCC_v5.51 [asm=111]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:30:08,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2580.00  2598.00  |  4842.00  4848.00  | 10578.00 10602.00  | //: 32-bit, MSC_v9.00 [asm=111]
+      2_ ||  1299.00  1302.00  |  2445.00  2445.00  |  5277.00  5283.00  | //: 32-bit, MSC_v9.00 [asm=111]
+      4_ ||   648.00   648.00  |  1213.50  1215.00  |  2644.50  2649.00  | //: 32-bit, MSC_v9.00 [asm=111]
+      8_ ||   324.00   324.75  |   610.50   610.50  |  1322.25  1323.00  | //: 32-bit, MSC_v9.00 [asm=111]
+     10_ ||   259.80   259.80  |   484.20   484.20  |  1059.60  1060.20  | //: 32-bit, MSC_v9.00 [asm=111]
+     16_ ||   162.00   162.38  |   302.63   302.63  |   660.38   662.63  | //: 32-bit, MSC_v9.00 [asm=111]
+     32_ ||    80.81    81.00  |   141.56   141.56  |   308.63   308.63  | //: 32-bit, MSC_v9.00 [asm=111]
+     64_ ||    54.38    54.47  |    70.41    70.41  |   154.41   154.59  | //: 32-bit, MSC_v9.00 [asm=111]
+    100_ ||    57.18    57.24  |    66.42    66.48  |    98.40    98.46  | //: 32-bit, MSC_v9.00 [asm=111]
+    128_ ||    48.28    48.28  |    51.75    51.75  |    76.97    77.02  | //: 32-bit, MSC_v9.00 [asm=111]
+    256_ ||    39.05    39.05  |    42.45    42.47  |    56.95    56.95  | //: 32-bit, MSC_v9.00 [asm=111]
+    512_ ||    36.09    36.11  |    37.65    37.66  |    47.05    47.06  | //: 32-bit, MSC_v9.00 [asm=111]
+   1000_ ||    35.56    35.59  |    35.96    35.96  |    42.79    42.80  | //: 32-bit, MSC_v9.00 [asm=111]
+   1024_ ||    34.62    34.63  |    35.28    35.28  |    41.47    41.47  | //: 32-bit, MSC_v9.00 [asm=111]
+   2048_ ||    33.91    33.91  |    34.00    34.08  |    39.33    39.33  | //: 32-bit, MSC_v9.00 [asm=111]
+   4096_ ||    33.38    33.66  |    33.49    33.49  |    38.04    38.23  | //: 32-bit, MSC_v9.00 [asm=111]
+   8192_ ||    33.15    33.23  |    32.76    33.07  |    37.21    37.22  | //: 32-bit, MSC_v9.00 [asm=111]
+  10000_ ||    33.69    36.50  |    33.29    33.42  |    37.98    41.34  | //: 32-bit, MSC_v9.00 [asm=111]
+  16384_ ||    33.07    35.17  |    33.08    34.97  |    37.10    38.12  | //: 32-bit, MSC_v9.00 [asm=111]
+  32768_ ||    34.35    34.53  |    33.80    34.05  |    38.21    40.46  | //: 32-bit, MSC_v9.00 [asm=111]
+ 100000_ ||    33.96    34.57  |    33.93    35.69  |    38.04    38.20  | //: 32-bit, MSC_v9.00 [asm=111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=111]
+  Block  ||         1276 bytes |         2532 bytes |         4983 bytes | //: 32-bit, MSC_v9.00 [asm=111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:13,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2514.00  2514.00  |  4836.00  4836.00  | 10392.00 10398.00  | //: 32-bit, GCC_v3.42 [asm=111]
+      2_ ||  1254.00  1260.00  |  2409.00  2412.00  |  5181.00  5184.00  | //: 32-bit, GCC_v3.42 [asm=111]
+      4_ ||   628.50   628.50  |  1204.50  1204.50  |  2596.50  2598.00  | //: 32-bit, GCC_v3.42 [asm=111]
+      8_ ||   312.75   312.75  |   602.25   603.00  |  1298.25  1299.00  | //: 32-bit, GCC_v3.42 [asm=111]
+     10_ ||   250.80   251.40  |   482.40   483.00  |  1035.00  1035.60  | //: 32-bit, GCC_v3.42 [asm=111]
+     16_ ||   157.50   157.50  |   302.25   302.63  |   652.50   652.50  | //: 32-bit, GCC_v3.42 [asm=111]
+     32_ ||    78.19    78.38  |   151.88   152.06  |   326.81   326.81  | //: 32-bit, GCC_v3.42 [asm=111]
+     64_ ||    57.09    57.19  |    75.47    75.47  |   163.31   163.41  | //: 32-bit, GCC_v3.42 [asm=111]
+    100_ ||    60.06    60.06  |    71.22    71.28  |   104.58   104.58  | //: 32-bit, GCC_v3.42 [asm=111]
+    128_ ||    46.83    46.88  |    55.45    55.50  |    81.33    81.38  | //: 32-bit, GCC_v3.42 [asm=111]
+    256_ ||    41.32    41.34  |    45.47    45.49  |    59.91    59.91  | //: 32-bit, GCC_v3.42 [asm=111]
+    512_ ||    38.51    38.52  |    40.16    40.16  |    49.49    49.49  | //: 32-bit, GCC_v3.42 [asm=111]
+   1000_ ||    37.92    37.93  |    38.60    38.60  |    45.40    45.41  | //: 32-bit, GCC_v3.42 [asm=111]
+   1024_ ||    37.08    37.08  |    37.93    38.33  |    45.24    45.25  | //: 32-bit, GCC_v3.42 [asm=111]
+   2048_ ||    36.55    36.56  |    36.88    36.88  |    42.42    42.49  | //: 32-bit, GCC_v3.42 [asm=111]
+   4096_ ||    35.77    35.77  |    33.56    37.02  |    37.73    39.65  | //: 32-bit, GCC_v3.42 [asm=111]
+   8192_ ||    32.68    34.17  |    33.13    33.19  |    38.41    40.47  | //: 32-bit, GCC_v3.42 [asm=111]
+  10000_ ||    35.92    36.59  |    35.00    36.14  |    37.65    39.24  | //: 32-bit, GCC_v3.42 [asm=111]
+  16384_ ||    33.37    34.20  |    32.77    33.93  |    36.86    37.94  | //: 32-bit, GCC_v3.42 [asm=111]
+  32768_ ||    34.22    34.41  |    33.82    34.06  |    37.39    37.74  | //: 32-bit, GCC_v3.42 [asm=111]
+ 100000_ ||    34.23    34.34  |    33.81    34.20  |    37.34    37.86  | //: 32-bit, GCC_v3.42 [asm=111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [asm=111]
+  Block  ||         1276 bytes |         2532 bytes |         4983 bytes | //: 32-bit, GCC_v3.42 [asm=111]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:30:17,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   678.00   678.00  |  1098.00  1098.00  |  2034.00  2040.00  | //: 64-bit, MSC_v9.00 [asm=111]
+      2_ ||   339.00   339.00  |   546.00   546.00  |  1017.00  1020.00  | //: 64-bit, MSC_v9.00 [asm=111]
+      4_ ||   168.00   169.50  |   273.00   273.00  |   510.00   511.50  | //: 64-bit, MSC_v9.00 [asm=111]
+      8_ ||    81.75    82.50  |   134.25   134.25  |   254.25   255.00  | //: 64-bit, MSC_v9.00 [asm=111]
+     10_ ||    66.60    66.60  |   109.80   109.80  |   204.00   204.00  | //: 64-bit, MSC_v9.00 [asm=111]
+     16_ ||    40.88    40.88  |    66.75    67.13  |   127.50   127.50  | //: 64-bit, MSC_v9.00 [asm=111]
+     32_ ||    20.25    20.44  |    33.56    33.56  |    63.56    63.56  | //: 64-bit, MSC_v9.00 [asm=111]
+     64_ ||    14.91    15.00  |    16.50    16.50  |    31.69    31.69  | //: 64-bit, MSC_v9.00 [asm=111]
+    100_ ||    15.48    16.68  |    16.98    16.98  |    22.38    22.38  | //: 64-bit, MSC_v9.00 [asm=111]
+    128_ ||    12.80    12.80  |    12.94    12.94  |    15.84    15.89  | //: 64-bit, MSC_v9.00 [asm=111]
+    256_ ||     9.84     9.84  |     9.33     9.33  |    11.60    11.63  | //: 64-bit, MSC_v9.00 [asm=111]
+    512_ ||     8.75     8.79  |     8.53     8.57  |     9.36     9.38  | //: 64-bit, MSC_v9.00 [asm=111]
+   1000_ ||     8.45     8.45  |     7.93     7.93  |     8.39     8.39  | //: 64-bit, MSC_v9.00 [asm=111]
+   1024_ ||     8.25     8.25  |     7.14     7.14  |     8.19     8.19  | //: 64-bit, MSC_v9.00 [asm=111]
+   2048_ ||     8.00     8.00  |     6.77     7.33  |     7.58     7.58  | //: 64-bit, MSC_v9.00 [asm=111]
+   4096_ ||     7.88     7.88  |     6.58     6.58  |     7.29     7.29  | //: 64-bit, MSC_v9.00 [asm=111]
+   8192_ ||     7.81     7.81  |     6.49     6.49  |     7.13     7.15  | //: 64-bit, MSC_v9.00 [asm=111]
+  10000_ ||     7.81     7.81  |     6.50     6.50  |     7.18     7.18  | //: 64-bit, MSC_v9.00 [asm=111]
+  16384_ ||     7.79     7.79  |     6.42     6.42  |     7.04     7.04  | //: 64-bit, MSC_v9.00 [asm=111]
+  32768_ ||     7.77     7.77  |     6.40     6.40  |     7.03     7.03  | //: 64-bit, MSC_v9.00 [asm=111]
+ 100000_ ||     8.08     8.09  |     6.40     6.71  |     6.98     7.21  | //: 64-bit, MSC_v9.00 [asm=111]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=111]
+  Block  ||          664 bytes |         1074 bytes |         2221 bytes | //: 64-bit, MSC_v9.00 [asm=111]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:19,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  4272.00  4296.00  |  7974.00  7980.00  | 17484.00 17496.00  | //: 32-bit, GCC_v3.42 [ C =332]
+      2_ ||  2139.00  2154.00  |  3981.00  3996.00  |  8736.00  8754.00  | //: 32-bit, GCC_v3.42 [ C =332]
+      4_ ||  1069.50  1071.00  |  1995.00  2002.50  |  4377.00  4378.50  | //: 32-bit, GCC_v3.42 [ C =332]
+      8_ ||   536.25   538.50  |   998.25  1000.50  |  2183.25  2186.25  | //: 32-bit, GCC_v3.42 [ C =332]
+     10_ ||   429.00   430.20  |   798.60   807.60  |  1749.60  1752.00  | //: 32-bit, GCC_v3.42 [ C =332]
+     16_ ||   267.75   270.00  |   498.00   499.88  |  1092.00  1093.13  | //: 32-bit, GCC_v3.42 [ C =332]
+     32_ ||   132.75   133.50  |   249.19   249.75  |   546.38   547.50  | //: 32-bit, GCC_v3.42 [ C =332]
+     64_ ||    98.44    99.00  |   123.94   124.03  |   272.25   272.34  | //: 32-bit, GCC_v3.42 [ C =332]
+    100_ ||   103.08   103.08  |   117.96   117.96  |   174.24   174.42  | //: 32-bit, GCC_v3.42 [ C =332]
+    128_ ||    80.72   121.13  |    92.34   133.22  |   137.06   137.39  | //: 32-bit, GCC_v3.42 [ C =332]
+    256_ ||    71.91    72.21  |    75.84    76.01  |   101.93   102.09  | //: 32-bit, GCC_v3.42 [ C =332]
+    512_ ||    67.50    67.59  |    67.62    67.75  |    83.95    84.47  | //: 32-bit, GCC_v3.42 [ C =332]
+   1000_ ||    66.71    67.00  |    64.95    65.28  |    77.12    77.20  | //: 32-bit, GCC_v3.42 [ C =332]
+   1024_ ||    64.89    64.96  |    63.19    63.23  |    74.67    74.67  | //: 32-bit, GCC_v3.42 [ C =332]
+   2048_ ||    63.35    63.36  |    61.13    61.14  |    70.19    70.19  | //: 32-bit, GCC_v3.42 [ C =332]
+   4096_ ||    62.80    62.80  |    60.11    60.12  |    62.58    62.65  | //: 32-bit, GCC_v3.42 [ C =332]
+   8192_ ||    57.83    59.01  |    55.02    60.12  |    66.75    66.86  | //: 32-bit, GCC_v3.42 [ C =332]
+  10000_ ||    62.69    62.87  |    59.76    59.87  |    67.20    67.63  | //: 32-bit, GCC_v3.42 [ C =332]
+  16384_ ||    62.50    62.75  |    55.96    59.35  |    62.39    63.28  | //: 32-bit, GCC_v3.42 [ C =332]
+  32768_ ||    58.66    59.69  |    56.17    56.62  |    61.97    63.07  | //: 32-bit, GCC_v3.42 [ C =332]
+ 100000_ ||    59.31    59.99  |    56.30    57.46  |    62.79    63.27  | //: 32-bit, GCC_v3.42 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [ C =332]
+  Block  ||         6640 bytes |        13040 bytes |        18448 bytes | //: 32-bit, GCC_v3.42 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:30:25,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2988.00  2994.00  |  6240.00  6246.00  | 13794.00 13800.00  | //: 32-bit, MSC_v9.00 [ C =332]
+      2_ ||  1488.00  1503.00  |  3120.00  3126.00  |  6900.00  6903.00  | //: 32-bit, MSC_v9.00 [ C =332]
+      4_ ||   744.00   751.50  |  1560.00  1560.00  |  3445.50  3447.00  | //: 32-bit, MSC_v9.00 [ C =332]
+      8_ ||   372.00   372.75  |   777.75   779.25  |  1723.50  1723.50  | //: 32-bit, MSC_v9.00 [ C =332]
+     10_ ||   297.60   299.40  |   623.40   624.00  |  1379.40  1380.00  | //: 32-bit, MSC_v9.00 [ C =332]
+     16_ ||   186.38   186.38  |   389.25   389.63  |   861.75   861.75  | //: 32-bit, MSC_v9.00 [ C =332]
+     32_ ||    92.44    92.44  |   195.38   195.56  |   431.25   431.44  | //: 32-bit, MSC_v9.00 [ C =332]
+     64_ ||    67.59    67.78  |    97.03    97.13  |   215.53   215.63  | //: 32-bit, MSC_v9.00 [ C =332]
+    100_ ||    70.26    70.32  |    91.92    91.92  |   138.00   138.06  | //: 32-bit, MSC_v9.00 [ C =332]
+    128_ ||    54.98    55.08  |    71.44    71.48  |   107.58   107.58  | //: 32-bit, MSC_v9.00 [ C =332]
+    256_ ||    48.68    48.70  |    58.57    58.57  |    79.83    79.83  | //: 32-bit, MSC_v9.00 [ C =332]
+    512_ ||    45.43    45.46  |    52.22    52.23  |    65.98    66.01  | //: 32-bit, MSC_v9.00 [ C =332]
+   1000_ ||    44.88    44.89  |    50.20    50.20  |    60.44    60.45  | //: 32-bit, MSC_v9.00 [ C =332]
+   1024_ ||    43.81    43.81  |    48.98    48.99  |    59.00    59.00  | //: 32-bit, MSC_v9.00 [ C =332]
+   2048_ ||    43.00    43.00  |    47.36    47.37  |    55.50    55.50  | //: 32-bit, MSC_v9.00 [ C =332]
+   4096_ ||    42.59    42.59  |    46.56    46.57  |    53.75    53.75  | //: 32-bit, MSC_v9.00 [ C =332]
+   8192_ ||    42.38    42.39  |    46.16    46.16  |    52.87    52.87  | //: 32-bit, MSC_v9.00 [ C =332]
+  10000_ ||    42.42    42.42  |    46.30    46.31  |    53.29    53.31  | //: 32-bit, MSC_v9.00 [ C =332]
+  16384_ ||    42.28    42.60  |    45.96    46.75  |    52.45    52.52  | //: 32-bit, MSC_v9.00 [ C =332]
+  32768_ ||    42.25    42.36  |    45.84    45.85  |    52.30    52.32  | //: 32-bit, MSC_v9.00 [ C =332]
+ 100000_ ||    42.21    42.50  |    43.60    45.77  |    49.55    50.03  | //: 32-bit, MSC_v9.00 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =332]
+  Block  ||         4560 bytes |         9232 bytes |        12560 bytes | //: 32-bit, MSC_v9.00 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:31,Oct  7 2008  by  'MSC_v6.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  9054.00  9060.00  | 17406.00 17514.00  | 36888.00 37032.00  | //: 32-bit, MSC_v6.00 [ C =332]
+      2_ ||  4341.00  4341.00  |  9129.00  9168.00  | 18273.00 18423.00  | //: 32-bit, MSC_v6.00 [ C =332]
+      4_ ||  2169.00  2170.50  |  4590.00  4636.50  |  9240.00  9334.50  | //: 32-bit, MSC_v6.00 [ C =332]
+      8_ ||  1083.75  1084.50  |  2223.00  2243.25  |  4581.75  4663.50  | //: 32-bit, MSC_v6.00 [ C =332]
+     10_ ||   867.00   867.60  |  1776.60  1790.40  |  3648.00  3672.00  | //: 32-bit, MSC_v6.00 [ C =332]
+     16_ ||   541.13   541.50  |  1044.75  1051.50  |  2274.38  2278.50  | //: 32-bit, MSC_v6.00 [ C =332]
+     32_ ||   271.13   271.88  |   567.56   573.75  |  1139.25  1140.00  | //: 32-bit, MSC_v6.00 [ C =332]
+     64_ ||   201.09   201.09  |   270.84   272.81  |   569.72   571.59  | //: 32-bit, MSC_v6.00 [ C =332]
+    100_ ||   212.70   213.24  |   261.12   262.02  |   365.16   365.28  | //: 32-bit, MSC_v6.00 [ C =332]
+    128_ ||   166.08   166.45  |   204.84   205.41  |   284.48   288.80  | //: 32-bit, MSC_v6.00 [ C =332]
+    256_ ||   148.69   149.34  |   169.59   169.95  |   221.65   221.79  | //: 32-bit, MSC_v6.00 [ C =332]
+    512_ ||   140.47   140.53  |   148.24   148.48  |   179.11   179.11  | //: 32-bit, MSC_v6.00 [ C =332]
+   1000_ ||   139.66   139.66  |   139.37   139.82  |   163.58   165.01  | //: 32-bit, MSC_v6.00 [ C =332]
+   1024_ ||   136.26   136.27  |   141.22   141.49  |   157.43   158.56  | //: 32-bit, MSC_v6.00 [ C =332]
+   2048_ ||   134.25   134.25  |   135.90   137.12  |   151.73   152.42  | //: 32-bit, MSC_v6.00 [ C =332]
+   4096_ ||   133.89   134.06  |   131.19   134.61  |   147.72   150.40  | //: 32-bit, MSC_v6.00 [ C =332]
+   8192_ ||   132.87   134.43  |   134.17   135.04  |   143.82   148.00  | //: 32-bit, MSC_v6.00 [ C =332]
+  10000_ ||   133.42   134.41  |   124.60   130.22  |   137.58   138.21  | //: 32-bit, MSC_v6.00 [ C =332]
+  16384_ ||   131.73   132.54  |   121.35   122.08  |   135.51   139.44  | //: 32-bit, MSC_v6.00 [ C =332]
+  32768_ ||   124.97   134.41  |   128.74   129.78  |   142.57   143.53  | //: 32-bit, MSC_v6.00 [ C =332]
+ 100000_ ||   126.77   134.34  |   126.62   129.40  |   135.08   139.54  | //: 32-bit, MSC_v6.00 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1486 bytes |         1348 bytes |         1445 bytes | //: 32-bit, MSC_v6.00 [ C =332]
+  Block  ||         6038 bytes |        13395 bytes |        15975 bytes | //: 32-bit, MSC_v6.00 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:42,Oct  7 2008  by  'MSC_v4.20', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5436.00  5436.00  |  9474.00  9474.00  | 20430.00 20442.00  | //: 32-bit, MSC_v4.20 [ C =332]
+      2_ ||  2715.00  2718.00  |  4731.00  4734.00  | 10215.00 10218.00  | //: 32-bit, MSC_v4.20 [ C =332]
+      4_ ||  1359.00  1360.50  |  2364.00  2364.00  |  5098.50  5103.00  | //: 32-bit, MSC_v4.20 [ C =332]
+      8_ ||   680.25   680.25  |  1182.00  1182.75  |  2549.25  2550.00  | //: 32-bit, MSC_v4.20 [ C =332]
+     10_ ||   538.80   543.60  |   945.60   946.20  |  2042.40  2043.00  | //: 32-bit, MSC_v4.20 [ C =332]
+     16_ ||   339.38   339.75  |   590.63   591.00  |  1274.63  1275.00  | //: 32-bit, MSC_v4.20 [ C =332]
+     32_ ||   166.69   166.69  |   295.31   295.50  |   637.13   637.50  | //: 32-bit, MSC_v4.20 [ C =332]
+     64_ ||   123.47   123.66  |   147.28   147.38  |   318.56   318.75  | //: 32-bit, MSC_v4.20 [ C =332]
+    100_ ||   130.62   130.74  |   140.28   140.76  |   203.94   204.00  | //: 32-bit, MSC_v4.20 [ C =332]
+    128_ ||   101.44   101.67  |   109.31   109.45  |   159.14   159.33  | //: 32-bit, MSC_v4.20 [ C =332]
+    256_ ||    90.70    90.75  |    90.52    90.56  |   118.66   118.83  | //: 32-bit, MSC_v4.20 [ C =332]
+    512_ ||    85.00    85.03  |    80.81    80.87  |    98.46    98.46  | //: 32-bit, MSC_v4.20 [ C =332]
+   1000_ ||    84.00    84.01  |    77.78    77.78  |    90.31    90.32  | //: 32-bit, MSC_v4.20 [ C =332]
+   1024_ ||    81.99    82.13  |    75.93    75.93  |    88.18    88.18  | //: 32-bit, MSC_v4.20 [ C =332]
+   2048_ ||    80.66    80.68  |    73.43    73.43  |    83.02    83.02  | //: 32-bit, MSC_v4.20 [ C =332]
+   4096_ ||    79.98    80.38  |    72.23    72.27  |    80.45    80.45  | //: 32-bit, MSC_v4.20 [ C =332]
+   8192_ ||    79.63    80.15  |    71.66    71.73  |    79.15    79.22  | //: 32-bit, MSC_v4.20 [ C =332]
+  10000_ ||    79.65    80.07  |    71.85    72.33  |    79.79    79.82  | //: 32-bit, MSC_v4.20 [ C =332]
+  16384_ ||    79.66    79.71  |    71.40    71.41  |    78.77    78.85  | //: 32-bit, MSC_v4.20 [ C =332]
+  32768_ ||    75.39    79.68  |    67.83    71.25  |    78.23    78.50  | //: 32-bit, MSC_v4.20 [ C =332]
+ 100000_ ||    75.49    77.32  |    67.60    67.87  |    74.33    75.55  | //: 32-bit, MSC_v4.20 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1152 bytes |         1024 bytes |         1088 bytes | //: 32-bit, MSC_v4.20 [ C =332]
+  Block  ||         4736 bytes |         8976 bytes |        12896 bytes | //: 32-bit, MSC_v4.20 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:30:49,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   780.00   798.00  |  1920.00  1920.00  |  3732.00  3732.00  | //: 64-bit, MSC_v9.00 [ C =332]
+      2_ ||   387.00   387.00  |   951.00   951.00  |  1866.00  1869.00  | //: 64-bit, MSC_v9.00 [ C =332]
+      4_ ||   199.50   199.50  |   477.00   480.00  |   936.00   936.00  | //: 64-bit, MSC_v9.00 [ C =332]
+      8_ ||    95.25    96.00  |   231.75   235.50  |   467.25   468.00  | //: 64-bit, MSC_v9.00 [ C =332]
+     10_ ||    76.80    76.80  |   189.00   191.40  |   402.60   402.60  | //: 64-bit, MSC_v9.00 [ C =332]
+     16_ ||    51.75    51.75  |   127.13   128.63  |   234.75   254.25  | //: 64-bit, MSC_v9.00 [ C =332]
+     32_ ||    23.63    23.63  |    58.13    58.31  |   115.69   115.69  | //: 64-bit, MSC_v9.00 [ C =332]
+     64_ ||    16.69    16.88  |    28.88    28.97  |    58.31    58.31  | //: 64-bit, MSC_v9.00 [ C =332]
+    100_ ||    17.10    17.16  |    27.66    27.90  |    37.62    37.62  | //: 64-bit, MSC_v9.00 [ C =332]
+    128_ ||    12.98    13.13  |    21.14    21.47  |    29.16    29.16  | //: 64-bit, MSC_v9.00 [ C =332]
+    256_ ||    11.27    11.30  |    17.04    17.18  |    21.66    21.73  | //: 64-bit, MSC_v9.00 [ C =332]
+    512_ ||    10.20    10.20  |    16.21    16.21  |    17.79    17.82  | //: 64-bit, MSC_v9.00 [ C =332]
+   1000_ ||     9.98    10.12  |    14.23    14.25  |    16.13    16.13  | //: 64-bit, MSC_v9.00 [ C =332]
+   1024_ ||     9.73    10.54  |    13.88    13.89  |    15.73    15.73  | //: 64-bit, MSC_v9.00 [ C =332]
+   2048_ ||     9.48     9.48  |    13.51    13.51  |    14.70    14.70  | //: 64-bit, MSC_v9.00 [ C =332]
+   4096_ ||     9.35     9.36  |    13.21    13.22  |    14.16    14.16  | //: 64-bit, MSC_v9.00 [ C =332]
+   8192_ ||     9.25     9.25  |    13.08    13.08  |    13.93    13.93  | //: 64-bit, MSC_v9.00 [ C =332]
+  10000_ ||     9.27     9.28  |    12.89    12.99  |    13.98    13.98  | //: 64-bit, MSC_v9.00 [ C =332]
+  16384_ ||     9.26     9.28  |    12.77    12.89  |    13.74    13.74  | //: 64-bit, MSC_v9.00 [ C =332]
+  32768_ ||     9.23     9.25  |    12.83    13.09  |    13.77    14.27  | //: 64-bit, MSC_v9.00 [ C =332]
+ 100000_ ||     9.32     9.56  |    13.12    13.19  |    14.15    14.23  | //: 64-bit, MSC_v9.00 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =332]
+  Block  ||         1200 bytes |         2928 bytes |         5008 bytes | //: 64-bit, MSC_v9.00 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:30:52,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5958.00  5958.00  | 10182.00 10188.00  | 21522.00 21522.00  | //: 32-bit, BCC_v5.51 [ C =332]
+      2_ ||  3228.00  3228.00  |  5091.00  5091.00  | 10761.00 10761.00  | //: 32-bit, BCC_v5.51 [ C =332]
+      4_ ||  1491.00  1491.00  |  2544.00  2544.00  |  5374.50  5821.50  | //: 32-bit, BCC_v5.51 [ C =332]
+      8_ ||   744.75   745.50  |  1272.00  1272.00  |  2686.50  2686.50  | //: 32-bit, BCC_v5.51 [ C =332]
+     10_ ||   595.80   595.80  |  1017.60  1102.80  |  2151.60  2152.20  | //: 32-bit, BCC_v5.51 [ C =332]
+     16_ ||   372.00   372.38  |   636.00   636.00  |  1343.25  1343.62  | //: 32-bit, BCC_v5.51 [ C =332]
+     32_ ||   184.69   184.69  |   317.62   317.81  |   671.62   671.62  | //: 32-bit, BCC_v5.51 [ C =332]
+     64_ ||   135.56   135.56  |   171.56   171.56  |   335.72   335.81  | //: 32-bit, BCC_v5.51 [ C =332]
+    100_ ||   141.90   141.90  |   150.18   150.18  |   232.68   232.68  | //: 32-bit, BCC_v5.51 [ C =332]
+    128_ ||   119.67   119.67  |   126.84   126.84  |   181.50   181.55  | //: 32-bit, BCC_v5.51 [ C =332]
+    256_ ||   106.15   106.15  |   104.25   104.27  |   135.00   135.02  | //: 32-bit, BCC_v5.51 [ C =332]
+    512_ ||    99.42    99.46  |    93.00    93.01  |   111.75   111.76  | //: 32-bit, BCC_v5.51 [ C =332]
+   1000_ ||    98.08    98.08  |    89.47    89.47  |   102.47   102.49  | //: 32-bit, BCC_v5.51 [ C =332]
+   1024_ ||    95.70    95.75  |    87.33    87.33  |   100.03   100.04  | //: 32-bit, BCC_v5.51 [ C =332]
+   2048_ ||    86.60    86.77  |    78.01    78.01  |    86.94    86.94  | //: 32-bit, BCC_v5.51 [ C =332]
+   4096_ ||    85.95    85.96  |    76.71    76.71  |    84.25    84.25  | //: 32-bit, BCC_v5.51 [ C =332]
+   8192_ ||    85.56    92.63  |    76.07    82.40  |    89.80    90.33  | //: 32-bit, BCC_v5.51 [ C =332]
+  10000_ ||    86.68    90.16  |    79.43    81.88  |    89.21    90.98  | //: 32-bit, BCC_v5.51 [ C =332]
+  16384_ ||    87.25    92.78  |    82.30    82.31  |    89.34    89.41  | //: 32-bit, BCC_v5.51 [ C =332]
+  32768_ ||    92.38    92.56  |    81.90    82.46  |    89.01    89.03  | //: 32-bit, BCC_v5.51 [ C =332]
+ 100000_ ||    88.00    88.57  |    78.54    81.88  |    84.53    88.34  | //: 32-bit, BCC_v5.51 [ C =332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [ C =332]
+  Block  ||         4340 bytes |         7660 bytes |        10408 bytes | //: 32-bit, BCC_v5.51 [ C =332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:31:00,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2784.00  2784.00  |  5094.00  5100.00  | 10800.00 10806.00  | //: 32-bit, BCC_v5.51 [asm=332]
+      2_ ||  1374.00  1377.00  |  2538.00  2538.00  |  5370.00  5373.00  | //: 32-bit, BCC_v5.51 [asm=332]
+      4_ ||   687.00   687.00  |  1267.50  1267.50  |  2695.50  2697.00  | //: 32-bit, BCC_v5.51 [asm=332]
+      8_ ||   341.25   341.25  |   633.00   633.75  |  1348.50  1348.50  | //: 32-bit, BCC_v5.51 [asm=332]
+     10_ ||   278.40   282.00  |   512.40   513.00  |  1089.00  1089.00  | //: 32-bit, BCC_v5.51 [asm=332]
+     16_ ||   172.50   172.50  |   320.25   320.25  |   678.00   679.12  | //: 32-bit, BCC_v5.51 [asm=332]
+     32_ ||    85.88    85.88  |   159.38   159.56  |   339.75   339.75  | //: 32-bit, BCC_v5.51 [asm=332]
+     64_ ||    61.03    61.12  |    79.31    79.41  |   169.78   169.78  | //: 32-bit, BCC_v5.51 [asm=332]
+    100_ ||    62.40    62.46  |    73.92    73.98  |   108.48   108.72  | //: 32-bit, BCC_v5.51 [asm=332]
+    128_ ||    48.28    48.28  |    57.47    57.47  |    84.14    84.14  | //: 32-bit, BCC_v5.51 [asm=332]
+    256_ ||    42.00    42.02  |    46.03    46.31  |    61.90    61.90  | //: 32-bit, BCC_v5.51 [asm=332]
+    512_ ||    37.96    39.39  |    41.17    41.18  |    51.16    51.33  | //: 32-bit, BCC_v5.51 [asm=332]
+   1000_ ||    37.45    37.74  |    38.66    62.12  |    45.85    69.53  | //: 32-bit, BCC_v5.51 [asm=332]
+   1024_ ||    36.34    58.89  |    37.42    60.09  |    44.31    44.62  | //: 32-bit, BCC_v5.51 [asm=332]
+   2048_ ||    35.77    36.18  |    36.06    36.34  |    42.06    42.07  | //: 32-bit, BCC_v5.51 [asm=332]
+   4096_ ||    35.04    35.44  |    35.33    35.33  |    40.48    48.07  | //: 32-bit, BCC_v5.51 [asm=332]
+   8192_ ||    34.80    43.64  |    35.91    35.92  |    40.46    40.66  | //: 32-bit, BCC_v5.51 [asm=332]
+  10000_ ||    35.05    36.40  |    35.54    37.36  |    41.01    54.16  | //: 32-bit, BCC_v5.51 [asm=332]
+  16384_ ||    34.92    36.14  |    35.74    40.79  |    40.28    43.83  | //: 32-bit, BCC_v5.51 [asm=332]
+  32768_ ||    35.39    38.29  |    35.19    37.31  |    39.88    40.94  | //: 32-bit, BCC_v5.51 [asm=332]
+ 100000_ ||    36.40    38.36  |    35.18    37.16  |    40.05    40.36  | //: 32-bit, BCC_v5.51 [asm=332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [asm=332]
+  Block  ||         3060 bytes |         6300 bytes |         8835 bytes | //: 32-bit, BCC_v5.51 [asm=332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:31:04,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2598.00  2604.00  |  4866.00  4878.00  | 10614.00 10632.00  | //: 32-bit, MSC_v9.00 [asm=332]
+      2_ ||  1290.00  1299.00  |  2451.00  2460.00  |  5331.00  5334.00  | //: 32-bit, MSC_v9.00 [asm=332]
+      4_ ||   648.00   649.50  |  1222.50  1222.50  |  2647.50  2656.50  | //: 32-bit, MSC_v9.00 [asm=332]
+      8_ ||   326.25   327.75  |   612.00   614.25  |  1330.50  1332.75  | //: 32-bit, MSC_v9.00 [asm=332]
+     10_ ||   260.40   261.00  |   490.20   490.20  |  1067.40  1067.40  | //: 32-bit, MSC_v9.00 [asm=332]
+     16_ ||   162.38   162.38  |   306.00   306.00  |   661.50   661.88  | //: 32-bit, MSC_v9.00 [asm=332]
+     32_ ||    80.81    80.81  |   153.94   153.94  |   333.75   333.94  | //: 32-bit, MSC_v9.00 [asm=332]
+     64_ ||    58.78    58.78  |    76.13    76.41  |   166.88   166.88  | //: 32-bit, MSC_v9.00 [asm=332]
+    100_ ||    60.78    60.78  |    72.00    72.00  |   106.86   106.92  | //: 32-bit, MSC_v9.00 [asm=332]
+    128_ ||    47.58    47.63  |    55.92    55.92  |    83.16    83.20  | //: 32-bit, MSC_v9.00 [asm=332]
+    256_ ||    42.05    42.05  |    45.75    45.75  |    61.59    61.64  | //: 32-bit, MSC_v9.00 [asm=332]
+    512_ ||    39.18    39.19  |    40.82    41.02  |    50.93    51.02  | //: 32-bit, MSC_v9.00 [asm=332]
+   1000_ ||    38.38    38.42  |    39.17    39.19  |    46.49    46.61  | //: 32-bit, MSC_v9.00 [asm=332]
+   1024_ ||    37.38    37.78  |    38.02    60.78  |    67.69    68.44  | //: 32-bit, MSC_v9.00 [asm=332]
+   2048_ ||    36.80    48.25  |    36.66    48.20  |    42.67    42.81  | //: 32-bit, MSC_v9.00 [asm=332]
+   4096_ ||    36.57    36.59  |    36.25    36.26  |    41.31    41.40  | //: 32-bit, MSC_v9.00 [asm=332]
+   8192_ ||    36.21    36.30  |    35.84    38.76  |    40.68    40.71  | //: 32-bit, MSC_v9.00 [asm=332]
+  10000_ ||    40.98    47.69  |    35.81    35.86  |    40.96    43.93  | //: 32-bit, MSC_v9.00 [asm=332]
+  16384_ ||    36.27    38.04  |    35.77    43.58  |    40.33    43.27  | //: 32-bit, MSC_v9.00 [asm=332]
+  32768_ ||    36.04    41.09  |    35.57    35.89  |    40.17    40.36  | //: 32-bit, MSC_v9.00 [asm=332]
+ 100000_ ||    34.46    36.34  |    34.07    37.16  |    39.60    43.18  | //: 32-bit, MSC_v9.00 [asm=332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=332]
+  Block  ||         3060 bytes |         6300 bytes |         8835 bytes | //: 32-bit, MSC_v9.00 [asm=332]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:31:10,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2550.00  2568.00  |  4896.00  4902.00  | 10662.00 10728.00  | //: 32-bit, GCC_v3.42 [asm=332]
+      2_ ||  1275.00  1290.00  |  2445.00  2457.00  |  5355.00  5379.00  | //: 32-bit, GCC_v3.42 [asm=332]
+      4_ ||   640.50   645.00  |  1224.00  1225.50  |  2655.00  2668.50  | //: 32-bit, GCC_v3.42 [asm=332]
+      8_ ||   318.75   320.25  |   609.75   610.50  |  1328.25  1332.75  | //: 32-bit, GCC_v3.42 [asm=332]
+     10_ ||   254.40   257.40  |   488.40   490.20  |  1070.40  1074.60  | //: 32-bit, GCC_v3.42 [asm=332]
+     16_ ||   161.25   162.00  |   306.38   307.13  |   669.00   671.25  | //: 32-bit, GCC_v3.42 [asm=332]
+     32_ ||    79.88    80.06  |   153.56   153.75  |   333.00   335.63  | //: 32-bit, GCC_v3.42 [asm=332]
+     64_ ||    58.50    58.69  |    76.50    76.59  |   166.69   167.34  | //: 32-bit, GCC_v3.42 [asm=332]
+    100_ ||    60.78    61.02  |    72.36    73.08  |   107.04   107.58  | //: 32-bit, GCC_v3.42 [asm=332]
+    128_ ||    47.39    47.44  |    56.06    56.30  |    83.44    83.63  | //: 32-bit, GCC_v3.42 [asm=332]
+    256_ ||    41.79    41.88  |    46.03    46.10  |    61.71    61.92  | //: 32-bit, GCC_v3.42 [asm=332]
+    512_ ||    39.20    39.33  |    40.96    41.10  |    51.23    51.30  | //: 32-bit, GCC_v3.42 [asm=332]
+   1000_ ||    38.40    38.57  |    39.23    39.26  |    46.83    47.06  | //: 32-bit, GCC_v3.42 [asm=332]
+   1024_ ||    37.53    37.72  |    38.27    38.33  |    45.78    46.00  | //: 32-bit, GCC_v3.42 [asm=332]
+   2048_ ||    36.94    37.00  |    37.03    37.15  |    43.10    56.43  | //: 32-bit, GCC_v3.42 [asm=332]
+   4096_ ||    41.78    53.31  |    36.01    40.53  |    40.97    41.13  | //: 32-bit, GCC_v3.42 [asm=332]
+   8192_ ||    35.90    36.04  |    35.84    48.31  |    40.53    40.55  | //: 32-bit, GCC_v3.42 [asm=332]
+  10000_ ||    36.42    36.48  |    35.85    46.01  |    40.60    40.74  | //: 32-bit, GCC_v3.42 [asm=332]
+  16384_ ||    36.20    39.37  |    35.61    38.72  |    40.15    41.33  | //: 32-bit, GCC_v3.42 [asm=332]
+  32768_ ||    36.47    40.53  |    35.81    39.15  |    40.13    41.96  | //: 32-bit, GCC_v3.42 [asm=332]
+ 100000_ ||    36.70    43.77  |    35.89    37.72  |    40.20    44.66  | //: 32-bit, GCC_v3.42 [asm=332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [asm=332]
+  Block  ||         3060 bytes |         6300 bytes |         8835 bytes | //: 32-bit, GCC_v3.42 [asm=332]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:31:14,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   684.00   690.00  |  1104.00  1104.00  |  2028.00  2034.00  | //: 64-bit, MSC_v9.00 [asm=332]
+      2_ ||   339.00   339.00  |   549.00   549.00  |  1014.00  1017.00  | //: 64-bit, MSC_v9.00 [asm=332]
+      4_ ||   168.00   169.50  |   276.00   276.00  |   507.00   508.50  | //: 64-bit, MSC_v9.00 [asm=332]
+      8_ ||    81.75    81.75  |   135.00   146.25  |   273.75   273.75  | //: 64-bit, MSC_v9.00 [asm=332]
+     10_ ||    70.80    70.80  |   120.00   120.00  |   219.00   219.00  | //: 64-bit, MSC_v9.00 [asm=332]
+     16_ ||    44.25    44.25  |    74.25    74.25  |   126.00   126.38  | //: 64-bit, MSC_v9.00 [asm=332]
+     32_ ||    20.06    20.25  |    33.75    33.75  |    63.00    63.00  | //: 64-bit, MSC_v9.00 [asm=332]
+     64_ ||    14.53    14.53  |    16.69    16.97  |    34.13    34.13  | //: 64-bit, MSC_v9.00 [asm=332]
+    100_ ||    15.72    15.72  |    16.74    16.74  |    22.20    22.20  | //: 64-bit, MSC_v9.00 [asm=332]
+    128_ ||    11.06    11.11  |    11.77    11.81  |    15.70    15.70  | //: 64-bit, MSC_v9.00 [asm=332]
+    256_ ||     9.52     9.52  |     9.05     9.07  |    12.38    12.40  | //: 64-bit, MSC_v9.00 [asm=332]
+    512_ ||     9.35     9.35  |     7.72     7.72  |     9.26     9.26  | //: 64-bit, MSC_v9.00 [asm=332]
+   1000_ ||     8.42     8.42  |     7.22     7.22  |     8.30     8.30  | //: 64-bit, MSC_v9.00 [asm=332]
+   1024_ ||     8.19     8.87  |     7.62     7.63  |     8.12     8.12  | //: 64-bit, MSC_v9.00 [asm=332]
+   2048_ ||     7.97     7.97  |     7.25     7.38  |     7.52     8.15  | //: 64-bit, MSC_v9.00 [asm=332]
+   4096_ ||     7.86     7.88  |     6.54     7.09  |     7.84    11.52  | //: 64-bit, MSC_v9.00 [asm=332]
+   8192_ ||     8.49    11.80  |     9.78    10.72  |     7.05    10.38  | //: 64-bit, MSC_v9.00 [asm=332]
+  10000_ ||     7.85     8.51  |     6.58     6.58  |     7.11     7.12  | //: 64-bit, MSC_v9.00 [asm=332]
+  16384_ ||     7.86     7.88  |     6.41     6.41  |     7.00     7.01  | //: 64-bit, MSC_v9.00 [asm=332]
+  32768_ ||     7.89     9.85  |     6.50     7.00  |     6.94     6.97  | //: 64-bit, MSC_v9.00 [asm=332]
+ 100000_ ||     7.80     9.43  |     6.90     7.71  |     7.18     8.48  | //: 64-bit, MSC_v9.00 [asm=332]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=332]
+  Block  ||         1288 bytes |         2182 bytes |         3449 bytes | //: 64-bit, MSC_v9.00 [asm=332]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:31:16,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  3954.00  3966.00  |  7350.00  7350.00  | 40698.00 40704.00  | //: 32-bit, GCC_v3.42 [ C =335]
+      2_ ||  1977.00  1977.00  |  3678.00  3678.00  | 22035.00 24258.00  | //: 32-bit, GCC_v3.42 [ C =335]
+      4_ ||  1072.50  1072.50  |  1837.50  1839.00  | 10161.00 11007.00  | //: 32-bit, GCC_v3.42 [ C =335]
+      8_ ||   495.00   495.75  |   993.75   999.75  |  5456.25  8527.50  | //: 32-bit, GCC_v3.42 [ C =335]
+     10_ ||   429.60   430.80  |   730.80   801.60  |  4380.00  5951.40  | //: 32-bit, GCC_v3.42 [ C =335]
+     16_ ||   249.00   269.25  |   499.13   502.13  |  2741.63  4381.13  | //: 32-bit, GCC_v3.42 [ C =335]
+     32_ ||   133.31   135.19  |   249.19   251.25  |  1369.69  2140.69  | //: 32-bit, GCC_v3.42 [ C =335]
+     64_ ||    90.84    99.00  |   114.47   123.94  |   635.06   635.16  | //: 32-bit, GCC_v3.42 [ C =335]
+    100_ ||    95.28   103.56  |   108.90   109.44  |   406.50   580.20  | //: 32-bit, GCC_v3.42 [ C =335]
+    128_ ||    74.44    80.44  |    85.50    91.88  |   317.25   317.30  | //: 32-bit, GCC_v3.42 [ C =335]
+    256_ ||    66.00    71.48  |    69.70    69.80  |   237.12   237.12  | //: 32-bit, GCC_v3.42 [ C =335]
+    512_ ||    66.96    66.98  |    67.36    67.39  |   224.53   224.58  | //: 32-bit, GCC_v3.42 [ C =335]
+   1000_ ||    66.20    66.21  |    64.66    64.73  |   205.97   206.02  | //: 32-bit, GCC_v3.42 [ C =335]
+   1024_ ||    64.61    87.60  |    63.19    63.20  |   175.51   194.46  | //: 32-bit, GCC_v3.42 [ C =335]
+   2048_ ||    58.87    66.35  |    56.44    60.15  |   165.23   193.47  | //: 32-bit, GCC_v3.42 [ C =335]
+   4096_ ||    58.09    71.92  |    55.50    55.51  |   168.27   197.84  | //: 32-bit, GCC_v3.42 [ C =335]
+   8192_ ||    57.83    61.97  |    55.02    64.00  |   173.49   203.78  | //: 32-bit, GCC_v3.42 [ C =335]
+  10000_ ||    62.69    63.71  |    59.52    62.18  |   176.01   194.46  | //: 32-bit, GCC_v3.42 [ C =335]
+  16384_ ||    62.78    65.42  |    59.37    63.71  |   182.36   201.21  | //: 32-bit, GCC_v3.42 [ C =335]
+  32768_ ||    62.48    70.41  |    59.61    63.00  |   184.14   189.59  | //: 32-bit, GCC_v3.42 [ C =335]
+ 100000_ ||    61.14    68.82  |    59.72    62.87  |   190.22   202.45  | //: 32-bit, GCC_v3.42 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [ C =335]
+  Block  ||         6640 bytes |        13040 bytes |        41968 bytes | //: 32-bit, GCC_v3.42 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:31:27,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2994.00  2994.00  |  6240.00  6240.00  | 14598.00 14604.00  | //: 32-bit, MSC_v9.00 [ C =335]
+      2_ ||  1488.00  1494.00  |  3123.00  3126.00  |  7308.00  7311.00  | //: 32-bit, MSC_v9.00 [ C =335]
+      4_ ||   744.00   745.50  |  1558.50  1558.50  |  3646.50  3648.00  | //: 32-bit, MSC_v9.00 [ C =335]
+      8_ ||   372.00   372.00  |   779.25   780.00  |  1827.75  1827.75  | //: 32-bit, MSC_v9.00 [ C =335]
+     10_ ||   300.60   301.20  |   624.00   624.60  |  1459.20  1461.00  | //: 32-bit, MSC_v9.00 [ C =335]
+     16_ ||   187.13   187.88  |   389.63   389.63  |   913.50   913.88  | //: 32-bit, MSC_v9.00 [ C =335]
+     32_ ||    92.63    93.19  |   195.38   195.56  |   456.56   456.56  | //: 32-bit, MSC_v9.00 [ C =335]
+     64_ ||    67.69    67.78  |    97.03    97.13  |   228.66   228.75  | //: 32-bit, MSC_v9.00 [ C =335]
+    100_ ||    70.62    70.68  |    91.86    91.92  |   146.10   146.16  | //: 32-bit, MSC_v9.00 [ C =335]
+    128_ ||    54.84    55.27  |    71.48    71.48  |   112.88   112.92  | //: 32-bit, MSC_v9.00 [ C =335]
+    256_ ||    48.49    48.54  |    58.48    58.50  |    83.37    83.48  | //: 32-bit, MSC_v9.00 [ C =335]
+    512_ ||    45.42    45.42  |    52.07    52.23  |    68.57    68.60  | //: 32-bit, MSC_v9.00 [ C =335]
+   1000_ ||    44.65    44.65  |    50.20    50.20  |    62.74    62.76  | //: 32-bit, MSC_v9.00 [ C =335]
+   1024_ ||    43.80    43.80  |    48.98    48.99  |    61.13    61.14  | //: 32-bit, MSC_v9.00 [ C =335]
+   2048_ ||    43.00    43.00  |    47.36    47.37  |    57.45    57.47  | //: 32-bit, MSC_v9.00 [ C =335]
+   4096_ ||    42.33    42.34  |    46.57    46.57  |    55.59    55.60  | //: 32-bit, MSC_v9.00 [ C =335]
+   8192_ ||    42.21    42.25  |    46.16    46.17  |    54.66    54.67  | //: 32-bit, MSC_v9.00 [ C =335]
+  10000_ ||    42.16    42.42  |    46.31    46.73  |    55.11    55.13  | //: 32-bit, MSC_v9.00 [ C =335]
+  16384_ ||    42.28    42.29  |    46.21    46.24  |    54.20    54.24  | //: 32-bit, MSC_v9.00 [ C =335]
+  32768_ ||    42.35    42.36  |    45.95    46.10  |    50.90    51.80  | //: 32-bit, MSC_v9.00 [ C =335]
+ 100000_ ||    40.09    40.55  |    45.76    45.97  |    51.00    53.08  | //: 32-bit, MSC_v9.00 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [ C =335]
+  Block  ||         4560 bytes |         9232 bytes |        29280 bytes | //: 32-bit, MSC_v9.00 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:31:35,Oct  7 2008  by  'MSC_v6.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  8670.00  8682.00  | 17376.00 17436.00  | 55890.00 56922.00  | //: 32-bit, MSC_v6.00 [ C =335]
+      2_ ||  4344.00  4347.00  |  9315.00  9414.00  | 27747.00 27771.00  | //: 32-bit, MSC_v6.00 [ C =335]
+      4_ ||  2164.50  2164.50  |  4500.00  4522.50  | 13807.50 13896.00  | //: 32-bit, MSC_v6.00 [ C =335]
+      8_ ||  1082.25  1088.25  |  2235.00  2245.50  |  6796.50  6931.50  | //: 32-bit, MSC_v6.00 [ C =335]
+     10_ ||   866.40   871.80  |  1800.60  1811.40  |  5465.40  5560.80  | //: 32-bit, MSC_v6.00 [ C =335]
+     16_ ||   548.25   548.25  |  1124.25  1129.88  |  3447.00  3447.75  | //: 32-bit, MSC_v6.00 [ C =335]
+     32_ ||   272.25   272.63  |   560.44   563.81  |  1723.50  1723.88  | //: 32-bit, MSC_v6.00 [ C =335]
+     64_ ||   202.59   203.44  |   282.47   283.50  |   837.00   840.09  | //: 32-bit, MSC_v6.00 [ C =335]
+    100_ ||   214.38   215.88  |   269.28   269.82  |   532.74   532.86  | //: 32-bit, MSC_v6.00 [ C =335]
+    128_ ||   167.63   167.67  |   208.64   210.70  |   418.03   421.73  | //: 32-bit, MSC_v6.00 [ C =335]
+    256_ ||   149.41   150.16  |   173.23   173.79  |   317.27   320.23  | //: 32-bit, MSC_v6.00 [ C =335]
+    512_ ||   147.41   147.48  |   148.73   148.88  |   265.04   265.10  | //: 32-bit, MSC_v6.00 [ C =335]
+   1000_ ||   140.53   140.53  |   144.00   144.20  |   244.22   250.33  | //: 32-bit, MSC_v6.00 [ C =335]
+   1024_ ||   141.71   141.73  |   142.42   142.59  |   235.18   235.96  | //: 32-bit, MSC_v6.00 [ C =335]
+   2048_ ||   135.36   135.38  |   137.08   137.16  |   220.27   221.11  | //: 32-bit, MSC_v6.00 [ C =335]
+   4096_ ||   124.41   128.93  |   123.98   126.91  |   200.32   204.77  | //: 32-bit, MSC_v6.00 [ C =335]
+   8192_ ||   124.35   126.84  |   124.37   130.27  |   204.17   219.92  | //: 32-bit, MSC_v6.00 [ C =335]
+  10000_ ||   126.45   133.37  |   133.76   134.30  |   217.48   218.27  | //: 32-bit, MSC_v6.00 [ C =335]
+  16384_ ||   133.14   135.72  |   128.15   128.86  |   191.67   194.97  | //: 32-bit, MSC_v6.00 [ C =335]
+  32768_ ||   129.20   133.13  |   127.94   129.85  |   202.98   210.08  | //: 32-bit, MSC_v6.00 [ C =335]
+ 100000_ ||   130.83   133.01  |   121.08   129.21  |   192.14   200.80  | //: 32-bit, MSC_v6.00 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1486 bytes |         1348 bytes |         1445 bytes | //: 32-bit, MSC_v6.00 [ C =335]
+  Block  ||         6038 bytes |        13395 bytes |        37221 bytes | //: 32-bit, MSC_v6.00 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:31:48,Oct  7 2008  by  'MSC_v4.20', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5430.00  5436.00  |  9474.00  9474.00  | 21330.00 21366.00  | //: 32-bit, MSC_v4.20 [ C =335]
+      2_ ||  2706.00  2724.00  |  4731.00  4734.00  | 10662.00 10683.00  | //: 32-bit, MSC_v4.20 [ C =335]
+      4_ ||  1359.00  1359.00  |  2364.00  2365.50  |  5322.00  5332.50  | //: 32-bit, MSC_v4.20 [ C =335]
+      8_ ||   675.00   679.50  |  1182.00  1182.75  |  2661.75  2666.25  | //: 32-bit, MSC_v4.20 [ C =335]
+     10_ ||   540.00   543.60  |   945.60   945.60  |  2132.40  2136.00  | //: 32-bit, MSC_v4.20 [ C =335]
+     16_ ||   337.50   339.75  |   590.63   591.00  |  1330.50  1332.75  | //: 32-bit, MSC_v4.20 [ C =335]
+     32_ ||   166.69   167.25  |   295.31   295.50  |   665.25   666.38  | //: 32-bit, MSC_v4.20 [ C =335]
+     64_ ||   123.66   123.84  |   147.19   147.28  |   332.63   333.19  | //: 32-bit, MSC_v4.20 [ C =335]
+    100_ ||   130.62   130.98  |   140.28   140.76  |   212.88   213.24  | //: 32-bit, MSC_v4.20 [ C =335]
+    128_ ||   101.44   101.72  |   109.31   109.45  |   166.08   166.41  | //: 32-bit, MSC_v4.20 [ C =335]
+    256_ ||    90.54    90.73  |    90.54    90.56  |   124.99   125.11  | //: 32-bit, MSC_v4.20 [ C =335]
+    512_ ||    84.93    85.02  |    80.95    80.95  |   101.98   101.98  | //: 32-bit, MSC_v4.20 [ C =335]
+   1000_ ||    84.00    84.02  |    77.78    77.79  |    93.61    93.62  | //: 32-bit, MSC_v4.20 [ C =335]
+   1024_ ||    81.96    82.10  |    75.93    75.93  |    91.37    91.39  | //: 32-bit, MSC_v4.20 [ C =335]
+   2048_ ||    80.68    80.69  |    73.49    73.49  |    85.58    85.59  | //: 32-bit, MSC_v4.20 [ C =335]
+   4096_ ||    79.98    80.00  |    72.23    72.24  |    82.21    82.57  | //: 32-bit, MSC_v4.20 [ C =335]
+   8192_ ||    79.62    80.01  |    71.61    72.15  |    80.57    81.37  | //: 32-bit, MSC_v4.20 [ C =335]
+  10000_ ||    79.72    80.04  |    71.86    71.92  |    81.67    81.67  | //: 32-bit, MSC_v4.20 [ C =335]
+  16384_ ||    79.47    79.72  |    67.22    67.80  |    76.81    77.22  | //: 32-bit, MSC_v4.20 [ C =335]
+  32768_ ||    75.20    79.32  |    67.07    68.60  |    75.91    78.02  | //: 32-bit, MSC_v4.20 [ C =335]
+ 100000_ ||    75.38    75.82  |    67.48    69.43  |    74.87    77.52  | //: 32-bit, MSC_v4.20 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1152 bytes |         1024 bytes |         1088 bytes | //: 32-bit, MSC_v4.20 [ C =335]
+  Block  ||         4736 bytes |         8976 bytes |        28880 bytes | //: 32-bit, MSC_v4.20 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:31:56,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   780.00   798.00  |  1890.00  1920.00  |  3498.00  3498.00  | //: 64-bit, MSC_v9.00 [ C =335]
+      2_ ||   387.00   387.00  |   951.00   969.00  |  1752.00  1752.00  | //: 64-bit, MSC_v9.00 [ C =335]
+      4_ ||   201.00   201.00  |   477.00   478.50  |   877.50   877.50  | //: 64-bit, MSC_v9.00 [ C =335]
+      8_ ||    95.25    96.00  |   234.75   237.00  |   440.25   441.00  | //: 64-bit, MSC_v9.00 [ C =335]
+     10_ ||    77.40    78.00  |   190.80   195.00  |   350.40   379.20  | //: 64-bit, MSC_v9.00 [ C =335]
+     16_ ||    52.13    52.13  |   126.75   127.13  |   237.00   237.00  | //: 64-bit, MSC_v9.00 [ C =335]
+     32_ ||    23.81    25.69  |    58.69    59.25  |   109.88   109.88  | //: 64-bit, MSC_v9.00 [ C =335]
+     64_ ||    16.97    16.97  |    29.34    29.72  |    54.75    54.75  | //: 64-bit, MSC_v9.00 [ C =335]
+    100_ ||    17.10    17.10  |    27.72    28.08  |    35.28    35.28  | //: 64-bit, MSC_v9.00 [ C =335]
+    128_ ||    13.03    13.03  |    21.19    21.52  |    27.33    27.33  | //: 64-bit, MSC_v9.00 [ C =335]
+    256_ ||    11.20    11.23  |    17.02    17.16  |    20.23    20.25  | //: 64-bit, MSC_v9.00 [ C =335]
+    512_ ||    10.21    10.22  |    14.95    16.15  |    16.56    17.94  | //: 64-bit, MSC_v9.00 [ C =335]
+   1000_ ||     9.95    10.00  |    14.23    14.24  |    15.09    15.10  | //: 64-bit, MSC_v9.00 [ C =335]
+   1024_ ||     9.71    10.50  |    13.91    13.91  |    14.68    14.68  | //: 64-bit, MSC_v9.00 [ C =335]
+   2048_ ||     9.48     9.73  |    13.43    13.51  |    13.73    13.75  | //: 64-bit, MSC_v9.00 [ C =335]
+   4096_ ||     9.36     9.36  |    13.21    13.22  |    13.28    13.28  | //: 64-bit, MSC_v9.00 [ C =335]
+   8192_ ||     9.28     9.31  |    12.83    12.94  |    12.97    14.48  | //: 64-bit, MSC_v9.00 [ C =335]
+  10000_ ||     9.30    10.06  |    12.94    14.10  |    13.07    14.36  | //: 64-bit, MSC_v9.00 [ C =335]
+  16384_ ||     9.25     9.27  |    12.98    13.01  |    12.83    12.83  | //: 64-bit, MSC_v9.00 [ C =335]
+  32768_ ||     9.22     9.24  |    12.81    12.91  |    12.90    12.90  | //: 64-bit, MSC_v9.00 [ C =335]
+ 100000_ ||     9.33     9.58  |    13.94    13.95  |    13.24    13.92  | //: 64-bit, MSC_v9.00 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [ C =335]
+  Block  ||         1200 bytes |         2928 bytes |        10880 bytes | //: 64-bit, MSC_v9.00 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:32:00,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  5964.00  6462.00  | 10182.00 10188.00  | 21516.00 21516.00  | //: 32-bit, BCC_v5.51 [ C =335]
+      2_ ||  2979.00  2982.00  |  5091.00  5091.00  | 10758.00 10761.00  | //: 32-bit, BCC_v5.51 [ C =335]
+      4_ ||  1489.50  1489.50  |  2545.50  2757.00  |  5374.50  5374.50  | //: 32-bit, BCC_v5.51 [ C =335]
+      8_ ||   744.75   745.50  |  1272.00  1272.00  |  2687.25  2687.25  | //: 32-bit, BCC_v5.51 [ C =335]
+     10_ ||   595.20   595.20  |  1017.60  1017.60  |  2151.60  2152.20  | //: 32-bit, BCC_v5.51 [ C =335]
+     16_ ||   372.38   372.38  |   636.00   636.00  |  1343.25  1343.62  | //: 32-bit, BCC_v5.51 [ C =335]
+     32_ ||   200.25   200.25  |   317.62   317.81  |   671.44   672.00  | //: 32-bit, BCC_v5.51 [ C =335]
+     64_ ||   135.56   135.56  |   158.34   158.34  |   335.72   335.72  | //: 32-bit, BCC_v5.51 [ C =335]
+    100_ ||   141.78   141.84  |   150.18   150.18  |   214.74   214.74  | //: 32-bit, BCC_v5.51 [ C =335]
+    128_ ||   110.44   110.44  |   117.05   117.09  |   167.53   167.53  | //: 32-bit, BCC_v5.51 [ C =335]
+    256_ ||   106.15   106.15  |   104.25   104.27  |   135.00   135.02  | //: 32-bit, BCC_v5.51 [ C =335]
+    512_ ||    99.46    99.50  |    93.00    93.01  |   111.62   111.63  | //: 32-bit, BCC_v5.51 [ C =335]
+   1000_ ||    98.07    98.17  |    89.48    89.48  |   102.39   102.39  | //: 32-bit, BCC_v5.51 [ C =335]
+   1024_ ||    95.84    95.84  |    87.35    87.35  |    99.96    99.96  | //: 32-bit, BCC_v5.51 [ C =335]
+   2048_ ||    93.84    94.01  |    84.51    84.52  |    94.10    94.10  | //: 32-bit, BCC_v5.51 [ C =335]
+   4096_ ||    93.12    93.13  |    83.10    83.10  |    84.16    91.18  | //: 32-bit, BCC_v5.51 [ C =335]
+   8192_ ||    92.65    93.16  |    82.39    82.91  |    89.71    89.79  | //: 32-bit, BCC_v5.51 [ C =335]
+  10000_ ||    92.75    93.17  |    77.57    78.30  |    83.49    85.13  | //: 32-bit, BCC_v5.51 [ C =335]
+  16384_ ||    87.43    88.16  |    77.83    79.58  |    84.22    84.89  | //: 32-bit, BCC_v5.51 [ C =335]
+  32768_ ||    87.33    88.36  |    77.63    78.37  |    84.49    86.56  | //: 32-bit, BCC_v5.51 [ C =335]
+ 100000_ ||    87.96    89.42  |    77.90    78.17  |    84.30    85.04  | //: 32-bit, BCC_v5.51 [ C =335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [ C =335]
+  Block  ||         4340 bytes |         7660 bytes |        24192 bytes | //: 32-bit, BCC_v5.51 [ C =335]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:32:08,Oct  7 2008  by  'BCC_v5.51', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2718.00  2718.00  |  5076.00  5082.00  | 10746.00 10752.00  | //: 32-bit, BCC_v5.51 [asm=335]
+      2_ ||  1359.00  1362.00  |  2499.00  2499.00  |  5373.00  5376.00  | //: 32-bit, BCC_v5.51 [asm=335]
+      4_ ||   679.50   681.00  |  1245.00  1251.00  |  2673.00  2674.50  | //: 32-bit, BCC_v5.51 [asm=335]
+      8_ ||   340.50   340.50  |   622.50   625.50  |  1335.75  1336.50  | //: 32-bit, BCC_v5.51 [asm=335]
+     10_ ||   269.40   269.40  |   499.20   499.80  |  1075.80  1075.80  | //: 32-bit, BCC_v5.51 [asm=335]
+     16_ ||   170.62   170.62  |   310.88   311.25  |   667.50   667.88  | //: 32-bit, BCC_v5.51 [asm=335]
+     32_ ||    82.88    83.06  |   156.00   156.00  |   334.31   334.31  | //: 32-bit, BCC_v5.51 [asm=335]
+     64_ ||    59.25    59.34  |    77.25    77.34  |   166.50   167.53  | //: 32-bit, BCC_v5.51 [asm=335]
+    100_ ||    60.48    60.54  |    72.30    72.36  |   107.16   107.22  | //: 32-bit, BCC_v5.51 [asm=335]
+    128_ ||    47.02    47.02  |    56.30    56.30  |    83.53    84.66  | //: 32-bit, BCC_v5.51 [asm=335]
+    256_ ||    42.21    42.23  |    46.59    46.62  |    62.53    62.53  | //: 32-bit, BCC_v5.51 [asm=335]
+    512_ ||    39.22    39.23  |    41.00    41.00  |    51.18    51.18  | //: 32-bit, BCC_v5.51 [asm=335]
+   1000_ ||    38.02    38.03  |    39.37    39.37  |    46.66    46.67  | //: 32-bit, BCC_v5.51 [asm=335]
+   1024_ ||    34.83    34.83  |    35.55    35.56  |    42.06    42.12  | //: 32-bit, BCC_v5.51 [asm=335]
+   2048_ ||    33.65    33.65  |    34.07    34.07  |    39.49    39.49  | //: 32-bit, BCC_v5.51 [asm=335]
+   4096_ ||    33.34    33.40  |    33.47    33.68  |    38.18    38.18  | //: 32-bit, BCC_v5.51 [asm=335]
+   8192_ ||    32.90    33.36  |    32.87    33.00  |    37.36    38.25  | //: 32-bit, BCC_v5.51 [asm=335]
+  10000_ ||    32.68    33.54  |    33.04    33.37  |    37.54    39.47  | //: 32-bit, BCC_v5.51 [asm=335]
+  16384_ ||    31.89    36.04  |    35.74    35.85  |    40.30    40.32  | //: 32-bit, BCC_v5.51 [asm=335]
+  32768_ ||    35.05    36.16  |    35.79    35.94  |    40.28    40.31  | //: 32-bit, BCC_v5.51 [asm=335]
+ 100000_ ||    34.30    35.27  |    33.66    34.18  |    38.25    39.54  | //: 32-bit, BCC_v5.51 [asm=335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          996 bytes |         1000 bytes |         1068 bytes | //: 32-bit, BCC_v5.51 [asm=335]
+  Block  ||         3060 bytes |         6300 bytes |        20391 bytes | //: 32-bit, BCC_v5.51 [asm=335]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:32:11,Oct  7 2008  by  'MSC_v9.00', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2586.00  2592.00  |  4896.00  4902.00  | 10668.00 10668.00  | //: 32-bit, MSC_v9.00 [asm=335]
+      2_ ||  1311.00  1317.00  |  2448.00  2451.00  |  5340.00  5343.00  | //: 32-bit, MSC_v9.00 [asm=335]
+      4_ ||   652.50   654.00  |  1224.00  1227.00  |  2665.50  2665.50  | //: 32-bit, MSC_v9.00 [asm=335]
+      8_ ||   327.00   328.50  |   613.50   614.25  |  1332.75  1332.75  | //: 32-bit, MSC_v9.00 [asm=335]
+     10_ ||   263.40   263.40  |   489.60   489.60  |  1069.20  1069.80  | //: 32-bit, MSC_v9.00 [asm=335]
+     16_ ||   163.88   163.88  |   306.00   306.38  |   666.38   666.75  | //: 32-bit, MSC_v9.00 [asm=335]
+     32_ ||    81.00    81.00  |   154.13   154.31  |   334.31   334.50  | //: 32-bit, MSC_v9.00 [asm=335]
+     64_ ||    58.88    58.97  |    76.41    76.59  |   167.16   167.25  | //: 32-bit, MSC_v9.00 [asm=335]
+    100_ ||    61.08    61.14  |    72.30    72.36  |   107.04   107.10  | //: 32-bit, MSC_v9.00 [asm=335]
+    128_ ||    47.81    47.86  |    56.16    56.20  |    83.34    83.34  | //: 32-bit, MSC_v9.00 [asm=335]
+    256_ ||    42.14    42.16  |    45.89    45.89  |    61.64    61.76  | //: 32-bit, MSC_v9.00 [asm=335]
+    512_ ||    36.23    36.23  |    37.66    37.66  |    46.98    46.99  | //: 32-bit, MSC_v9.00 [asm=335]
+   1000_ ||    35.57    35.57  |    36.11    36.12  |    43.07    43.12  | //: 32-bit, MSC_v9.00 [asm=335]
+   1024_ ||    34.85    34.85  |    35.24    35.24  |    42.05    42.06  | //: 32-bit, MSC_v9.00 [asm=335]
+   2048_ ||    34.17    34.17  |    33.88    34.06  |    39.46    39.53  | //: 32-bit, MSC_v9.00 [asm=335]
+   4096_ ||    33.74    33.85  |    33.38    33.46  |    38.32    38.32  | //: 32-bit, MSC_v9.00 [asm=335]
+   8192_ ||    33.65    33.67  |    33.17    34.20  |    37.70    37.71  | //: 32-bit, MSC_v9.00 [asm=335]
+  10000_ ||    33.68    34.51  |    33.29    36.32  |    37.91    39.80  | //: 32-bit, MSC_v9.00 [asm=335]
+  16384_ ||    33.26    35.05  |    32.98    35.06  |    37.34    39.35  | //: 32-bit, MSC_v9.00 [asm=335]
+  32768_ ||    36.34    36.38  |    35.62    35.76  |    40.21    41.08  | //: 32-bit, MSC_v9.00 [asm=335]
+ 100000_ ||    36.32    36.43  |    35.91    35.98  |    38.02    38.19  | //: 32-bit, MSC_v9.00 [asm=335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          864 bytes |          704 bytes |          720 bytes | //: 32-bit, MSC_v9.00 [asm=335]
+  Block  ||         3060 bytes |         6300 bytes |        20391 bytes | //: 32-bit, MSC_v9.00 [asm=335]
+
+Skein performance, in clks per byte, dtMin =   36 clks.
+         [compiled 14:32:16,Oct  7 2008  by  'GCC_v3.42', 32-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||  2562.00  2562.00  |  4866.00  4866.00  | 10698.00 10722.00  | //: 32-bit, GCC_v3.42 [asm=335]
+      2_ ||  1269.00  1275.00  |  2436.00  2439.00  |  5343.00  5355.00  | //: 32-bit, GCC_v3.42 [asm=335]
+      4_ ||   645.00   645.00  |  1222.50  1224.00  |  2667.00  2676.00  | //: 32-bit, GCC_v3.42 [asm=335]
+      8_ ||   322.50   323.25  |   610.50   610.50  |  1332.75  1338.00  | //: 32-bit, GCC_v3.42 [asm=335]
+     10_ ||   254.40   255.60  |   486.60   487.20  |  1070.40  1072.20  | //: 32-bit, GCC_v3.42 [asm=335]
+     16_ ||   161.63   162.00  |   306.00   306.00  |   668.25   669.38  | //: 32-bit, GCC_v3.42 [asm=335]
+     32_ ||    73.69    73.69  |   141.56   141.94  |   307.31   307.50  | //: 32-bit, GCC_v3.42 [asm=335]
+     64_ ||    54.28    54.38  |    70.59    70.59  |   153.66   153.75  | //: 32-bit, GCC_v3.42 [asm=335]
+    100_ ||    56.16    56.40  |    66.66    66.66  |    98.40   107.22  | //: 32-bit, GCC_v3.42 [asm=335]
+    128_ ||    47.67    47.81  |    56.16    56.20  |    83.25    83.25  | //: 32-bit, GCC_v3.42 [asm=335]
+    256_ ||    41.72    41.86  |    45.84    45.87  |    61.48    61.52  | //: 32-bit, GCC_v3.42 [asm=335]
+    512_ ||    38.66    38.68  |    40.70    40.70  |    50.68    50.71  | //: 32-bit, GCC_v3.42 [asm=335]
+   1000_ ||    38.09    38.57  |    38.98    38.99  |    46.42    46.43  | //: 32-bit, GCC_v3.42 [asm=335]
+   1024_ ||    37.16    37.17  |    38.10    38.10  |    45.29    45.30  | //: 32-bit, GCC_v3.42 [asm=335]
+   2048_ ||    36.50    36.78  |    36.76    36.76  |    42.45    42.54  | //: 32-bit, GCC_v3.42 [asm=335]
+   4096_ ||    36.23    36.52  |    36.06    36.06  |    41.19    41.21  | //: 32-bit, GCC_v3.42 [asm=335]
+   8192_ ||    33.71    35.10  |    33.02    33.08  |    37.40    37.47  | //: 32-bit, GCC_v3.42 [asm=335]
+  10000_ ||    33.41    33.76  |    33.24    33.24  |    37.68    38.45  | //: 32-bit, GCC_v3.42 [asm=335]
+  16384_ ||    33.63    34.72  |    33.12    35.81  |    40.09    40.41  | //: 32-bit, GCC_v3.42 [asm=335]
+  32768_ ||    33.96    34.18  |    33.53    33.86  |    37.88    38.29  | //: 32-bit, GCC_v3.42 [asm=335]
+ 100000_ ||    35.61    36.38  |    34.24    35.18  |    37.98    38.64  | //: 32-bit, GCC_v3.42 [asm=335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||         1568 bytes |         1264 bytes |         1472 bytes | //: 32-bit, GCC_v3.42 [asm=335]
+  Block  ||         3060 bytes |         6300 bytes |        20391 bytes | //: 32-bit, GCC_v3.42 [asm=335]
+
+Skein performance, in clks per byte, dtMin =   24 clks.
+         [compiled 14:32:20,Oct  7 2008  by  'MSC_v9.00', 64-bit]
+         =================================================================
+         ||                       Skein block size                       |
+         ||--------------------------------------------------------------|
+ Message ||       256 bits     |       512 bits     |      1024 bits     |
+ Length  ||====================|====================|====================|
+ (bytes) ||     min    median  |     min    median  |     min    median  |
+=========||====================|====================|====================|
+      1_ ||   684.00   690.00  |  1104.00  1104.00  |  2022.00  2022.00  | //: 64-bit, MSC_v9.00 [asm=335]
+      2_ ||   339.00   342.00  |   549.00   549.00  |  1011.00  1014.00  | //: 64-bit, MSC_v9.00 [asm=335]
+      4_ ||   168.00   169.50  |   277.50   277.50  |   505.50   505.50  | //: 64-bit, MSC_v9.00 [asm=335]
+      8_ ||    81.00    81.75  |   135.00   135.00  |   252.00   252.00  | //: 64-bit, MSC_v9.00 [asm=335]
+     10_ ||    65.40    65.40  |   109.80   109.80  |   201.60   202.20  | //: 64-bit, MSC_v9.00 [asm=335]
+     16_ ||    40.88    40.88  |    67.13    67.50  |   126.00   126.00  | //: 64-bit, MSC_v9.00 [asm=335]
+     32_ ||    20.06    20.25  |    33.56    33.75  |    62.81    63.00  | //: 64-bit, MSC_v9.00 [asm=335]
+     64_ ||    14.53    14.63  |    18.19    18.28  |    33.84    33.94  | //: 64-bit, MSC_v9.00 [asm=335]
+    100_ ||    15.78    15.78  |    16.80    16.80  |    22.02    22.08  | //: 64-bit, MSC_v9.00 [asm=335]
+    128_ ||    11.11    11.11  |    11.77    11.77  |    15.61    15.66  | //: 64-bit, MSC_v9.00 [asm=335]
+    256_ ||     9.52     9.52  |     9.07     9.09  |    11.41    11.41  | //: 64-bit, MSC_v9.00 [asm=335]
+    512_ ||     8.63     8.64  |     7.72     7.72  |     9.15     9.16  | //: 64-bit, MSC_v9.00 [asm=335]
+   1000_ ||     8.41     8.42  |     7.21     7.22  |     8.24     8.26  | //: 64-bit, MSC_v9.00 [asm=335]
+   1024_ ||     8.89     8.90  |     7.62     7.63  |     8.08     8.08  | //: 64-bit, MSC_v9.00 [asm=335]
+   2048_ ||     8.00     8.00  |     6.69     7.25  |     7.50     7.50  | //: 64-bit, MSC_v9.00 [asm=335]
+   4096_ ||     7.89     7.89  |     6.52     6.52  |     7.22     7.22  | //: 64-bit, MSC_v9.00 [asm=335]
+   8192_ ||     7.84     7.84  |     6.44     6.44  |     7.07     7.07  | //: 64-bit, MSC_v9.00 [asm=335]
+  10000_ ||     7.84     7.84  |     6.45     6.50  |     7.12     7.12  | //: 64-bit, MSC_v9.00 [asm=335]
+  16384_ ||     7.82     7.82  |     6.40     6.40  |     6.99     7.01  | //: 64-bit, MSC_v9.00 [asm=335]
+  32768_ ||     7.79     7.80  |     6.37     6.37  |     6.96     6.96  | //: 64-bit, MSC_v9.00 [asm=335]
+ 100000_ ||     8.11     8.11  |     6.49     6.74  |     6.95     7.26  | //: 64-bit, MSC_v9.00 [asm=335]
+=========||====================|====================|====================|
+Code Size||                    |                    |                    |
+=========||====================|====================|====================|
+    API  ||          992 bytes |         1312 bytes |          864 bytes | //: 64-bit, MSC_v9.00 [asm=335]
+  Block  ||         1288 bytes |         2182 bytes |         7133 bytes | //: 64-bit, MSC_v9.00 [asm=335]
diff --git a/Additional_Implementations/skein_rot_search2.c b/Additional_Implementations/skein_rot_search2.c
new file mode 100644
index 000000000000..a47f5c81d3e3
--- /dev/null
+++ b/Additional_Implementations/skein_rot_search2.c
@@ -0,0 +1,2538 @@
+/***********************************************************************
+**
+** Generate Skein rotation constant candidate sets and test them.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <time.h>
+#include <ctype.h>
+#include <math.h>
+#include <assert.h>
+#include "brg_types.h"                  /* get Brian Gladman's platform-specific definitions */
+
+#define uint    unsigned int
+#define u08b    uint_8t
+#define u32b    uint_32t
+#define u64b    uint_64t
+
+/* Threefish algorithm parameters */
+#ifndef BITS_PER_WORD
+#define BITS_PER_WORD           (64)    /* number of bits in each word of a Threefish block */
+#endif
+
+#define ROUNDS_PER_CYCLE         (8)    /* when do we inject keys and start reusing rotation constants? */
+#define MAX_BITS_PER_BLK      (1024)
+
+#define MAX_WORDS_PER_BLK       (MAX_BITS_PER_BLK/BITS_PER_WORD) 
+#define MAX_ROTS_PER_CYCLE      (MAX_WORDS_PER_BLK*(ROUNDS_PER_CYCLE/2))  
+
+/* default search parameters for different block sizes */
+#define DEFAULT_GEN_CNT_4     (5500)
+#define DEFAULT_ROUND_CNT_4     ( 8)
+#define MIN_HW_OR_4             (50)
+#define MAX_SAT_ROUNDS_4        ( 9)
+
+#define DEFAULT_GEN_CNT_8     (1600)
+#define DEFAULT_ROUND_CNT_8     ( 8)
+#define MIN_HW_OR_8             (36)
+#define MAX_SAT_ROUNDS_8        (10)
+
+#define DEFAULT_GEN_CNT_16     (400)    /* the 1024-bit search is slower, so search for fewer iterations :-( */
+#define DEFAULT_ROUND_CNT_16    ( 9)
+#define MIN_HW_OR_16            (40)
+#define MAX_SAT_ROUNDS_16       (11)
+
+#define MAX_ROT_VER_CNT         ( 4)
+#define MAX_ROT_VER_MASK        ((1 << MAX_ROT_VER_CNT ) - 1)
+
+#define MAX_POP_CNT           (1024)    /* size of population */
+#define MIN_POP_CNT           (  32)
+#define DEFAULT_POP_CNT       (MAX_POP_CNT)
+
+#define ID_RECALC_BIT_NUM       (16)
+#define TWIDDLE_CNT_BIT0        (17)
+#define TWIDDLE_CNT_MASK    ((1 << TWIDDLE_CNT_BIT0  ) - 1)
+#define ID_RECALC_BIT       ( 1 << ID_RECALC_BIT_NUM )
+#define ID_NUM_MASK         ((1 << ID_RECALC_BIT_NUM ) - 1)
+
+#if     BITS_PER_WORD == 64
+typedef u64b    Word;
+#elif   BITS_PER_WORD == 32
+typedef u32b    Word;
+#else
+#error  "Invalid BITS_PER_WORD"
+#endif
+
+/* tstFlag bits */
+#define TST_FLG_SHOW        (1u << 0)
+#define TST_FLG_SHOW_HIST   (1u << 1)
+#define TST_FLG_VERBOSE     (1u << 2)
+#define TST_FLG_STDERR      (1u << 3)
+#define TST_FLG_QUICK_EXIT  (1u << 4)
+#define TST_FLG_USE_ABS     (1u << 5)
+#define TST_FLG_KEEP_MIN_HW (1u << 6)
+#define TST_FLG_WEIGHT_REP  (1u << 7)
+#define TST_FLG_CHECK_ONE   (1u << 8)
+#define TST_FLG_DO_RAND     (1u << 9)
+
+/* parameters for ShowSearchRec */
+#define SHOW_ROTS_FINAL     (4)          
+#define SHOW_ROTS_H         (3)
+#define SHOW_ROTS_PRELIM    (2)
+#define SHOW_ROTS           (1)
+#define SHOW_NONE           (0)
+
+typedef struct { Word x[MAX_WORDS_PER_BLK]; } Block;
+
+typedef void cycle_func(Word *b, const u08b *rotates, int rounds);
+
+typedef struct                          /* record for dealing with rotation searches */
+    {
+    u08b rotList[MAX_ROTS_PER_CYCLE];   /* rotation constants */
+    uint CRC;                           /* CRC of rotates[] -- use as a quick "ID" */
+    uint ID;                            /* (get_rotation index) + (TwiddleCnt << TWIDDLE_CNT_BIT0) */
+    uint parentCRC;                     /* CRC of the parent (allows us to track genealogy a bit) */
+    uint rWorst;                        /* "worst" min bit-to-bit differential */
+    u08b hw_OR[MAX_ROT_VER_CNT];        /* min hamming weights (over all words), using OR */
+    } rSearchRec;
+
+typedef struct                          /* pass a bunch of parameters to RunSearch */
+    {
+    uint    tstFlags;
+    uint    rounds;
+    uint    minHW_or;
+    uint    minOffs;
+    uint    diffBits;
+    uint    genCntMax;
+    uint    sampleCnt;
+    uint    maxSatRnds;
+    uint    seed0;
+    uint    rotVerMask;
+    uint    popCnt;
+    uint    runHours;                   /* 0 ==> never */
+    uint    dupRotMask;                 /* zero --> allow dup rots within the same round */
+    uint    regradeCnt;                 /* default = 3 */
+    u64b    goodRotCntMask;             /* which rotation values are ok? */
+    } testParms;
+
+/* globals */
+cycle_func *fwd_cycle       =   NULL;
+cycle_func *rev_cycle       =   NULL;
+cycle_func *fwd_cycle_or    =   NULL;   /* slow but steady */
+cycle_func *rev_cycle_or    =   NULL;
+cycle_func *fwd_cycle_or_rN =   NULL;   /* optimized for the current # rounds (for speed) */
+cycle_func *rev_cycle_or_rN =   NULL;
+const char *rotFileName     =   NULL;   /* read from file instead of generate random? */
+uint        bitsPerBlock    =      0;   /* default is to process all block sizes */
+uint        rotsPerCycle;
+uint        wordsPerBlock;
+
+/* macro "functions" */
+#define RotCnt_Bad(rotCnt) (((t.goodRotCntMask >> (rotCnt)) & 1) == 0)
+#define  left_rot(a,N)     (((a) << (N)) | ((a) >> (BITS_PER_WORD - (N))))
+#define right_rot(a,N)     (((a) >> (N)) | ((a) << (BITS_PER_WORD - (N))))
+#define DUP_64(w32)        ((w32) | (((u64b) (w32)) << 32))
+
+/********************** use RC4 to generate test data ******************/
+/* Note: this works identically on all platforms (big/little-endian)   */
+static struct
+    {
+    uint I,J;                           /* RC4 vars */
+    u08b state[256];
+    } prng;
+
+void RandBytes(void *dst,uint byteCnt)
+    {
+    u08b a,b;
+    u08b *d = (u08b *) dst;
+
+    for (;byteCnt;byteCnt--,d++)        /* run RC4  */
+        {
+        prng.I  = (prng.I+1) & 0xFF;
+        a       =  prng.state[prng.I];
+        prng.J  = (prng.J+a) & 0xFF;
+        b       =  prng.state[prng.J];
+        prng.state[prng.I] = b;
+        prng.state[prng.J] = a;
+        *d      =  prng.state[(a+b) & 0xFF];
+        }
+    }
+
+/* get a pseudo-random 8-bit integer */
+uint Rand08(void)
+    {
+    u08b b;
+    RandBytes(&b,1);
+    return (uint) b;
+    }
+
+/* get a pseudo-random 32-bit integer in a portable way */
+uint Rand32(void)
+    {
+    uint i,n;
+    u08b tmp[sizeof(uint)];
+
+    RandBytes(tmp,sizeof(tmp));
+
+    for (i=n=0;i<sizeof(tmp);i++)
+        n = n*256 + tmp[i];
+    
+    return n;
+    }
+
+/* get a pseudo-random 64-bit integer in a portable way */
+u64b Rand64(void)
+    {
+    uint i;
+    u64b n;
+    u08b tmp[sizeof(u64b)];
+
+    RandBytes(tmp,sizeof(tmp));
+
+    n=0;
+    for (i=0;i<sizeof(tmp);i++)
+        n = n*256 + tmp[i];
+    
+    return n;
+    }
+
+/* init the (RC4-based) prng */
+void Rand_Init(u64b seed)
+    {
+    uint i,j;
+    u08b tmp[4*256];
+
+    /* init the "key" in an endian-independent fashion */
+    for (i=0;i<8;i++)
+        tmp[i] = (u08b) (seed >> (8*i));
+
+    /* initialize the permutation */
+    for (i=0;i<256;i++)
+        prng.state[i]=(u08b) i;
+
+    /* now run the RC4 key schedule */
+    for (i=j=0;i<256;i++)
+        {                   
+        j = (j + prng.state[i] + tmp[i%8]) & 0xFF;
+        tmp[256]      = prng.state[i];
+        prng.state[i] = prng.state[j];
+        prng.state[j] = tmp[256];
+        }
+    prng.I = prng.J = 0;  /* init I,J variables for RC4 */
+    
+    /* discard some initial RC4 keystream before returning */
+    RandBytes(tmp,sizeof(tmp));
+    }
+
+/* implementations of Skein round functions for various block sizes */
+void fwd_cycle_16(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 0]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 1]); b[ 3] ^= b[ 2];
+        b[ 4] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[ 2]); b[ 5] ^= b[ 4];
+        b[ 6] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[ 3]); b[ 7] ^= b[ 6];
+        b[ 8] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[ 4]); b[ 9] ^= b[ 8];
+        b[10] += b[11]; b[11] = left_rot(b[11], rotates[ 5]); b[11] ^= b[10];
+        b[12] += b[13]; b[13] = left_rot(b[13], rotates[ 6]); b[13] ^= b[12];
+        b[14] += b[15]; b[15] = left_rot(b[15], rotates[ 7]); b[15] ^= b[14];
+        if (rounds == 1) break;                           
+                                                          
+        b[ 0] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[ 8]); b[ 9] ^= b[ 0];
+        b[ 2] += b[13]; b[13] = left_rot(b[13], rotates[ 9]); b[13] ^= b[ 2];
+        b[ 6] += b[11]; b[11] = left_rot(b[11], rotates[10]); b[11] ^= b[ 6];
+        b[ 4] += b[15]; b[15] = left_rot(b[15], rotates[11]); b[15] ^= b[ 4];
+        b[10] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[12]); b[ 7] ^= b[10];
+        b[12] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[13]); b[ 3] ^= b[12];
+        b[14] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[14]); b[ 5] ^= b[14];
+        b[ 8] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[15]); b[ 1] ^= b[ 8];
+        if (rounds == 2) break;                           
+                                                          
+        b[ 0] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[16]); b[ 7] ^= b[ 0];
+        b[ 2] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[17]); b[ 5] ^= b[ 2];
+        b[ 4] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[18]); b[ 3] ^= b[ 4];
+        b[ 6] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[19]); b[ 1] ^= b[ 6];
+        b[12] += b[15]; b[15] = left_rot(b[15], rotates[20]); b[15] ^= b[12];
+        b[14] += b[13]; b[13] = left_rot(b[13], rotates[21]); b[13] ^= b[14];
+        b[ 8] += b[11]; b[11] = left_rot(b[11], rotates[22]); b[11] ^= b[ 8];
+        b[10] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[23]); b[ 9] ^= b[10];
+        if (rounds == 3) break;                           
+                                                          
+        b[ 0] += b[15]; b[15] = left_rot(b[15], rotates[24]); b[15] ^= b[ 0];
+        b[ 2] += b[11]; b[11] = left_rot(b[11], rotates[25]); b[11] ^= b[ 2];
+        b[ 6] += b[13]; b[13] = left_rot(b[13], rotates[26]); b[13] ^= b[ 6];
+        b[ 4] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[27]); b[ 9] ^= b[ 4];
+        b[14] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[28]); b[ 1] ^= b[14];
+        b[ 8] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[29]); b[ 5] ^= b[ 8];
+        b[10] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[30]); b[ 3] ^= b[10];
+        b[12] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[31]); b[ 7] ^= b[12];
+        if (rounds == 4) break;                           
+                                                          
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[32]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[33]); b[ 3] ^= b[ 2];
+        b[ 4] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[34]); b[ 5] ^= b[ 4];
+        b[ 6] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[35]); b[ 7] ^= b[ 6];
+        b[ 8] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[36]); b[ 9] ^= b[ 8];
+        b[10] += b[11]; b[11] = left_rot(b[11], rotates[37]); b[11] ^= b[10];
+        b[12] += b[13]; b[13] = left_rot(b[13], rotates[38]); b[13] ^= b[12];
+        b[14] += b[15]; b[15] = left_rot(b[15], rotates[39]); b[15] ^= b[14];
+        if (rounds == 5) break;                           
+                                                          
+        b[ 0] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[40]); b[ 9] ^= b[ 0];
+        b[ 2] += b[13]; b[13] = left_rot(b[13], rotates[41]); b[13] ^= b[ 2];
+        b[ 6] += b[11]; b[11] = left_rot(b[11], rotates[42]); b[11] ^= b[ 6];
+        b[ 4] += b[15]; b[15] = left_rot(b[15], rotates[43]); b[15] ^= b[ 4];
+        b[10] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[44]); b[ 7] ^= b[10];
+        b[12] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[45]); b[ 3] ^= b[12];
+        b[14] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[46]); b[ 5] ^= b[14];
+        b[ 8] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[47]); b[ 1] ^= b[ 8];
+        if (rounds == 6) break;                           
+                                                          
+        b[ 0] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[48]); b[ 7] ^= b[ 0];
+        b[ 2] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[49]); b[ 5] ^= b[ 2];
+        b[ 4] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[50]); b[ 3] ^= b[ 4];
+        b[ 6] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[51]); b[ 1] ^= b[ 6];
+        b[12] += b[15]; b[15] = left_rot(b[15], rotates[52]); b[15] ^= b[12];
+        b[14] += b[13]; b[13] = left_rot(b[13], rotates[53]); b[13] ^= b[14];
+        b[ 8] += b[11]; b[11] = left_rot(b[11], rotates[54]); b[11] ^= b[ 8];
+        b[10] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[55]); b[ 9] ^= b[10];
+        if (rounds == 7) break;                           
+                                                          
+        b[ 0] += b[15]; b[15] = left_rot(b[15], rotates[56]); b[15] ^= b[ 0];
+        b[ 2] += b[11]; b[11] = left_rot(b[11], rotates[57]); b[11] ^= b[ 2];
+        b[ 6] += b[13]; b[13] = left_rot(b[13], rotates[58]); b[13] ^= b[ 6];
+        b[ 4] += b[ 9]; b[ 9] = left_rot(b[ 9], rotates[59]); b[ 9] ^= b[ 4];
+        b[14] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[60]); b[ 1] ^= b[14];
+        b[ 8] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[61]); b[ 5] ^= b[ 8];
+        b[10] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[62]); b[ 3] ^= b[10];
+        b[12] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[63]); b[ 7] ^= b[12];
+        }
+    }
+
+void fwd_cycle_8(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 0]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 1]); b[ 3] ^= b[ 2];
+        b[ 4] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[ 2]); b[ 5] ^= b[ 4];
+        b[ 6] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[ 3]); b[ 7] ^= b[ 6];
+        if (rounds == 1) break;
+
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 4]); b[ 1] ^= b[ 2];
+        b[ 4] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[ 5]); b[ 7] ^= b[ 4];
+        b[ 6] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[ 6]); b[ 5] ^= b[ 6];
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 7]); b[ 3] ^= b[ 0];
+        if (rounds == 2) break;
+
+        b[ 4] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 8]); b[ 1] ^= b[ 4];
+        b[ 6] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 9]); b[ 3] ^= b[ 6];
+        b[ 0] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[10]); b[ 5] ^= b[ 0];
+        b[ 2] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[11]); b[ 7] ^= b[ 2];
+        if (rounds == 3) break;
+
+        b[ 6] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[12]); b[ 1] ^= b[ 6];
+        b[ 0] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[13]); b[ 7] ^= b[ 0];
+        b[ 2] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[14]); b[ 5] ^= b[ 2];
+        b[ 4] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[15]); b[ 3] ^= b[ 4];
+        if (rounds == 4) break;
+
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[16]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[17]); b[ 3] ^= b[ 2];
+        b[ 4] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[18]); b[ 5] ^= b[ 4];
+        b[ 6] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[19]); b[ 7] ^= b[ 6];
+        if (rounds == 5) break;
+
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[20]); b[ 1] ^= b[ 2];
+        b[ 4] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[21]); b[ 7] ^= b[ 4];
+        b[ 6] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[22]); b[ 5] ^= b[ 6];
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[23]); b[ 3] ^= b[ 0];
+        if (rounds == 6) break;
+
+        b[ 4] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[24]); b[ 1] ^= b[ 4];
+        b[ 6] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[25]); b[ 3] ^= b[ 6];
+        b[ 0] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[26]); b[ 5] ^= b[ 0];
+        b[ 2] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[27]); b[ 7] ^= b[ 2];
+        if (rounds == 7) break;
+
+        b[ 6] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[28]); b[ 1] ^= b[ 6];
+        b[ 0] += b[ 7]; b[ 7] = left_rot(b[ 7], rotates[29]); b[ 7] ^= b[ 0];
+        b[ 2] += b[ 5]; b[ 5] = left_rot(b[ 5], rotates[30]); b[ 5] ^= b[ 2];
+        b[ 4] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[31]); b[ 3] ^= b[ 4];
+        }
+    }
+
+void fwd_cycle_4(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 0]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 1]); b[ 3] ^= b[ 2];
+        if (rounds == 1) break;
+
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 2]); b[ 3] ^= b[ 0];
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 3]); b[ 1] ^= b[ 2];
+        if (rounds == 2) break;
+
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 4]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 5]); b[ 3] ^= b[ 2];
+        if (rounds == 3) break;
+
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 6]); b[ 3] ^= b[ 0];
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 7]); b[ 1] ^= b[ 2];
+        if (rounds == 4) break;
+
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[ 8]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[ 9]); b[ 3] ^= b[ 2];
+        if (rounds == 5) break;
+
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[10]); b[ 3] ^= b[ 0];
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[11]); b[ 1] ^= b[ 2];
+        if (rounds == 6) break;
+
+        b[ 0] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[12]); b[ 1] ^= b[ 0];
+        b[ 2] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[13]); b[ 3] ^= b[ 2];
+        if (rounds == 7) break;
+
+        b[ 0] += b[ 3]; b[ 3] = left_rot(b[ 3], rotates[14]); b[ 3] ^= b[ 0];
+        b[ 2] += b[ 1]; b[ 1] = left_rot(b[ 1], rotates[15]); b[ 1] ^= b[ 2];
+        }
+    }
+
+/* reverse versions of the cipher */
+void rev_cycle_16(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds = (rounds-1) & ~7)
+        {
+        switch (rounds & 7)
+            {
+            case 0:
+                    b[ 7] ^= b[12]; b[ 7] = right_rot(b[ 7], rotates[63]); b[12] -= b[ 7]; 
+                    b[ 3] ^= b[10]; b[ 3] = right_rot(b[ 3], rotates[62]); b[10] -= b[ 3]; 
+                    b[ 5] ^= b[ 8]; b[ 5] = right_rot(b[ 5], rotates[61]); b[ 8] -= b[ 5]; 
+                    b[ 1] ^= b[14]; b[ 1] = right_rot(b[ 1], rotates[60]); b[14] -= b[ 1]; 
+                    b[ 9] ^= b[ 4]; b[ 9] = right_rot(b[ 9], rotates[59]); b[ 4] -= b[ 9]; 
+                    b[13] ^= b[ 6]; b[13] = right_rot(b[13], rotates[58]); b[ 6] -= b[13]; 
+                    b[11] ^= b[ 2]; b[11] = right_rot(b[11], rotates[57]); b[ 2] -= b[11]; 
+                    b[15] ^= b[ 0]; b[15] = right_rot(b[15], rotates[56]); b[ 0] -= b[15];
+            case 7:                                                                       
+                    b[ 9] ^= b[10]; b[ 9] = right_rot(b[ 9], rotates[55]); b[10] -= b[ 9];
+                    b[11] ^= b[ 8]; b[11] = right_rot(b[11], rotates[54]); b[ 8] -= b[11];
+                    b[13] ^= b[14]; b[13] = right_rot(b[13], rotates[53]); b[14] -= b[13];
+                    b[15] ^= b[12]; b[15] = right_rot(b[15], rotates[52]); b[12] -= b[15];
+                    b[ 1] ^= b[ 6]; b[ 1] = right_rot(b[ 1], rotates[51]); b[ 6] -= b[ 1];
+                    b[ 3] ^= b[ 4]; b[ 3] = right_rot(b[ 3], rotates[50]); b[ 4] -= b[ 3];
+                    b[ 5] ^= b[ 2]; b[ 5] = right_rot(b[ 5], rotates[49]); b[ 2] -= b[ 5];
+                    b[ 7] ^= b[ 0]; b[ 7] = right_rot(b[ 7], rotates[48]); b[ 0] -= b[ 7];
+            case 6:                                                                       
+                    b[ 1] ^= b[ 8]; b[ 1] = right_rot(b[ 1], rotates[47]); b[ 8] -= b[ 1];
+                    b[ 5] ^= b[14]; b[ 5] = right_rot(b[ 5], rotates[46]); b[14] -= b[ 5];
+                    b[ 3] ^= b[12]; b[ 3] = right_rot(b[ 3], rotates[45]); b[12] -= b[ 3];
+                    b[ 7] ^= b[10]; b[ 7] = right_rot(b[ 7], rotates[44]); b[10] -= b[ 7];
+                    b[15] ^= b[ 4]; b[15] = right_rot(b[15], rotates[43]); b[ 4] -= b[15];
+                    b[11] ^= b[ 6]; b[11] = right_rot(b[11], rotates[42]); b[ 6] -= b[11];
+                    b[13] ^= b[ 2]; b[13] = right_rot(b[13], rotates[41]); b[ 2] -= b[13];
+                    b[ 9] ^= b[ 0]; b[ 9] = right_rot(b[ 9], rotates[40]); b[ 0] -= b[ 9];
+            case 5:                                                                       
+                    b[15] ^= b[14]; b[15] = right_rot(b[15], rotates[39]); b[14] -= b[15];
+                    b[13] ^= b[12]; b[13] = right_rot(b[13], rotates[38]); b[12] -= b[13];
+                    b[11] ^= b[10]; b[11] = right_rot(b[11], rotates[37]); b[10] -= b[11];
+                    b[ 9] ^= b[ 8]; b[ 9] = right_rot(b[ 9], rotates[36]); b[ 8] -= b[ 9];
+                    b[ 7] ^= b[ 6]; b[ 7] = right_rot(b[ 7], rotates[35]); b[ 6] -= b[ 7];
+                    b[ 5] ^= b[ 4]; b[ 5] = right_rot(b[ 5], rotates[34]); b[ 4] -= b[ 5];
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[33]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[32]); b[ 0] -= b[ 1];
+            case 4:                                                                       
+                    b[ 7] ^= b[12]; b[ 7] = right_rot(b[ 7], rotates[31]); b[12] -= b[ 7];
+                    b[ 3] ^= b[10]; b[ 3] = right_rot(b[ 3], rotates[30]); b[10] -= b[ 3];
+                    b[ 5] ^= b[ 8]; b[ 5] = right_rot(b[ 5], rotates[29]); b[ 8] -= b[ 5];
+                    b[ 1] ^= b[14]; b[ 1] = right_rot(b[ 1], rotates[28]); b[14] -= b[ 1];
+                    b[ 9] ^= b[ 4]; b[ 9] = right_rot(b[ 9], rotates[27]); b[ 4] -= b[ 9];
+                    b[13] ^= b[ 6]; b[13] = right_rot(b[13], rotates[26]); b[ 6] -= b[13];
+                    b[11] ^= b[ 2]; b[11] = right_rot(b[11], rotates[25]); b[ 2] -= b[11];
+                    b[15] ^= b[ 0]; b[15] = right_rot(b[15], rotates[24]); b[ 0] -= b[15];
+            case 3:                                                                       
+                    b[ 9] ^= b[10]; b[ 9] = right_rot(b[ 9], rotates[23]); b[10] -= b[ 9];
+                    b[11] ^= b[ 8]; b[11] = right_rot(b[11], rotates[22]); b[ 8] -= b[11];
+                    b[13] ^= b[14]; b[13] = right_rot(b[13], rotates[21]); b[14] -= b[13];
+                    b[15] ^= b[12]; b[15] = right_rot(b[15], rotates[20]); b[12] -= b[15];
+                    b[ 1] ^= b[ 6]; b[ 1] = right_rot(b[ 1], rotates[19]); b[ 6] -= b[ 1];
+                    b[ 3] ^= b[ 4]; b[ 3] = right_rot(b[ 3], rotates[18]); b[ 4] -= b[ 3];
+                    b[ 5] ^= b[ 2]; b[ 5] = right_rot(b[ 5], rotates[17]); b[ 2] -= b[ 5];
+                    b[ 7] ^= b[ 0]; b[ 7] = right_rot(b[ 7], rotates[16]); b[ 0] -= b[ 7];
+            case 2:                                                                       
+                    b[ 1] ^= b[ 8]; b[ 1] = right_rot(b[ 1], rotates[15]); b[ 8] -= b[ 1];
+                    b[ 5] ^= b[14]; b[ 5] = right_rot(b[ 5], rotates[14]); b[14] -= b[ 5];
+                    b[ 3] ^= b[12]; b[ 3] = right_rot(b[ 3], rotates[13]); b[12] -= b[ 3];
+                    b[ 7] ^= b[10]; b[ 7] = right_rot(b[ 7], rotates[12]); b[10] -= b[ 7];
+                    b[15] ^= b[ 4]; b[15] = right_rot(b[15], rotates[11]); b[ 4] -= b[15];
+                    b[11] ^= b[ 6]; b[11] = right_rot(b[11], rotates[10]); b[ 6] -= b[11];
+                    b[13] ^= b[ 2]; b[13] = right_rot(b[13], rotates[ 9]); b[ 2] -= b[13];
+                    b[ 9] ^= b[ 0]; b[ 9] = right_rot(b[ 9], rotates[ 8]); b[ 0] -= b[ 9];
+            case 1:                                                                       
+                    b[15] ^= b[14]; b[15] = right_rot(b[15], rotates[ 7]); b[14] -= b[15];
+                    b[13] ^= b[12]; b[13] = right_rot(b[13], rotates[ 6]); b[12] -= b[13];
+                    b[11] ^= b[10]; b[11] = right_rot(b[11], rotates[ 5]); b[10] -= b[11];
+                    b[ 9] ^= b[ 8]; b[ 9] = right_rot(b[ 9], rotates[ 4]); b[ 8] -= b[ 9];
+                    b[ 7] ^= b[ 6]; b[ 7] = right_rot(b[ 7], rotates[ 3]); b[ 6] -= b[ 7];
+                    b[ 5] ^= b[ 4]; b[ 5] = right_rot(b[ 5], rotates[ 2]); b[ 4] -= b[ 5];
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[ 1]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[ 0]); b[ 0] -= b[ 1];
+            }                                                                             
+                                                                                          
+        }                                                                                 
+    }                                                                                     
+                                                                                          
+void rev_cycle_8(Word *b, const u08b *rotates, int rounds)                                
+    {                                                                                     
+    for (;rounds > 0;rounds = (rounds-1) & ~7)                                            
+        {                                                                                 
+        switch (rounds & 7)                                                               
+            {                                                                             
+            case 0:                                                                       
+                    b[ 3] ^= b[ 4]; b[ 3] = right_rot(b[ 3], rotates[31]); b[ 4] -= b[ 3];
+                    b[ 5] ^= b[ 2]; b[ 5] = right_rot(b[ 5], rotates[30]); b[ 2] -= b[ 5];
+                    b[ 7] ^= b[ 0]; b[ 7] = right_rot(b[ 7], rotates[29]); b[ 0] -= b[ 7];
+                    b[ 1] ^= b[ 6]; b[ 1] = right_rot(b[ 1], rotates[28]); b[ 6] -= b[ 1];
+            case 7:                                                                       
+                    b[ 7] ^= b[ 2]; b[ 7] = right_rot(b[ 7], rotates[27]); b[ 2] -= b[ 7];
+                    b[ 5] ^= b[ 0]; b[ 5] = right_rot(b[ 5], rotates[26]); b[ 0] -= b[ 5];
+                    b[ 3] ^= b[ 6]; b[ 3] = right_rot(b[ 3], rotates[25]); b[ 6] -= b[ 3];
+                    b[ 1] ^= b[ 4]; b[ 1] = right_rot(b[ 1], rotates[24]); b[ 4] -= b[ 1];
+            case 6:                                                                       
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[23]); b[ 0] -= b[ 3];
+                    b[ 5] ^= b[ 6]; b[ 5] = right_rot(b[ 5], rotates[22]); b[ 6] -= b[ 5];
+                    b[ 7] ^= b[ 4]; b[ 7] = right_rot(b[ 7], rotates[21]); b[ 4] -= b[ 7];
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[20]); b[ 2] -= b[ 1];
+            case 5:                                                                       
+                    b[ 7] ^= b[ 6]; b[ 7] = right_rot(b[ 7], rotates[19]); b[ 6] -= b[ 7];
+                    b[ 5] ^= b[ 4]; b[ 5] = right_rot(b[ 5], rotates[18]); b[ 4] -= b[ 5];
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[17]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[16]); b[ 0] -= b[ 1];
+            case 4:                                                                       
+                    b[ 3] ^= b[ 4]; b[ 3] = right_rot(b[ 3], rotates[15]); b[ 4] -= b[ 3];
+                    b[ 5] ^= b[ 2]; b[ 5] = right_rot(b[ 5], rotates[14]); b[ 2] -= b[ 5];
+                    b[ 7] ^= b[ 0]; b[ 7] = right_rot(b[ 7], rotates[13]); b[ 0] -= b[ 7];
+                    b[ 1] ^= b[ 6]; b[ 1] = right_rot(b[ 1], rotates[12]); b[ 6] -= b[ 1];
+            case 3:                                                                       
+                    b[ 7] ^= b[ 2]; b[ 7] = right_rot(b[ 7], rotates[11]); b[ 2] -= b[ 7];
+                    b[ 5] ^= b[ 0]; b[ 5] = right_rot(b[ 5], rotates[10]); b[ 0] -= b[ 5];
+                    b[ 3] ^= b[ 6]; b[ 3] = right_rot(b[ 3], rotates[ 9]); b[ 6] -= b[ 3];
+                    b[ 1] ^= b[ 4]; b[ 1] = right_rot(b[ 1], rotates[ 8]); b[ 4] -= b[ 1];
+            case 2:                                                                       
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[ 7]); b[ 0] -= b[ 3];
+                    b[ 5] ^= b[ 6]; b[ 5] = right_rot(b[ 5], rotates[ 6]); b[ 6] -= b[ 5];
+                    b[ 7] ^= b[ 4]; b[ 7] = right_rot(b[ 7], rotates[ 5]); b[ 4] -= b[ 7];
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[ 4]); b[ 2] -= b[ 1];
+            case 1:                                                                       
+                    b[ 7] ^= b[ 6]; b[ 7] = right_rot(b[ 7], rotates[ 3]); b[ 6] -= b[ 7];
+                    b[ 5] ^= b[ 4]; b[ 5] = right_rot(b[ 5], rotates[ 2]); b[ 4] -= b[ 5];
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[ 1]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[ 0]); b[ 0] -= b[ 1];
+            }                                                                             
+        }                                                                                 
+    }                                                                                     
+                                                                                          
+void rev_cycle_4(Word *b, const u08b *rotates, int rounds)                                
+    {                                                                                     
+    for (;rounds > 0;rounds = (rounds-1) & ~7)                                            
+        {                                                                                 
+        switch (rounds & 7)                                                               
+            {                                                                             
+            case 0:                                                                       
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[15]); b[ 2] -= b[ 1];
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[14]); b[ 0] -= b[ 3];
+            case 7:                                                                       
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[13]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[12]); b[ 0] -= b[ 1];
+            case 6:                                                                       
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[11]); b[ 2] -= b[ 1];
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[10]); b[ 0] -= b[ 3];
+            case 5:                                                                       
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[ 9]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[ 8]); b[ 0] -= b[ 1];
+            case 4:                                                                       
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[ 7]); b[ 2] -= b[ 1];
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[ 6]); b[ 0] -= b[ 3];
+            case 3:                                                                       
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[ 5]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[ 4]); b[ 0] -= b[ 1];
+            case 2:                                                                       
+                    b[ 1] ^= b[ 2]; b[ 1] = right_rot(b[ 1], rotates[ 3]); b[ 2] -= b[ 1];
+                    b[ 3] ^= b[ 0]; b[ 3] = right_rot(b[ 3], rotates[ 2]); b[ 0] -= b[ 3];
+            case 1:                                                                       
+                    b[ 3] ^= b[ 2]; b[ 3] = right_rot(b[ 3], rotates[ 1]); b[ 2] -= b[ 3];
+                    b[ 1] ^= b[ 0]; b[ 1] = right_rot(b[ 1], rotates[ 0]); b[ 0] -= b[ 1];
+            }
+        }
+    }
+
+#ifdef TEST_OR  /* enable this to simplify testing, since OR is not invertible */
+#define AddOp(I,J) b[I] += b[J]
+#define SubOp(I,J) b[I] -= b[J]
+#define XorOp(I,J) b[I] ^= b[J]
+#else           /* this is the "real" OR version */
+#define AddOp(I,J) b[I] |= b[J]
+#define SubOp(I,J) b[I] |= b[J]
+#define XorOp(I,J) b[I] |= b[J]
+#endif
+
+/* "OR" versions of the cipher: replace ADD, XOR with OR */
+void fwd_cycle_16_or(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[ 2]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[ 3]); XorOp( 7, 6);
+        AddOp( 8, 9); b[ 9] = left_rot(b[ 9], rotates[ 4]); XorOp( 9, 8);
+        AddOp(10,11); b[11] = left_rot(b[11], rotates[ 5]); XorOp(11,10);
+        AddOp(12,13); b[13] = left_rot(b[13], rotates[ 6]); XorOp(13,12);
+        AddOp(14,15); b[15] = left_rot(b[15], rotates[ 7]); XorOp(15,14);
+        if (rounds == 1) break;                         
+                                                        
+        AddOp( 0, 9); b[ 9] = left_rot(b[ 9], rotates[ 8]); XorOp( 9, 0);
+        AddOp( 2,13); b[13] = left_rot(b[13], rotates[ 9]); XorOp(13, 2);
+        AddOp( 6,11); b[11] = left_rot(b[11], rotates[10]); XorOp(11, 6);
+        AddOp( 4,15); b[15] = left_rot(b[15], rotates[11]); XorOp(15, 4);
+        AddOp(10, 7); b[ 7] = left_rot(b[ 7], rotates[12]); XorOp( 7,10);
+        AddOp(12, 3); b[ 3] = left_rot(b[ 3], rotates[13]); XorOp( 3,12);
+        AddOp(14, 5); b[ 5] = left_rot(b[ 5], rotates[14]); XorOp( 5,14);
+        AddOp( 8, 1); b[ 1] = left_rot(b[ 1], rotates[15]); XorOp( 1, 8);
+        if (rounds == 2) break;                         
+                                                        
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[16]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[17]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[18]); XorOp( 3, 4);
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[19]); XorOp( 1, 6);
+        AddOp(12,15); b[15] = left_rot(b[15], rotates[20]); XorOp(15,12);
+        AddOp(14,13); b[13] = left_rot(b[13], rotates[21]); XorOp(13,14);
+        AddOp( 8,11); b[11] = left_rot(b[11], rotates[22]); XorOp(11, 8);
+        AddOp(10, 9); b[ 9] = left_rot(b[ 9], rotates[23]); XorOp( 9,10);
+        if (rounds == 3) break;                         
+                                                        
+        AddOp( 0,15); b[15] = left_rot(b[15], rotates[24]); XorOp(15, 0);
+        AddOp( 2,11); b[11] = left_rot(b[11], rotates[25]); XorOp(11, 2);
+        AddOp( 6,13); b[13] = left_rot(b[13], rotates[26]); XorOp(13, 6);
+        AddOp( 4, 9); b[ 9] = left_rot(b[ 9], rotates[27]); XorOp( 9, 4);
+        AddOp(14, 1); b[ 1] = left_rot(b[ 1], rotates[28]); XorOp( 1,14);
+        AddOp( 8, 5); b[ 5] = left_rot(b[ 5], rotates[29]); XorOp( 5, 8);
+        AddOp(10, 3); b[ 3] = left_rot(b[ 3], rotates[30]); XorOp( 3,10);
+        AddOp(12, 7); b[ 7] = left_rot(b[ 7], rotates[31]); XorOp( 7,12);
+        if (rounds == 4) break;                         
+                                                        
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[32]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[33]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[34]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[35]); XorOp( 7, 6);
+        AddOp( 8, 9); b[ 9] = left_rot(b[ 9], rotates[36]); XorOp( 9, 8);
+        AddOp(10,11); b[11] = left_rot(b[11], rotates[37]); XorOp(11,10);
+        AddOp(12,13); b[13] = left_rot(b[13], rotates[38]); XorOp(13,12);
+        AddOp(14,15); b[15] = left_rot(b[15], rotates[39]); XorOp(15,14);
+        if (rounds == 5) break;                         
+                                                        
+        AddOp( 0, 9); b[ 9] = left_rot(b[ 9], rotates[40]); XorOp( 9, 0);
+        AddOp( 2,13); b[13] = left_rot(b[13], rotates[41]); XorOp(13, 2);
+        AddOp( 6,11); b[11] = left_rot(b[11], rotates[42]); XorOp(11, 6);
+        AddOp( 4,15); b[15] = left_rot(b[15], rotates[43]); XorOp(15, 4);
+        AddOp(10, 7); b[ 7] = left_rot(b[ 7], rotates[44]); XorOp( 7,10);
+        AddOp(12, 3); b[ 3] = left_rot(b[ 3], rotates[45]); XorOp( 3,12);
+        AddOp(14, 5); b[ 5] = left_rot(b[ 5], rotates[46]); XorOp( 5,14);
+        AddOp( 8, 1); b[ 1] = left_rot(b[ 1], rotates[47]); XorOp( 1, 8);
+        if (rounds == 6) break;                         
+                                                        
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[48]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[49]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[50]); XorOp( 3, 4);
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[51]); XorOp( 1, 6);
+        AddOp(12,15); b[15] = left_rot(b[15], rotates[52]); XorOp(15,12);
+        AddOp(14,13); b[13] = left_rot(b[13], rotates[53]); XorOp(13,14);
+        AddOp( 8,11); b[11] = left_rot(b[11], rotates[54]); XorOp(11, 8);
+        AddOp(10, 9); b[ 9] = left_rot(b[ 9], rotates[55]); XorOp( 9,10);
+        if (rounds == 7) break;                         
+                                                        
+        AddOp( 0,15); b[15] = left_rot(b[15], rotates[56]); XorOp(15, 0);
+        AddOp( 2,11); b[11] = left_rot(b[11], rotates[57]); XorOp(11, 2);
+        AddOp( 6,13); b[13] = left_rot(b[13], rotates[58]); XorOp(13, 6);
+        AddOp( 4, 9); b[ 9] = left_rot(b[ 9], rotates[59]); XorOp( 9, 4);
+        AddOp(14, 1); b[ 1] = left_rot(b[ 1], rotates[60]); XorOp( 1,14);
+        AddOp( 8, 5); b[ 5] = left_rot(b[ 5], rotates[61]); XorOp( 5, 8);
+        AddOp(10, 3); b[ 3] = left_rot(b[ 3], rotates[62]); XorOp( 3,10);
+        AddOp(12, 7); b[ 7] = left_rot(b[ 7], rotates[63]); XorOp( 7,12);
+        }
+    }
+
+void fwd_cycle_8_or(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[ 2]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[ 3]); XorOp( 7, 6);
+        if (rounds == 1) break;
+
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 4]); XorOp( 1, 2);
+        AddOp( 4, 7); b[ 7] = left_rot(b[ 7], rotates[ 5]); XorOp( 7, 4);
+        AddOp( 6, 5); b[ 5] = left_rot(b[ 5], rotates[ 6]); XorOp( 5, 6);
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 7]); XorOp( 3, 0);
+        if (rounds == 2) break;
+
+        AddOp( 4, 1); b[ 1] = left_rot(b[ 1], rotates[ 8]); XorOp( 1, 4);
+        AddOp( 6, 3); b[ 3] = left_rot(b[ 3], rotates[ 9]); XorOp( 3, 6);
+        AddOp( 0, 5); b[ 5] = left_rot(b[ 5], rotates[10]); XorOp( 5, 0);
+        AddOp( 2, 7); b[ 7] = left_rot(b[ 7], rotates[11]); XorOp( 7, 2);
+        if (rounds == 3) break;
+
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[12]); XorOp( 1, 6);
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[13]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[14]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[15]); XorOp( 3, 4);
+        if (rounds == 4) break;
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[16]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[17]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[18]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[19]); XorOp( 7, 6);
+        if (rounds == 5) break;
+
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[20]); XorOp( 1, 2);
+        AddOp( 4, 7); b[ 7] = left_rot(b[ 7], rotates[21]); XorOp( 7, 4);
+        AddOp( 6, 5); b[ 5] = left_rot(b[ 5], rotates[22]); XorOp( 5, 6);
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[23]); XorOp( 3, 0);
+        if (rounds == 6) break;
+
+        AddOp( 4, 1); b[ 1] = left_rot(b[ 1], rotates[24]); XorOp( 1, 4);
+        AddOp( 6, 3); b[ 3] = left_rot(b[ 3], rotates[25]); XorOp( 3, 6);
+        AddOp( 0, 5); b[ 5] = left_rot(b[ 5], rotates[26]); XorOp( 5, 0);
+        AddOp( 2, 7); b[ 7] = left_rot(b[ 7], rotates[27]); XorOp( 7, 2);
+        if (rounds == 7) break;
+
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[28]); XorOp( 1, 6);
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[29]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[30]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[31]); XorOp( 3, 4);
+        }
+    }
+
+void fwd_cycle_4_or(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds -=8)
+        {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        if (rounds == 1) break;
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 2]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 3]); XorOp( 1, 2);
+        if (rounds == 2) break;
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 4]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 5]); XorOp( 3, 2);
+        if (rounds == 3) break;
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 6]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 7]); XorOp( 1, 2);
+        if (rounds == 4) break;
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 8]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 9]); XorOp( 3, 2);
+        if (rounds == 5) break;
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[10]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[11]); XorOp( 1, 2);
+        if (rounds == 6) break;
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[12]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[13]); XorOp( 3, 2);
+        if (rounds == 7) break;
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[14]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[15]); XorOp( 1, 2);
+        }
+    }
+
+/* reverse versions of the cipher, using OR */
+void rev_cycle_16_or(Word *b, const u08b *rotates, int rounds)
+    {
+    for (;rounds > 0;rounds = (rounds-1) & ~7)
+        {
+        switch (rounds & 7)
+            {
+            case 0:
+                    XorOp( 7,12); b[ 7] = right_rot(b[ 7], rotates[63]); SubOp(12, 7); 
+                    XorOp( 3,10); b[ 3] = right_rot(b[ 3], rotates[62]); SubOp(10, 3); 
+                    XorOp( 5, 8); b[ 5] = right_rot(b[ 5], rotates[61]); SubOp( 8, 5); 
+                    XorOp( 1,14); b[ 1] = right_rot(b[ 1], rotates[60]); SubOp(14, 1); 
+                    XorOp( 9, 4); b[ 9] = right_rot(b[ 9], rotates[59]); SubOp( 4, 9); 
+                    XorOp(13, 6); b[13] = right_rot(b[13], rotates[58]); SubOp( 6,13); 
+                    XorOp(11, 2); b[11] = right_rot(b[11], rotates[57]); SubOp( 2,11); 
+                    XorOp(15, 0); b[15] = right_rot(b[15], rotates[56]); SubOp( 0,15);
+            case 7:
+                    XorOp( 9,10); b[ 9] = right_rot(b[ 9], rotates[55]); SubOp(10, 9); 
+                    XorOp(11, 8); b[11] = right_rot(b[11], rotates[54]); SubOp( 8,11); 
+                    XorOp(13,14); b[13] = right_rot(b[13], rotates[53]); SubOp(14,13); 
+                    XorOp(15,12); b[15] = right_rot(b[15], rotates[52]); SubOp(12,15); 
+                    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[51]); SubOp( 6, 1); 
+                    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[50]); SubOp( 4, 3); 
+                    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[49]); SubOp( 2, 5); 
+                    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[48]); SubOp( 0, 7);
+            case 6:
+                    XorOp( 1, 8); b[ 1] = right_rot(b[ 1], rotates[47]); SubOp( 8, 1); 
+                    XorOp( 5,14); b[ 5] = right_rot(b[ 5], rotates[46]); SubOp(14, 5); 
+                    XorOp( 3,12); b[ 3] = right_rot(b[ 3], rotates[45]); SubOp(12, 3); 
+                    XorOp( 7,10); b[ 7] = right_rot(b[ 7], rotates[44]); SubOp(10, 7); 
+                    XorOp(15, 4); b[15] = right_rot(b[15], rotates[43]); SubOp( 4,15); 
+                    XorOp(11, 6); b[11] = right_rot(b[11], rotates[42]); SubOp( 6,11); 
+                    XorOp(13, 2); b[13] = right_rot(b[13], rotates[41]); SubOp( 2,13); 
+                    XorOp( 9, 0); b[ 9] = right_rot(b[ 9], rotates[40]); SubOp( 0, 9);
+            case 5:
+                    XorOp(15,14); b[15] = right_rot(b[15], rotates[39]); SubOp(14,15); 
+                    XorOp(13,12); b[13] = right_rot(b[13], rotates[38]); SubOp(12,13); 
+                    XorOp(11,10); b[11] = right_rot(b[11], rotates[37]); SubOp(10,11); 
+                    XorOp( 9, 8); b[ 9] = right_rot(b[ 9], rotates[36]); SubOp( 8, 9); 
+                    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[35]); SubOp( 6, 7); 
+                    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[34]); SubOp( 4, 5); 
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[33]); SubOp( 2, 3); 
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[32]); SubOp( 0, 1);
+            case 4:
+                    XorOp( 7,12); b[ 7] = right_rot(b[ 7], rotates[31]); SubOp(12, 7); 
+                    XorOp( 3,10); b[ 3] = right_rot(b[ 3], rotates[30]); SubOp(10, 3); 
+                    XorOp( 5, 8); b[ 5] = right_rot(b[ 5], rotates[29]); SubOp( 8, 5); 
+                    XorOp( 1,14); b[ 1] = right_rot(b[ 1], rotates[28]); SubOp(14, 1); 
+                    XorOp( 9, 4); b[ 9] = right_rot(b[ 9], rotates[27]); SubOp( 4, 9); 
+                    XorOp(13, 6); b[13] = right_rot(b[13], rotates[26]); SubOp( 6,13); 
+                    XorOp(11, 2); b[11] = right_rot(b[11], rotates[25]); SubOp( 2,11); 
+                    XorOp(15, 0); b[15] = right_rot(b[15], rotates[24]); SubOp( 0,15);
+            case 3:                                                                       
+                    XorOp( 9,10); b[ 9] = right_rot(b[ 9], rotates[23]); SubOp(10, 9);
+                    XorOp(11, 8); b[11] = right_rot(b[11], rotates[22]); SubOp( 8,11);
+                    XorOp(13,14); b[13] = right_rot(b[13], rotates[21]); SubOp(14,13);
+                    XorOp(15,12); b[15] = right_rot(b[15], rotates[20]); SubOp(12,15);
+                    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[19]); SubOp( 6, 1);
+                    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[18]); SubOp( 4, 3);
+                    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[17]); SubOp( 2, 5);
+                    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[16]); SubOp( 0, 7);
+            case 2:                                                                       
+                    XorOp( 1, 8); b[ 1] = right_rot(b[ 1], rotates[15]); SubOp( 8, 1);
+                    XorOp( 5,14); b[ 5] = right_rot(b[ 5], rotates[14]); SubOp(14, 5);
+                    XorOp( 3,12); b[ 3] = right_rot(b[ 3], rotates[13]); SubOp(12, 3);
+                    XorOp( 7,10); b[ 7] = right_rot(b[ 7], rotates[12]); SubOp(10, 7);
+                    XorOp(15, 4); b[15] = right_rot(b[15], rotates[11]); SubOp( 4,15);
+                    XorOp(11, 6); b[11] = right_rot(b[11], rotates[10]); SubOp( 6,11);
+                    XorOp(13, 2); b[13] = right_rot(b[13], rotates[ 9]); SubOp( 2,13);
+                    XorOp( 9, 0); b[ 9] = right_rot(b[ 9], rotates[ 8]); SubOp( 0, 9);
+            case 1:                                                                       
+                    XorOp(15,14); b[15] = right_rot(b[15], rotates[ 7]); SubOp(14,15);
+                    XorOp(13,12); b[13] = right_rot(b[13], rotates[ 6]); SubOp(12,13);
+                    XorOp(11,10); b[11] = right_rot(b[11], rotates[ 5]); SubOp(10,11);
+                    XorOp( 9, 8); b[ 9] = right_rot(b[ 9], rotates[ 4]); SubOp( 8, 9);
+                    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[ 3]); SubOp( 6, 7);
+                    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[ 2]); SubOp( 4, 5);
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+            }                                                                             
+                                                                                          
+        }                                                                                 
+    }                                                                                     
+                                                                                          
+void rev_cycle_8_or(Word *b, const u08b *rotates, int rounds)                             
+    {                                                                                     
+    for (;rounds > 0;rounds = (rounds-1) & ~7)                                            
+        {                                                                                 
+        switch (rounds & 7)                                                               
+            {                                                                             
+            case 0:                                                                       
+                    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[31]); SubOp( 4, 3);
+                    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[30]); SubOp( 2, 5);
+                    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[29]); SubOp( 0, 7);
+                    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[28]); SubOp( 6, 1);
+            case 7:                                                                       
+                    XorOp( 7, 2); b[ 7] = right_rot(b[ 7], rotates[27]); SubOp( 2, 7);
+                    XorOp( 5, 0); b[ 5] = right_rot(b[ 5], rotates[26]); SubOp( 0, 5);
+                    XorOp( 3, 6); b[ 3] = right_rot(b[ 3], rotates[25]); SubOp( 6, 3);
+                    XorOp( 1, 4); b[ 1] = right_rot(b[ 1], rotates[24]); SubOp( 4, 1);
+            case 6:                                                                       
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[23]); SubOp( 0, 3);
+                    XorOp( 5, 6); b[ 5] = right_rot(b[ 5], rotates[22]); SubOp( 6, 5);
+                    XorOp( 7, 4); b[ 7] = right_rot(b[ 7], rotates[21]); SubOp( 4, 7);
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[20]); SubOp( 2, 1);
+            case 5:                                                                       
+                    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[19]); SubOp( 6, 7);
+                    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[18]); SubOp( 4, 5);
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[17]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[16]); SubOp( 0, 1);
+            case 4:                                                                       
+                    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[15]); SubOp( 4, 3);
+                    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[14]); SubOp( 2, 5);
+                    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[13]); SubOp( 0, 7);
+                    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[12]); SubOp( 6, 1);
+            case 3:                                                                       
+                    XorOp( 7, 2); b[ 7] = right_rot(b[ 7], rotates[11]); SubOp( 2, 7);
+                    XorOp( 5, 0); b[ 5] = right_rot(b[ 5], rotates[10]); SubOp( 0, 5);
+                    XorOp( 3, 6); b[ 3] = right_rot(b[ 3], rotates[ 9]); SubOp( 6, 3);
+                    XorOp( 1, 4); b[ 1] = right_rot(b[ 1], rotates[ 8]); SubOp( 4, 1);
+            case 2:                                                                       
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 7]); SubOp( 0, 3);
+                    XorOp( 5, 6); b[ 5] = right_rot(b[ 5], rotates[ 6]); SubOp( 6, 5);
+                    XorOp( 7, 4); b[ 7] = right_rot(b[ 7], rotates[ 5]); SubOp( 4, 7);
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 4]); SubOp( 2, 1);
+            case 1:                                                                       
+                    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[ 3]); SubOp( 6, 7);
+                    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[ 2]); SubOp( 4, 5);
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+            }                                                                             
+        }                                                                                 
+    }                                                                                     
+                                                                                          
+void rev_cycle_4_or(Word *b, const u08b *rotates, int rounds)                             
+    {                                                                                     
+    for (;rounds > 0;rounds = (rounds-1) & ~7)                                            
+        {                                                                                 
+        switch (rounds & 7)                                                               
+            {                                                                             
+            case 0:                                                                       
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[15]); SubOp( 2, 1);
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[14]); SubOp( 0, 3);
+            case 7:                                                                       
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[13]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[12]); SubOp( 0, 1);
+            case 6:                                                                       
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[11]); SubOp( 2, 1);
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[10]); SubOp( 0, 3);
+            case 5:                                                                       
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 9]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 8]); SubOp( 0, 1);
+            case 4:                                                                       
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 7]); SubOp( 2, 1);
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 6]); SubOp( 0, 3);
+            case 3:                                                                       
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 5]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 4]); SubOp( 0, 1);
+            case 2:                                                                       
+                    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 3]); SubOp( 2, 1);
+                    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 2]); SubOp( 0, 3);
+            case 1:                                                                       
+                    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+                    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+            }
+        }
+    }
+
+/* optimized versions for default round counts */
+#if   defined(__BORLANDC__)
+#pragma argsused
+#elif defined(_MSC_VER)
+#pragma warning(disable:4100)
+#endif
+void fwd_cycle_16_or_r9(Word *b, const u08b *rotates, int rounds)
+    {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[ 2]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[ 3]); XorOp( 7, 6);
+        AddOp( 8, 9); b[ 9] = left_rot(b[ 9], rotates[ 4]); XorOp( 9, 8);
+        AddOp(10,11); b[11] = left_rot(b[11], rotates[ 5]); XorOp(11,10);
+        AddOp(12,13); b[13] = left_rot(b[13], rotates[ 6]); XorOp(13,12);
+        AddOp(14,15); b[15] = left_rot(b[15], rotates[ 7]); XorOp(15,14);
+
+        AddOp( 0, 9); b[ 9] = left_rot(b[ 9], rotates[ 8]); XorOp( 9, 0);
+        AddOp( 2,13); b[13] = left_rot(b[13], rotates[ 9]); XorOp(13, 2);
+        AddOp( 6,11); b[11] = left_rot(b[11], rotates[10]); XorOp(11, 6);
+        AddOp( 4,15); b[15] = left_rot(b[15], rotates[11]); XorOp(15, 4);
+        AddOp(10, 7); b[ 7] = left_rot(b[ 7], rotates[12]); XorOp( 7,10);
+        AddOp(12, 3); b[ 3] = left_rot(b[ 3], rotates[13]); XorOp( 3,12);
+        AddOp(14, 5); b[ 5] = left_rot(b[ 5], rotates[14]); XorOp( 5,14);
+        AddOp( 8, 1); b[ 1] = left_rot(b[ 1], rotates[15]); XorOp( 1, 8);
+
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[16]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[17]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[18]); XorOp( 3, 4);
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[19]); XorOp( 1, 6);
+        AddOp(12,15); b[15] = left_rot(b[15], rotates[20]); XorOp(15,12);
+        AddOp(14,13); b[13] = left_rot(b[13], rotates[21]); XorOp(13,14);
+        AddOp( 8,11); b[11] = left_rot(b[11], rotates[22]); XorOp(11, 8);
+        AddOp(10, 9); b[ 9] = left_rot(b[ 9], rotates[23]); XorOp( 9,10);
+
+        AddOp( 0,15); b[15] = left_rot(b[15], rotates[24]); XorOp(15, 0);
+        AddOp( 2,11); b[11] = left_rot(b[11], rotates[25]); XorOp(11, 2);
+        AddOp( 6,13); b[13] = left_rot(b[13], rotates[26]); XorOp(13, 6);
+        AddOp( 4, 9); b[ 9] = left_rot(b[ 9], rotates[27]); XorOp( 9, 4);
+        AddOp(14, 1); b[ 1] = left_rot(b[ 1], rotates[28]); XorOp( 1,14);
+        AddOp( 8, 5); b[ 5] = left_rot(b[ 5], rotates[29]); XorOp( 5, 8);
+        AddOp(10, 3); b[ 3] = left_rot(b[ 3], rotates[30]); XorOp( 3,10);
+        AddOp(12, 7); b[ 7] = left_rot(b[ 7], rotates[31]); XorOp( 7,12);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[32]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[33]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[34]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[35]); XorOp( 7, 6);
+        AddOp( 8, 9); b[ 9] = left_rot(b[ 9], rotates[36]); XorOp( 9, 8);
+        AddOp(10,11); b[11] = left_rot(b[11], rotates[37]); XorOp(11,10);
+        AddOp(12,13); b[13] = left_rot(b[13], rotates[38]); XorOp(13,12);
+        AddOp(14,15); b[15] = left_rot(b[15], rotates[39]); XorOp(15,14);
+
+        AddOp( 0, 9); b[ 9] = left_rot(b[ 9], rotates[40]); XorOp( 9, 0);
+        AddOp( 2,13); b[13] = left_rot(b[13], rotates[41]); XorOp(13, 2);
+        AddOp( 6,11); b[11] = left_rot(b[11], rotates[42]); XorOp(11, 6);
+        AddOp( 4,15); b[15] = left_rot(b[15], rotates[43]); XorOp(15, 4);
+        AddOp(10, 7); b[ 7] = left_rot(b[ 7], rotates[44]); XorOp( 7,10);
+        AddOp(12, 3); b[ 3] = left_rot(b[ 3], rotates[45]); XorOp( 3,12);
+        AddOp(14, 5); b[ 5] = left_rot(b[ 5], rotates[46]); XorOp( 5,14);
+        AddOp( 8, 1); b[ 1] = left_rot(b[ 1], rotates[47]); XorOp( 1, 8);
+
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[48]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[49]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[50]); XorOp( 3, 4);
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[51]); XorOp( 1, 6);
+        AddOp(12,15); b[15] = left_rot(b[15], rotates[52]); XorOp(15,12);
+        AddOp(14,13); b[13] = left_rot(b[13], rotates[53]); XorOp(13,14);
+        AddOp( 8,11); b[11] = left_rot(b[11], rotates[54]); XorOp(11, 8);
+        AddOp(10, 9); b[ 9] = left_rot(b[ 9], rotates[55]); XorOp( 9,10);
+
+        AddOp( 0,15); b[15] = left_rot(b[15], rotates[56]); XorOp(15, 0);
+        AddOp( 2,11); b[11] = left_rot(b[11], rotates[57]); XorOp(11, 2);
+        AddOp( 6,13); b[13] = left_rot(b[13], rotates[58]); XorOp(13, 6);
+        AddOp( 4, 9); b[ 9] = left_rot(b[ 9], rotates[59]); XorOp( 9, 4);
+        AddOp(14, 1); b[ 1] = left_rot(b[ 1], rotates[60]); XorOp( 1,14);
+        AddOp( 8, 5); b[ 5] = left_rot(b[ 5], rotates[61]); XorOp( 5, 8);
+        AddOp(10, 3); b[ 3] = left_rot(b[ 3], rotates[62]); XorOp( 3,10);
+        AddOp(12, 7); b[ 7] = left_rot(b[ 7], rotates[63]); XorOp( 7,12);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[ 2]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[ 3]); XorOp( 7, 6);
+        AddOp( 8, 9); b[ 9] = left_rot(b[ 9], rotates[ 4]); XorOp( 9, 8);
+        AddOp(10,11); b[11] = left_rot(b[11], rotates[ 5]); XorOp(11,10);
+        AddOp(12,13); b[13] = left_rot(b[13], rotates[ 6]); XorOp(13,12);
+        AddOp(14,15); b[15] = left_rot(b[15], rotates[ 7]); XorOp(15,14);
+    }
+
+#if   defined(__BORLANDC__)
+#pragma argsused
+#endif
+void fwd_cycle_8_or_r8(Word *b, const u08b *rotates, int rounds)
+    {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[ 2]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[ 3]); XorOp( 7, 6);
+
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 4]); XorOp( 1, 2);
+        AddOp( 4, 7); b[ 7] = left_rot(b[ 7], rotates[ 5]); XorOp( 7, 4);
+        AddOp( 6, 5); b[ 5] = left_rot(b[ 5], rotates[ 6]); XorOp( 5, 6);
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 7]); XorOp( 3, 0);
+
+        AddOp( 4, 1); b[ 1] = left_rot(b[ 1], rotates[ 8]); XorOp( 1, 4);
+        AddOp( 6, 3); b[ 3] = left_rot(b[ 3], rotates[ 9]); XorOp( 3, 6);
+        AddOp( 0, 5); b[ 5] = left_rot(b[ 5], rotates[10]); XorOp( 5, 0);
+        AddOp( 2, 7); b[ 7] = left_rot(b[ 7], rotates[11]); XorOp( 7, 2);
+
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[12]); XorOp( 1, 6);
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[13]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[14]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[15]); XorOp( 3, 4);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[16]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[17]); XorOp( 3, 2);
+        AddOp( 4, 5); b[ 5] = left_rot(b[ 5], rotates[18]); XorOp( 5, 4);
+        AddOp( 6, 7); b[ 7] = left_rot(b[ 7], rotates[19]); XorOp( 7, 6);
+
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[20]); XorOp( 1, 2);
+        AddOp( 4, 7); b[ 7] = left_rot(b[ 7], rotates[21]); XorOp( 7, 4);
+        AddOp( 6, 5); b[ 5] = left_rot(b[ 5], rotates[22]); XorOp( 5, 6);
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[23]); XorOp( 3, 0);
+
+        AddOp( 4, 1); b[ 1] = left_rot(b[ 1], rotates[24]); XorOp( 1, 4);
+        AddOp( 6, 3); b[ 3] = left_rot(b[ 3], rotates[25]); XorOp( 3, 6);
+        AddOp( 0, 5); b[ 5] = left_rot(b[ 5], rotates[26]); XorOp( 5, 0);
+        AddOp( 2, 7); b[ 7] = left_rot(b[ 7], rotates[27]); XorOp( 7, 2);
+
+        AddOp( 6, 1); b[ 1] = left_rot(b[ 1], rotates[28]); XorOp( 1, 6);
+        AddOp( 0, 7); b[ 7] = left_rot(b[ 7], rotates[29]); XorOp( 7, 0);
+        AddOp( 2, 5); b[ 5] = left_rot(b[ 5], rotates[30]); XorOp( 5, 2);
+        AddOp( 4, 3); b[ 3] = left_rot(b[ 3], rotates[31]); XorOp( 3, 4);
+    }
+
+#ifdef __BORLANDC__
+#pragma argsused
+#endif
+void fwd_cycle_4_or_r8(Word *b, const u08b *rotates, int rounds)
+    {
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 0]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 1]); XorOp( 3, 2);
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 2]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 3]); XorOp( 1, 2);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 4]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 5]); XorOp( 3, 2);
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[ 6]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[ 7]); XorOp( 1, 2);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[ 8]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[ 9]); XorOp( 3, 2);
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[10]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[11]); XorOp( 1, 2);
+
+        AddOp( 0, 1); b[ 1] = left_rot(b[ 1], rotates[12]); XorOp( 1, 0);
+        AddOp( 2, 3); b[ 3] = left_rot(b[ 3], rotates[13]); XorOp( 3, 2);
+
+        AddOp( 0, 3); b[ 3] = left_rot(b[ 3], rotates[14]); XorOp( 3, 0);
+        AddOp( 2, 1); b[ 1] = left_rot(b[ 1], rotates[15]); XorOp( 1, 2);
+    }
+
+/* reverse versions of the cipher, using OR, for fixed round numbers */
+#ifdef __BORLANDC__
+#pragma argsused
+#endif
+void rev_cycle_16_or_r9(Word *b, const u08b *rotates, int rounds)
+    {
+    XorOp(15,14); b[15] = right_rot(b[15], rotates[ 7]); SubOp(14,15);
+    XorOp(13,12); b[13] = right_rot(b[13], rotates[ 6]); SubOp(12,13);
+    XorOp(11,10); b[11] = right_rot(b[11], rotates[ 5]); SubOp(10,11);
+    XorOp( 9, 8); b[ 9] = right_rot(b[ 9], rotates[ 4]); SubOp( 8, 9);
+    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[ 3]); SubOp( 6, 7);
+    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[ 2]); SubOp( 4, 5);
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+                                                     
+    XorOp( 7,12); b[ 7] = right_rot(b[ 7], rotates[63]); SubOp(12, 7); 
+    XorOp( 3,10); b[ 3] = right_rot(b[ 3], rotates[62]); SubOp(10, 3); 
+    XorOp( 5, 8); b[ 5] = right_rot(b[ 5], rotates[61]); SubOp( 8, 5); 
+    XorOp( 1,14); b[ 1] = right_rot(b[ 1], rotates[60]); SubOp(14, 1); 
+    XorOp( 9, 4); b[ 9] = right_rot(b[ 9], rotates[59]); SubOp( 4, 9); 
+    XorOp(13, 6); b[13] = right_rot(b[13], rotates[58]); SubOp( 6,13); 
+    XorOp(11, 2); b[11] = right_rot(b[11], rotates[57]); SubOp( 2,11); 
+    XorOp(15, 0); b[15] = right_rot(b[15], rotates[56]); SubOp( 0,15);
+                                                     
+    XorOp( 9,10); b[ 9] = right_rot(b[ 9], rotates[55]); SubOp(10, 9); 
+    XorOp(11, 8); b[11] = right_rot(b[11], rotates[54]); SubOp( 8,11); 
+    XorOp(13,14); b[13] = right_rot(b[13], rotates[53]); SubOp(14,13); 
+    XorOp(15,12); b[15] = right_rot(b[15], rotates[52]); SubOp(12,15); 
+    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[51]); SubOp( 6, 1); 
+    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[50]); SubOp( 4, 3); 
+    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[49]); SubOp( 2, 5); 
+    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[48]); SubOp( 0, 7);
+                                                     
+    XorOp( 1, 8); b[ 1] = right_rot(b[ 1], rotates[47]); SubOp( 8, 1); 
+    XorOp( 5,14); b[ 5] = right_rot(b[ 5], rotates[46]); SubOp(14, 5); 
+    XorOp( 3,12); b[ 3] = right_rot(b[ 3], rotates[45]); SubOp(12, 3); 
+    XorOp( 7,10); b[ 7] = right_rot(b[ 7], rotates[44]); SubOp(10, 7); 
+    XorOp(15, 4); b[15] = right_rot(b[15], rotates[43]); SubOp( 4,15); 
+    XorOp(11, 6); b[11] = right_rot(b[11], rotates[42]); SubOp( 6,11); 
+    XorOp(13, 2); b[13] = right_rot(b[13], rotates[41]); SubOp( 2,13); 
+    XorOp( 9, 0); b[ 9] = right_rot(b[ 9], rotates[40]); SubOp( 0, 9);
+                                                     
+    XorOp(15,14); b[15] = right_rot(b[15], rotates[39]); SubOp(14,15); 
+    XorOp(13,12); b[13] = right_rot(b[13], rotates[38]); SubOp(12,13); 
+    XorOp(11,10); b[11] = right_rot(b[11], rotates[37]); SubOp(10,11); 
+    XorOp( 9, 8); b[ 9] = right_rot(b[ 9], rotates[36]); SubOp( 8, 9); 
+    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[35]); SubOp( 6, 7); 
+    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[34]); SubOp( 4, 5); 
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[33]); SubOp( 2, 3); 
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[32]); SubOp( 0, 1);
+                                                     
+    XorOp( 7,12); b[ 7] = right_rot(b[ 7], rotates[31]); SubOp(12, 7); 
+    XorOp( 3,10); b[ 3] = right_rot(b[ 3], rotates[30]); SubOp(10, 3); 
+    XorOp( 5, 8); b[ 5] = right_rot(b[ 5], rotates[29]); SubOp( 8, 5); 
+    XorOp( 1,14); b[ 1] = right_rot(b[ 1], rotates[28]); SubOp(14, 1); 
+    XorOp( 9, 4); b[ 9] = right_rot(b[ 9], rotates[27]); SubOp( 4, 9); 
+    XorOp(13, 6); b[13] = right_rot(b[13], rotates[26]); SubOp( 6,13); 
+    XorOp(11, 2); b[11] = right_rot(b[11], rotates[25]); SubOp( 2,11); 
+    XorOp(15, 0); b[15] = right_rot(b[15], rotates[24]); SubOp( 0,15);
+                                                     
+    XorOp( 9,10); b[ 9] = right_rot(b[ 9], rotates[23]); SubOp(10, 9);
+    XorOp(11, 8); b[11] = right_rot(b[11], rotates[22]); SubOp( 8,11);
+    XorOp(13,14); b[13] = right_rot(b[13], rotates[21]); SubOp(14,13);
+    XorOp(15,12); b[15] = right_rot(b[15], rotates[20]); SubOp(12,15);
+    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[19]); SubOp( 6, 1);
+    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[18]); SubOp( 4, 3);
+    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[17]); SubOp( 2, 5);
+    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[16]); SubOp( 0, 7);
+                                                     
+    XorOp( 1, 8); b[ 1] = right_rot(b[ 1], rotates[15]); SubOp( 8, 1);
+    XorOp( 5,14); b[ 5] = right_rot(b[ 5], rotates[14]); SubOp(14, 5);
+    XorOp( 3,12); b[ 3] = right_rot(b[ 3], rotates[13]); SubOp(12, 3);
+    XorOp( 7,10); b[ 7] = right_rot(b[ 7], rotates[12]); SubOp(10, 7);
+    XorOp(15, 4); b[15] = right_rot(b[15], rotates[11]); SubOp( 4,15);
+    XorOp(11, 6); b[11] = right_rot(b[11], rotates[10]); SubOp( 6,11);
+    XorOp(13, 2); b[13] = right_rot(b[13], rotates[ 9]); SubOp( 2,13);
+    XorOp( 9, 0); b[ 9] = right_rot(b[ 9], rotates[ 8]); SubOp( 0, 9);
+                                                     
+    XorOp(15,14); b[15] = right_rot(b[15], rotates[ 7]); SubOp(14,15);
+    XorOp(13,12); b[13] = right_rot(b[13], rotates[ 6]); SubOp(12,13);
+    XorOp(11,10); b[11] = right_rot(b[11], rotates[ 5]); SubOp(10,11);
+    XorOp( 9, 8); b[ 9] = right_rot(b[ 9], rotates[ 4]); SubOp( 8, 9);
+    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[ 3]); SubOp( 6, 7);
+    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[ 2]); SubOp( 4, 5);
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+    }
+                                                                                          
+#ifdef __BORLANDC__
+#pragma argsused
+#endif
+void rev_cycle_8_or_r8(Word *b, const u08b *rotates, int rounds)                             
+    {                                                                                     
+    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[31]); SubOp( 4, 3);
+    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[30]); SubOp( 2, 5);
+    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[29]); SubOp( 0, 7);
+    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[28]); SubOp( 6, 1);
+
+    XorOp( 7, 2); b[ 7] = right_rot(b[ 7], rotates[27]); SubOp( 2, 7);
+    XorOp( 5, 0); b[ 5] = right_rot(b[ 5], rotates[26]); SubOp( 0, 5);
+    XorOp( 3, 6); b[ 3] = right_rot(b[ 3], rotates[25]); SubOp( 6, 3);
+    XorOp( 1, 4); b[ 1] = right_rot(b[ 1], rotates[24]); SubOp( 4, 1);
+
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[23]); SubOp( 0, 3);
+    XorOp( 5, 6); b[ 5] = right_rot(b[ 5], rotates[22]); SubOp( 6, 5);
+    XorOp( 7, 4); b[ 7] = right_rot(b[ 7], rotates[21]); SubOp( 4, 7);
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[20]); SubOp( 2, 1);
+
+    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[19]); SubOp( 6, 7);
+    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[18]); SubOp( 4, 5);
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[17]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[16]); SubOp( 0, 1);
+
+    XorOp( 3, 4); b[ 3] = right_rot(b[ 3], rotates[15]); SubOp( 4, 3);
+    XorOp( 5, 2); b[ 5] = right_rot(b[ 5], rotates[14]); SubOp( 2, 5);
+    XorOp( 7, 0); b[ 7] = right_rot(b[ 7], rotates[13]); SubOp( 0, 7);
+    XorOp( 1, 6); b[ 1] = right_rot(b[ 1], rotates[12]); SubOp( 6, 1);
+
+    XorOp( 7, 2); b[ 7] = right_rot(b[ 7], rotates[11]); SubOp( 2, 7);
+    XorOp( 5, 0); b[ 5] = right_rot(b[ 5], rotates[10]); SubOp( 0, 5);
+    XorOp( 3, 6); b[ 3] = right_rot(b[ 3], rotates[ 9]); SubOp( 6, 3);
+    XorOp( 1, 4); b[ 1] = right_rot(b[ 1], rotates[ 8]); SubOp( 4, 1);
+
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 7]); SubOp( 0, 3);
+    XorOp( 5, 6); b[ 5] = right_rot(b[ 5], rotates[ 6]); SubOp( 6, 5);
+    XorOp( 7, 4); b[ 7] = right_rot(b[ 7], rotates[ 5]); SubOp( 4, 7);
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 4]); SubOp( 2, 1);
+
+    XorOp( 7, 6); b[ 7] = right_rot(b[ 7], rotates[ 3]); SubOp( 6, 7);
+    XorOp( 5, 4); b[ 5] = right_rot(b[ 5], rotates[ 2]); SubOp( 4, 5);
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+    }                                                                                     
+                                                                                          
+#ifdef __BORLANDC__
+#pragma argsused
+#endif
+void rev_cycle_4_or_r8(Word *b, const u08b *rotates, int rounds)                             
+    {                                                                                     
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[15]); SubOp( 2, 1);
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[14]); SubOp( 0, 3);
+
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[13]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[12]); SubOp( 0, 1);
+
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[11]); SubOp( 2, 1);
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[10]); SubOp( 0, 3);
+
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 9]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 8]); SubOp( 0, 1);
+
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 7]); SubOp( 2, 1);
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 6]); SubOp( 0, 3);
+
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 5]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 4]); SubOp( 0, 1);
+
+    XorOp( 1, 2); b[ 1] = right_rot(b[ 1], rotates[ 3]); SubOp( 2, 1);
+    XorOp( 3, 0); b[ 3] = right_rot(b[ 3], rotates[ 2]); SubOp( 0, 3);
+
+    XorOp( 3, 2); b[ 3] = right_rot(b[ 3], rotates[ 1]); SubOp( 2, 3);
+    XorOp( 1, 0); b[ 1] = right_rot(b[ 1], rotates[ 0]); SubOp( 0, 1);
+    }
+
+
+/* test that fwd and rev ciphers are truly inverses */
+void InverseChecks(void)
+    {
+    uint  i,j,k,wCnt,tstCnt;
+    int   r,rN;
+    Block pt,ct,xt;
+    u08b  rots[MAX_ROTS_PER_CYCLE];
+    uint  TEST_CNT = (sizeof(size_t) == 8) ? 64 : 8;
+
+    cycle_func *fwd;
+    cycle_func *rev;
+    cycle_func *fwd_or;
+    cycle_func *fwd_or_rN;
+#ifdef TEST_OR
+    cycle_func *rev_or;
+    cycle_func *rev_or_rN;
+#endif
+    
+    Rand_Init(0);
+    for (wCnt=4;wCnt<=MAX_WORDS_PER_BLK;wCnt *= 2)
+        {
+        switch (wCnt)
+            {
+            case  4: fwd       = fwd_cycle_4        ; rev       = rev_cycle_4        ;
+                     fwd_or    = fwd_cycle_4_or     ; fwd_or_rN = fwd_cycle_4_or_r8  ; break;
+            case  8: fwd       = fwd_cycle_8        ; rev       = rev_cycle_8        ;
+                     fwd_or    = fwd_cycle_8_or     ; fwd_or_rN = fwd_cycle_8_or_r8  ; break;
+            default: fwd       = fwd_cycle_16       ; rev       = rev_cycle_16       ; 
+                     fwd_or    = fwd_cycle_16_or    ; fwd_or_rN = fwd_cycle_16_or_r9 ; break;
+            }
+#ifdef TEST_OR
+        switch (wCnt)
+            {
+            case  4: rev_or_rN = rev_cycle_4_or_r8  ; rev_or    = rev_cycle_4_or     ; break;
+            case  8: rev_or_rN = rev_cycle_8_or_r8  ; rev_or    = rev_cycle_8_or     ; break;
+            default: rev_or_rN = rev_cycle_16_or_r9 ; rev_or    = rev_cycle_16_or    ; break;
+            }
+#endif
+        for (tstCnt=0;tstCnt<TEST_CNT;tstCnt++)
+            {
+            if (tstCnt == 0)
+                {
+                memset(pt.x,0,sizeof(pt));      /* make the first test simple, for debug */
+                pt.x[0]++;
+                }
+            else
+                RandBytes(pt.x,wCnt*sizeof(pt.x[0]));
+
+            RandBytes(rots,sizeof(rots));       /* use random rotation constants */
+            for (i=0;i<MAX_ROTS_PER_CYCLE;i++)
+                rots[i] &= (BITS_PER_WORD-1);
+            for (r=1;r<32;r++)
+                {
+                ct=pt;
+                rev(ct.x,rots,r);
+                fwd(ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure: #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }
+                fwd(ct.x,rots,r);
+                rev(ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure: #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }
+#ifdef TEST_OR
+                fwd_or(ct.x,rots,r);
+                rev   (ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure (fwd_or): #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }
+                fwd   (ct.x,rots,r);
+                rev_or(ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure (rev_or): #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }                
+                if (r != ((wCnt == 16) ? 9 : 8))
+                    continue;
+                fwd_or_rN(ct.x,rots,r);
+                rev      (ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure (fwd_or_rN): #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }                
+                fwd      (ct.x,rots,r);
+                rev_or_rN(ct.x,rots,r);
+                if (memcmp(pt.x,ct.x,wCnt*sizeof(pt.x[0])))
+                    {
+                    printf("Inverse failure (rev_or_rN): #%03d: wCnt=%d. r=%2d",tstCnt,wCnt,r);
+                    exit(8);
+                    }                
+#else
+                /* validate that "quick" Hamming weight checks are ok, using OR */
+                for (i=0;i<wCnt;i++)
+                    {
+                    memset(ct.x,0,sizeof(ct.x));
+                    ct.x[i]=1;
+                    fwd_or(ct.x,rots,r);
+                    for (j=1;j<64;j++)
+                        {
+                        memset(xt.x,0,sizeof(xt.x));
+                        xt.x[i]=((u64b) 1) << j;
+                        fwd_or(xt.x,rots,r);
+                        for (k=0;k<wCnt;k++)
+                            if (left_rot(ct.x[k],j) != xt.x[k])
+                                {
+                                printf("Quick HW check failure: blk=%4d bits. r=%d. j=%d",wCnt*64,r,j);
+                                exit(2);
+                                }
+                        }
+                    }
+#endif
+                }
+            }
+        /* test the "hard coded" versions against variable versions of OR routines */
+        for (tstCnt=0;tstCnt<TEST_CNT;tstCnt++)
+            {
+            RandBytes(rots,sizeof(rots));
+            for (i=0;i<MAX_ROTS_PER_CYCLE;i++)
+                rots[i] &= (BITS_PER_WORD-1);
+            rN = (wCnt == 16) ? 9 : 8;
+            for (i=0;i<wCnt*64;i++)
+                {
+                memset(pt.x,0,sizeof(pt));
+                pt.x[i / 64] = ((u64b) 1) << (i % 64);
+                ct=pt;
+                xt=pt;
+                fwd_or   (ct.x,rots,rN);
+                fwd_or_rN(xt.x,rots,rN);
+                if (memcmp(xt.x,ct.x,wCnt*sizeof(xt.x[0])))
+                    {
+                    printf("OR failure: #%03d: wCnt=%d. i=%2d",tstCnt,wCnt,i);
+                    exit(8);
+                    }
+                }
+            }
+        }
+    }
+
+/* count the bits set in the word */
+uint HammingWeight(Word x)
+    {
+#if BITS_PER_WORD == 64
+    x = (x & DUP_64(0x55555555)) + ((x >> 1) & DUP_64(0x55555555));
+    x = (x & DUP_64(0x33333333)) + ((x >> 2) & DUP_64(0x33333333));
+    x = (x & DUP_64(0x0F0F0F0F)) + ((x >> 4) & DUP_64(0x0F0F0F0F));
+    x = (x & DUP_64(0x00FF00FF)) + ((x >> 8) & DUP_64(0x00FF00FF));
+    x = (x & DUP_64(0x0000FFFF)) + ((x >>16) & DUP_64(0x0000FFFF));
+    x = (x & DUP_64(0x000000FF)) + ((x >>32) & DUP_64(0x000000FF));
+#else
+    x = (x & 0x55555555) + ((x >> 1) & 0x55555555);
+    x = (x & 0x33333333) + ((x >> 2) & 0x33333333);
+    x = (x & 0x0F0F0F0F) + ((x >> 4) & 0x0F0F0F0F);
+    x = (x & 0x00FF00FF) + ((x >> 8) & 0x00FF00FF);
+    x = (x & 0x0000FFFF) + ((x >>16) & 0x000000FF);
+#endif
+    return (uint) x;
+    }
+
+
+/* use the CRC value as quick ID to help identify/verify rotation sets */
+void Set_CRC(rSearchRec *r)
+    {
+#define CRC_FDBK ((0x04C11DB7u >> 1) ^ 0x80000000u) /* CRC-32-IEEE-802.3 (from Wikipedia) */
+    uint i,h=~0u;
+
+    for (i=0;i<rotsPerCycle;i++)
+        {
+        h ^= r->rotList[i];
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        h = (h & 1) ? (h >> 1) ^ CRC_FDBK : (h >> 1);
+        }
+    r->CRC = h;
+    }
+
+/* qsort routine for search records: keep in descending order */
+int Compare_SearchRec_Descending(const void *aPtr,const void *bPtr)
+    {
+    uint wA = ((const rSearchRec *) aPtr)->rWorst;
+    uint wB = ((const rSearchRec *) bPtr)->rWorst;
+
+    if (wA < wB)
+        return +1;
+    if (wA > wB)
+        return -1;
+    else
+        {   /* equal metric. Sort by ID number */
+        wA = ((const rSearchRec *) aPtr)->ID;
+        wB = ((const rSearchRec *) bPtr)->ID;
+        if (wA < wB)
+            return -1;
+        if (wA > wB)
+            return +1;
+        return  0;
+        }
+    }
+
+const char *ASCII_TimeDate(void)
+    {
+    time_t t;
+    time(&t);   
+    return ctime(&t);
+    }
+
+/* test the rotation set for minimum hamming weight >= minHW */
+/*   [try to do it fast: rely on rotational symmetry using OR, */
+/*    and do an early exit if hamming weight is too low] */
+int Cycle_Min_HW(uint rounds, const u08b *rotList,uint minHW,uint verMask)
+    {
+    uint    i,j,v,hw,hMin;
+    u08b    rots[MAX_ROTS_PER_CYCLE];
+    Block   b;
+
+    hMin = BITS_PER_WORD;
+    for (v=0;v<MAX_ROT_VER_CNT;v++)
+        {
+        if ((verMask & (1 << v)) == 0)
+            continue;
+        if (v & 1)
+            { /* do it on the "half-cycle" */
+            for (i=0;i<rotsPerCycle;i++)
+                {
+                rots[i] = rotList[(i >= rotsPerCycle/2) ? i - rotsPerCycle/2 : i + rotsPerCycle/2];
+                }
+            }
+        else
+            memcpy(rots,rotList,rotsPerCycle*sizeof(rots[0]));
+        for (i=0;i<wordsPerBlock;i++)
+            {
+            memset(b.x,0,wordsPerBlock*sizeof(b.x[0]));
+            b.x[i] = 1;                     /* test propagation into one word */
+            if (minHW)
+                {       /* use the "_rN" versions for speed */
+                if (v & 2)
+                    rev_cycle_or_rN(b.x,rots,(int)rounds);
+                else
+                    fwd_cycle_or_rN(b.x,rots,(int)rounds);
+                }
+            else
+                {       /* saturation check */
+                if (v & 2)
+                    rev_cycle_or   (b.x,rots,(int)rounds);
+                else
+                    fwd_cycle_or   (b.x,rots,(int)rounds);
+                }
+            for (j=0;j<wordsPerBlock;j++)
+                {
+                hw = HammingWeight(b.x[j]);
+                if (minHW > hw)
+                    return 0;               /* stop if this isn't good enough */
+                if (hMin  > hw)             /* else keep track of min */
+                    hMin  = hw;
+                }
+            }
+        }
+    return hMin;
+    }
+
+/* compute/set the minimum hamming weight of the rotation set */
+/*   [more thorough check than Cycle_Min_HW] */
+uint Set_Min_hw_OR(rSearchRec *r,uint verMask,uint rounds)
+    { 
+    uint  i,j,v,hw,hwMin;
+    u08b  rots[MAX_ROTS_PER_CYCLE];
+    Block b;
+
+    Set_CRC(r);
+    hwMin = BITS_PER_WORD;
+    for (v=0;v<MAX_ROT_VER_CNT;v++)
+        {
+        r->hw_OR[v] = BITS_PER_WORD;
+        if ((verMask & (1 << v)) == 0)
+            continue;
+        if (v & 1)
+            { /* do it on the "half-cycle" */
+            for (i=0;i<rotsPerCycle;i++)
+                {
+                rots[i] = r->rotList[(i >= rotsPerCycle/2) ? i - rotsPerCycle/2 : i + rotsPerCycle/2];
+                }
+            }
+        else
+            memcpy(rots,r->rotList,rotsPerCycle*sizeof(rots[0]));
+        for (i=0;i<bitsPerBlock;i+=BITS_PER_WORD)
+            {
+            memset(b.x,0,sizeof(b.x));
+            b.x[i/BITS_PER_WORD] |= (((u64b) 1) << (i%BITS_PER_WORD));
+            if (v & 2)
+                rev_cycle_or(b.x,rots,(int) rounds);
+            else
+                fwd_cycle_or(b.x,rots,(int) rounds);
+            for (j=0;j<wordsPerBlock;j++)
+                {
+                hw = HammingWeight(b.x[j]);
+                if (hwMin > hw)
+                    hwMin = hw;
+                if (r->hw_OR[v] > (u08b) hw)
+                    r->hw_OR[v] = (u08b) hw;
+                }
+            }
+        }
+    return hwMin;
+    }
+
+/* show how the Hamming weight varies as a function of # rounds */
+void Show_HW_rounds(const u08b *rotates)
+    {
+    uint i,r,minHW,hw[4];
+
+    for (r=4;r<12;r++)
+        {  
+        minHW = bitsPerBlock;
+        for (i=0;i<4;i++)
+            {
+            hw[i]=Cycle_Min_HW(r,rotates,0,1 << i);
+            if (minHW > hw[i])
+                minHW = hw[i];
+            }
+        printf("%2d rounds: minHW = %2d  [",r,minHW);
+        for (i=0;i<4;i++)   /* show the different "versions" */
+            printf(" %2d",hw[i]);
+        printf(" ]\n");
+        }
+    }
+
+/* read rotations value from file */
+const u08b *get_rotation_file(const char *rfName)
+    {
+    enum   { MAX_LINE = 512 };
+    char   line[MAX_LINE+4];
+    uint   i,rotVal;
+    uint   rotShow=0;
+    static FILE *rf=NULL;
+    static u08b rotates[MAX_ROTS_PER_CYCLE];
+    static uint rotCnt =0;
+/**** sample format: 
++++++++++++++ Preliminary results: sampleCnt =  1024, block =  256 bits
+rMin = 0.425. #079C[*21] [CRC=D89E7C72. hw_OR=62. cnt= 1024. blkSize= 256]          
+   46   52
+   21   38
+   13   13
+   20   27
+   14   40
+   43   26
+   35   29
+   19   63
+rMin = 0.425. #0646[*17] [CRC=527174F3. hw_OR=61. cnt= 1024. blkSize= 256]          
+   26   24
+   50   48
+   40   25
+   36   55
+   10   20
+   10   16
+   60   55
+   18    7
+...
+****/
+    if (rfName[0] == '+')
+        {
+        rfName++;
+        rotShow = 1;
+        }
+    if (rf == NULL)
+        {
+        rf = fopen(rfName,"rt");
+        if (rf == NULL)
+            {
+            printf("Unable to open rotation file '%s'",rfName);
+            exit(2);
+            }
+        rotCnt=0;
+        for (;;)        /* skip to "preliminary results" section */
+            {
+            line[0]=0;
+            if (fgets(line,sizeof(line)-4,rf) == NULL || line[0] == 0)
+                {
+                fclose(rf);                 /* eof --> stop */
+                rf = NULL;
+                return NULL;
+                }
+            /* check for the header */
+            if (line[0] != '+' || line[1] != '+' || line[2] != '+' ||
+                strstr(line,"reliminary results:") == NULL)
+                continue;
+            /* now check for the correct block size */
+            for (i=strlen(line);i;i--)      /* start at eol and look backwards */
+                if (line[i-1] == '=')       /* check for '=' sign for block size */
+                    break;
+            if (i > 0 && sscanf(line+i,"%u bits",&i) == 1 && i == bitsPerBlock)
+                break;
+            }
+        }
+    /* now at the rMin line */
+    line[0]=0;
+    if (fgets(line,sizeof(line)-4,rf) == NULL || line[0] == 0 || strncmp(line,"rMin =",6))  
+        {
+        fclose(rf);
+        rf = NULL;
+        return NULL;
+        }
+
+    /* now read in all the rotation values */
+    for (i=0;i<rotsPerCycle;i++)
+        {
+        if (fscanf(rf,"%u",&rotVal) != 1 || rotVal >= bitsPerBlock)
+            {   /* Invalid rotation value */
+            fclose(rf);
+            rf = NULL;
+            return NULL;
+            }
+        rotates[i] = (u08b) rotVal;
+        }
+    if (fgets(line,sizeof(line)-4,rf) == NULL)          /* skip eol */
+        {
+        fclose(rf);
+        rf = NULL;
+        }
+    if (rotShow)
+        {   /* show the hamming weight profile */
+        printf("\n:::::::::::\n");
+        printf("Rot #%02d [%4d-bit blocks] read from file '%s':\n",rotCnt,bitsPerBlock,rfName);
+        for (i=0;i<rotsPerCycle;i++)
+            printf("%4d%s",rotates[i],((i+1)%(wordsPerBlock/2))?"":"\n");
+        Show_HW_rounds(rotates);     /* show HW results for different numbers of rounds */
+        printf(":::::::::::\n");
+        }
+    rotCnt++;
+    return rotates;
+    }
+
+/* generate a randomly chosen set of rotation constants of given minimum hamming weight (using OR) */
+/* (this may take a while, depending on minHW,rounds) */
+uint get_rotation(rSearchRec *r,testParms t)
+    {
+    static  u64b rCnt    = 1;
+    static  u64b rCntOK  = 0;
+    static  uint rScale  = BITS_PER_WORD;
+    static  uint hwBase  = 0;
+    static  uint rID     = 1;
+    uint    i,j,k,m,n,b,hw,q,qMask;
+    static  u08b rotates[MAX_ROTS_PER_CYCLE];   /*  last generated rotation set */
+    u08b    goodRots[BITS_PER_WORD];
+    uint    goodRotCnt;
+    
+    r->rWorst       =  0;
+    r->parentCRC    = ~0u;
+
+    if (rotFileName)                            /* get from search results file? */
+        {
+        const u08b *rf = get_rotation_file(rotFileName);
+        if (rf)
+            {
+            for (i=0;i<rotsPerCycle;i++)
+                r->rotList[i] = rf[i];
+            Set_Min_hw_OR(r,t.rotVerMask,t.rounds);
+            r->ID = rID++;
+            return 1;
+            }
+        /* here with file exhausted. Keep going with randomized values */
+        rotFileName = NULL;                     /* don't use file any more */
+        return 0;
+        }
+    for (i=goodRotCnt=0;i<BITS_PER_WORD;i++)
+        if (!RotCnt_Bad(i))
+            {
+            goodRots[goodRotCnt++] = (u08b) i;
+            }
+    
+    qMask   = ((wordsPerBlock/2)-1) & t.dupRotMask;     /* filter for dup rotate counts in the same round? */
+    for (;;rCnt++)
+        {
+        if (hwBase == 0)
+            {   /* pick a rotation set at random */
+            for (i=0;i<rotsPerCycle;)
+                {
+                rotates[i] = goodRots[Rand32() % goodRotCnt];
+                /* filter out unapproved rotation sets here */
+                for (q=i & ~qMask;q < i;q++)    /* check for dups in the same round */
+                    if (rotates[i] == rotates[q])
+                        break;
+                if (q >= i)                     /* no dup, value ok, so this value is ok */
+                    i++;
+                }
+            hw = Cycle_Min_HW(t.rounds,rotates,t.minHW_or-t.minOffs,t.rotVerMask);
+            if (hw == 0)                /* did we get close? */
+                continue;
+            rCntOK++;
+
+            hwBase = hw;
+            if (hw >= t.minHW_or)
+                if (Cycle_Min_HW(t.maxSatRnds, rotates,0,t.rotVerMask) == BITS_PER_WORD)
+                    {
+                    for (i=0;i<rotsPerCycle;i++)
+                        r->rotList[i] = rotates[i];
+                    rScale = 1;         /* set up for scaling below */
+                    }
+            }
+        /* use odd scaling for randomly generated rotations */
+        for (;rScale < BITS_PER_WORD;)
+            {
+            for (i=0;i<rotsPerCycle;i++)
+                {
+                r->rotList[i] = (rotates[i] * rScale) % BITS_PER_WORD;
+                if (RotCnt_Bad(r->rotList[i]))
+                    break;
+                }
+            rScale+=2;                  /* bump scale factor for next time */
+            if (i >= rotsPerCycle)
+                {   /* all values ok: this one's a keeper */
+                Set_Min_hw_OR(r,t.rotVerMask,t.rounds);
+                r->ID = rID++;
+                return 1;
+                }
+            }
+        /* Try nearby values to see if hw gets better: monotonic hill climb. */
+        /*      -- exhaustively try all possible values of pairs of changes  */
+        for (m=0;m<rotsPerCycle;m++)
+        for (b=0;b<BITS_PER_WORD ;b++)
+            {
+            k = rotsPerCycle-1-m;           /* work backwards, since we're already close */
+            rotates[k]++;
+            rotates[k] &= (BITS_PER_WORD-1);
+            if (RotCnt_Bad(rotates[k]))
+                continue;
+            for (q=k | qMask;q > k;q--)    /* check for dups in the same round */
+                if (rotates[k] == rotates[q])
+                    break;
+            if (q > k)      
+                continue;
+            for (i=m+1;i<rotsPerCycle;i++)
+                {
+                n = rotsPerCycle-1-i;   /* work backwards */
+                for (j=0;j<BITS_PER_WORD;j++)
+                    {
+                    rotates[n]++;       /* try another rotation value */
+                    rotates[n] &= (BITS_PER_WORD-1);
+                    if (RotCnt_Bad(rotates[n]))
+                        continue;
+                    for (q=n | qMask;q > n;q--)    /* check for dups in the same round */
+                        if (rotates[n] == rotates[q])
+                            break;
+                    if (q > n)      
+                        continue;  
+                    k  = (t.minHW_or > hwBase) ? t.minHW_or : hwBase;
+                    hw = Cycle_Min_HW(t.rounds,rotates,k,t.rotVerMask);
+                    if (hw > hwBase)
+                        if (Cycle_Min_HW(t.maxSatRnds, rotates,0,t.rotVerMask) == BITS_PER_WORD)
+                            {   /* must improve hw to accept this new rotation set */
+                            assert(hw >= t.minHW_or);
+                            hwBase = hw;
+                            rScale = 3; /* set up for scaling next time */
+                            for (i=0;i<rotsPerCycle;i++)
+                                r->rotList[i] = rotates[i];
+                            Set_Min_hw_OR(r,t.rotVerMask,t.rounds);
+                            r->ID = rID++;
+                            return 1;
+                            }
+                    }
+                }
+            }
+        hwBase = 0;                     /* back to random  */
+        }
+    }
+
+/* display a search record result */
+void ShowSearchRec(FILE *f,const rSearchRec *r,testParms t,uint showMode,char markCh,uint showNum)
+    {
+    uint  i,j,n,hwMin;
+    const char *s;
+    char  fStr[200];
+
+    hwMin=BITS_PER_WORD;
+    for (i=0;i<MAX_ROT_VER_CNT;i++)
+        if (hwMin > (uint) r->hw_OR[i])
+            hwMin = (uint) r->hw_OR[i];
+
+    switch (showMode)
+        {
+        case SHOW_ROTS_FINAL:  sprintf(fStr,".final:%02d " ,showNum); s = fStr; break;
+        case SHOW_ROTS_H:      s = ".format";  break;
+        case SHOW_ROTS_PRELIM: s = ".prelim";  break;
+        default:               s = "";         break;
+        }
+
+    fprintf(f,"rMin = %5.3f.%c [CRC=%08X. parent=%08X. ID=%08X. hw_OR=%2d. cnt=%5d. bits=%4u]%-10s%s%s\n",
+            r->rWorst/(double)t.sampleCnt,markCh,r->CRC,r->parentCRC,r->ID,
+            hwMin,t.sampleCnt,bitsPerBlock,s,
+            (t.tstFlags & TST_FLG_USE_ABS)?" useAbs":"",(r->ID & ID_RECALC_BIT)?" recalc":""
+           );
+
+    switch (showMode)
+        {
+        case SHOW_NONE:
+            break;
+        case SHOW_ROTS_H: /* format for "skein.h" */
+            for (j=n=0;j<rotsPerCycle/(wordsPerBlock/2);j++)
+                {
+                fprintf(f,"   ");
+                for (i=0;i<wordsPerBlock/2;i++)
+                    {
+                    fprintf(f,(wordsPerBlock == 16)?" R%04d":" R_%03d",wordsPerBlock*64);
+                    fprintf(f,"_%d_%d=%2d,",j,i,r->rotList[n++]);
+                    }
+                fprintf(f,"\n");
+                }
+            break;
+        default:
+            for (i=0;i<rotsPerCycle;i++)
+                fprintf(f,"   %2d%s",r->rotList[i],((i+1)%(wordsPerBlock/2))?"":"\n");
+            break;
+        }
+    }
+
+/* compute Skein differentials for a given rotation set */
+uint CheckDifferentials(rSearchRec *r,testParms t)
+    {
+    enum  { HIST_BINS =  20 };
+
+    uint    i,j,k,v,n,d,dMax,minCnt,maxCnt,vCnt,q;
+    uint    rMin,rMax,hwMin,hwMax,hw,rMinCnt,rMaxCnt,iMin,jMin,iMax,jMax;
+    uint    hist[HIST_BINS+1];
+    u08b    rots[MAX_ROTS_PER_CYCLE];
+    u64b    totSum,w,y,z,oMask;
+    double  fSum,fSqr,x,var,denom;
+    static  u64b onesCnt[3][MAX_BITS_PER_BLK][MAX_BITS_PER_BLK/8]; /* pack eight 8-bit counts into each u64b (for speed) */
+    u64b   *oPtr;
+    struct
+        {
+        Block pt,ct;
+        } a,b;
+
+    r->rWorst = t.sampleCnt;
+    dMax = 1u << (t.diffBits & (BITS_PER_WORD-1));
+    iMin = jMin = iMax = jMax = bitsPerBlock + 1;
+
+    for (v=vCnt=0;v < MAX_ROT_VER_CNT; v++)  
+        { /* different versions of rotation schedule, including "inverse" cipher */
+        if ((t.rotVerMask & (1 << v)) == 0)
+            continue;
+        vCnt++;     /* number of versions processed */
+        if (v & 1)
+            { /* do it on the "half-cycle" */
+            for (i=0;i<rotsPerCycle;i++)
+                {
+                rots[i] = r->rotList[(i >= rotsPerCycle/2) ? i - rotsPerCycle/2 : i + rotsPerCycle/2];
+                }
+            }
+        else
+            memcpy(rots,r->rotList,rotsPerCycle*sizeof(rots[0]));
+        for (d=1; d < dMax; d+=2)    /* multi-bit difference patterns (must start with a '1' bit)  */
+            {
+            hwMax=0;
+            hwMin=bitsPerBlock+1;
+            memset(onesCnt,0,sizeof(onesCnt));      /* clear stats before starting */
+                
+            oMask = DUP_64(0x01010101);             /* mask for adding, 8 bins at a time */
+            for (n=1;n<=t.sampleCnt;n++)
+                {
+                for (i=0;i<wordsPerBlock;i++)       /* generate input blocks in a portable way */
+                    a.pt.x[i] = Rand64();
+                a.ct = a.pt;
+                if (v & 2)
+                    rev_cycle(a.ct.x,rots,t.rounds);
+                else
+                    fwd_cycle(a.ct.x,rots,t.rounds);
+                for (i=0;i<bitsPerBlock;i++)
+                    {
+                    b.pt = a.pt;
+                    b.pt.x[i/BITS_PER_WORD] ^= left_rot((u64b)d,(i%BITS_PER_WORD));  /* inject input difference  */
+                    b.ct = b.pt;
+                    if (t.tstFlags & TST_FLG_DO_RAND)
+                        RandBytes(b.ct.x,sizeof(b.ct.x));       /* random results as a comparison point */
+                    else if (v & 2)
+                        rev_cycle(b.ct.x,rots,t.rounds);        /* let Skein do the mixing */
+                    else
+                        fwd_cycle(b.ct.x,rots,t.rounds);        /* let Skein do the mixing */
+                    z  = 0;                                     /* accumulate total hamming weight in z */
+                    oPtr = onesCnt[0][i];
+                    for (j=0;j<wordsPerBlock;j++)
+                        {                                       /* inner-most loop: unroll it fully */
+                        w = b.ct.x[j] ^ a.ct.x[j];              /* xor difference in each ciphertext word */
+                        y = (w     ) & oMask; oPtr[0] += y; z += y;   /* sum 8 bins at a time (bits 0,8,16,24...,56) */
+                        y = (w >> 1) & oMask; oPtr[1] += y; z += y;
+                        y = (w >> 2) & oMask; oPtr[2] += y; z += y;   /* do it 8 times to cover all bits in w */
+                        y = (w >> 3) & oMask; oPtr[3] += y; z += y;
+                                                                                    
+                        y = (w >> 4) & oMask; oPtr[4] += y; z += y;
+                        y = (w >> 5) & oMask; oPtr[5] += y; z += y;
+                        y = (w >> 6) & oMask; oPtr[6] += y; z += y;
+                        y = (w >> 7) & oMask; oPtr[7] += y; z += y;
+                        oPtr += 8;
+                        }
+                    /* sum up the total hamming weight bins (very carefully) */
+                    z = (z & DUP_64(0x00FF00FF)) + ((z >> 8) & DUP_64(0x00FF00FF));
+                    hw  = (uint) (z + (z >> 16) + (z >> 32) + (z >> 48)) & 0xFFFF;
+                    if (hwMin > hw) hwMin = hw;                 /* update total hw min/max stats */
+                    if (hwMax < hw) hwMax = hw;
+                    }
+                if ((n & 0x7F) == 0)
+                    {   /* prevent onesCnt[0] overflow by "transferring" MSBs of 8-bit bytes into onesCnt[1] */
+                    for (i=0;i<bitsPerBlock  ;i++)
+                    for (j=0;j<bitsPerBlock/8;j++)
+                        {   /* add the MSB (bit 7) of each byte into onesCnt[1], then mask it off in onesCnt[0] */
+                        onesCnt[1][i][j] += (onesCnt[0][i][j] >> 7) & oMask;
+                        onesCnt[0][i][j] &= ~(oMask << 7);
+                        }
+                    if ((n & 0x3FFF) == 0)
+                        {   /* propagate overflow into onesCnt[2] (occasionally, as needed) */
+                        for (i=0;i<bitsPerBlock  ;i++)
+                        for (j=0;j<bitsPerBlock/8;j++)
+                            { 
+                            onesCnt[2][i][j] += (onesCnt[1][i][j] >> 7) & oMask;
+                            onesCnt[1][i][j] &= ~(oMask << 7);
+                            }
+                        }
+                    }
+                if (n == 32 && d == 1 && (t.tstFlags & TST_FLG_QUICK_EXIT))
+                    {   /* quick exit if not even close to random looking after a few samples */
+                    for (i=0;i<bitsPerBlock  ;i++)
+                    for (j=0;j<bitsPerBlock/8;j++)
+                        {
+                        if ((onesCnt[0][i][j] & ~oMask) == 0)  /* any count less than 2? */
+                            {
+                            /** Since an ideal random function has prob=0.5 each for input/output bit 
+                             ** pair, the expected distribution of onesCnt[i][j] is binomial. 
+                             ** Thus, at this point, the probability of onesCnt[i][j] < 2 is:
+                             **     ((1+32)/2)/(2**-32)
+                             ** This probability is roughly 2**(-27), so when we observe such an
+                             ** occurrence, we exit immediately to save taking a lot of stats just
+                             ** to fail later. This filter significantly speeds up the search, at a
+                             ** very low probability of improperly dismissing a "good" rotation set.
+                             **/
+                            if (t.tstFlags & TST_FLG_SHOW && vCnt > 1)
+                                {   /* show why we stopped, if we already showed something */
+                                printf("%23s/* quick exit: %d/%d */\n","",(uint)onesCnt[0][i][j],n);
+                                }
+                            return r->rWorst = 0;   /* not a good result */
+                            }
+                        }
+                    }
+                }
+            /* now process the stats from the samples we just generated */
+            assert(t.sampleCnt < (1 << 22));            /* 2**22 is big enough not to worry! */
+            memset(hist,0,sizeof(hist));
+            fSum  = fSqr = 0.0;
+            denom = 1.0 / (double) t.sampleCnt;
+            rMin  = minCnt = ~0u;
+            totSum= rMax = rMinCnt = rMaxCnt = maxCnt = 0;
+            for (i=0;i<bitsPerBlock;i++)
+                {
+                for (j=0;j<bitsPerBlock/8;j++)
+                    {
+                    w = onesCnt[0][i][j];               /* 7+ bits here */
+                    y = onesCnt[1][i][j];               /* 7+ bits here */
+                    z = onesCnt[2][i][j];               /* 8  bits here.  Total = 22 bits */
+                    for (k=0;k<8;k++,w >>= 8,y >>= 8,z >>= 8)
+                        {
+                        q = (uint) ((w & 0xFF) + ((y & 0xFF) << 7) + ((z & 0xFF) << 14));
+                        if (maxCnt < q) { maxCnt = q; iMax = i; jMax = j; if (rMax < q) { rMax = q; rMaxCnt = 0; } }
+                        if (minCnt > q) { minCnt = q; iMin = i; jMin = j; if (rMin > q) { rMin = q; rMinCnt = 0; } }
+                        if (rMin == minCnt) rMinCnt++;
+                        if (rMax == maxCnt) rMaxCnt++;
+                        if (t.tstFlags & TST_FLG_SHOW)
+                            {   /* compute more extensive stats only if showing results below */
+                            totSum  += q;
+                            x        = q*denom;                 /* update stats for stdDev  */
+                            fSum    += x;
+                            fSqr    += x*x;
+                            hist[(uint)floor(x*HIST_BINS)]++;   /* track histogram  */
+                            }
+                        }
+                    }
+                }
+            if (t.tstFlags & TST_FLG_USE_ABS && rMin > t.sampleCnt - rMax)
+                {
+                rMin = t.sampleCnt - rMax;                      /* use max variation from 1/2 */
+                iMin = iMax;
+                jMin = jMax;
+                }
+            if (r->rWorst > rMin)
+                {
+                r->rWorst = rMin;
+                if (rMin == 0)
+                    {  /* if far worse than current best, stop now (to speed up the search) */
+                    if (t.tstFlags & TST_FLG_SHOW && (d > 1 || vCnt > 1)) /* show why we stopped, if we already showed something */
+                        printf("%23s/* early exit */\n","");
+                    return r->rWorst = 0;
+                    }
+                }
+            if (t.tstFlags & TST_FLG_SHOW)
+                {         /* show some detailed results of the test */
+                if (d == 1)
+                    {     /* put out the rotation info the first time thru */
+                    if ((t.tstFlags & TST_FLG_DO_RAND) == 0)
+                        {
+                        printf("Rotation set [CRC=%08X. hw_OR=%2d. sampleCnt=%5d. block=%4d bits. v=%d]:\n",
+                               r->CRC,r->hw_OR[v],t.sampleCnt,bitsPerBlock,v);
+                        if (vCnt == 0)
+                            for (i=0;i<rotsPerCycle;i++)
+                                printf("   %2d%s",r->rotList[i],((i+1)%(wordsPerBlock/2))?"":"\n");
+                        }
+                    }
+                printf("rnds=%2d,cnt=%5d",t.rounds,t.sampleCnt);
+                x  =  fSum/(bitsPerBlock*bitsPerBlock);
+                var= (fSqr/(bitsPerBlock*bitsPerBlock)) - x*x;
+                printf(" min=%5.3f.[%c] max=%5.3f.[%c]  hw=%3d..%3d.  avg=%7.5f. std=%6.4f. d=%X. [%3d,%3d]",
+                       rMin*denom,(rMinCnt > 9) ? '+' : '0'+rMinCnt,
+                       rMax*denom,(rMaxCnt > 9) ? '+' : '0'+rMaxCnt,
+                       hwMin,hwMax,
+                       (totSum*denom)/(bitsPerBlock*bitsPerBlock),sqrt(var),(uint)d,iMin,jMin);
+                if (t.tstFlags & TST_FLG_SHOW_HIST)
+                    { /* very wide histogram display */
+                    for (i=0;i<=HIST_BINS;i++)
+                        if (hist[i])
+                            printf(" %7.5f",hist[i]/(double)(bitsPerBlock*bitsPerBlock));
+                        else
+                            printf("  _     ");
+                    }
+                if (t.tstFlags & TST_FLG_DO_RAND)
+                    printf(" [RANDOM] ");
+                printf("\n");
+                fflush(stdout);
+                }
+            if (t.tstFlags & TST_FLG_DO_RAND)
+                break;        /* no need to do more than one random setting per rotation set */
+            }   /* for (d=1;d<dMax;d+=2) */
+        if (t.tstFlags & TST_FLG_DO_RAND)
+            break;        /* no need to do more than one random setting per rotation set */
+        }
+    return r->rWorst;
+    }
+
+/* twiddle a bit with an entry, but keep maxSatRounds satisfied */
+void Twiddle(rSearchRec *r,testParms t)
+    {
+    enum { MAX_TWIDDLE_CNT = 100, MAX_ROT_CNT = 6 };
+    uint i,j,k,n,v[MAX_ROT_CNT];
+    u08b old[MAX_ROT_CNT];
+    u64b usedBitmap;
+    u08b goodRots[BITS_PER_WORD];
+    uint goodRotCnt;
+
+    assert(rotsPerCycle <= sizeof(usedBitmap)*8);
+    r->ID += (1 << TWIDDLE_CNT_BIT0);           /* bump count of number of times twiddled */
+    r->ID &= ~ID_RECALC_BIT;                    /* show this one hasn't been had recalc yet */
+    r->parentCRC = r->CRC;                      /* track genealogy */
+
+    for (i=goodRotCnt=0;i<BITS_PER_WORD;i++)
+        if (!RotCnt_Bad(i))
+            {
+            goodRots[goodRotCnt++] = (u08b) i;
+            }
+
+    n = 1 + (Rand08() % MAX_ROT_CNT);
+    for (i=0;i<4;i++)
+        {
+        usedBitmap = 0; 
+        for (j=0;j<n;j++)
+            {               /* pick which set of n rotation constants to change */
+            do  {
+                v[j] = Rand08() % rotsPerCycle; /* rotation index */
+                }
+            while ((usedBitmap >> v[j]) & 1);   /* make sure all v[j] values are unique */
+            usedBitmap |= (((u64b) 1) << v[j]);
+            old[j] = r->rotList[v[j]];          /* save current value */
+            }
+        for (k=0;k<MAX_TWIDDLE_CNT/4;k++)
+            {  /* here with n rotation indices (v[0..n-1]) to be changed */
+            for (j=0;j<n;j++)
+                {
+                do  {
+                    r->rotList[v[j]] = goodRots[Rand32() % goodRotCnt];
+                    }   /* make sure new rotation value changes */
+                while (r->rotList[v[j]] == old[j]);
+                }
+            if (Cycle_Min_HW(t.maxSatRnds,r->rotList,0,t.rotVerMask) == BITS_PER_WORD)
+                {
+                if (i >= 2 || !(t.tstFlags & TST_FLG_KEEP_MIN_HW) ||
+                    Cycle_Min_HW(t.rounds,r->rotList,t.minHW_or,t.rotVerMask) >= (int) t.minHW_or)
+                    {
+                    Set_Min_hw_OR(r,t.rotVerMask,t.rounds);
+                    return;
+                    }
+                }
+            for (j=0;j<n;j++)   /* didn't work: go back to the old values */
+                r->rotList[v[j]] = old[j];
+            }
+        }
+    /* twiddling failed to produce a valid set (very rare). Select a brand new one */
+    get_rotation(r,t);
+    }
+
+/* run a full search */
+void RunSearch(testParms t)
+    {
+    enum        { KEEP_DIV = 16, KEEP_REP = 10, SHOW_CNT = 8 };
+    rSearchRec  popList[MAX_POP_CNT+2];
+    uint        i,j,k,n,repCnt,genCnt,keepCnt,prevBest[SHOW_CNT],showMask;
+    const       char *timeStr;
+    time_t      t0,t1;
+
+    Rand_Init(t.seed0 + (((u64b) bitsPerBlock) << 32));
+    memset(prevBest,0,sizeof(prevBest));
+
+    /* now set up the globals according to selected Skein blocksize */
+    switch (bitsPerBlock)
+        {
+        case  256:
+            t.genCntMax      = (t.genCntMax) ? t.genCntMax    : DEFAULT_GEN_CNT_4  ;
+            t.rounds         = (t.rounds)    ? t.rounds       : DEFAULT_ROUND_CNT_4;
+            t.minHW_or       = (t.minHW_or)  ? t.minHW_or     :         MIN_HW_OR_4;
+            t.maxSatRnds     = (t.maxSatRnds)? t.maxSatRnds   :    MAX_SAT_ROUNDS_4;
+            fwd_cycle_or_rN  = (t.rounds!=8) ? fwd_cycle_4_or :  fwd_cycle_4_or_r8 ;
+            rev_cycle_or_rN  = (t.rounds!=8) ? rev_cycle_4_or :  rev_cycle_4_or_r8 ;
+            fwd_cycle_or     = fwd_cycle_4_or;
+            rev_cycle_or     = fwd_cycle_4_or;
+            fwd_cycle        = fwd_cycle_4;
+            rev_cycle        = rev_cycle_4;
+            showMask         = 7;
+            break;
+        case  512:
+            t.genCntMax      = (t.genCntMax) ? t.genCntMax    : DEFAULT_GEN_CNT_8  ;
+            t.rounds         = (t.rounds)    ? t.rounds       : DEFAULT_ROUND_CNT_8;
+            t.minHW_or       = (t.minHW_or)  ? t.minHW_or     :         MIN_HW_OR_8;
+            t.maxSatRnds     = (t.maxSatRnds)? t.maxSatRnds   :    MAX_SAT_ROUNDS_8;
+            fwd_cycle_or_rN  = (t.rounds!=8) ? fwd_cycle_8_or :  fwd_cycle_8_or_r8 ;
+            rev_cycle_or_rN  = (t.rounds!=8) ? rev_cycle_8_or :  rev_cycle_8_or_r8 ;
+            fwd_cycle_or     = fwd_cycle_8_or;
+            rev_cycle_or     = rev_cycle_8_or;
+            fwd_cycle        = fwd_cycle_8;
+            rev_cycle        = rev_cycle_8;
+            showMask         = 3;
+            break;
+        case 1024:
+            t.genCntMax      = (t.genCntMax) ? t.genCntMax    : DEFAULT_GEN_CNT_16  ;
+            t.rounds         = (t.rounds)    ? t.rounds       : DEFAULT_ROUND_CNT_16;
+            t.minHW_or       = (t.minHW_or)  ? t.minHW_or     :         MIN_HW_OR_16;
+            t.maxSatRnds     = (t.maxSatRnds)? t.maxSatRnds   :    MAX_SAT_ROUNDS_16;
+            fwd_cycle_or_rN  = (t.rounds!=9) ? fwd_cycle_16_or: fwd_cycle_16_or_r9  ;
+            rev_cycle_or_rN  = (t.rounds!=9) ? rev_cycle_16_or: rev_cycle_16_or_r9  ;
+            fwd_cycle_or     = fwd_cycle_16_or;
+            rev_cycle_or     = rev_cycle_16_or;
+            fwd_cycle        = fwd_cycle_16;
+            rev_cycle        = rev_cycle_16;
+            showMask         = 1;
+            break;
+        default:
+            printf("Invalid block size!");
+            exit(2);
+        }
+    if (t.popCnt > MAX_POP_CNT)
+        t.popCnt = MAX_POP_CNT;
+    if (t.popCnt < MIN_POP_CNT)
+        t.popCnt = MIN_POP_CNT;
+    wordsPerBlock =   bitsPerBlock /      BITS_PER_WORD;
+    rotsPerCycle  = (wordsPerBlock / 2) * ROUNDS_PER_CYCLE;
+
+    keepCnt = t.popCnt/KEEP_DIV;
+    assert(keepCnt*(1+KEEP_REP) <= t.popCnt);
+    
+    printf("******************************************************************\n");
+    printf("Random seed = %u. BlockSize =%4d bits. sampleCnt =%6d. rounds = %2d. minHW_or=%d. CPU = %d-bit\n",
+                       t.seed0,bitsPerBlock,t.sampleCnt,t.rounds,t.minHW_or,(uint)sizeof(size_t)*8);
+    printf("Population  = %d. keepCnt = %d. repCnt = %d. rest = %d. keepMinHW = %d\n",
+            t.popCnt,keepCnt,KEEP_REP,t.popCnt-keepCnt*(1+KEEP_REP),(t.tstFlags & TST_FLG_KEEP_MIN_HW)?1:0); 
+    timeStr = ASCII_TimeDate();
+    if (t.tstFlags & TST_FLG_STDERR)
+        {
+        fprintf(stderr,"Start: %sBlock size = %d bits. popCnt = %d. sampleCnt = %d. keepMinHW = %d",
+                        timeStr,bitsPerBlock,t.popCnt,t.sampleCnt,(t.tstFlags & TST_FLG_KEEP_MIN_HW)?1:0);
+        if (t.runHours)
+            fprintf(stderr,". run time = %d hours",t.runHours);
+        fprintf(stderr,"\n");
+        }
+    else
+        showMask = 0;
+    printf("Start: %s  \n",timeStr);
+    time(&t0);
+    fflush(stdout);
+
+    for (n=0;n<t.popCnt;n++)
+        {   /* initialize the population with rotations that have "reasonable" hw_OR */
+        if (t.tstFlags & TST_FLG_STDERR)
+            fprintf(stderr,"\rGetRot: %04X    \r",t.popCnt-n);
+        if (get_rotation(&popList[n],t) == 0)
+            t.popCnt = n;               /* stop after end of file read in */
+        }
+    if (t.tstFlags & TST_FLG_STDERR)
+        fprintf(stderr,"\r%25s\r","");
+
+    for (genCnt=0;genCnt < t.genCntMax;genCnt++)
+        {   /* advance to the next generation */
+        for (i=0;i<t.popCnt;i++)
+            {   /* generate stats for all entries (this loop is where all the time is spent!) */
+            if ((i & showMask) == 1)
+                fprintf(stderr,"#%04X \r",t.popCnt-i);
+            if (genCnt == 0 || i >= keepCnt)
+                {
+                CheckDifferentials(&popList[i],t);
+                }
+            else if (i <= keepCnt/2 && (popList[i].ID & ID_RECALC_BIT) == 0)
+                {   /* recalc with bigger sampleCnt for better accuracy */
+                t.sampleCnt <<= 2;
+                CheckDifferentials(&popList[i],t);
+                t.sampleCnt >>= 2;
+                popList[i].rWorst = (popList[i].rWorst + 2) / 4;
+                popList[i].ID |= ID_RECALC_BIT;
+                }
+            }
+        qsort(popList,t.popCnt,sizeof(popList[0]),Compare_SearchRec_Descending);
+        if (t.genCntMax == 1)
+            { keepCnt = t.popCnt; break; }  /* allow quick processing from file */
+        /* now update the population for the next generation */
+        n = t.popCnt-1;                 /* start discarding at the end of the list */
+        for (i=0;i<keepCnt;i++)
+            {
+            if (t.tstFlags & TST_FLG_WEIGHT_REP)
+                repCnt = (i < keepCnt/2) ? KEEP_REP+2 : KEEP_REP-2 ;
+            else
+                repCnt = KEEP_REP;
+            for (j=0;j<repCnt;j++,n--)
+                {                       /* replicate the best ones, replacing the worst ones */
+                popList[n] = popList[i];
+                if (j == 0)
+                    {   /* splice two together, but only if they are from the same initial rotation set */
+                    k = Rand32() %  keepCnt;    
+                    if (((popList[n].ID ^ popList[k].ID) & ID_NUM_MASK) == 0)
+                        memcpy(popList[n].rotList,
+                               popList[k].rotList,
+                               rotsPerCycle*sizeof(popList[n].rotList[0])/2);
+                    }
+                Twiddle(&popList[n],t); /* tweak the replicate entry a bit */
+                assert(n >= keepCnt);   /* sanity check  */
+                }
+            }
+        for (;n>=keepCnt;n--)           /* just tweak the rest */
+            {
+            Twiddle(&popList[n],t);
+            }
+        time(&t1);
+        /* show current best */
+        if (t.tstFlags & TST_FLG_STDERR)
+            {   /* first to stderr (assuming redirected stdout */
+            fprintf(stderr,"\r%4d: ",genCnt+1);
+            for (i=j=0;i<SHOW_CNT;i++)
+                {
+                fprintf(stderr," %5.3f%c",popList[i].rWorst/(double)t.sampleCnt,(popList[i].ID & ID_RECALC_BIT)?'r':' ');
+                j |= (popList[i].rWorst ^ prevBest[i]); /* track changes */
+                prevBest[i] = popList[i].rWorst;
+                }
+            fprintf(stderr,"  {%6d sec%c}\n",(uint)(t1-t0),(j) ? '*':' ');
+            }
+        if (t.tstFlags & TST_FLG_VERBOSE)
+            {   /* then more details to stdout */
+            printf("::::: Gen =%5d. Best =%6.3f. PopCnt =%5d. SampleCnt =%5d. time=%6d.\n",
+                   genCnt+1,popList[0].rWorst/(double)t.sampleCnt,t.popCnt,t.sampleCnt,(uint)(t1-t0));
+            for (i=0;i<keepCnt;i++)
+                ShowSearchRec(stdout,&popList[i],t,SHOW_ROTS_PRELIM,(i)?' ':'-',i+1);
+            fflush(stdout);
+            }
+        if (t.runHours && t.runHours*3600 < (uint) (t1 - t0))
+            break;      /* timeout? */
+        }
+
+    /* re-grade the top entries using larger sampleCnt values */
+    printf("\n+++++++++++++ Preliminary results: sampleCnt = %5d, block = %4d bits\n",t.sampleCnt,bitsPerBlock);
+    qsort(popList,keepCnt,sizeof(popList[0]),Compare_SearchRec_Descending);
+    for (i=0;i<keepCnt;i++)
+        ShowSearchRec(stdout,&popList[i],t,SHOW_ROTS_PRELIM,' ',i+1);
+
+    /* re-run several times, since there will be statistical variations */
+    t.rotVerMask = MAX_ROT_VER_MASK;
+    t.diffBits   = (t.diffBits & 0x100) ? t.diffBits : 3;
+    t.sampleCnt *= 2;
+    t.tstFlags  |= TST_FLG_SHOW;
+    t.tstFlags  &= (TST_FLG_STDERR | TST_FLG_SHOW | TST_FLG_USE_ABS | TST_FLG_CHECK_ONE | TST_FLG_SHOW_HIST);
+
+    for (j=0;j < ((t.tstFlags & TST_FLG_CHECK_ONE) ? 1u:2u) ;j++)
+        {   /* do it twice, once with and once without USE_ABS, unless TST_FLG_CHECK_ONE set */
+        if (!(t.tstFlags & TST_FLG_CHECK_ONE))
+            t.tstFlags  ^= TST_FLG_USE_ABS;
+        for (n=0;n<t.regradeCnt;n++)
+            {
+            t.sampleCnt *= 2;
+            printf("+++ Re-running differentials with sampleCnt = %d, blockSize = %4d bits.%s\n",
+                   t.sampleCnt,bitsPerBlock,(t.tstFlags & TST_FLG_USE_ABS)?" absDiff":"" );
+            for (i=0;i<keepCnt;i++)
+                {
+                if (t.tstFlags & TST_FLG_STDERR)
+                    fprintf(stderr,"       Re-run: samples=%d, blk=%4d. #%02d.%s    \r",
+                            t.sampleCnt,bitsPerBlock,keepCnt-i,(t.tstFlags & TST_FLG_USE_ABS)?" absDiff":"" );
+                CheckDifferentials(&popList[i],t);
+                fflush(stdout);
+                }
+            if (keepCnt == 1)
+                {   /* show random comparison for final values */
+                printf("        RANDOM OUTPUT: /* useful stats for comparison to 'ideal' */\n");
+                t.tstFlags |=  TST_FLG_DO_RAND;
+                for (i=0;i<2;i++)
+                    {
+                    popList[keepCnt] =  popList[keepCnt-1];
+                    CheckDifferentials(&popList[keepCnt],t);
+                    }
+                t.tstFlags &= ~TST_FLG_DO_RAND;
+                }
+            /* sort per new stats */
+            if (t.tstFlags & TST_FLG_STDERR)
+                fprintf(stderr,"\r%60s\r","");
+            printf("\n+++++++++++++ Final results: sampleCnt = %5d, blockSize = %4d bits.%s\n",
+                   t.sampleCnt,bitsPerBlock,(t.tstFlags & TST_FLG_USE_ABS)?" absDiff":"" );
+            qsort(popList,keepCnt,sizeof(popList[0]),Compare_SearchRec_Descending);
+            for (i=keepCnt;i;i--)
+                ShowSearchRec(stdout,&popList[i-1],t,SHOW_ROTS_FINAL,(i==1)?'-':' ',i);
+            fflush(stdout);
+            }
+        printf("\n+++++++++++++ Formatted results: sampleCnt = %5d, blockSize = %4d bits. %s\n",
+               t.sampleCnt,bitsPerBlock,(t.tstFlags & TST_FLG_USE_ABS)?" absDiff":"" );
+        for (i=keepCnt;i;i--)
+            {
+            ShowSearchRec(stdout,&popList[i-1],t,SHOW_ROTS_H,' ',i);
+            printf("\n");
+            Show_HW_rounds(popList[i-1].rotList);
+            printf("\n");
+            }
+        fflush(stdout);
+        t.sampleCnt >>= n;  /* revert to original sampleCnt */
+        }
+
+    time(&t1);
+    printf("End:   %s\n",ASCII_TimeDate());
+    printf("Elapsed time = %6.3f hours\n\n",(t1-t0)/(double)3600.0);
+    if (t.tstFlags & TST_FLG_STDERR)
+        fprintf(stderr,"\r%60s\n","");    /* clear the screen if needed */
+    fflush(stdout);
+    }
+
+void GiveHelp(void)
+    {
+    printf("Usage:   skein_rot_search [options/flags]\n"
+           "Options: -Bnn   = set Skein block size in bits (default=512)\n"
+           "         -Cnn   = set count of random differentials taken\n"
+           "         -Dnn   = set number bits of difference pattern tested (default=1)\n"
+           "         -Gnn   = set min invalid rotation value (default 0)\n"
+           "         -Inn   = set rotation version mask\n"
+           "         -Onn   = set Hamming weight offset\n"
+           "         -Pnn   = set population count\n"
+           "         -Rnn   = set round count\n"
+           "         -Snn   = set initial random seed (0 --> randomize)\n"
+           "         -Tnn   = set max time to run (in hours)\n"
+           "         -Wnn   = set minimum hamming weight\n"
+           "         -Xnn   = set max test rotation count\n"
+           "         -Znn   = set max rounds needed for saturation using OR\n"
+           "         @file  = read rotations from file\n"
+           "Flags:   -A     = use min, not absolute difference\n"
+           "         -E     = no stderr output\n"
+           "         -H     = show histogram (very wide)\n"
+           "         -K     = keep minHW_or during twiddling\n"
+           "         -Q     = disable quick exit in search\n"
+           "         -U     = weighted repeat count (repeat best more frequently)\n"
+           "         -V     = verbose mode\n"
+          );
+    exit(0);
+    }
+
+int main(int argc,char *argv[])
+    {
+    uint        i,bMin,bMax;
+    testParms   t;
+    uint chkInv =        1;   /* check inverse functions at startup (slow for debbuging) */
+    uint goodRot=        2;   /* first allowed rotation value (+/-) */
+    uint seed   =        1;   /* 0 = randomize based on time, else use specified seed */
+    uint do8    =        0;   /* optimize 8-bit CPU performance */
+
+    t.rounds    =        0;   /* number of Skein rounds to test */
+    t.minHW_or  =        0;   /* minHW (using OR) required */
+    t.minOffs   =        4;   /* heuristic used to speed up rotation search */
+    t.diffBits  =        1;   /* # consecutive bits of differential inputs tested */
+    t.sampleCnt =     1024;   /* number of differential pairs tested */
+    t.genCntMax =        0;   /* number of "generations" tested */
+    t.maxSatRnds=        0;   /* number of rounds to Hamming weight "saturation" */
+    t.rotVerMask=        3;   /* mask of which versions to run */
+    t.runHours  =        0;   /* stop searching after this many hours */
+    t.dupRotMask=        0;   /* default is to allow same rotation value in a round */
+    t.regradeCnt=        3;   /* how many scaled up counts to try */
+    t.popCnt    = DEFAULT_POP_CNT;                      /* size of population */
+    t.tstFlags  = TST_FLG_STDERR | TST_FLG_VERBOSE | TST_FLG_USE_ABS | TST_FLG_CHECK_ONE; /* default flags */
+
+    for (i=1;i<(uint)argc;i++)
+        {   /* parse command line args */
+        if (argv[i][0] == '?')
+            GiveHelp();
+        else if (argv[i][0] == '-' || argv[i][0] == '+')
+            {
+#define arg_toi(s) atoi(s + ((s[2] == '=') ? 3 : 2))
+            switch (toupper(argv[i][1]))
+                {
+                case '?': GiveHelp();                            break;
+                                                                 
+                case 'A': t.tstFlags   &= ~TST_FLG_USE_ABS;      break;
+                case 'E': t.tstFlags   &= ~TST_FLG_STDERR;       break;
+                case 'H': t.tstFlags   |=  TST_FLG_SHOW_HIST;    break;
+                case 'K': t.tstFlags   |=  TST_FLG_KEEP_MIN_HW;  break;
+                case 'Q': t.tstFlags   |=  TST_FLG_QUICK_EXIT;   break;
+                case 'U': t.tstFlags   |=  TST_FLG_WEIGHT_REP;   break;
+                case 'V': t.tstFlags   &= ~TST_FLG_VERBOSE;      break;
+                case '1': t.tstFlags   &= ~TST_FLG_CHECK_ONE;    break;
+
+                case 'B': bitsPerBlock  =  arg_toi(argv[i]);     break;
+                case 'C': t.sampleCnt   =  arg_toi(argv[i]);     break;
+                case 'D': t.diffBits    =  arg_toi(argv[i]);     break;
+                case 'G': goodRot       =  arg_toi(argv[i]);     break;
+                case 'I': t.rotVerMask  =  arg_toi(argv[i]);     break;
+                case 'J': t.regradeCnt  =  arg_toi(argv[i]);     break;
+                case 'O': t.minOffs     =  arg_toi(argv[i]);     break;
+                case 'P': t.popCnt      =  arg_toi(argv[i]);     break;
+                case 'R': t.rounds      =  arg_toi(argv[i]);     break;
+                case 'S': seed          =  arg_toi(argv[i]);     break;
+                case 'T': t.runHours    =  arg_toi(argv[i]);     break;
+                case 'W': t.minHW_or    =  arg_toi(argv[i]);     break;
+                case 'X': t.genCntMax   =  arg_toi(argv[i]);     break;
+                case 'Z': t.maxSatRnds  =  arg_toi(argv[i]);     break;
+                case '2': t.dupRotMask  = ~0u;                   break;
+                case '0': chkInv        =  0;                    break;
+                case '8': do8           =  1;                    break;
+
+                default : printf("Unknown option: %s\n",argv[i]); GiveHelp();     break;
+                }
+            }
+        else if (argv[i][0] == '@')
+            {
+            rotFileName = argv[i]+1;
+            t.genCntMax = 1;            /* stop after one generation */
+            }
+        }
+
+    if (chkInv)
+        InverseChecks();    /* check fwd vs. rev transforms (slow in debugger) */
+
+    t.goodRotCntMask = 0;
+    for (i=goodRot; i <= BITS_PER_WORD - goodRot ;i++)
+        t.goodRotCntMask |= (((u64b) 1) << i);
+    if (do8) 
+        t.goodRotCntMask = (((u64b) 0x03838383) << 32) | 0x83838380;
+
+    if (bitsPerBlock == 0)
+        {
+        printf("Running search for all Skein block sizes (256, 512, and 1024)\n");
+        t.rounds   = 0;   /* use defaults, since otherwise it makes little sense */
+        t.minHW_or = 0;
+        }
+
+    bMin = (bitsPerBlock) ? bitsPerBlock :  256;
+    bMax = (bitsPerBlock) ? bitsPerBlock : 1024;
+
+    for (bitsPerBlock=bMin;bitsPerBlock<=bMax;bitsPerBlock*=2)
+        {
+        t.seed0 = (seed) ? seed : (uint) time(NULL);   /* randomize based on time if -s0 is given */
+        RunSearch(t);
+        }
+    
+    return 0;
+    }
diff --git a/Additional_Implementations/skein_test.c b/Additional_Implementations/skein_test.c
new file mode 100644
index 000000000000..9d999e0d49c0
--- /dev/null
+++ b/Additional_Implementations/skein_test.c
@@ -0,0 +1,1380 @@
+/***********************************************************************
+**
+** Test/verification code for the Skein block functions.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+** Testing:
+**   - buffering of incremental calls (random cnt steps)
+**   - partial input byte handling
+**   - output sample hash results (for comparison of ref vs. optimized)
+**   - performance
+**
+***********************************************************************/
+
+#include <ctype.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdarg.h>
+#include <time.h>
+#include <assert.h>
+
+#include "skein.h"
+#include "SHA3api_ref.h"
+
+static const uint_t HASH_BITS[] =    /* list of hash hash lengths to test */
+        { 160,224,256,384,512,1024, 256+8,512+8,1024+8,2048+8 };
+
+#define HASH_BITS_CNT   (sizeof(HASH_BITS)/sizeof(HASH_BITS[0]))
+
+/* bits of the verbose flag word */
+#define V_KAT_LONG      (1u << 0)
+#define V_KAT_SHORT     (1u << 1)
+#define V_KAT_NO_TREE   (1u << 2)
+#define V_KAT_NO_SEQ    (1u << 3)
+#define V_KAT_NO_3FISH  (1u << 4)
+#define V_KAT_DO_3FISH  (1u << 5)
+
+/* automatic compiler version number detection */
+#if !defined(CompilerVersion)
+
+#if   defined(_MSC_VER) && (_MSC_VER >= 1400)
+#define CompilerVersion (900)
+#elif defined(_MSC_VER) && (_MSC_VER >= 1200)
+#define CompilerVersion (600)
+#elif defined(_MSC_VER) && (_MSC_VER >= 1000)
+#define CompilerVersion (420)
+#elif defined(__GNUC__) && defined(__GNUC_MINOR__) && defined(__GNUC_PATCHLEVEL__)
+#define CompilerVersion (100*__GNUC__ + 10*__GNUC_MINOR__ + __GNUC_PATCHLEVEL__)
+#elif defined(__BORLANDC__) /* this is in hex */
+#define CompilerVersion (100*(__BORLANDC__ >> 8) + 10*((__BORLANDC__ >> 4) & 0xF) + (__BORLANDC__ & 0xF))
+#endif
+
+#endif
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF) 
+/* external functions to determine code size (in bytes) */
+size_t  Skein_256_Process_Block_CodeSize(void);
+size_t  Skein_512_Process_Block_CodeSize(void);
+size_t  Skein1024_Process_Block_CodeSize(void);
+size_t  Skein_256_API_CodeSize(void);
+size_t  Skein_512_API_CodeSize(void);
+size_t  Skein1024_API_CodeSize(void);
+uint_t  Skein_256_Unroll_Cnt(void);
+uint_t  Skein_512_Unroll_Cnt(void);
+uint_t  Skein1024_Unroll_Cnt(void);
+#elif defined(SKEIN_LOOP)
+uint_t  Skein_256_Unroll_Cnt(void) { return (SKEIN_LOOP / 100) % 10; }
+uint_t  Skein_512_Unroll_Cnt(void) { return (SKEIN_LOOP /  10) % 10; }
+uint_t  Skein1024_Unroll_Cnt(void) { return (SKEIN_LOOP      ) % 10; }
+#else
+uint_t  Skein_256_Unroll_Cnt(void) { return 0; }
+uint_t  Skein_512_Unroll_Cnt(void) { return 0; }
+uint_t  Skein1024_Unroll_Cnt(void) { return 0; }
+#endif
+
+/* External function to process blkCnt (nonzero) full block(s) of data. */
+void    Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void    Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void    Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+
+/********************** debug i/o helper routines **********************/
+void FatalError(const char *s,...)
+    { /* print out a msg and exit with an error code */
+    va_list ap;
+    va_start(ap,s);
+    vprintf(s,ap);
+    va_end(ap);
+    printf("\n");
+    exit(2);
+    }
+
+static uint_t _quiet_   =   0;  /* quiet processing? */
+static uint_t verbose   =   0;  /* verbose flag bits */
+static uint_t katHash   = ~0u;  /* use as a quick check on KAT results */
+
+void ShowBytes(uint_t cnt,const u08b_t *b)
+    { /* formatted output of byte array */
+    uint_t i;
+
+    for (i=0;i < cnt;i++)
+        {
+        if (i %16 ==  0) printf("    ");
+        else if (i % 4 == 0) printf(" ");
+        printf(" %02X",b[i]);
+        katHash = (katHash ^ b[i]) * 0xDEADBEEF;
+        katHash = (katHash ^ (katHash >> 23) ^ (katHash >> 17) ^ (katHash >> 9)) * 0xCAFEF00D;
+        if (i %16 == 15 || i==cnt-1) printf("\n");
+        }
+    }
+
+#ifndef SKEIN_DEBUG
+uint_t skein_DebugFlag     =   0;     /* dummy flags (if not defined elsewhere) */
+#endif
+
+#define SKEIN_DEBUG_SHORT   (SKEIN_DEBUG_HDR | SKEIN_DEBUG_STATE | SKEIN_DEBUG_TWEAK | SKEIN_DEBUG_KEY | SKEIN_DEBUG_INPUT_08 | SKEIN_DEBUG_FINAL)
+#define SKEIN_DEBUG_DEFAULT (SKEIN_DEBUG_SHORT)
+
+void Show_Debug(const char *s,...)
+    {
+    if (skein_DebugFlag)              /* are we showing debug info? */
+        {
+        va_list ap;
+        va_start(ap,s);
+        vprintf(s,ap);
+        va_end(ap);
+        }
+    }
+
+/************** Timing routine (for performance measurements) ***********/
+/* unfortunately, this is generally assembly code and not very portable */
+
+#if defined(_M_IX86) || defined(__i386) || defined(_i386) || defined(__i386__) || defined(i386) || \
+    defined(_X86_)   || defined(__x86_64__) || defined(_M_X64) || defined(__x86_64)
+#define _Is_X86_    1
+#endif
+
+#if  defined(_Is_X86_) && (!defined(__STRICT_ANSI__)) && (defined(__GNUC__) || !defined(__STDC__)) && \
+    (defined(__BORLANDC__) || defined(_MSC_VER) || defined(__MINGW_H) || defined(__GNUC__))
+#define HI_RES_CLK_OK         1          /* it's ok to use RDTSC opcode */
+
+#if defined(_MSC_VER) && defined(_M_X64)
+#include <intrin.h>
+#pragma intrinsic(__rdtsc)
+#endif
+
+#endif
+
+uint_32t HiResTime(void)
+    {
+#if defined(HI_RES_CLK_OK)
+    uint_32t x[2];
+#if   defined(__BORLANDC__)
+#define COMPILER_ID "BCC"
+    _asm { push edx };
+    __emit__(0x0F,0x31);            /* RDTSC instruction */
+    _asm { pop  edx };
+    _asm { mov x[0],eax };
+#elif defined(_MSC_VER)
+#define COMPILER_ID "MSC"
+#if defined(_MSC_VER) && defined(_M_X64)
+    x[0] = (uint_32t) __rdtsc();
+#else
+    _asm { push  edx };
+    _asm { _emit 0fh }; _asm { _emit 031h };
+    _asm { pop   edx };
+    _asm { mov x[0],eax };
+#endif
+#elif defined(__MINGW_H) || defined(__GNUC__)
+#define COMPILER_ID "GCC"
+    asm volatile("rdtsc" : "=a"(x[0]), "=d"(x[1]));
+#else
+#error  "HI_RES_CLK_OK -- but no assembler code for this platform (?)"
+#endif
+    return x[0];
+#else
+    /* avoid annoying MSVC 9.0 compiler warning #4720 in ANSI mode! */
+#if (!defined(_MSC_VER)) || (!defined(__STDC__)) || (_MSC_VER < 1300)
+    FatalError("No support for RDTSC on this CPU platform\n");
+#endif
+    return 0;
+#endif /* defined(HI_RES_CLK_OK) */
+    }
+
+/******** OS-specific calls for setting priorities and sleeping ******/
+#if (defined(_MSC_VER) && (_MSC_VER >= 1300) && !defined(__STRICT_ANSI__) && !defined(__STDC__)) \
+    && defined(_M_X64)
+#include <Windows.h>
+#include <WinBase.h>
+
+#ifdef  SKEIN_FORCE_LOCK_CPU            /* NielsF says this is not a good way to do things */
+#define SKEIN_LOCK_CPU_OK (1)
+int Lock_CPU(void)
+    {   /* lock this process to this CPU for perf timing */
+        /*   -- thanks to Brian Gladman for this code    */
+    HANDLE ph;
+    DWORD_PTR afp;
+    DWORD_PTR afs;
+    ph = GetCurrentProcess();
+    if(GetProcessAffinityMask(ph, &afp, &afs))
+        {
+        afp &= (((size_t)1u) << GetCurrentProcessorNumber());
+        if(!SetProcessAffinityMask(ph, afp))
+            return 1;
+        }
+    else
+        {
+        return 2;
+        }
+    return 0;   /* success */
+    }
+#endif
+
+#define _GOT_OS_SLEEP        (1)
+void OS_Sleep(uint_t msec)
+    {
+    Sleep(msec);
+    }
+
+#define _GOT_OS_SET_PRIORITY (1)
+int OS_Set_High_Priority(void)
+    {
+    if(!SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_HIGHEST))
+        return 1;
+#ifdef SKEIN_LOCK_CPU_OK    
+    if (Lock_CPU())
+        return 2;
+#endif
+    return 0;
+    }
+
+int OS_Set_Normal_Priority(void)
+    {
+    if(!SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_NORMAL))
+        return 1;
+    return 0;
+    }
+#endif
+
+#if defined(__linux) || defined(__linux__) || defined(linux) || defined(__gnu_linux__)
+#include <unistd.h>
+#define _GOT_OS_SLEEP        (1)
+void OS_Sleep(uint_t mSec)
+    {
+    usleep(mSec*1000);
+    }
+#endif
+
+#ifndef _GOT_OS_SET_PRIORITY
+/* dummy routines if nothing is available */
+int OS_Set_High_Priority(void)
+    {
+    return 0;
+    }
+int OS_Set_Normal_Priority(void)
+    {
+    return 0;
+    }
+#endif
+
+#ifndef _GOT_OS_SLEEP
+uint_32t OS_Sleep(uint_32t mSec)
+    {
+    return mSec;    /* avoid compiler warnings */
+    }
+#endif
+   
+#ifndef COMPILER_ID
+#define COMPILER_ID "(unknown)"
+#endif
+/********************** use RC4 to generate test data ******************/
+/* Note: this works identically on all platforms (big/little-endian)   */
+static struct
+    {
+    uint_t I,J;                         /* RC4 vars */
+    u08b_t state[256];
+    } prng;
+
+void RandBytes(void *dst,uint_t byteCnt)
+    {
+    u08b_t a,b;
+    u08b_t *d = (u08b_t *) dst;
+
+    for (;byteCnt;byteCnt--,d++)        /* run RC4  */
+        {
+        prng.I  = (prng.I+1) & 0xFF;
+        a       =  prng.state[prng.I];
+        prng.J  = (prng.J+a) & 0xFF;
+        b       =  prng.state[prng.J];
+        prng.state[prng.I] = b;
+        prng.state[prng.J] = a;
+        *d      =  prng.state[(a+b) & 0xFF];
+        }
+    }
+
+/* get a pseudo-random 32-bit integer in a portable way */
+uint_t Rand32(void)
+    {
+    uint_t i,n;
+    u08b_t tmp[4];
+
+    RandBytes(tmp,sizeof(tmp));
+
+    for (i=n=0;i<sizeof(tmp);i++)
+        n = n*256 + tmp[i];
+    
+    return n;
+    }
+
+/* init the (RC4-based) prng */
+void Rand_Init(u64b_t seed)
+    {
+    uint_t i,j;
+    u08b_t tmp[512];
+
+    /* init the "key" in an endian-independent fashion */
+    for (i=0;i<8;i++)
+        tmp[i] = (u08b_t) (seed >> (8*i));
+
+    /* initialize the permutation */
+    for (i=0;i<256;i++)
+        prng.state[i]=(u08b_t) i;
+
+    /* now run the RC4 key schedule */
+    for (i=j=0;i<256;i++)
+        {                   
+        j = (j + prng.state[i] + tmp[i%8]) & 0xFF;
+        tmp[256]      = prng.state[i];
+        prng.state[i] = prng.state[j];
+        prng.state[j] = tmp[256];
+        }
+    prng.I = prng.J = 0;  /* init I,J variables for RC4 */
+    
+    /* discard initial keystream before returning */
+    RandBytes(tmp,sizeof(tmp));
+    }
+    
+/***********************************************************************/
+/* An AHS-like API that allows explicit setting of block size          */
+/*    [i.e., the AHS API selects a block size based solely on the ]    */
+/*    [hash result length, while Skein allows independent hash    ]    */
+/*    [result size and block size                                 ]    */
+/***********************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* select the context size and init the context */
+int Skein_Init(int blkSize,hashState *state, int hashbitlen)
+    {
+    switch (blkSize)
+        {
+        case  256:
+            state->statebits = 64*SKEIN_256_STATE_WORDS;
+            return Skein_256_Init(&state->u.ctx_256,(size_t) hashbitlen);
+        case  512:
+            state->statebits = 64*SKEIN_512_STATE_WORDS;
+            return Skein_512_Init(&state->u.ctx_512,(size_t) hashbitlen);
+        case 1024:
+            state->statebits = 64*SKEIN1024_STATE_WORDS;
+            return Skein1024_Init(&state->u.ctx1024,(size_t) hashbitlen);
+        default:
+            return SKEIN_FAIL;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* select the context size and init (extended) the context */
+int Skein_InitExt(int blkSize,hashState *state, int hashbitlen,u64b_t treeInfo,const u08b_t *key,size_t keyBytes)
+    {
+    switch (blkSize)
+        {
+        case  256:
+            state->statebits = 64*SKEIN_256_STATE_WORDS;
+            return Skein_256_InitExt(&state->u.ctx_256,(size_t) hashbitlen,treeInfo,key,keyBytes);
+        case  512:
+            state->statebits = 64*SKEIN_512_STATE_WORDS;
+            return Skein_512_InitExt(&state->u.ctx_512,(size_t) hashbitlen,treeInfo,key,keyBytes);
+        case 1024:
+            state->statebits = 64*SKEIN1024_STATE_WORDS;
+            return Skein1024_InitExt(&state->u.ctx1024,(size_t) hashbitlen,treeInfo,key,keyBytes);
+        default:
+            return SKEIN_FAIL;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process data to be hashed */
+int Skein_Update(hashState *state, const BitSequence *data, DataLength databitlen)
+    {
+    /* only the final Update() call is allowed do partial bytes, else assert an error */
+    Skein_Assert((state->u.h.T[1] & SKEIN_T1_FLAG_BIT_PAD) == 0 || databitlen == 0, FAIL);
+
+    if ((databitlen & 7) == 0)
+        {
+        switch (state->statebits)
+            {
+            case  512:  return Skein_512_Update(&state->u.ctx_512,data,databitlen >> 3);
+            case  256:  return Skein_256_Update(&state->u.ctx_256,data,databitlen >> 3);
+            case 1024:  return Skein1024_Update(&state->u.ctx1024,data,databitlen >> 3);
+            default: return SKEIN_FAIL;
+            }
+        }
+    else
+        {
+        size_t bCnt = (databitlen >> 3) + 1;                  /* number of bytes to handle */
+        u08b_t mask,*p;
+
+#if (!defined(_MSC_VER)) || (MSC_VER >= 1200)                 /* MSC v4.2 gives (invalid) warning here!!  */
+        Skein_assert(&state->u.h == &state->u.ctx_256.h);     /* sanity checks: allow u.h --> all contexts */
+        Skein_assert(&state->u.h == &state->u.ctx_512.h);
+        Skein_assert(&state->u.h == &state->u.ctx1024.h);
+#endif
+        switch (state->statebits)
+            {
+            case  512: Skein_512_Update(&state->u.ctx_512,data,bCnt);
+                       p    = state->u.ctx_512.b;
+                       break;
+            case  256: Skein_256_Update(&state->u.ctx_256,data,bCnt);
+                       p    = state->u.ctx_256.b;
+                       break;
+            case 1024: Skein1024_Update(&state->u.ctx1024,data,bCnt);
+                       p    = state->u.ctx1024.b;
+                       break;
+            default:   return FAIL;
+            }
+        Skein_Set_Bit_Pad_Flag(state->u.h);                     /* set tweak flag for the final call */
+        /* now "pad" the final partial byte the way NIST likes */
+        bCnt = state->u.h.bCnt;         /* get the bCnt value (same location for all block sizes) */
+        Skein_assert(bCnt != 0);        /* internal sanity check: there IS a partial byte in the buffer! */
+        mask = (u08b_t) (1u << (7 - (databitlen & 7)));         /* partial byte bit mask */
+        p[bCnt-1]  = (u08b_t)((p[bCnt-1] & (0-mask)) | mask);   /* apply bit padding on final byte (in the buffer) */
+        
+        return SUCCESS;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize hash computation and output the result (hashbitlen bits) */
+int Skein_Final(hashState *state, BitSequence *hashval)
+    {
+    switch (state->statebits)
+        {
+        case  512:  return Skein_512_Final(&state->u.ctx_512,hashval);
+        case  256:  return Skein_256_Final(&state->u.ctx_256,hashval);
+        case 1024:  return Skein1024_Final(&state->u.ctx1024,hashval);
+        default:    return SKEIN_FAIL;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* all-in-one hash function */
+int Skein_Hash(int blkSize,int hashbitlen, const BitSequence *data, /* all-in-one call */
+                DataLength databitlen,BitSequence *hashval)
+    {
+    hashState  state;
+    int r = Skein_Init(blkSize,&state,hashbitlen);
+    if (r == SKEIN_SUCCESS)
+        { /* these calls do not fail when called properly */
+        r = Skein_Update(&state,data,databitlen);
+        Skein_Final(&state,hashval);
+        }
+    return r;
+    }
+
+/***********************************************************************/
+/* various self-consistency checks */
+uint_t Skein_Test(uint_t blkSize,uint_t maxLen,uint_t hashLen,uint_t nStep,uint_t oneBlk)
+    {
+    enum        { MAX_BUF=1024 };
+    u08b_t      b[MAX_BUF+4],hashVal[2][MAX_BUF+4];
+    uint_t      i,j,k,n,bCnt,useAHS,step,bitLen,testCnt=0;
+    hashState   s[2];
+                
+    assert(blkSize > 0 && blkSize <= 1024 && (blkSize % 256) == 0);
+    assert((hashLen % 8) == 0);
+
+    if (maxLen  > MAX_BUF*8)     /* keep things reasonably small */
+        maxLen  = MAX_BUF*8;
+    if (hashLen > MAX_BUF*8)
+        hashLen = MAX_BUF*8;
+    if (maxLen  == 0)            /* default sizes */
+        maxLen  = blkSize*2;
+    if (hashLen == 0)
+        hashLen = blkSize;
+
+    if (oneBlk)
+        {
+        if (oneBlk > MAX_BUF*8)
+            oneBlk = MAX_BUF*8;
+        for (i=0;i<oneBlk/8;i++)
+            b[i] = (u08b_t) i;
+        if (Skein_Hash(blkSize,hashLen,b,oneBlk,hashVal[0]) != SKEIN_SUCCESS)
+            FatalError("Skein_Hash != SUCCESS");
+        return 1;
+        }
+
+    if (nStep == 0)
+        {
+        printf("Testing Skein: blkSize = %4d bits. hashLen=%4d bits. maxMsgLen = %4d bits.\n",
+               blkSize,hashLen,maxLen);
+        nStep = 1;
+        }
+
+    n = skein_DebugFlag;
+    skein_DebugFlag = 0;        /* turn of debug display for this "fake" AHS call */
+    if (Init(&s[0],hashLen) != SUCCESS) /* just see if AHS API supports this <blkSize,hashLen> pair */
+        FatalError("AHS_API Init() error!");
+    skein_DebugFlag = n;        /* restore debug display status */
+
+    useAHS = (s[0].statebits == blkSize);  /* does this <blkSize,hashLen> pair work via AHS_API? */
+    
+    bCnt = (maxLen + 7) / 8;    /* convert maxLen to bytes */
+    for (n=0;n < bCnt;n+=nStep) /* process all the data lengths (# bytes = n+1)*/
+        {
+        RandBytes(b,maxLen);    /* get something to hash */
+        for (j=8;j>0;j--)       /* j = # bits in final byte */
+            {
+            testCnt++;
+            memset(hashVal,0,sizeof(hashVal));
+            Show_Debug("\n*** Single Hash() call (%d bits)\n",8*n+j);
+            if (Skein_Hash(blkSize,hashLen,b,8*n+j,hashVal[0]) != SKEIN_SUCCESS)
+                FatalError("Skein_Hash != SUCCESS");
+            for (k=hashLen/8;k<=MAX_BUF;k++)
+                if (hashVal[0][k] != 0)
+                    FatalError("Skein hash output overrun!: hashLen = %d bits",hashLen);
+            if (useAHS)         /* compare using AHS API, if supported */
+                {      
+                Show_Debug("\n*** Single AHS API Hash() call\n");
+                if (Hash(hashLen,b,8*n+j,hashVal[1]) != SUCCESS)
+                    FatalError("Skein_Hash != SUCCESS");
+                for (k=hashLen/8;k<=MAX_BUF;k++)
+                    if (hashVal[1][k] != 0)
+                        FatalError("Skein AHS_API hash output overrun!: hashLen = %d bits",hashLen);
+                if (memcmp(hashVal[1],hashVal[0],hashLen/8))
+                    FatalError("Skein vs. AHS API miscompare");
+                }
+            /* now try (randomized) steps thru entire input block */
+            for (i=0;i<4;i++) 
+                {  
+                Show_Debug("\n*** Multiple Update() calls [%s]",(i)?"random steps":"step==1");
+                if (i >= 2)
+                    {
+                    Show_Debug("  [re-use precomputed state]");
+                    s[0] = s[1]; 
+                    }
+                else
+                    {
+                    k = (i) ? Skein_Init   (blkSize,&s[0],hashLen) :
+                              Skein_InitExt(blkSize,&s[0],hashLen,SKEIN_CFG_TREE_INFO_SEQUENTIAL,NULL,0);
+                    if (k != SKEIN_SUCCESS)
+                        FatalError("Skein_Init != SUCCESS");
+                    s[1] = s[0];            /* make a copy for next time */
+                    }
+                Show_Debug("\n");
+                for (k=0;k<n+1;k+=step)     /* step thru with variable sized steps */
+                    {/* for i == 0, step one byte at a time. for i>0, randomly */
+                    step = (i == 0) ? 1 : 1 + (Rand32() % (n+1-k));     /* # bytes to process */
+                    bitLen = (k+step >= n+1) ? 8*(step-1) + j: 8*step;  /* partial final byte handling */
+                    if (Skein_Update(&s[0],&b[k],bitLen) != SKEIN_SUCCESS)
+                        FatalError("Skein_Update != SUCCESS");
+                    }
+                if (Skein_Final(&s[0],hashVal[1]) != SKEIN_SUCCESS)
+                    FatalError("Skein_Final != SUCCESS");
+                for (k=hashLen/8;k<=MAX_BUF;k++)
+                    if (hashVal[0][k] != 0)
+                        FatalError("Skein hash output overrun!: hashLen = %d bits",hashLen);
+                if (memcmp(hashVal[1],hashVal[0],hashLen/8))
+                    FatalError("Skein Hash() vs. Update() miscompare!");
+                }
+            }
+        }
+    return testCnt;
+    }
+
+/* filter out <blkSize,hashBits> pairs in short KAT mode */
+uint_t Short_KAT_OK(uint_t blkSize,uint_t hashBits)
+    {
+    switch (blkSize)
+        {
+        case  256:
+            if (hashBits != 256 && hashBits != 224)
+                return 0;
+            break;
+        case  512:
+            if (hashBits != 256 && hashBits != 384 && hashBits != 512)
+                return 0;
+            break;
+        case 1024:
+            if (hashBits != 384 && hashBits != 512 && hashBits != 1024)
+                return 0;
+            break;
+        default:
+            return 0;
+        }
+    return 1;
+    }
+
+#if SKEIN_TREE_HASH
+#define MAX_TREE_MSG_LEN  (1 << 12)
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* pad final block, no OUTPUT stage */
+int Skein_Final_Pad(hashState *state, BitSequence *hashval)
+    {
+    switch (state->statebits)
+        {
+        case  512:  return Skein_512_Final_Pad(&state->u.ctx_512,hashval);
+        case  256:  return Skein_256_Final_Pad(&state->u.ctx_256,hashval);
+        case 1024:  return Skein1024_Final_Pad(&state->u.ctx1024,hashval);
+        default:    return SKEIN_FAIL;
+        }
+    }
+/* just the OUTPUT stage */
+int Skein_Output(hashState *state, BitSequence *hashval)
+    {
+    switch (state->statebits)
+        {
+        case  512:  return Skein_512_Output(&state->u.ctx_512,hashval);
+        case  256:  return Skein_256_Output(&state->u.ctx_256,hashval);
+        case 1024:  return Skein1024_Output(&state->u.ctx1024,hashval);
+        default:    return SKEIN_FAIL;
+        }
+    }
+
+/* generate a KAT test for the given data and tree parameters. */
+/* This is an "all-in-one" call. It is not intended to represent */
+/* how a real multi-processor version would be implemented, but  */
+/* the results will be the same */
+void Skein_TreeHash
+    (uint_t blkSize,uint_t hashBits,const u08b_t *msg,size_t msgBytes,
+     uint_t leaf   ,uint_t node    ,uint_t maxLevel  ,u08b_t *hashRes)
+    {
+    enum      { MAX_HEIGHT = 32 };          /* how deep we can go here */
+    uint_t    height;
+    uint_t    blkBytes  = blkSize/8;
+    uint_t    saveDebug = skein_DebugFlag;
+    size_t    n,nodeLen,srcOffs,dstOffs,bCnt;
+    u64b_t    treeInfo;
+    u08b_t    M[MAX_TREE_MSG_LEN+4];
+    hashState G,s;
+
+    assert(node < 256 && leaf < 256 && maxLevel < 256);
+    assert(node >  0  && leaf >  0  && maxLevel >  1 );
+    assert(blkSize == 256 || blkSize == 512 || blkSize == 1024);
+    assert(blkBytes <= sizeof(M));
+    assert(msgBytes <= sizeof(M));
+
+    /* precompute the config block result G for multiple uses below */
+#ifdef SKEIN_DEBUG
+    if (skein_DebugFlag)
+        skein_DebugFlag |= SKEIN_DEBUG_CONFIG;
+#endif
+    treeInfo = SKEIN_CFG_TREE_INFO(leaf,node,maxLevel);
+    if (Skein_InitExt(blkSize,&G,hashBits,treeInfo,NULL,0) != SKEIN_SUCCESS)
+        FatalError("Skein_InitExt() fails in tree");
+    skein_DebugFlag = saveDebug;
+
+    bCnt = msgBytes;
+    memcpy(M,msg,bCnt);
+    for (height=0;;height++)            /* walk up the tree */
+        {
+        if (height && (bCnt==blkBytes)) /* are we done (with only one block left)? */
+            break;
+        if (height+1 == maxLevel)       /* is this the final allowed level? */
+            {                           /* if so, do it as one big hash */
+            s = G;
+            Skein_Set_Tree_Level(s.u.h,height+1);
+            Skein_Update   (&s,M,bCnt*8);
+            Skein_Final_Pad(&s,M);
+            break;
+            }
+        nodeLen = blkBytes << ((height) ? node : leaf);
+        for (srcOffs=dstOffs=0;srcOffs <= bCnt;)
+            {
+            n = bCnt - srcOffs;         /* number of bytes left at this level */
+            if (n > nodeLen)            /* limit to node size */
+                n = nodeLen;
+            s = G;
+            s.u.h.T[0] = srcOffs;       /* nonzero initial offset in tweak! */
+            Skein_Set_Tree_Level(s.u.h,height+1);
+            Skein_Update   (&s,M+srcOffs,n*8);
+            Skein_Final_Pad(&s,M+dstOffs);  /* finish up this node, output intermediate result to M[]*/
+            dstOffs+=blkBytes;
+            srcOffs+=n;
+            if (srcOffs >= bCnt)        /* special logic to handle (msgBytes == 0) case */
+                break;
+            }
+        bCnt = dstOffs;
+        }
+
+    /* output the result */
+    Skein_Output(&s,hashRes);
+    }
+
+/*
+** Generate tree-mode hash KAT vectors.
+** Note:
+**    Tree vectors are different enough from non-tree vectors that it 
+**    makes sense to separate this out into a different function, rather 
+**    than shoehorn it into the same KAT logic as the other modes.
+**/
+void Skein_GenKAT_Tree(uint_t blkSize)
+    {
+    static const struct
+        {
+        uint_t leaf,node,maxLevel,levels;
+        }
+        TREE_PARMS[] = { {2,2,2,2}, {1,2,3,2}, {2,1,0xFF,3} };
+#define TREE_PARM_CNT (sizeof(TREE_PARMS)/sizeof(TREE_PARMS[0]))
+
+    u08b_t  msg[MAX_TREE_MSG_LEN+4],hashVal[MAX_TREE_MSG_LEN+4];
+    uint_t  i,j,k,n,p,q,hashBits,node,leaf,leafBytes,msgBytes,byteCnt,levels,maxLevel;
+
+    assert(blkSize == 256 || blkSize == 512 || blkSize == 1024);
+    for (i=0;i<MAX_TREE_MSG_LEN;i+=2)
+        {   /* generate "incrementing" tree hash input msg data */
+        msg[i  ] = (u08b_t) ((i ^ blkSize) ^ (i >> 16));
+        msg[i+1] = (u08b_t) ((i ^ blkSize) >> 8);
+        }
+    for (k=q=n=0;k < HASH_BITS_CNT;k++)
+        {
+        hashBits = HASH_BITS[k];
+        if (!Short_KAT_OK(blkSize,hashBits))
+            continue;
+        if ((verbose & V_KAT_SHORT) && (hashBits != blkSize))
+            continue;
+        for (p=0;p <TREE_PARM_CNT;p++)
+            {
+            if (p && (verbose & V_KAT_SHORT))
+                continue;           /* keep short KATs short */
+            if (p && hashBits != blkSize)
+                continue;           /* we only need one "non-full" size */
+
+            leaf      = TREE_PARMS[p].leaf;
+            node      = TREE_PARMS[p].node;
+            maxLevel  = TREE_PARMS[p].maxLevel;
+            levels    = TREE_PARMS[p].levels;
+            leafBytes = (blkSize/8) << leaf;    /* number of bytes in a "full" leaf */
+
+            for (j=0;j<4;j++)       /* different numbers of leaf results */
+                {
+                if ((verbose & V_KAT_SHORT) && (j != 3) && (j != 0))
+                    continue;
+                if (j && (hashBits != blkSize))
+                    break;
+                switch (j)
+                    {
+                    case 0: n = 1;                                break;
+                    case 1: n = 2;                                break;         
+                    case 2: n = (1 << (node * (levels-2)))*3/2;
+                            if (n <= 2) continue;                 break;
+                    case 3: n = (1 << (node * (levels-1)));       break;
+                    }
+                byteCnt = n*leafBytes;
+                assert(byteCnt > 0);
+                if (byteCnt > MAX_TREE_MSG_LEN)
+                    continue;
+                q = (q+1) % leafBytes;
+                msgBytes = byteCnt - q;
+                switch (blkSize)
+                    {
+                    case  256: printf("\n:Skein-256: "); break;
+                    case  512: printf("\n:Skein-512: "); break;
+                    case 1024: printf("\n:Skein-1024:"); break;
+                    }
+                printf(" %4d-bit hash, msgLen =%6d bits",hashBits,msgBytes*8);
+                printf(". Tree: leaf=%02X, node=%02X, maxLevels=%02X\n",leaf,node,maxLevel);
+                printf("\nMessage data:\n");
+                if (msgBytes == 0)
+                    printf("    (none)\n");
+                else
+                    ShowBytes(msgBytes,msg);
+                
+                Skein_TreeHash(blkSize,hashBits,msg,msgBytes,leaf,node,maxLevel,hashVal);
+                
+                printf("Result:\n");
+                ShowBytes((hashBits+7)/8,hashVal);
+                printf("--------------------------------\n");
+                }
+            }
+        }
+    }
+#endif
+
+/*
+** Output some KAT values. This output is generally re-directed to a file and
+** can be compared across platforms to help validate an implementation on a
+** new platform (or compare reference vs. optimized code, for example). The
+** file will be provided as part of the Skein submission package to NIST.
+**
+** When used in conjunction with the debug flag, this will output a VERY long
+** result. The verbose flag is used to output even more combinations of
+**      <blkSize,hashSize,msgLen>
+**
+** Note: this function does NOT output the NIST AHS KAT format.
+*/
+void Skein_ShowKAT(uint_t blkSizeMask)
+    {
+    enum
+        {
+        DATA_TYPE_ZERO  = 0,
+        DATA_TYPE_INC,
+        DATA_TYPE_RAND,
+        DATA_TYPE_MAC,
+        DATA_TYPE_TREE,
+        DATA_TYPE_CNT,
+
+        MAX_BYTES = 3*1024/8
+        };
+    static const char *TYPE_NAMES[] = { "zero","incrementing","random","random+MAC","tree",NULL };
+    static const uint_t  MSG_BITS[] =
+                { 0,1,2,3,4,5,6,7,8,9,10,32,64,128,192,
+                   256-1, 256, 256+1,  384,
+                   512-1, 512, 512+1,  768,
+                  1024-1,1024,1024+1,
+                  2048-1,2048,2048+1
+                };
+#define MSG_BITS_CNT (sizeof(MSG_BITS)/sizeof(MSG_BITS[0]))
+
+    uint_t      i,j,k,blkSize,dataType,hashBits,msgBits,keyBytes,blkBytes,keyType;
+    u08b_t      data[MAX_BYTES+4],key[MAX_BYTES+4],hashVal[MAX_BYTES+4];
+    const char *msgType;
+    hashState   s;
+
+    Rand_Init(SKEIN_MK_64(0xDEADBEEF,0)); /* init PRNG with repeatable value */
+    katHash = ~0u;
+    keyType =  0;
+
+#ifdef SKEIN_DEBUG
+    /* first, show some "raw" Threefish + feedforward block calls, with round-by-round debug info if enabled */
+    if (skein_DebugFlag && !(verbose & V_KAT_NO_3FISH))
+        {
+        k = skein_DebugFlag;                                        /* save debug flag value */
+        skein_DebugFlag  = THREEFISH_DEBUG_ALL & ~ SKEIN_DEBUG_HDR; /* turn on full debug detail, use Threefish name */
+        skein_DebugFlag |= (k & SKEIN_DEBUG_PERMUTE);
+#else
+    if (verbose & V_KAT_DO_3FISH)                                   /* allow non-SKEIN_DEBUG testing */
+        {
+#endif
+        for (blkSize = 256;blkSize <= 1024; blkSize*=2)
+            {
+            if (blkSizeMask && (blkSize & blkSizeMask) == 0)
+                continue;
+            for (dataType=DATA_TYPE_ZERO; dataType <= DATA_TYPE_INC; dataType++)
+                {
+                switch (dataType)
+                    {
+                    case DATA_TYPE_ZERO:
+                            memset(data,0,sizeof(data));
+                            memset(key ,0,sizeof(key));
+                            break;
+                    case DATA_TYPE_INC:
+                            for (i=0;i<MAX_BYTES;i++)
+                                {
+                                key [i] = (u08b_t)      i ;
+                                data[i] = (u08b_t) ~key[i];
+                                }
+                            break;
+                    default:
+                        continue;
+                    }
+#ifdef SKEIN_DEBUG
+                switch (blkSize)
+                    {
+                    case  256: printf("\n:Threefish-256: "); break;
+                    case  512: printf("\n:Threefish-512: "); break;
+                    case 1024: printf("\n:Threefish-1024:"); break;
+                    }
+                printf(" encryption + plaintext feedforward (round-by-round):\n");
+#endif
+                memset(&s,0,sizeof(s));
+                s.u.h.hashBitLen = blkSize;
+                Skein_Get64_LSB_First(s.u.h.T      ,key,2);               /* init T[] */
+                Skein_Get64_LSB_First(s.u.ctx1024.X,key+2*8,blkSize/64);  /* init X[] */
+                switch (blkSize)
+                    {
+                    case  256: Skein_256_Process_Block(&s.u.ctx_256,data,1,0); break;
+                    case  512: Skein_512_Process_Block(&s.u.ctx_512,data,1,0); break;
+                    case 1024: Skein1024_Process_Block(&s.u.ctx1024,data,1,0); break;
+                    }
+#ifdef SKEIN_DEBUG
+                printf("++++++++++++++++++++++++++++++++++++++\n");
+#endif
+                }
+            }
+#ifdef SKEIN_DEBUG
+        skein_DebugFlag = k;
+#endif
+        }
+
+    for (dataType=DATA_TYPE_ZERO; dataType < DATA_TYPE_CNT; dataType++)
+        {
+        msgType = TYPE_NAMES[dataType];
+        switch (dataType)
+            {
+            case DATA_TYPE_ZERO:
+                    memset(data,0,sizeof(data));
+                    memset(key ,0,sizeof(key));
+                    break;
+            case DATA_TYPE_INC:
+                    for (i=0;i<MAX_BYTES;i++)
+                        {
+                        key [i] = (u08b_t)      i ;
+                        data[i] = (u08b_t) ~key[i];
+                        }
+                    break;
+            case DATA_TYPE_MAC:
+                    RandBytes(key ,sizeof(key ));
+            case DATA_TYPE_RAND:
+                    RandBytes(data,sizeof(data));
+                    break;
+            case DATA_TYPE_TREE:
+                    if (verbose & V_KAT_NO_TREE)
+                        continue;
+                    break;
+            default:    /* should never get here */
+                    FatalError("Invalid data type: %d --> '%s'",dataType,msgType);
+                    break;
+            }
+        
+        for (blkSize = 256;blkSize <= 1024; blkSize*=2)
+            {
+            if (blkSizeMask && (blkSize & blkSizeMask) == 0)
+                continue;
+            if (dataType == DATA_TYPE_TREE)
+                {
+#if SKEIN_TREE_HASH
+                Skein_GenKAT_Tree(blkSize);
+#endif
+                continue;
+                }
+            if (verbose & V_KAT_NO_SEQ)
+                continue;
+            blkBytes = blkSize/8;
+            for (j=0;j <  MSG_BITS_CNT;j++)
+            for (k=0;k < HASH_BITS_CNT;k++)
+                {
+                msgBits  =  MSG_BITS[j];  /* message length   */
+                hashBits = HASH_BITS[k];  /* hash result size */
+                assert(MAX_BYTES*8 >= hashBits && MAX_BYTES*8 >= msgBits);
+                if (msgBits != 1024 && hashBits != blkSize && !(verbose & V_KAT_LONG))
+                    continue;   /* keep the output size reasonable, unless verbose */
+                if (verbose & V_KAT_SHORT)
+                    {           /* -v2 ==> generate "short" KAT set by filtering out most vectors */
+                    if (dataType != DATA_TYPE_INC)
+                        continue;
+                    if (msgBits != 8 && msgBits != blkSize && msgBits != 2*blkSize)
+                        continue;
+                    if (!Short_KAT_OK(blkSize,hashBits))
+                        continue;
+                    }
+                switch (blkSize)
+                    {
+                    case  256: printf("\n:Skein-256: "); break;
+                    case  512: printf("\n:Skein-512: "); break;
+                    case 1024: printf("\n:Skein-1024:"); break;
+                    }
+                printf(" %4d-bit hash, msgLen =%6d bits",hashBits,msgBits);
+                if (!(verbose & V_KAT_SHORT))
+                    printf(", data = '%s'",msgType);
+                printf("\n\nMessage data:\n");
+                if (msgBits == 0)
+                    printf("    (none)\n");
+                else
+                    ShowBytes((msgBits+7)/8,data);
+                switch (dataType)
+                    {
+                    default:                            /* straight hash value */
+                        if (Skein_Hash(blkSize,hashBits,data,msgBits,hashVal) != SKEIN_SUCCESS)
+                            FatalError("Skein_Hash() error!");
+                        break;
+                    case DATA_TYPE_MAC:                 /* include some MAC computations in KAT file */
+                        switch (keyType++)              /* sequence thru different MAC key lengths */
+                            {           
+                            case 0: keyBytes = blkBytes/2;   break;
+                            case 1: keyBytes = blkBytes;     break;
+                            case 2: keyBytes = blkBytes  +1; break;
+                            case 3: keyBytes = blkBytes*2+1; break;
+                            default:keyBytes = 0;       /* not actually a MAC this time, but use InitExt() */
+                                    keyType  = 0;       /* start the cycle again next time */
+                            }
+                        printf("MAC key = %4d bytes:\n",keyBytes);
+                        if (keyBytes)                   /* show MAC key, if any */
+                            ShowBytes(keyBytes,key);
+                        else
+                            printf("    (none)          /* use InitExt() call */\n");
+
+                        if (Skein_InitExt(blkSize,&s,hashBits,SKEIN_CFG_TREE_INFO_SEQUENTIAL,key,keyBytes) != SKEIN_SUCCESS)
+                            FatalError("Skein_InitExt() error!");
+                        if (Skein_Update(&s,data,msgBits) != SKEIN_SUCCESS)
+                            FatalError("Skein_Update() error!");
+                        if (Skein_Final(&s,hashVal) != SKEIN_SUCCESS)
+                            FatalError("Skein_Final() error!");
+                        break;
+                    case DATA_TYPE_TREE:
+                        assert(0);
+                        break;
+                    }
+                printf("Result:\n");
+                ShowBytes((hashBits+7)/8,hashVal);
+                printf("--------------------------------\n");
+                }
+            }
+        }
+    if (!_quiet_)
+        fprintf(stderr,"katHash = %08X\n",katHash ^ 0x150183D2);
+    }
+
+/* generate pre-computed IVs for inclusion in Skein C code */
+void Skein_GenerateIV(void)
+    {
+    static const struct
+        { uint_t blkSize,hashBits; }
+            IV_TAB[] = /* which pairs to precompute */
+                { { 256, 128 }, { 256, 160 }, { 256, 224 }, { 256, 256 },
+                  { 512, 128 }, { 512, 160 }, { 512, 224 }, { 512, 256 },
+                  { 512, 384 }, { 512, 512 },
+                  {1024, 384 }, {1024, 512 }, {1024,1024 }
+                };
+    uint_t       i,j,blkSize,hashBits;
+    hashState    state;
+    const u64b_t *w;
+    const char   *s;
+
+    printf("#ifndef _SKEIN_IV_H_\n"
+           "#define _SKEIN_IV_H_\n\n"
+           "#include \"skein.h\"    /* get Skein macros and types */\n\n"
+           "/*\n"
+           "***************** Pre-computed Skein IVs *******************\n"
+           "**\n"
+           "** NOTE: these values are not \"magic\" constants, but\n"
+           "** are generated using the Threefish block function.\n"
+           "** They are pre-computed here only for speed; i.e., to\n"
+           "** avoid the need for a Threefish call during Init().\n"
+           "**\n"
+           "** The IV for any fixed hash length may be pre-computed.\n"
+           "** Only the most common values are included here.\n"
+           "**\n"
+           "************************************************************\n"
+           "**/\n\n"
+           "#define MK_64 SKEIN_MK_64\n\n"
+          );
+    for (i=0;i < sizeof(IV_TAB)/sizeof(IV_TAB[0]); i++)
+        {
+        blkSize  = IV_TAB[i].blkSize;
+        hashBits = IV_TAB[i].hashBits;
+        switch (blkSize)
+            {
+            case  256:  w = state.u.ctx_256.X;  s = "_256"; break;
+            case  512:  w = state.u.ctx_512.X;  s = "_512"; break;
+            case 1024:  w = state.u.ctx1024.X;  s = "1024"; break;
+            default:    FatalError("Invalid blkSize");
+                        continue; /* should never happen, but avoids gcc warning */
+            }
+        if (Skein_Init(blkSize,&state,hashBits) != SKEIN_SUCCESS)
+            FatalError("Error generating IV: blkSize=%d, hashBits=%d",blkSize,hashBits);
+        printf("/* blkSize = %4d bits. hashSize = %4d bits */\n",blkSize,hashBits);
+        printf("const u64b_t SKEIN%s_IV_%d[] =\n    {\n",s,hashBits);
+        for (j=0;j<blkSize/64;j++)
+            printf("    MK_64(0x%08X,0x%08X)%s\n",
+                   (uint_32t)(w[j] >> 32),(uint_32t)w[j],(j+1 == blkSize/64)?"":",");
+        printf("    };\n\n");
+        }
+    printf("#endif /* _SKEIN_IV_H_ */\n");
+    }
+
+/* qsort routine */
+int compare_uint_32t(const void *aPtr,const void *bPtr)
+    {
+    uint_32t a = * ((uint_32t *) aPtr);
+    uint_32t b = * ((uint_32t *) bPtr);
+    
+    if (a > b) return  1;
+    if (a < b) return -1;
+    return 0;
+    }
+
+void ShowCompiler(const char *CVER)
+    {
+    printf(" //:");
+#if defined(SKEIN_XMM)
+    printf(" 32-XMM, ");
+#else
+    printf(" %2u-bit, ",(uint_t)(8*sizeof(size_t)));
+#endif
+    printf("%s%s",COMPILER_ID,CVER);
+
+    /* do we need to show unroll amount? */
+#if defined(SKEIN_USE_ASM) && SKEIN_USE_ASM
+    printf(" [asm=");
+#define _SC_DO_LOOP_ (1)
+#elif defined(SKEIN_LOOP)
+    printf(" [ C =");
+#define _SC_DO_LOOP_ (1)
+#endif
+
+#ifdef  _SC_DO_LOOP_
+    printf("%c",(Skein_256_Unroll_Cnt())?'0'+Skein_256_Unroll_Cnt():'.');
+    printf("%c",(Skein_512_Unroll_Cnt())?'0'+Skein_512_Unroll_Cnt():'.');
+    printf("%c",(Skein1024_Unroll_Cnt())?'0'+Skein1024_Unroll_Cnt():'.');
+    printf("]");
+#endif
+    }
+
+/* measure the speed (in CPU clks/byte) for a Skein implementation */
+void Skein_MeasurePerformance(const char *target)
+    {
+    const uint_t MSG_BYTES[] = {1,2,4,8,10,16,32,64,100,128,256,512,1000,1024,2048,4096,8192,10000,16384,32768,100000,0};
+    enum     {  TIMER_SAMPLE_CNT = 13, MAX_BUFFER=1024*100, PERF_TIMEOUT_CLKS = 500000 };
+    enum     {  _256 = 256, _512 = 512 };
+    uint_32t dt[24][3][TIMER_SAMPLE_CNT],t0,t1;
+    uint_32t dtMin = ~0u;
+    uint_t   targetSize = 0;
+    uint_t   repCnt     = 1;
+    uint_t   i,k,n,r,blkSize,msgBytes;
+    u08b_t   b[MAX_BUFFER],hashVal[SKEIN1024_BLOCK_BYTES*4];
+    hashState s;
+#ifdef CompilerVersion
+    char     CVER[20];                      /* avoid ANSI compiler warnings for sprintf()! :-(( */
+    n          = CompilerVersion;
+    CVER[0]    = '_';
+    CVER[1]    = 'v';
+    CVER[2]    = (char)('0'+((n /100)%10));
+    CVER[3]    = '.';
+    CVER[4]    = (char)('0'+((n / 10)%10));
+    CVER[5]    = (char)('0'+((n /  1)%10));
+    CVER[6]    = 0;
+#else
+#define CVER ""
+#endif      
+    if (target && target[0])
+        {
+        targetSize = atoi(target);
+        for (i=0;target[i];i++)
+            if (target[i] == '.')
+                {
+                repCnt = atoi(target+i+1);
+                break;
+                }
+        if (repCnt == 0)
+            repCnt = 1;
+        }
+
+    assert(sizeof(dt)/(3*TIMER_SAMPLE_CNT*sizeof(dt[0][0][0])) >=
+           sizeof(MSG_BYTES)/sizeof(MSG_BYTES[0]));
+    if (OS_Set_High_Priority())
+        printf("Unable to set thread to high priority\n");
+    fflush(stdout);                     /* let things calm down */
+    OS_Sleep(200);                      /* let things settle down for a bit */
+    memset(dt,0,sizeof(dt));
+    RandBytes(b,sizeof(b));             /* use random data for testing */
+    for (i=0;i<4*TIMER_SAMPLE_CNT;i++)  /* calibrate the overhead for measuring time */
+        {
+        t0 = HiResTime();
+        t1 = HiResTime();
+        if (dtMin > t1-t0)              /* keep only the minimum time */
+            dtMin = t1-t0;
+        }
+    for (r=0;r<repCnt;r++)
+        {
+        /* first take all the data and store it in dt, with no printf() activity */
+        for (n=0;n < sizeof(MSG_BYTES)/sizeof(MSG_BYTES[0]);n++)
+            {
+            msgBytes = MSG_BYTES[n];        /* pick the message size (in bits) */
+            if (msgBytes > MAX_BUFFER || msgBytes == 0)
+                break;
+            if (targetSize && targetSize != msgBytes)
+                continue;
+            for (k=0;k<3;k++)
+                {                           /* cycle thru the different block sizes */
+                blkSize=256 << k;
+                t0=HiResTime();
+                t1=HiResTime();
+#define OneTest(BITS)                                           \
+                Skein##BITS##_Init  (&s.u.ctx##BITS,BITS);      \
+                Skein##BITS##_Update(&s.u.ctx##BITS,b,msgBytes);\
+                Skein##BITS##_Final (&s.u.ctx##BITS,hashVal);
+
+                OS_Sleep(0);                        /* yield the time slice to OS */
+                for (i=0;i<TIMER_SAMPLE_CNT;i++)
+                    {
+                    HiResTime();                    /* prime the pump */
+                    switch (blkSize)
+                        {
+                        case  256:
+                            OneTest(_256);          /* prime the pump */
+                            t0 = HiResTime();
+                            OneTest(_256);          /* do it twice for some averaging */
+                            OneTest(_256);
+                            t1 = HiResTime();
+                            break;
+                        case  512:
+                            OneTest(_512);
+                            t0 = HiResTime();
+                            OneTest(_512);
+                            OneTest(_512);
+                            t1 = HiResTime();
+                            break;
+                        case 1024:
+                            OneTest(1024);
+                            t0 = HiResTime();
+                            OneTest(1024);
+                            OneTest(1024);
+                            t1 = HiResTime();
+                            break;
+                        }
+                    dt[n][k][i] = ((t1 - t0) - dtMin)/2; /* adjust for HiResTime() overhead */
+                    }
+                }
+            }
+        OS_Set_Normal_Priority();
+
+        if (targetSize == 0)
+            {
+            printf("\nSkein performance, in clks per byte, dtMin = %4d clks.\n",dtMin);
+            printf("         [compiled %s,%s  by  '%s%s', %u-bit]\n",__TIME__,__DATE__,COMPILER_ID,CVER,(uint_t)(8*sizeof(size_t)));
+            printf("         =================================================================\n");
+            printf("         ||                       Skein block size                       |\n");
+            printf("         ||--------------------------------------------------------------|\n");
+            printf(" Message ||       256 bits     |       512 bits     |      1024 bits     |\n");
+            printf(" Length  ||====================|====================|====================|\n");
+            printf(" (bytes) ||     min    median  |     min    median  |     min    median  |\n"); 
+            printf("=========||====================|====================|====================|\n");
+            }
+
+        /* now display the results */
+        for (n=0;n < sizeof(MSG_BYTES)/sizeof(MSG_BYTES[0]);n++)
+            {
+            msgBytes = MSG_BYTES[n];       /* pick the message size (in bits) */
+            if (msgBytes > MAX_BUFFER || msgBytes == 0)
+                break;
+            if (targetSize && targetSize != msgBytes)
+                continue;
+            printf("%7d_ ||",msgBytes);
+            for (k=0;k<3;k++)              /* cycle thru the different Skein block sizes */
+                {   /* here with dt[n][k][] full of time differences */
+                    /* discard high/low, then show min/median of the rest, in clks/byte */
+                qsort(dt[n][k],TIMER_SAMPLE_CNT,sizeof(dt[0][0][0]),compare_uint_32t);
+                printf(" %8.2f %8.2f  |",dt[n][k][1]/(double)msgBytes,dt[n][k][TIMER_SAMPLE_CNT/2]/(double)msgBytes);
+                }
+            ShowCompiler(CVER);
+            printf("\n");
+            if (targetSize == 0 && target && target[0] && repCnt == 1)
+                {   /* show the details */
+                for (k=0;k<3;k++)
+                    {
+                    printf("%4d: ",256 << k);
+                    for (i=0;i<TIMER_SAMPLE_CNT;i++)
+                        printf("%8d",dt[n][k][i]);
+                    printf("\n");
+                    }
+                }
+            }
+        }
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+    if (targetSize == 0)
+        {
+        printf("=========||====================|====================|====================|\n");
+        printf("Code Size||                    |                    |                    |\n");
+        printf("=========||====================|====================|====================|\n");
+        printf("    API  || %12d bytes | %12d bytes | %12d bytes |",
+               (int) Skein_256_API_CodeSize(),
+               (int) Skein_512_API_CodeSize(),
+               (int) Skein1024_API_CodeSize());
+        ShowCompiler(CVER);
+        printf("\n");
+        printf("  Block  || %12d bytes | %12d bytes | %12d bytes |",
+               (int) Skein_256_Process_Block_CodeSize(),
+               (int) Skein_512_Process_Block_CodeSize(),
+               (int) Skein1024_Process_Block_CodeSize());
+        ShowCompiler(CVER);
+        printf("\n");
+        }
+#endif
+    }
+
+void GiveHelp(void)
+    {
+    printf("Syntax:  skein_test [options]\n"
+           "Options: -bNN  = set Skein block size to NN bits\n"
+           "         -lNN  = set max test length  to NN bits\n"
+           "         -tNN  = set Skein hash length to NN bits\n"
+           "         -sNN  = set initial random seed\n"
+           "         -g    = generate precomputed IV values to stdout\n"
+           "         -k    = output KAT results to stdout\n"
+           "         -p    = output performance (clks/byte)\n"
+          );
+    exit(2);
+    }
+                   
+int main(int argc,char *argv[])
+    {
+    int    i,n;
+    uint_t testCnt;
+    uint_t doKAT   =    0;   /* generate KAT vectors?    */
+    uint_t blkSize =    0;   /* Skein state size in bits */
+    uint_t maxLen  = 1024;   /* max block size   in bits */
+    uint_t hashLen =    0;   /* hash length      in bits (0 --> all) */
+    uint_t seed0   = (uint_t) time(NULL); /* randomize based on time */
+    uint_t oneBlk  =    0;   /* test block size */
+
+    for (i=1;i<argc;i++)
+        {   /* process command-line switches */
+        if (argv[i][0] == '-')
+            {
+            switch(toupper(argv[i][1]))
+                {
+                case '?': GiveHelp();                         break;
+                case 'B': blkSize       |= atoi(argv[i]+2);   break;
+                case 'L': maxLen         = atoi(argv[i]+2);   break;
+                case 'S': seed0          = atoi(argv[i]+2);   break;
+                case 'T': hashLen        = atoi(argv[i]+2);   break;
+                case 'K': doKAT          = 1;                 break;
+                case 'V': verbose       |= (argv[i][2]) ? atoi(argv[i]+2) : V_KAT_LONG; break;
+                case 'G': Skein_GenerateIV();                 return 0;
+                case 'P': Skein_MeasurePerformance(argv[i]+2);return 0;
+                case 'Q': _quiet_        = 1;                 break;
+                case 'D': switch (toupper(argv[i][2]))
+                              {
+#ifdef SKEIN_DEBUG
+                              case  0 : skein_DebugFlag |= SKEIN_DEBUG_DEFAULT; break;
+                              case '-': skein_DebugFlag |= SKEIN_DEBUG_SHORT;   break;
+                              case '+': skein_DebugFlag |= SKEIN_DEBUG_ALL;     break;
+                              case 'P': skein_DebugFlag |= SKEIN_DEBUG_PERMUTE; break;
+                              case 'I': skein_DebugFlag |= SKEIN_DEBUG_SHORT |  SKEIN_DEBUG_INJECT; break;
+                              case 'C': skein_DebugFlag |= SKEIN_DEBUG_SHORT & ~SKEIN_DEBUG_CONFIG; break;
+#endif
+                              default : skein_DebugFlag |= atoi(argv[i]+2);     break;
+                              }
+                          break;
+                default:  FatalError("Unsupported command-line option: %s",argv[i]);
+                          break;
+                }
+            }
+        else if (argv[i][0] == '?')
+            GiveHelp();
+        else if (isdigit(argv[i][0]))
+            oneBlk = atoi(argv[i]);
+        }
+
+    if (blkSize == 0)                     /* default is all block sizes */
+        blkSize = 256 | 512 | 1024;
+    if (doKAT)
+        {
+        Skein_ShowKAT(blkSize);
+        }
+    else
+        {
+        if (oneBlk == 0)
+            printf("Seed0 = %d. Compiler = %s\n",seed0,COMPILER_ID);
+        Rand_Init(SKEIN_MK_64(0xDEADBEEF,seed0)); /* init PRNG for test data */
+
+        testCnt=0;
+        for (i=256;i<=1024;i*=2)
+            {
+            if (blkSize & i)
+                {
+                if (hashLen == 0)              /* use all hash sizes? */
+                    {
+                    for (n=0;n < HASH_BITS_CNT;n++)
+                        testCnt += Skein_Test(i,maxLen,HASH_BITS[n],0,oneBlk);
+                    }
+                else
+                    testCnt += Skein_Test(i,maxLen,hashLen,0,oneBlk);
+                }
+            }
+        if (oneBlk)
+            return 0;
+        if (testCnt)
+            printf("Success: %4d tests\n",testCnt);
+        }
+    /* do a quick final self-consistentcy check test to make sure nothing is broken */
+    skein_DebugFlag = 0;        /* no debug output here */
+    for (blkSize = 256;blkSize <= 1024; blkSize*=2)
+        {
+        Skein_Test(blkSize,16,0,1,0);
+        }
+
+    return 0;
+    }
diff --git a/Optimized_32bit/SHA3api_ref.c b/Optimized_32bit/SHA3api_ref.c
new file mode 100644
index 000000000000..6861a3e4bffb
--- /dev/null
+++ b/Optimized_32bit/SHA3api_ref.c
@@ -0,0 +1,115 @@
+/***********************************************************************
+**
+** Implementation of the AHS API using the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+** 
+************************************************************************/
+
+#include <string.h>     /* get the memcpy/memset functions */
+#include "skein.h"      /* get the Skein API definitions   */
+#include "SHA3api_ref.h"/* get the  AHS  API definitions   */
+
+/******************************************************************/
+/*     AHS API code                                               */
+/******************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* select the context size and init the context */
+HashReturn Init(hashState *state, int hashbitlen)
+    {
+#if SKEIN_256_NIST_MAX_HASH_BITS
+    if (hashbitlen <= SKEIN_256_NIST_MAX_HASHBITS)
+        {
+        Skein_Assert(hashbitlen > 0,BAD_HASHLEN);
+        state->statebits = 64*SKEIN_256_STATE_WORDS;
+        return Skein_256_Init(&state->u.ctx_256,(size_t) hashbitlen);
+        }
+#endif
+    if (hashbitlen <= SKEIN_512_NIST_MAX_HASHBITS)
+        {
+        state->statebits = 64*SKEIN_512_STATE_WORDS;
+        return Skein_512_Init(&state->u.ctx_512,(size_t) hashbitlen);
+        }
+    else
+        {
+        state->statebits = 64*SKEIN1024_STATE_WORDS;
+        return Skein1024_Init(&state->u.ctx1024,(size_t) hashbitlen);
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process data to be hashed */
+HashReturn Update(hashState *state, const BitSequence *data, DataLength databitlen)
+    {
+    /* only the final Update() call is allowed do partial bytes, else assert an error */
+    Skein_Assert((state->u.h.T[1] & SKEIN_T1_FLAG_BIT_PAD) == 0 || databitlen == 0, FAIL);
+
+    Skein_Assert(state->statebits % 256 == 0 && (state->statebits-256) < 1024,FAIL);
+    if ((databitlen & 7) == 0)  /* partial bytes? */
+        {
+        switch ((state->statebits >> 8) & 3)
+            {
+            case 2:  return Skein_512_Update(&state->u.ctx_512,data,databitlen >> 3);
+            case 1:  return Skein_256_Update(&state->u.ctx_256,data,databitlen >> 3);
+            case 0:  return Skein1024_Update(&state->u.ctx1024,data,databitlen >> 3);
+            default: return FAIL;
+            }
+        }
+    else
+        {   /* handle partial final byte */
+        size_t bCnt = (databitlen >> 3) + 1;                  /* number of bytes to handle (nonzero here!) */
+        u08b_t b,mask;
+
+        mask = (u08b_t) (1u << (7 - (databitlen & 7)));       /* partial byte bit mask */
+        b    = (u08b_t) ((data[bCnt-1] & (0-mask)) | mask);   /* apply bit padding on final byte */
+
+        switch ((state->statebits >> 8) & 3)
+            {
+            case 2:  Skein_512_Update(&state->u.ctx_512,data,bCnt-1); /* process all but the final byte    */
+                     Skein_512_Update(&state->u.ctx_512,&b  ,  1   ); /* process the (masked) partial byte */
+                     break;
+            case 1:  Skein_256_Update(&state->u.ctx_256,data,bCnt-1); /* process all but the final byte    */
+                     Skein_256_Update(&state->u.ctx_256,&b  ,  1   ); /* process the (masked) partial byte */
+                     break;
+            case 0:  Skein1024_Update(&state->u.ctx1024,data,bCnt-1); /* process all but the final byte    */
+                     Skein1024_Update(&state->u.ctx1024,&b  ,  1   ); /* process the (masked) partial byte */
+                     break;
+            default: return FAIL;
+            }
+        Skein_Set_Bit_Pad_Flag(state->u.h);                    /* set tweak flag for the final call */
+        
+        return SUCCESS;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize hash computation and output the result (hashbitlen bits) */
+HashReturn Final(hashState *state, BitSequence *hashval)
+    {
+    Skein_Assert(state->statebits % 256 == 0 && (state->statebits-256) < 1024,FAIL);
+    switch ((state->statebits >> 8) & 3)
+        {
+        case 2:  return Skein_512_Final(&state->u.ctx_512,hashval);
+        case 1:  return Skein_256_Final(&state->u.ctx_256,hashval);
+        case 0:  return Skein1024_Final(&state->u.ctx1024,hashval);
+        default: return FAIL;
+        }
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* all-in-one hash function */
+HashReturn Hash(int hashbitlen, const BitSequence *data, /* all-in-one call */
+                DataLength databitlen,BitSequence *hashval)
+    {
+    hashState  state;
+    HashReturn r = Init(&state,hashbitlen);
+    if (r == SUCCESS)
+        { /* these calls do not fail when called properly */
+        r = Update(&state,data,databitlen);
+        Final(&state,hashval);
+        }
+    return r;
+    }
diff --git a/Optimized_32bit/SHA3api_ref.h b/Optimized_32bit/SHA3api_ref.h
new file mode 100644
index 000000000000..6d62304e59b7
--- /dev/null
+++ b/Optimized_32bit/SHA3api_ref.h
@@ -0,0 +1,66 @@
+#ifndef _AHS_API_H_
+#define _AHS_API_H_
+
+/***********************************************************************
+**
+** Interface declarations of the AHS API using the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+** 
+************************************************************************/
+
+#include "skein.h"
+
+typedef enum
+    {
+    SUCCESS     = SKEIN_SUCCESS,
+    FAIL        = SKEIN_FAIL,
+    BAD_HASHLEN = SKEIN_BAD_HASHLEN
+    }
+    HashReturn;
+
+typedef size_t   DataLength;                /* bit count  type */
+typedef u08b_t   BitSequence;               /* bit stream type */
+
+typedef struct
+    {
+    uint_t  statebits;                      /* 256, 512, or 1024 */
+    union
+        {
+        Skein_Ctxt_Hdr_t h;                 /* common header "overlay" */
+        Skein_256_Ctxt_t ctx_256;
+        Skein_512_Ctxt_t ctx_512;
+        Skein1024_Ctxt_t ctx1024;
+        } u;
+    }
+    hashState;
+
+/* "incremental" hashing API */
+HashReturn Init  (hashState *state, int hashbitlen);
+HashReturn Update(hashState *state, const BitSequence *data, DataLength databitlen);
+HashReturn Final (hashState *state,       BitSequence *hashval);
+
+/* "all-in-one" call */
+HashReturn Hash  (int hashbitlen,   const BitSequence *data, 
+                  DataLength databitlen,  BitSequence *hashval);
+
+
+/*
+** Re-define the compile-time constants below to change the selection
+** of the Skein state size in the Init() function in SHA3api_ref.c.
+**
+** That is, the NIST API does not allow for explicit selection of the
+** Skein block size, so it must be done implicitly in the Init() function.
+** The selection is controlled by these constants.
+*/
+#ifndef SKEIN_256_NIST_MAX_HASHBITS
+#define SKEIN_256_NIST_MAX_HASHBITS (0)
+#endif
+
+#ifndef SKEIN_512_NIST_MAX_HASHBITS
+#define SKEIN_512_NIST_MAX_HASHBITS (512)
+#endif
+
+#endif  /* ifdef _AHS_API_H_ */
diff --git a/Optimized_32bit/brg_endian.h b/Optimized_32bit/brg_endian.h
new file mode 100644
index 000000000000..978eb33f08cf
--- /dev/null
+++ b/Optimized_32bit/brg_endian.h
@@ -0,0 +1,148 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 2003, Dr Brian Gladman, Worcester, UK.   All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+   1. distributions of this source code include the above copyright
+      notice, this list of conditions and the following disclaimer;
+
+   2. distributions in binary form include the above copyright
+      notice, this list of conditions and the following disclaimer
+      in the documentation and/or other associated materials;
+
+   3. the copyright holder's name is not used to endorse products
+      built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 20/10/2006
+*/
+
+#ifndef BRG_ENDIAN_H
+#define BRG_ENDIAN_H
+
+#define IS_BIG_ENDIAN      4321 /* byte 0 is most significant (mc68k) */
+#define IS_LITTLE_ENDIAN   1234 /* byte 0 is least significant (i386) */
+
+/* Include files where endian defines and byteswap functions may reside */
+#if defined( __FreeBSD__ ) || defined( __OpenBSD__ ) || defined( __NetBSD__ )
+#  include <sys/endian.h>
+#elif defined( BSD ) && ( BSD >= 199103 ) || defined( __APPLE__ ) || \
+      defined( __CYGWIN32__ ) || defined( __DJGPP__ ) || defined( __osf__ )
+#  include <machine/endian.h>
+#elif defined( __linux__ ) || defined( __GNUC__ ) || defined( __GNU_LIBRARY__ )
+#  if !defined( __MINGW32__ ) && !defined(AVR)
+#    include <endian.h>
+#    if !defined( __BEOS__ )
+#      include <byteswap.h>
+#    endif
+#  endif
+#endif
+
+/* Now attempt to set the define for platform byte order using any  */
+/* of the four forms SYMBOL, _SYMBOL, __SYMBOL & __SYMBOL__, which  */
+/* seem to encompass most endian symbol definitions                 */
+
+#if defined( BIG_ENDIAN ) && defined( LITTLE_ENDIAN )
+#  if defined( BYTE_ORDER ) && BYTE_ORDER == BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( BYTE_ORDER ) && BYTE_ORDER == LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( _BIG_ENDIAN ) && defined( _LITTLE_ENDIAN )
+#  if defined( _BYTE_ORDER ) && _BYTE_ORDER == _BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( _BYTE_ORDER ) && _BYTE_ORDER == _LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( _BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( _LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN ) && defined( __LITTLE_ENDIAN )
+#  if defined( __BYTE_ORDER ) && __BYTE_ORDER == __BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( __BYTE_ORDER ) && __BYTE_ORDER == __LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( __BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN__ ) && defined( __LITTLE_ENDIAN__ )
+#  if defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __BIG_ENDIAN__
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __LITTLE_ENDIAN__
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( __BIG_ENDIAN__ )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN__ )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+/*  if the platform byte order could not be determined, then try to */
+/*  set this define using common machine defines                    */
+#if !defined(PLATFORM_BYTE_ORDER)
+
+#if   defined( __alpha__ ) || defined( __alpha ) || defined( i386 )       || \
+      defined( __i386__ )  || defined( _M_I86 )  || defined( _M_IX86 )    || \
+      defined( __OS2__ )   || defined( sun386 )  || defined( __TURBOC__ ) || \
+      defined( vax )       || defined( vms )     || defined( VMS )        || \
+      defined( __VMS )     || defined( _M_X64 )  || defined( AVR )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+
+#elif defined( AMIGA )   || defined( applec )    || defined( __AS400__ )  || \
+      defined( _CRAY )   || defined( __hppa )    || defined( __hp9000 )   || \
+      defined( ibm370 )  || defined( mc68000 )   || defined( m68k )       || \
+      defined( __MRC__ ) || defined( __MVS__ )   || defined( __MWERKS__ ) || \
+      defined( sparc )   || defined( __sparc)    || defined( SYMANTEC_C ) || \
+      defined( __VOS__ ) || defined( __TIGCC__ ) || defined( __TANDEM )   || \
+      defined( THINK_C ) || defined( __VMCMS__ )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+
+#elif 0     /* **** EDIT HERE IF NECESSARY **** */
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#elif 0     /* **** EDIT HERE IF NECESSARY **** */
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#else
+#  error Please edit lines 126 or 128 in brg_endian.h to set the platform byte order
+#endif
+#endif
+
+/* special handler for IA64, which may be either endianness (?)  */
+/* here we assume little-endian, but this may need to be changed */
+#if defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
+#  define PLATFORM_MUST_ALIGN (1)
+#ifndef PLATFORM_BYTE_ORDER
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+#endif
+
+#ifndef   PLATFORM_MUST_ALIGN
+#  define PLATFORM_MUST_ALIGN (0)
+#endif
+
+#endif  /* ifndef BRG_ENDIAN_H */
diff --git a/Optimized_32bit/brg_types.h b/Optimized_32bit/brg_types.h
new file mode 100644
index 000000000000..d6d6cdab9fbf
--- /dev/null
+++ b/Optimized_32bit/brg_types.h
@@ -0,0 +1,188 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 1998-2006, Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+   1. distributions of this source code include the above copyright
+      notice, this list of conditions and the following disclaimer;
+
+   2. distributions in binary form include the above copyright
+      notice, this list of conditions and the following disclaimer
+      in the documentation and/or other associated materials;
+
+   3. the copyright holder's name is not used to endorse products
+      built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 09/09/2006
+
+ The unsigned integer types defined here are of the form uint_<nn>t where
+ <nn> is the length of the type; for example, the unsigned 32-bit type is
+ 'uint_32t'.  These are NOT the same as the 'C99 integer types' that are
+ defined in the inttypes.h and stdint.h headers since attempts to use these
+ types have shown that support for them is still highly variable.  However,
+ since the latter are of the form uint<nn>_t, a regular expression search
+ and replace (in VC++ search on 'uint_{:z}t' and replace with 'uint\1_t')
+ can be used to convert the types used here to the C99 standard types.
+*/
+
+#ifndef BRG_TYPES_H
+#define BRG_TYPES_H
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#include <limits.h>
+
+#ifndef BRG_UI8
+#  define BRG_UI8
+#  if UCHAR_MAX == 255u
+     typedef unsigned char uint_8t;
+#  else
+#    error Please define uint_8t as an 8-bit unsigned integer type in brg_types.h
+#  endif
+#endif
+
+#ifndef BRG_UI16
+#  define BRG_UI16
+#  if USHRT_MAX == 65535u
+     typedef unsigned short uint_16t;
+#  else
+#    error Please define uint_16t as a 16-bit unsigned short type in brg_types.h
+#  endif
+#endif
+
+#ifndef BRG_UI32
+#  define BRG_UI32
+#  if UINT_MAX == 4294967295u
+#    define li_32(h) 0x##h##u
+     typedef unsigned int uint_32t;
+#  elif ULONG_MAX == 4294967295u
+#    define li_32(h) 0x##h##ul
+     typedef unsigned long uint_32t;
+#  elif defined( _CRAY )
+#    error This code needs 32-bit data types, which Cray machines do not provide
+#  else
+#    error Please define uint_32t as a 32-bit unsigned integer type in brg_types.h
+#  endif
+#endif
+
+#ifndef BRG_UI64
+#  if defined( __BORLANDC__ ) && !defined( __MSDOS__ )
+#    define BRG_UI64
+#    define li_64(h) 0x##h##ui64
+     typedef unsigned __int64 uint_64t;
+#  elif defined( _MSC_VER ) && ( _MSC_VER < 1300 )    /* 1300 == VC++ 7.0 */
+#    define BRG_UI64
+#    define li_64(h) 0x##h##ui64
+     typedef unsigned __int64 uint_64t;
+#  elif defined( __sun ) && defined(ULONG_MAX) && ULONG_MAX == 0xfffffffful
+#    define BRG_UI64
+#    define li_64(h) 0x##h##ull
+     typedef unsigned long long uint_64t;
+#  elif defined( UINT_MAX ) && UINT_MAX > 4294967295u
+#    if UINT_MAX == 18446744073709551615u
+#      define BRG_UI64
+#      define li_64(h) 0x##h##u
+       typedef unsigned int uint_64t;
+#    endif
+#  elif defined( ULONG_MAX ) && ULONG_MAX > 4294967295u
+#    if ULONG_MAX == 18446744073709551615ul
+#      define BRG_UI64
+#      define li_64(h) 0x##h##ul
+       typedef unsigned long uint_64t;
+#    endif
+#  elif defined( ULLONG_MAX ) && ULLONG_MAX > 4294967295u
+#    if ULLONG_MAX == 18446744073709551615ull
+#      define BRG_UI64
+#      define li_64(h) 0x##h##ull
+       typedef unsigned long long uint_64t;
+#    endif
+#  elif defined( ULONG_LONG_MAX ) && ULONG_LONG_MAX > 4294967295u
+#    if ULONG_LONG_MAX == 18446744073709551615ull
+#      define BRG_UI64
+#      define li_64(h) 0x##h##ull
+       typedef unsigned long long uint_64t;
+#    endif
+#  elif defined(__GNUC__)  /* DLW: avoid mingw problem with -ansi */
+#      define BRG_UI64
+#      define li_64(h) 0x##h##ull
+       typedef unsigned long long uint_64t;
+#  endif
+#endif
+
+#if defined( NEED_UINT_64T ) && !defined( BRG_UI64 )
+#  error Please define uint_64t as an unsigned 64 bit type in brg_types.h
+#endif
+
+#ifndef RETURN_VALUES
+#  define RETURN_VALUES
+#  if defined( DLL_EXPORT )
+#    if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+#      define VOID_RETURN    __declspec( dllexport ) void __stdcall
+#      define INT_RETURN     __declspec( dllexport ) int  __stdcall
+#    elif defined( __GNUC__ )
+#      define VOID_RETURN    __declspec( __dllexport__ ) void
+#      define INT_RETURN     __declspec( __dllexport__ ) int
+#    else
+#      error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+#    endif
+#  elif defined( DLL_IMPORT )
+#    if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+#      define VOID_RETURN    __declspec( dllimport ) void __stdcall
+#      define INT_RETURN     __declspec( dllimport ) int  __stdcall
+#    elif defined( __GNUC__ )
+#      define VOID_RETURN    __declspec( __dllimport__ ) void
+#      define INT_RETURN     __declspec( __dllimport__ ) int
+#    else
+#      error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+#    endif
+#  elif defined( __WATCOMC__ )
+#    define VOID_RETURN  void __cdecl
+#    define INT_RETURN   int  __cdecl
+#  else
+#    define VOID_RETURN  void
+#    define INT_RETURN   int
+#  endif
+#endif
+
+/*  These defines are used to declare buffers in a way that allows
+    faster operations on longer variables to be used.  In all these
+    defines 'size' must be a power of 2 and >= 8
+
+    dec_unit_type(size,x)       declares a variable 'x' of length 
+                                'size' bits
+
+    dec_bufr_type(size,bsize,x) declares a buffer 'x' of length 'bsize' 
+                                bytes defined as an array of variables
+                                each of 'size' bits (bsize must be a 
+                                multiple of size / 8)
+
+    ptr_cast(x,size)            casts a pointer to a pointer to a 
+                                varaiable of length 'size' bits
+*/
+
+#define ui_type(size)               uint_##size##t
+#define dec_unit_type(size,x)       typedef ui_type(size) x
+#define dec_bufr_type(size,bsize,x) typedef ui_type(size) x[bsize / (size >> 3)]
+#define ptr_cast(x,size)            ((ui_type(size)*)(x))
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif
diff --git a/Optimized_32bit/skein.c b/Optimized_32bit/skein.c
new file mode 100644
index 000000000000..c9289cd49e8e
--- /dev/null
+++ b/Optimized_32bit/skein.c
@@ -0,0 +1,753 @@
+/***********************************************************************
+**
+** Implementation of the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+** 
+************************************************************************/
+
+#define  SKEIN_PORT_CODE /* instantiate any code in skein_port.h */
+
+#include <string.h>      /* get the memcpy/memset functions */
+#include "skein.h"       /* get the Skein API definitions   */
+#include "skein_iv.h"    /* get precomputed IVs */
+
+/*****************************************************************/
+/* External function to process blkCnt (nonzero) full block(s) of data. */
+void    Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void    Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void    Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+
+/*****************************************************************/
+/*     256-bit Skein                                             */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation  */
+int Skein_256_Init(Skein_256_Ctxt_t *ctx, size_t hashBitLen)
+    {
+    union
+        {
+        u08b_t  b[SKEIN_256_STATE_BYTES];
+        u64b_t  w[SKEIN_256_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    ctx->h.hashBitLen = hashBitLen;         /* output hash bit count */
+
+    switch (hashBitLen)
+        {             /* use pre-computed values, where available */
+#ifndef SKEIN_NO_PRECOMP
+        case  256: memcpy(ctx->X,SKEIN_256_IV_256,sizeof(ctx->X));  break;
+        case  224: memcpy(ctx->X,SKEIN_256_IV_224,sizeof(ctx->X));  break;
+        case  160: memcpy(ctx->X,SKEIN_256_IV_160,sizeof(ctx->X));  break;
+        case  128: memcpy(ctx->X,SKEIN_256_IV_128,sizeof(ctx->X));  break;
+#endif
+        default:
+            /* here if there is no precomputed IV value available */
+            /* build/process the config block, type == CONFIG (could be precomputed) */
+            Skein_Start_New_Type(ctx,CFG_FINAL);        /* set tweaks: T0=0; T1=CFG | FINAL */
+
+            cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);  /* set the schema, version */
+            cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+            cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+            memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+            /* compute the initial chaining values from config block */
+            memset(ctx->X,0,sizeof(ctx->X));            /* zero the chaining variables */
+            Skein_256_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+            break;
+        }
+    /* The chaining vars ctx->X are now initialized for the given hashBitLen. */
+    /* Set up to process the data message portion of the hash (default) */
+    Skein_Start_New_Type(ctx,MSG);              /* T0=0, T1= MSG type */
+
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a MAC and/or tree hash operation */
+/* [identical to Skein_256_Init() when keyBytes == 0 && treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL] */
+int Skein_256_InitExt(Skein_256_Ctxt_t *ctx,size_t hashBitLen,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+    {
+    union
+        {
+        u08b_t  b[SKEIN_256_STATE_BYTES];
+        u64b_t  w[SKEIN_256_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    Skein_Assert(keyBytes == 0 || key != NULL,SKEIN_FAIL);
+
+    /* compute the initial chaining values ctx->X[], based on key */
+    if (keyBytes == 0)                          /* is there a key? */
+        {                                   
+        memset(ctx->X,0,sizeof(ctx->X));        /* no key: use all zeroes as key for config block */
+        }
+    else                                        /* here to pre-process a key */
+        {
+        Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+        /* do a mini-Init right here */
+        ctx->h.hashBitLen=8*sizeof(ctx->X);     /* set output hash bit count = state size */
+        Skein_Start_New_Type(ctx,KEY);          /* set tweaks: T0 = 0; T1 = KEY type */
+        memset(ctx->X,0,sizeof(ctx->X));        /* zero the initial chaining variables */
+        Skein_256_Update(ctx,key,keyBytes);     /* hash the key */
+        Skein_256_Final_Pad(ctx,cfg.b);         /* put result into cfg.b[] */
+        memcpy(ctx->X,cfg.b,sizeof(cfg.b));     /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+        {
+        uint_t i;
+        for (i=0;i<SKEIN_256_STATE_WORDS;i++)   /* convert key bytes to context words */
+            ctx->X[i] = Skein_Swap64(ctx->X[i]);
+        }
+#endif
+        }
+    /* build/process the config block, type == CONFIG (could be precomputed for each key) */
+    ctx->h.hashBitLen = hashBitLen;             /* output hash bit count */
+    Skein_Start_New_Type(ctx,CFG_FINAL);
+
+    memset(&cfg.w,0,sizeof(cfg.w));             /* pre-pad cfg.w[] with zeroes */
+    cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+    cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+    cfg.w[2] = Skein_Swap64(treeInfo);          /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+    Skein_Show_Key(256,&ctx->h,key,keyBytes);
+
+    /* compute the initial chaining values from config block */
+    Skein_256_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+
+    /* The chaining vars ctx->X are now initialized */
+    /* Set up to process the data message portion of the hash (default) */
+    ctx->h.bCnt = 0;                            /* buffer b[] starts out empty */
+    Skein_Start_New_Type(ctx,MSG);
+    
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process the input bytes */
+int Skein_256_Update(Skein_256_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt)
+    {
+    size_t n;
+
+    Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* process full blocks, if any */
+    if (msgByteCnt + ctx->h.bCnt > SKEIN_256_BLOCK_BYTES)
+        {
+        if (ctx->h.bCnt)                              /* finish up any buffered message data */
+            {
+            n = SKEIN_256_BLOCK_BYTES - ctx->h.bCnt;  /* # bytes free in buffer b[] */
+            if (n)
+                {
+                Skein_assert(n < msgByteCnt);         /* check on our logic here */
+                memcpy(&ctx->b[ctx->h.bCnt],msg,n);
+                msgByteCnt  -= n;
+                msg         += n;
+                ctx->h.bCnt += n;
+                }
+            Skein_assert(ctx->h.bCnt == SKEIN_256_BLOCK_BYTES);
+            Skein_256_Process_Block(ctx,ctx->b,1,SKEIN_256_BLOCK_BYTES);
+            ctx->h.bCnt = 0;
+            }
+        /* now process any remaining full blocks, directly from input message data */
+        if (msgByteCnt > SKEIN_256_BLOCK_BYTES)
+            {
+            n = (msgByteCnt-1) / SKEIN_256_BLOCK_BYTES;   /* number of full blocks to process */
+            Skein_256_Process_Block(ctx,msg,n,SKEIN_256_BLOCK_BYTES);
+            msgByteCnt -= n * SKEIN_256_BLOCK_BYTES;
+            msg        += n * SKEIN_256_BLOCK_BYTES;
+            }
+        Skein_assert(ctx->h.bCnt == 0);
+        }
+
+    /* copy any remaining source message data bytes into b[] */
+    if (msgByteCnt)
+        {
+        Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES);
+        memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
+        ctx->h.bCnt += msgByteCnt;
+        }
+
+    return SKEIN_SUCCESS;
+    }
+   
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the result */
+int Skein_256_Final(Skein_256_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN_256_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;                 /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES)            /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+
+    Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);  /* process the final block */
+    
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;             /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN_256_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN_256_BLOCK_BYTES)
+            n  = SKEIN_256_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN_256_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(256,&ctx->h,n,hashVal+i*SKEIN_256_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_256_API_CodeSize(void)
+    {
+    return ((u08b_t *) Skein_256_API_CodeSize) -
+           ((u08b_t *) Skein_256_Init);
+    }
+#endif
+
+/*****************************************************************/
+/*     512-bit Skein                                             */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation  */
+int Skein_512_Init(Skein_512_Ctxt_t *ctx, size_t hashBitLen)
+    {
+    union
+        {
+        u08b_t  b[SKEIN_512_STATE_BYTES];
+        u64b_t  w[SKEIN_512_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    ctx->h.hashBitLen = hashBitLen;         /* output hash bit count */
+
+    switch (hashBitLen)
+        {             /* use pre-computed values, where available */
+#ifndef SKEIN_NO_PRECOMP
+        case  512: memcpy(ctx->X,SKEIN_512_IV_512,sizeof(ctx->X));  break;
+        case  384: memcpy(ctx->X,SKEIN_512_IV_384,sizeof(ctx->X));  break;
+        case  256: memcpy(ctx->X,SKEIN_512_IV_256,sizeof(ctx->X));  break;
+        case  224: memcpy(ctx->X,SKEIN_512_IV_224,sizeof(ctx->X));  break;
+#endif
+        default:
+            /* here if there is no precomputed IV value available */
+            /* build/process the config block, type == CONFIG (could be precomputed) */
+            Skein_Start_New_Type(ctx,CFG_FINAL);        /* set tweaks: T0=0; T1=CFG | FINAL */
+
+            cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);  /* set the schema, version */
+            cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+            cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+            memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+            /* compute the initial chaining values from config block */
+            memset(ctx->X,0,sizeof(ctx->X));            /* zero the chaining variables */
+            Skein_512_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+            break;
+        }
+
+    /* The chaining vars ctx->X are now initialized for the given hashBitLen. */
+    /* Set up to process the data message portion of the hash (default) */
+    Skein_Start_New_Type(ctx,MSG);              /* T0=0, T1= MSG type */
+
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a MAC and/or tree hash operation */
+/* [identical to Skein_512_Init() when keyBytes == 0 && treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL] */
+int Skein_512_InitExt(Skein_512_Ctxt_t *ctx,size_t hashBitLen,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+    {
+    union
+        {
+        u08b_t  b[SKEIN_512_STATE_BYTES];
+        u64b_t  w[SKEIN_512_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    Skein_Assert(keyBytes == 0 || key != NULL,SKEIN_FAIL);
+
+    /* compute the initial chaining values ctx->X[], based on key */
+    if (keyBytes == 0)                          /* is there a key? */
+        {                                   
+        memset(ctx->X,0,sizeof(ctx->X));        /* no key: use all zeroes as key for config block */
+        }
+    else                                        /* here to pre-process a key */
+        {
+        Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+        /* do a mini-Init right here */
+        ctx->h.hashBitLen=8*sizeof(ctx->X);     /* set output hash bit count = state size */
+        Skein_Start_New_Type(ctx,KEY);          /* set tweaks: T0 = 0; T1 = KEY type */
+        memset(ctx->X,0,sizeof(ctx->X));        /* zero the initial chaining variables */
+        Skein_512_Update(ctx,key,keyBytes);     /* hash the key */
+        Skein_512_Final_Pad(ctx,cfg.b);         /* put result into cfg.b[] */
+        memcpy(ctx->X,cfg.b,sizeof(cfg.b));     /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+        {
+        uint_t i;
+        for (i=0;i<SKEIN_512_STATE_WORDS;i++)   /* convert key bytes to context words */
+            ctx->X[i] = Skein_Swap64(ctx->X[i]);
+        }
+#endif
+        }
+    /* build/process the config block, type == CONFIG (could be precomputed for each key) */
+    ctx->h.hashBitLen = hashBitLen;             /* output hash bit count */
+    Skein_Start_New_Type(ctx,CFG_FINAL);
+
+    memset(&cfg.w,0,sizeof(cfg.w));             /* pre-pad cfg.w[] with zeroes */
+    cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+    cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+    cfg.w[2] = Skein_Swap64(treeInfo);          /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+    Skein_Show_Key(512,&ctx->h,key,keyBytes);
+
+    /* compute the initial chaining values from config block */
+    Skein_512_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+
+    /* The chaining vars ctx->X are now initialized */
+    /* Set up to process the data message portion of the hash (default) */
+    ctx->h.bCnt = 0;                            /* buffer b[] starts out empty */
+    Skein_Start_New_Type(ctx,MSG);
+    
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process the input bytes */
+int Skein_512_Update(Skein_512_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt)
+    {
+    size_t n;
+
+    Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* process full blocks, if any */
+    if (msgByteCnt + ctx->h.bCnt > SKEIN_512_BLOCK_BYTES)
+        {
+        if (ctx->h.bCnt)                              /* finish up any buffered message data */
+            {
+            n = SKEIN_512_BLOCK_BYTES - ctx->h.bCnt;  /* # bytes free in buffer b[] */
+            if (n)
+                {
+                Skein_assert(n < msgByteCnt);         /* check on our logic here */
+                memcpy(&ctx->b[ctx->h.bCnt],msg,n);
+                msgByteCnt  -= n;
+                msg         += n;
+                ctx->h.bCnt += n;
+                }
+            Skein_assert(ctx->h.bCnt == SKEIN_512_BLOCK_BYTES);
+            Skein_512_Process_Block(ctx,ctx->b,1,SKEIN_512_BLOCK_BYTES);
+            ctx->h.bCnt = 0;
+            }
+        /* now process any remaining full blocks, directly from input message data */
+        if (msgByteCnt > SKEIN_512_BLOCK_BYTES)
+            {
+            n = (msgByteCnt-1) / SKEIN_512_BLOCK_BYTES;   /* number of full blocks to process */
+            Skein_512_Process_Block(ctx,msg,n,SKEIN_512_BLOCK_BYTES);
+            msgByteCnt -= n * SKEIN_512_BLOCK_BYTES;
+            msg        += n * SKEIN_512_BLOCK_BYTES;
+            }
+        Skein_assert(ctx->h.bCnt == 0);
+        }
+
+    /* copy any remaining source message data bytes into b[] */
+    if (msgByteCnt)
+        {
+        Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES);
+        memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
+        ctx->h.bCnt += msgByteCnt;
+        }
+
+    return SKEIN_SUCCESS;
+    }
+   
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the result */
+int Skein_512_Final(Skein_512_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN_512_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;                 /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES)            /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+
+    Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);  /* process the final block */
+    
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;             /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN_512_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN_512_BLOCK_BYTES)
+            n  = SKEIN_512_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN_512_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(512,&ctx->h,n,hashVal+i*SKEIN_512_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_512_API_CodeSize(void)
+    {
+    return ((u08b_t *) Skein_512_API_CodeSize) -
+           ((u08b_t *) Skein_512_Init);
+    }
+#endif
+
+/*****************************************************************/
+/*    1024-bit Skein                                             */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation  */
+int Skein1024_Init(Skein1024_Ctxt_t *ctx, size_t hashBitLen)
+    {
+    union
+        {
+        u08b_t  b[SKEIN1024_STATE_BYTES];
+        u64b_t  w[SKEIN1024_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    ctx->h.hashBitLen = hashBitLen;         /* output hash bit count */
+
+    switch (hashBitLen)
+        {              /* use pre-computed values, where available */
+#ifndef SKEIN_NO_PRECOMP
+        case  512: memcpy(ctx->X,SKEIN1024_IV_512 ,sizeof(ctx->X)); break;
+        case  384: memcpy(ctx->X,SKEIN1024_IV_384 ,sizeof(ctx->X)); break;
+        case 1024: memcpy(ctx->X,SKEIN1024_IV_1024,sizeof(ctx->X)); break;
+#endif
+        default:
+            /* here if there is no precomputed IV value available */
+            /* build/process the config block, type == CONFIG (could be precomputed) */
+            Skein_Start_New_Type(ctx,CFG_FINAL);        /* set tweaks: T0=0; T1=CFG | FINAL */
+
+            cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);  /* set the schema, version */
+            cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+            cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+            memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+            /* compute the initial chaining values from config block */
+            memset(ctx->X,0,sizeof(ctx->X));            /* zero the chaining variables */
+            Skein1024_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+            break;
+        }
+
+    /* The chaining vars ctx->X are now initialized for the given hashBitLen. */
+    /* Set up to process the data message portion of the hash (default) */
+    Skein_Start_New_Type(ctx,MSG);              /* T0=0, T1= MSG type */
+
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a MAC and/or tree hash operation */
+/* [identical to Skein1024_Init() when keyBytes == 0 && treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL] */
+int Skein1024_InitExt(Skein1024_Ctxt_t *ctx,size_t hashBitLen,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+    {
+    union
+        {
+        u08b_t  b[SKEIN1024_STATE_BYTES];
+        u64b_t  w[SKEIN1024_STATE_WORDS];
+        } cfg;                              /* config block */
+        
+    Skein_Assert(hashBitLen > 0,SKEIN_BAD_HASHLEN);
+    Skein_Assert(keyBytes == 0 || key != NULL,SKEIN_FAIL);
+
+    /* compute the initial chaining values ctx->X[], based on key */
+    if (keyBytes == 0)                          /* is there a key? */
+        {                                   
+        memset(ctx->X,0,sizeof(ctx->X));        /* no key: use all zeroes as key for config block */
+        }
+    else                                        /* here to pre-process a key */
+        {
+        Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+        /* do a mini-Init right here */
+        ctx->h.hashBitLen=8*sizeof(ctx->X);     /* set output hash bit count = state size */
+        Skein_Start_New_Type(ctx,KEY);          /* set tweaks: T0 = 0; T1 = KEY type */
+        memset(ctx->X,0,sizeof(ctx->X));        /* zero the initial chaining variables */
+        Skein1024_Update(ctx,key,keyBytes);     /* hash the key */
+        Skein1024_Final_Pad(ctx,cfg.b);         /* put result into cfg.b[] */
+        memcpy(ctx->X,cfg.b,sizeof(cfg.b));     /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+        {
+        uint_t i;
+        for (i=0;i<SKEIN1024_STATE_WORDS;i++)   /* convert key bytes to context words */
+            ctx->X[i] = Skein_Swap64(ctx->X[i]);
+        }
+#endif
+        }
+    /* build/process the config block, type == CONFIG (could be precomputed for each key) */
+    ctx->h.hashBitLen = hashBitLen;             /* output hash bit count */
+    Skein_Start_New_Type(ctx,CFG_FINAL);
+
+    memset(&cfg.w,0,sizeof(cfg.w));             /* pre-pad cfg.w[] with zeroes */
+    cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+    cfg.w[1] = Skein_Swap64(hashBitLen);        /* hash result length in bits */
+    cfg.w[2] = Skein_Swap64(treeInfo);          /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+    Skein_Show_Key(1024,&ctx->h,key,keyBytes);
+
+    /* compute the initial chaining values from config block */
+    Skein1024_Process_Block(ctx,cfg.b,1,SKEIN_CFG_STR_LEN);
+
+    /* The chaining vars ctx->X are now initialized */
+    /* Set up to process the data message portion of the hash (default) */
+    ctx->h.bCnt = 0;                            /* buffer b[] starts out empty */
+    Skein_Start_New_Type(ctx,MSG);
+    
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process the input bytes */
+int Skein1024_Update(Skein1024_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt)
+    {
+    size_t n;
+
+    Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* process full blocks, if any */
+    if (msgByteCnt + ctx->h.bCnt > SKEIN1024_BLOCK_BYTES)
+        {
+        if (ctx->h.bCnt)                              /* finish up any buffered message data */
+            {
+            n = SKEIN1024_BLOCK_BYTES - ctx->h.bCnt;  /* # bytes free in buffer b[] */
+            if (n)
+                {
+                Skein_assert(n < msgByteCnt);         /* check on our logic here */
+                memcpy(&ctx->b[ctx->h.bCnt],msg,n);
+                msgByteCnt  -= n;
+                msg         += n;
+                ctx->h.bCnt += n;
+                }
+            Skein_assert(ctx->h.bCnt == SKEIN1024_BLOCK_BYTES);
+            Skein1024_Process_Block(ctx,ctx->b,1,SKEIN1024_BLOCK_BYTES);
+            ctx->h.bCnt = 0;
+            }
+        /* now process any remaining full blocks, directly from input message data */
+        if (msgByteCnt > SKEIN1024_BLOCK_BYTES)
+            {
+            n = (msgByteCnt-1) / SKEIN1024_BLOCK_BYTES;   /* number of full blocks to process */
+            Skein1024_Process_Block(ctx,msg,n,SKEIN1024_BLOCK_BYTES);
+            msgByteCnt -= n * SKEIN1024_BLOCK_BYTES;
+            msg        += n * SKEIN1024_BLOCK_BYTES;
+            }
+        Skein_assert(ctx->h.bCnt == 0);
+        }
+
+    /* copy any remaining source message data bytes into b[] */
+    if (msgByteCnt)
+        {
+        Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES);
+        memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
+        ctx->h.bCnt += msgByteCnt;
+        }
+
+    return SKEIN_SUCCESS;
+    }
+   
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the result */
+int Skein1024_Final(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN1024_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;                 /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES)            /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+
+    Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);  /* process the final block */
+    
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;             /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN1024_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN1024_BLOCK_BYTES)
+            n  = SKEIN1024_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN1024_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(1024,&ctx->h,n,hashVal+i*SKEIN1024_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein1024_API_CodeSize(void)
+    {
+    return ((u08b_t *) Skein1024_API_CodeSize) -
+           ((u08b_t *) Skein1024_Init);
+    }
+#endif
+
+/**************** Functions to support MAC/tree hashing ***************/
+/*   (this code is identical for Optimized and Reference versions)    */
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int Skein_256_Final_Pad(Skein_256_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;        /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES)   /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+    Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);    /* process the final block */
+    
+    Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_256_BLOCK_BYTES);   /* "output" the state bytes */
+    
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int Skein_512_Final_Pad(Skein_512_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;        /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES)   /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+    Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);    /* process the final block */
+    
+    Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_512_BLOCK_BYTES);   /* "output" the state bytes */
+    
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;        /* tag as the final block */
+    if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES)   /* zero pad b[] if necessary */
+        memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+    Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt);    /* process the final block */
+    
+    Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN1024_BLOCK_BYTES);   /* "output" the state bytes */
+    
+    return SKEIN_SUCCESS;
+    }
+
+#if SKEIN_TREE_HASH
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage                                       */
+int Skein_256_Output(Skein_256_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN_256_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;    /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN_256_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN_256_BLOCK_BYTES)
+            n  = SKEIN_256_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN_256_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(256,&ctx->h,n,hashVal+i*SKEIN_256_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage                                       */
+int Skein_512_Output(Skein_512_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN_512_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;    /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN_512_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN_512_BLOCK_BYTES)
+            n  = SKEIN_512_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN_512_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(256,&ctx->h,n,hashVal+i*SKEIN_512_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage                                       */
+int Skein1024_Output(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+    {
+    size_t i,n,byteCnt;
+    u64b_t X[SKEIN1024_STATE_WORDS];
+    Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL);    /* catch uninitialized context */
+
+    /* now output the result */
+    byteCnt = (ctx->h.hashBitLen + 7) >> 3;    /* total number of output bytes */
+
+    /* run Threefish in "counter mode" to generate output */
+    memset(ctx->b,0,sizeof(ctx->b));  /* zero out b[], so it can hold the counter */
+    memcpy(X,ctx->X,sizeof(X));       /* keep a local copy of counter mode "key" */
+    for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+        {
+        ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+        Skein_Start_New_Type(ctx,OUT_FINAL);
+        Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+        n = byteCnt - i*SKEIN1024_BLOCK_BYTES;   /* number of output bytes left to go */
+        if (n >= SKEIN1024_BLOCK_BYTES)
+            n  = SKEIN1024_BLOCK_BYTES;
+        Skein_Put64_LSB_First(hashVal+i*SKEIN1024_BLOCK_BYTES,ctx->X,n);   /* "output" the ctr mode bytes */
+        Skein_Show_Final(256,&ctx->h,n,hashVal+i*SKEIN1024_BLOCK_BYTES);
+        memcpy(ctx->X,X,sizeof(X));   /* restore the counter mode key for next time */
+        }
+    return SKEIN_SUCCESS;
+    }
+#endif
diff --git a/Optimized_32bit/skein.h b/Optimized_32bit/skein.h
new file mode 100644
index 000000000000..721c9bc9ce0d
--- /dev/null
+++ b/Optimized_32bit/skein.h
@@ -0,0 +1,327 @@
+#ifndef _SKEIN_H_
+#define _SKEIN_H_     1
+/**************************************************************************
+**
+** Interface declarations and internal definitions for Skein hashing.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+***************************************************************************
+** 
+** The following compile-time switches may be defined to control some
+** tradeoffs between speed, code size, error checking, and security.
+**
+** The "default" note explains what happens when the switch is not defined.
+**
+**  SKEIN_DEBUG            -- make callouts from inside Skein code
+**                            to examine/display intermediate values.
+**                            [default: no callouts (no overhead)]
+**
+**  SKEIN_ERR_CHECK        -- how error checking is handled inside Skein
+**                            code. If not defined, most error checking 
+**                            is disabled (for performance). Otherwise, 
+**                            the switch value is interpreted as:
+**                                0: use assert()      to flag errors
+**                                1: return SKEIN_FAIL to flag errors
+**
+***************************************************************************/
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+#include <stddef.h>                          /* get size_t definition */
+#include "skein_port.h"                      /* get platform-specific definitions */
+
+enum
+    {
+    SKEIN_SUCCESS         =      0,          /* return codes from Skein calls */
+    SKEIN_FAIL            =      1,
+    SKEIN_BAD_HASHLEN     =      2
+    };
+
+#define  SKEIN_MODIFIER_WORDS  ( 2)          /* number of modifier (tweak) words */
+
+#define  SKEIN_256_STATE_WORDS ( 4)
+#define  SKEIN_512_STATE_WORDS ( 8)
+#define  SKEIN1024_STATE_WORDS (16)
+#define  SKEIN_MAX_STATE_WORDS (16)
+
+#define  SKEIN_256_STATE_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define  SKEIN_512_STATE_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define  SKEIN1024_STATE_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+#define  SKEIN_256_STATE_BITS  (64*SKEIN_256_STATE_WORDS)
+#define  SKEIN_512_STATE_BITS  (64*SKEIN_512_STATE_WORDS)
+#define  SKEIN1024_STATE_BITS  (64*SKEIN1024_STATE_WORDS)
+
+#define  SKEIN_256_BLOCK_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define  SKEIN_512_BLOCK_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define  SKEIN1024_BLOCK_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+typedef struct
+    {
+    size_t  hashBitLen;                      /* size of hash result, in bits */
+    size_t  bCnt;                            /* current byte count in buffer b[] */
+    u64b_t  T[SKEIN_MODIFIER_WORDS];         /* tweak words: T[0]=byte cnt, T[1]=flags */
+    } Skein_Ctxt_Hdr_t;
+
+typedef struct                               /*  256-bit Skein hash context structure */
+    {
+    Skein_Ctxt_Hdr_t h;                      /* common header context variables */
+    u64b_t  X[SKEIN_256_STATE_WORDS];        /* chaining variables */
+    u08b_t  b[SKEIN_256_BLOCK_BYTES];        /* partial block buffer (8-byte aligned) */
+    } Skein_256_Ctxt_t;
+
+typedef struct                               /*  512-bit Skein hash context structure */
+    {
+    Skein_Ctxt_Hdr_t h;                      /* common header context variables */
+    u64b_t  X[SKEIN_512_STATE_WORDS];        /* chaining variables */
+    u08b_t  b[SKEIN_512_BLOCK_BYTES];        /* partial block buffer (8-byte aligned) */
+    } Skein_512_Ctxt_t;
+
+typedef struct                               /* 1024-bit Skein hash context structure */
+    {
+    Skein_Ctxt_Hdr_t h;                      /* common header context variables */
+    u64b_t  X[SKEIN1024_STATE_WORDS];        /* chaining variables */
+    u08b_t  b[SKEIN1024_BLOCK_BYTES];        /* partial block buffer (8-byte aligned) */
+    } Skein1024_Ctxt_t;
+
+/*   Skein APIs for (incremental) "straight hashing" */
+int  Skein_256_Init  (Skein_256_Ctxt_t *ctx, size_t hashBitLen);
+int  Skein_512_Init  (Skein_512_Ctxt_t *ctx, size_t hashBitLen);
+int  Skein1024_Init  (Skein1024_Ctxt_t *ctx, size_t hashBitLen);
+
+int  Skein_256_Update(Skein_256_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int  Skein_512_Update(Skein_512_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int  Skein1024_Update(Skein1024_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+
+int  Skein_256_Final (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein_512_Final (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein1024_Final (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+/*
+**   Skein APIs for "extended" initialization: MAC keys, tree hashing.
+**   After an InitExt() call, just use Update/Final calls as with Init().
+**
+**   Notes: Same parameters as _Init() calls, plus treeInfo/key/keyBytes.
+**          When keyBytes == 0 and treeInfo == SKEIN_SEQUENTIAL, 
+**              the results of InitExt() are identical to calling Init().
+**          The function Init() may be called once to "precompute" the IV for
+**              a given hashBitLen value, then by saving a copy of the context
+**              the IV computation may be avoided in later calls.
+**          Similarly, the function InitExt() may be called once per MAC key 
+**              to precompute the MAC IV, then a copy of the context saved and
+**              reused for each new MAC computation.
+**/
+int  Skein_256_InitExt(Skein_256_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int  Skein_512_InitExt(Skein_512_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int  Skein1024_InitExt(Skein1024_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+
+/*
+**   Skein APIs for MAC and tree hash:
+**      Final_Pad:  pad, do final block, but no OUTPUT type
+**      Output:     do just the output stage
+*/
+int  Skein_256_Final_Pad(Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein_512_Final_Pad(Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+#ifndef SKEIN_TREE_HASH
+#define SKEIN_TREE_HASH (1)
+#endif
+#if  SKEIN_TREE_HASH
+int  Skein_256_Output   (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein_512_Output   (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int  Skein1024_Output   (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+#endif
+
+/*****************************************************************
+** "Internal" Skein definitions
+**    -- not needed for sequential hashing API, but will be 
+**           helpful for other uses of Skein (e.g., tree hash mode).
+**    -- included here so that they can be shared between
+**           reference and optimized code.
+******************************************************************/
+
+/* tweak word T[1]: bit field starting positions */
+#define SKEIN_T1_BIT(BIT)       ((BIT) - 64)            /* offset 64 because it's the second word  */
+                                
+#define SKEIN_T1_POS_TREE_LVL   SKEIN_T1_BIT(112)       /* bits 112..118: level in hash tree       */
+#define SKEIN_T1_POS_BIT_PAD    SKEIN_T1_BIT(119)       /* bit  119     : partial final input byte */
+#define SKEIN_T1_POS_BLK_TYPE   SKEIN_T1_BIT(120)       /* bits 120..125: type field               */
+#define SKEIN_T1_POS_FIRST      SKEIN_T1_BIT(126)       /* bits 126     : first block flag         */
+#define SKEIN_T1_POS_FINAL      SKEIN_T1_BIT(127)       /* bit  127     : final block flag         */
+                                
+/* tweak word T[1]: flag bit definition(s) */
+#define SKEIN_T1_FLAG_FIRST     (((u64b_t)  1 ) << SKEIN_T1_POS_FIRST)
+#define SKEIN_T1_FLAG_FINAL     (((u64b_t)  1 ) << SKEIN_T1_POS_FINAL)
+#define SKEIN_T1_FLAG_BIT_PAD   (((u64b_t)  1 ) << SKEIN_T1_POS_BIT_PAD)
+                                
+/* tweak word T[1]: tree level bit field mask */
+#define SKEIN_T1_TREE_LVL_MASK  (((u64b_t)0x7F) << SKEIN_T1_POS_TREE_LVL)
+#define SKEIN_T1_TREE_LEVEL(n)  (((u64b_t) (n)) << SKEIN_T1_POS_TREE_LVL)
+
+/* tweak word T[1]: block type field */
+#define SKEIN_BLK_TYPE_KEY      ( 0)                    /* key, for MAC and KDF */
+#define SKEIN_BLK_TYPE_CFG      ( 4)                    /* configuration block */
+#define SKEIN_BLK_TYPE_PERS     ( 8)                    /* personalization string */
+#define SKEIN_BLK_TYPE_PK       (12)                    /* public key (for digital signature hashing) */
+#define SKEIN_BLK_TYPE_KDF      (16)                    /* key identifier for KDF */
+#define SKEIN_BLK_TYPE_NONCE    (20)                    /* nonce for PRNG */
+#define SKEIN_BLK_TYPE_MSG      (48)                    /* message processing */
+#define SKEIN_BLK_TYPE_OUT      (63)                    /* output stage */
+#define SKEIN_BLK_TYPE_MASK     (63)                    /* bit field mask */
+
+#define SKEIN_T1_BLK_TYPE(T)   (((u64b_t) (SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
+#define SKEIN_T1_BLK_TYPE_KEY   SKEIN_T1_BLK_TYPE(KEY)  /* key, for MAC and KDF */
+#define SKEIN_T1_BLK_TYPE_CFG   SKEIN_T1_BLK_TYPE(CFG)  /* configuration block */
+#define SKEIN_T1_BLK_TYPE_PERS  SKEIN_T1_BLK_TYPE(PERS) /* personalization string */
+#define SKEIN_T1_BLK_TYPE_PK    SKEIN_T1_BLK_TYPE(PK)   /* public key (for digital signature hashing) */
+#define SKEIN_T1_BLK_TYPE_KDF   SKEIN_T1_BLK_TYPE(KDF)  /* key identifier for KDF */
+#define SKEIN_T1_BLK_TYPE_NONCE SKEIN_T1_BLK_TYPE(NONCE)/* nonce for PRNG */
+#define SKEIN_T1_BLK_TYPE_MSG   SKEIN_T1_BLK_TYPE(MSG)  /* message processing */
+#define SKEIN_T1_BLK_TYPE_OUT   SKEIN_T1_BLK_TYPE(OUT)  /* output stage */
+#define SKEIN_T1_BLK_TYPE_MASK  SKEIN_T1_BLK_TYPE(MASK) /* field bit mask */
+
+#define SKEIN_T1_BLK_TYPE_CFG_FINAL       (SKEIN_T1_BLK_TYPE_CFG | SKEIN_T1_FLAG_FINAL)
+#define SKEIN_T1_BLK_TYPE_OUT_FINAL       (SKEIN_T1_BLK_TYPE_OUT | SKEIN_T1_FLAG_FINAL)
+
+#define SKEIN_VERSION           (1)
+
+#ifndef SKEIN_ID_STRING_LE      /* allow compile-time personalization */
+#define SKEIN_ID_STRING_LE      (0x33414853)            /* "SHA3" (little-endian)*/
+#endif
+
+#define SKEIN_MK_64(hi32,lo32)  ((lo32) + (((u64b_t) (hi32)) << 32))
+#define SKEIN_SCHEMA_VER        SKEIN_MK_64(SKEIN_VERSION,SKEIN_ID_STRING_LE)
+#define SKEIN_KS_PARITY         SKEIN_MK_64(0x1BD11BDA,0xA9FC1A22)
+
+#define SKEIN_CFG_STR_LEN       (4*8)
+
+/* bit field definitions in config block treeInfo word */
+#define SKEIN_CFG_TREE_LEAF_SIZE_POS  ( 0)
+#define SKEIN_CFG_TREE_NODE_SIZE_POS  ( 8)
+#define SKEIN_CFG_TREE_MAX_LEVEL_POS  (16)
+
+#define SKEIN_CFG_TREE_LEAF_SIZE_MSK  (((u64b_t) 0xFF) << SKEIN_CFG_TREE_LEAF_SIZE_POS)
+#define SKEIN_CFG_TREE_NODE_SIZE_MSK  (((u64b_t) 0xFF) << SKEIN_CFG_TREE_NODE_SIZE_POS)
+#define SKEIN_CFG_TREE_MAX_LEVEL_MSK  (((u64b_t) 0xFF) << SKEIN_CFG_TREE_MAX_LEVEL_POS)
+
+#define SKEIN_CFG_TREE_INFO(leaf,node,maxLvl)                   \
+    ( (((u64b_t)(leaf  )) << SKEIN_CFG_TREE_LEAF_SIZE_POS) |    \
+      (((u64b_t)(node  )) << SKEIN_CFG_TREE_NODE_SIZE_POS) |    \
+      (((u64b_t)(maxLvl)) << SKEIN_CFG_TREE_MAX_LEVEL_POS) )
+
+#define SKEIN_CFG_TREE_INFO_SEQUENTIAL SKEIN_CFG_TREE_INFO(0,0,0) /* use as treeInfo in InitExt() call for sequential processing */
+
+/*
+**   Skein macros for getting/setting tweak words, etc.
+**   These are useful for partial input bytes, hash tree init/update, etc.
+**/
+#define Skein_Get_Tweak(ctxPtr,TWK_NUM)         ((ctxPtr)->h.T[TWK_NUM])
+#define Skein_Set_Tweak(ctxPtr,TWK_NUM,tVal)    {(ctxPtr)->h.T[TWK_NUM] = (tVal);}
+
+#define Skein_Get_T0(ctxPtr)    Skein_Get_Tweak(ctxPtr,0)
+#define Skein_Get_T1(ctxPtr)    Skein_Get_Tweak(ctxPtr,1)
+#define Skein_Set_T0(ctxPtr,T0) Skein_Set_Tweak(ctxPtr,0,T0)
+#define Skein_Set_T1(ctxPtr,T1) Skein_Set_Tweak(ctxPtr,1,T1)
+
+/* set both tweak words at once */
+#define Skein_Set_T0_T1(ctxPtr,T0,T1)           \
+    {                                           \
+    Skein_Set_T0(ctxPtr,(T0));                  \
+    Skein_Set_T1(ctxPtr,(T1));                  \
+    }
+
+#define Skein_Set_Type(ctxPtr,BLK_TYPE)         \
+    Skein_Set_T1(ctxPtr,SKEIN_T1_BLK_TYPE_##BLK_TYPE)
+
+/* set up for starting with a new type: h.T[0]=0; h.T[1] = NEW_TYPE; h.bCnt=0; */
+#define Skein_Start_New_Type(ctxPtr,BLK_TYPE)   \
+    { Skein_Set_T0_T1(ctxPtr,0,SKEIN_T1_FLAG_FIRST | SKEIN_T1_BLK_TYPE_##BLK_TYPE); (ctxPtr)->h.bCnt=0; }
+
+#define Skein_Clear_First_Flag(hdr)      { (hdr).T[1] &= ~SKEIN_T1_FLAG_FIRST;       }
+#define Skein_Set_Bit_Pad_Flag(hdr)      { (hdr).T[1] |=  SKEIN_T1_FLAG_BIT_PAD;     }
+
+#define Skein_Set_Tree_Level(hdr,height) { (hdr).T[1] |= SKEIN_T1_TREE_LEVEL(height);}
+
+/*****************************************************************
+** "Internal" Skein definitions for debugging and error checking
+******************************************************************/
+#ifdef  SKEIN_DEBUG             /* examine/display intermediate values? */
+#include "skein_debug.h"
+#else                           /* default is no callouts */
+#define Skein_Show_Block(bits,ctx,X,blkPtr,wPtr,ksEvenPtr,ksOddPtr)
+#define Skein_Show_Round(bits,ctx,r,X)
+#define Skein_Show_R_Ptr(bits,ctx,r,X_ptr)
+#define Skein_Show_Final(bits,ctx,cnt,outPtr)
+#define Skein_Show_Key(bits,ctx,key,keyBytes)
+#endif
+
+#ifndef SKEIN_ERR_CHECK        /* run-time checks (e.g., bad params, uninitialized context)? */
+#define Skein_Assert(x,retCode)/* default: ignore all Asserts, for performance */
+#define Skein_assert(x)
+#elif   defined(SKEIN_ASSERT)
+#include <assert.h>     
+#define Skein_Assert(x,retCode) assert(x) 
+#define Skein_assert(x)         assert(x) 
+#else
+#include <assert.h>     
+#define Skein_Assert(x,retCode) { if (!(x)) return retCode; } /*  caller  error */
+#define Skein_assert(x)         assert(x)                     /* internal error */
+#endif
+
+/*****************************************************************
+** Skein block function constants (shared across Ref and Opt code)
+******************************************************************/
+enum    
+    {   
+        /* Skein_256 round rotation constants */
+    R_256_0_0=14, R_256_0_1=16,
+    R_256_1_0=52, R_256_1_1=57,
+    R_256_2_0=23, R_256_2_1=40,
+    R_256_3_0= 5, R_256_3_1=37,
+    R_256_4_0=25, R_256_4_1=33,
+    R_256_5_0=46, R_256_5_1=12,
+    R_256_6_0=58, R_256_6_1=22,
+    R_256_7_0=32, R_256_7_1=32,
+
+        /* Skein_512 round rotation constants */
+    R_512_0_0=46, R_512_0_1=36, R_512_0_2=19, R_512_0_3=37,
+    R_512_1_0=33, R_512_1_1=27, R_512_1_2=14, R_512_1_3=42,
+    R_512_2_0=17, R_512_2_1=49, R_512_2_2=36, R_512_2_3=39,
+    R_512_3_0=44, R_512_3_1= 9, R_512_3_2=54, R_512_3_3=56,
+    R_512_4_0=39, R_512_4_1=30, R_512_4_2=34, R_512_4_3=24,
+    R_512_5_0=13, R_512_5_1=50, R_512_5_2=10, R_512_5_3=17,
+    R_512_6_0=25, R_512_6_1=29, R_512_6_2=39, R_512_6_3=43,
+    R_512_7_0= 8, R_512_7_1=35, R_512_7_2=56, R_512_7_3=22,
+
+        /* Skein1024 round rotation constants */
+    R1024_0_0=24, R1024_0_1=13, R1024_0_2= 8, R1024_0_3=47, R1024_0_4= 8, R1024_0_5=17, R1024_0_6=22, R1024_0_7=37,
+    R1024_1_0=38, R1024_1_1=19, R1024_1_2=10, R1024_1_3=55, R1024_1_4=49, R1024_1_5=18, R1024_1_6=23, R1024_1_7=52,
+    R1024_2_0=33, R1024_2_1= 4, R1024_2_2=51, R1024_2_3=13, R1024_2_4=34, R1024_2_5=41, R1024_2_6=59, R1024_2_7=17,
+    R1024_3_0= 5, R1024_3_1=20, R1024_3_2=48, R1024_3_3=41, R1024_3_4=47, R1024_3_5=28, R1024_3_6=16, R1024_3_7=25,
+    R1024_4_0=41, R1024_4_1= 9, R1024_4_2=37, R1024_4_3=31, R1024_4_4=12, R1024_4_5=47, R1024_4_6=44, R1024_4_7=30,
+    R1024_5_0=16, R1024_5_1=34, R1024_5_2=56, R1024_5_3=51, R1024_5_4= 4, R1024_5_5=53, R1024_5_6=42, R1024_5_7=41,
+    R1024_6_0=31, R1024_6_1=44, R1024_6_2=47, R1024_6_3=46, R1024_6_4=19, R1024_6_5=42, R1024_6_6=44, R1024_6_7=25,
+    R1024_7_0= 9, R1024_7_1=48, R1024_7_2=35, R1024_7_3=52, R1024_7_4=23, R1024_7_5=31, R1024_7_6=37, R1024_7_7=20
+    };
+
+#ifndef SKEIN_ROUNDS
+#define SKEIN_256_ROUNDS_TOTAL (72)          /* number of rounds for the different block sizes */
+#define SKEIN_512_ROUNDS_TOTAL (72)
+#define SKEIN1024_ROUNDS_TOTAL (80)
+#else                                        /* allow command-line define in range 8*(5..14)   */
+#define SKEIN_256_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/100) + 5) % 10) + 5))
+#define SKEIN_512_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/ 10) + 5) % 10) + 5))
+#define SKEIN1024_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS    ) + 5) % 10) + 5))
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* ifndef _SKEIN_H_ */
diff --git a/Optimized_32bit/skein_block.c b/Optimized_32bit/skein_block.c
new file mode 100644
index 000000000000..bfd29d1eee2d
--- /dev/null
+++ b/Optimized_32bit/skein_block.c
@@ -0,0 +1,689 @@
+/***********************************************************************
+**
+** Implementation of the Skein block functions.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+** Compile-time switches:
+**
+**  SKEIN_USE_ASM  -- set bits (256/512/1024) to select which
+**                    versions use ASM code for block processing
+**                    [default: use C for all block sizes]
+**
+************************************************************************/
+
+#include <string.h>
+#include "skein.h"
+
+#ifndef SKEIN_USE_ASM
+#define SKEIN_USE_ASM   (0)                     /* default is all C code (no ASM) */
+#endif
+
+#ifndef SKEIN_LOOP
+#define SKEIN_LOOP 001                          /* default: unroll 256 and 512, but not 1024 */
+#endif
+
+#define BLK_BITS        (WCNT*64)               /* some useful definitions for code here */
+#define KW_TWK_BASE     (0)
+#define KW_KEY_BASE     (3)
+#define ks              (kw + KW_KEY_BASE)                
+#define ts              (kw + KW_TWK_BASE)
+
+#ifdef SKEIN_DEBUG
+#define DebugSaveTweak(ctx) { ctx->h.T[0] = ts[0]; ctx->h.T[1] = ts[1]; }
+#else
+#define DebugSaveTweak(ctx)
+#endif
+
+/*****************************  Skein_256 ******************************/
+#if !(SKEIN_USE_ASM & 256)
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd)
+    { /* do it in C */
+    enum
+        {
+        WCNT = SKEIN_256_STATE_WORDS
+        };
+#undef  RCNT
+#define RCNT  (SKEIN_256_ROUNDS_TOTAL/8)
+
+#ifdef  SKEIN_LOOP                              /* configure how much to unroll the loop */
+#define SKEIN_UNROLL_256 (((SKEIN_LOOP)/100)%10)
+#else
+#define SKEIN_UNROLL_256 (0)
+#endif
+
+#if SKEIN_UNROLL_256
+#if (RCNT % SKEIN_UNROLL_256)
+#error "Invalid SKEIN_UNROLL_256"               /* sanity check on unroll count */
+#endif
+    size_t  r;
+    u64b_t  kw[WCNT+4+RCNT*2];                  /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+    u64b_t  kw[WCNT+4];                         /* key schedule words : chaining vars + tweak */
+#endif
+    u64b_t  X0,X1,X2,X3;                        /* local copy of context vars, for speed */
+    u64b_t  w [WCNT];                           /* local copy of input block */
+#ifdef SKEIN_DEBUG
+    const u64b_t *Xptr[4];                      /* use for debugging (help compiler put Xn in registers) */
+    Xptr[0] = &X0;  Xptr[1] = &X1;  Xptr[2] = &X2;  Xptr[3] = &X3;
+#endif
+    Skein_assert(blkCnt != 0);                  /* never call with blkCnt == 0! */
+    ts[0] = ctx->h.T[0];
+    ts[1] = ctx->h.T[1];
+    do  {
+        /* this implementation only supports 2**64 input bytes (no carry out here) */
+        ts[0] += byteCntAdd;                    /* update processed length */
+
+        /* precompute the key schedule for this block */
+        ks[0] = ctx->X[0];     
+        ks[1] = ctx->X[1];
+        ks[2] = ctx->X[2];
+        ks[3] = ctx->X[3];
+        ks[4] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^ SKEIN_KS_PARITY;
+
+        ts[2] = ts[0] ^ ts[1];
+
+        Skein_Get64_LSB_First(w,blkPtr,WCNT);   /* get input block in little-endian format */
+        DebugSaveTweak(ctx);
+        Skein_Show_Block(BLK_BITS,&ctx->h,ctx->X,blkPtr,w,ks,ts);
+
+        X0 = w[0] + ks[0];                      /* do the first full key injection */
+        X1 = w[1] + ks[1] + ts[0];
+        X2 = w[2] + ks[2] + ts[1];
+        X3 = w[3] + ks[3];
+
+        Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INITIAL,Xptr);    /* show starting state values */
+
+        blkPtr += SKEIN_256_BLOCK_BYTES;
+
+        /* run the rounds */
+
+#define Round256(p0,p1,p2,p3,ROT,rNum)                              \
+    X##p0 += X##p1; X##p1 = RotL_64(X##p1,ROT##_0); X##p1 ^= X##p0; \
+    X##p2 += X##p3; X##p3 = RotL_64(X##p3,ROT##_1); X##p3 ^= X##p2; \
+
+#if SKEIN_UNROLL_256 == 0                       
+#define R256(p0,p1,p2,p3,ROT,rNum)           /* fully unrolled */   \
+    Round256(p0,p1,p2,p3,ROT,rNum)                                  \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,rNum,Xptr);
+
+#define I256(R)                                                     \
+    X0   += ks[((R)+1) % 5];    /* inject the key schedule value */ \
+    X1   += ks[((R)+2) % 5] + ts[((R)+1) % 3];                      \
+    X2   += ks[((R)+3) % 5] + ts[((R)+2) % 3];                      \
+    X3   += ks[((R)+4) % 5] +     (R)+1;                            \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr);
+#else                                       /* looping version */
+#define R256(p0,p1,p2,p3,ROT,rNum)                                  \
+    Round256(p0,p1,p2,p3,ROT,rNum)                                  \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,4*(r-1)+rNum,Xptr);
+
+#define I256(R)                                                     \
+    X0   += ks[r+(R)+0];        /* inject the key schedule value */ \
+    X1   += ks[r+(R)+1] + ts[r+(R)+0];                              \
+    X2   += ks[r+(R)+2] + ts[r+(R)+1];                              \
+    X3   += ks[r+(R)+3] +    r+(R)   ;                              \
+    ks[r + (R)+4    ]   = ks[r+(R)-1];     /* rotate key schedule */\
+    ts[r + (R)+2    ]   = ts[r+(R)-1];                              \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr);
+
+    for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_256)  /* loop thru it */
+#endif  
+        {    
+#define R256_8_rounds(R)                  \
+        R256(0,1,2,3,R_256_0,8*(R) + 1);  \
+        R256(0,3,2,1,R_256_1,8*(R) + 2);  \
+        R256(0,1,2,3,R_256_2,8*(R) + 3);  \
+        R256(0,3,2,1,R_256_3,8*(R) + 4);  \
+        I256(2*(R));                      \
+        R256(0,1,2,3,R_256_4,8*(R) + 5);  \
+        R256(0,3,2,1,R_256_5,8*(R) + 6);  \
+        R256(0,1,2,3,R_256_6,8*(R) + 7);  \
+        R256(0,3,2,1,R_256_7,8*(R) + 8);  \
+        I256(2*(R)+1);
+
+        R256_8_rounds( 0);
+
+#define R256_Unroll_R(NN) ((SKEIN_UNROLL_256 == 0 && SKEIN_256_ROUNDS_TOTAL/8 > (NN)) || (SKEIN_UNROLL_256 > (NN)))
+
+  #if   R256_Unroll_R( 1)
+        R256_8_rounds( 1);
+  #endif
+  #if   R256_Unroll_R( 2)
+        R256_8_rounds( 2);
+  #endif
+  #if   R256_Unroll_R( 3)
+        R256_8_rounds( 3);
+  #endif
+  #if   R256_Unroll_R( 4)
+        R256_8_rounds( 4);
+  #endif
+  #if   R256_Unroll_R( 5)
+        R256_8_rounds( 5);
+  #endif
+  #if   R256_Unroll_R( 6)
+        R256_8_rounds( 6);
+  #endif
+  #if   R256_Unroll_R( 7)
+        R256_8_rounds( 7);
+  #endif
+  #if   R256_Unroll_R( 8)
+        R256_8_rounds( 8);
+  #endif
+  #if   R256_Unroll_R( 9)
+        R256_8_rounds( 9);
+  #endif
+  #if   R256_Unroll_R(10)
+        R256_8_rounds(10);
+  #endif
+  #if   R256_Unroll_R(11)
+        R256_8_rounds(11);
+  #endif
+  #if   R256_Unroll_R(12)
+        R256_8_rounds(12);
+  #endif
+  #if   R256_Unroll_R(13)
+        R256_8_rounds(13);
+  #endif
+  #if   R256_Unroll_R(14)
+        R256_8_rounds(14);
+  #endif
+  #if  (SKEIN_UNROLL_256 > 14)
+#error  "need more unrolling in Skein_256_Process_Block"
+  #endif
+        }
+        /* do the final "feedforward" xor, update context chaining vars */
+        ctx->X[0] = X0 ^ w[0];
+        ctx->X[1] = X1 ^ w[1];
+        ctx->X[2] = X2 ^ w[2];
+        ctx->X[3] = X3 ^ w[3];
+
+        Skein_Show_Round(BLK_BITS,&ctx->h,SKEIN_RND_FEED_FWD,ctx->X);
+
+        ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+        }
+    while (--blkCnt);
+    ctx->h.T[0] = ts[0];
+    ctx->h.T[1] = ts[1];
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_256_Process_Block_CodeSize(void)
+    {
+    return ((u08b_t *) Skein_256_Process_Block_CodeSize) -
+           ((u08b_t *) Skein_256_Process_Block);
+    }
+uint_t Skein_256_Unroll_Cnt(void)
+    {
+    return SKEIN_UNROLL_256;
+    }
+#endif
+#endif
+
+/*****************************  Skein_512 ******************************/
+#if !(SKEIN_USE_ASM & 512)
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd)
+    { /* do it in C */
+    enum
+        {
+        WCNT = SKEIN_512_STATE_WORDS
+        };
+#undef  RCNT
+#define RCNT  (SKEIN_512_ROUNDS_TOTAL/8)
+
+#ifdef  SKEIN_LOOP                              /* configure how much to unroll the loop */
+#define SKEIN_UNROLL_512 (((SKEIN_LOOP)/10)%10)
+#else
+#define SKEIN_UNROLL_512 (0)
+#endif
+
+#if SKEIN_UNROLL_512
+#if (RCNT % SKEIN_UNROLL_512)
+#error "Invalid SKEIN_UNROLL_512"               /* sanity check on unroll count */
+#endif
+    size_t  r;
+    u64b_t  kw[WCNT+4+RCNT*2];                  /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+    u64b_t  kw[WCNT+4];                         /* key schedule words : chaining vars + tweak */
+#endif
+    u64b_t  X0,X1,X2,X3,X4,X5,X6,X7;            /* local copy of vars, for speed */
+    u64b_t  w [WCNT];                           /* local copy of input block */
+#ifdef SKEIN_DEBUG
+    const u64b_t *Xptr[8];                      /* use for debugging (help compiler put Xn in registers) */
+    Xptr[0] = &X0;  Xptr[1] = &X1;  Xptr[2] = &X2;  Xptr[3] = &X3;
+    Xptr[4] = &X4;  Xptr[5] = &X5;  Xptr[6] = &X6;  Xptr[7] = &X7;
+#endif
+
+    Skein_assert(blkCnt != 0);                  /* never call with blkCnt == 0! */
+    ts[0] = ctx->h.T[0];
+    ts[1] = ctx->h.T[1];
+    do  {
+        /* this implementation only supports 2**64 input bytes (no carry out here) */
+        ts[0] += byteCntAdd;                    /* update processed length */
+
+        /* precompute the key schedule for this block */
+        ks[0] = ctx->X[0];
+        ks[1] = ctx->X[1];
+        ks[2] = ctx->X[2];
+        ks[3] = ctx->X[3];
+        ks[4] = ctx->X[4];
+        ks[5] = ctx->X[5];
+        ks[6] = ctx->X[6];
+        ks[7] = ctx->X[7];
+        ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^ 
+                ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;
+
+        ts[2] = ts[0] ^ ts[1];
+
+        Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+        DebugSaveTweak(ctx);
+        Skein_Show_Block(BLK_BITS,&ctx->h,ctx->X,blkPtr,w,ks,ts);
+
+        X0   = w[0] + ks[0];                    /* do the first full key injection */
+        X1   = w[1] + ks[1];
+        X2   = w[2] + ks[2];
+        X3   = w[3] + ks[3];
+        X4   = w[4] + ks[4];
+        X5   = w[5] + ks[5] + ts[0];
+        X6   = w[6] + ks[6] + ts[1];
+        X7   = w[7] + ks[7];
+
+        blkPtr += SKEIN_512_BLOCK_BYTES;
+
+        Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INITIAL,Xptr);
+        /* run the rounds */
+#define Round512(p0,p1,p2,p3,p4,p5,p6,p7,ROT,rNum)                  \
+    X##p0 += X##p1; X##p1 = RotL_64(X##p1,ROT##_0); X##p1 ^= X##p0; \
+    X##p2 += X##p3; X##p3 = RotL_64(X##p3,ROT##_1); X##p3 ^= X##p2; \
+    X##p4 += X##p5; X##p5 = RotL_64(X##p5,ROT##_2); X##p5 ^= X##p4; \
+    X##p6 += X##p7; X##p7 = RotL_64(X##p7,ROT##_3); X##p7 ^= X##p6; \
+
+#if SKEIN_UNROLL_512 == 0                       
+#define R512(p0,p1,p2,p3,p4,p5,p6,p7,ROT,rNum)      /* unrolled */  \
+    Round512(p0,p1,p2,p3,p4,p5,p6,p7,ROT,rNum)                      \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,rNum,Xptr);
+
+#define I512(R)                                                     \
+    X0   += ks[((R)+1) % 9];   /* inject the key schedule value */  \
+    X1   += ks[((R)+2) % 9];                                        \
+    X2   += ks[((R)+3) % 9];                                        \
+    X3   += ks[((R)+4) % 9];                                        \
+    X4   += ks[((R)+5) % 9];                                        \
+    X5   += ks[((R)+6) % 9] + ts[((R)+1) % 3];                      \
+    X6   += ks[((R)+7) % 9] + ts[((R)+2) % 3];                      \
+    X7   += ks[((R)+8) % 9] +     (R)+1;                            \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr);
+#else                                       /* looping version */
+#define R512(p0,p1,p2,p3,p4,p5,p6,p7,ROT,rNum)                      \
+    Round512(p0,p1,p2,p3,p4,p5,p6,p7,ROT,rNum)                      \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,4*(r-1)+rNum,Xptr);
+
+#define I512(R)                                                     \
+    X0   += ks[r+(R)+0];        /* inject the key schedule value */ \
+    X1   += ks[r+(R)+1];                                            \
+    X2   += ks[r+(R)+2];                                            \
+    X3   += ks[r+(R)+3];                                            \
+    X4   += ks[r+(R)+4];                                            \
+    X5   += ks[r+(R)+5] + ts[r+(R)+0];                              \
+    X6   += ks[r+(R)+6] + ts[r+(R)+1];                              \
+    X7   += ks[r+(R)+7] +    r+(R)   ;                              \
+    ks[r +       (R)+8] = ks[r+(R)-1];  /* rotate key schedule */   \
+    ts[r +       (R)+2] = ts[r+(R)-1];                              \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr);
+
+    for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_512)   /* loop thru it */
+#endif                         /* end of looped code definitions */
+        {
+#define R512_8_rounds(R)  /* do 8 full rounds */  \
+        R512(0,1,2,3,4,5,6,7,R_512_0,8*(R)+ 1);   \
+        R512(2,1,4,7,6,5,0,3,R_512_1,8*(R)+ 2);   \
+        R512(4,1,6,3,0,5,2,7,R_512_2,8*(R)+ 3);   \
+        R512(6,1,0,7,2,5,4,3,R_512_3,8*(R)+ 4);   \
+        I512(2*(R));                              \
+        R512(0,1,2,3,4,5,6,7,R_512_4,8*(R)+ 5);   \
+        R512(2,1,4,7,6,5,0,3,R_512_5,8*(R)+ 6);   \
+        R512(4,1,6,3,0,5,2,7,R_512_6,8*(R)+ 7);   \
+        R512(6,1,0,7,2,5,4,3,R_512_7,8*(R)+ 8);   \
+        I512(2*(R)+1);        /* and key injection */
+
+        R512_8_rounds( 0);
+
+#define R512_Unroll_R(NN) ((SKEIN_UNROLL_512 == 0 && SKEIN_512_ROUNDS_TOTAL/8 > (NN)) || (SKEIN_UNROLL_512 > (NN)))
+
+  #if   R512_Unroll_R( 1)
+        R512_8_rounds( 1);
+  #endif
+  #if   R512_Unroll_R( 2)
+        R512_8_rounds( 2);
+  #endif
+  #if   R512_Unroll_R( 3)
+        R512_8_rounds( 3);
+  #endif
+  #if   R512_Unroll_R( 4)
+        R512_8_rounds( 4);
+  #endif
+  #if   R512_Unroll_R( 5)
+        R512_8_rounds( 5);
+  #endif
+  #if   R512_Unroll_R( 6)
+        R512_8_rounds( 6);
+  #endif
+  #if   R512_Unroll_R( 7)
+        R512_8_rounds( 7);
+  #endif
+  #if   R512_Unroll_R( 8)
+        R512_8_rounds( 8);
+  #endif
+  #if   R512_Unroll_R( 9)
+        R512_8_rounds( 9);
+  #endif
+  #if   R512_Unroll_R(10)
+        R512_8_rounds(10);
+  #endif
+  #if   R512_Unroll_R(11)
+        R512_8_rounds(11);
+  #endif
+  #if   R512_Unroll_R(12)
+        R512_8_rounds(12);
+  #endif
+  #if   R512_Unroll_R(13)
+        R512_8_rounds(13);
+  #endif
+  #if   R512_Unroll_R(14)
+        R512_8_rounds(14);
+  #endif
+  #if  (SKEIN_UNROLL_512 > 14)
+#error  "need more unrolling in Skein_512_Process_Block"
+  #endif
+        }
+
+        /* do the final "feedforward" xor, update context chaining vars */
+        ctx->X[0] = X0 ^ w[0];
+        ctx->X[1] = X1 ^ w[1];
+        ctx->X[2] = X2 ^ w[2];
+        ctx->X[3] = X3 ^ w[3];
+        ctx->X[4] = X4 ^ w[4];
+        ctx->X[5] = X5 ^ w[5];
+        ctx->X[6] = X6 ^ w[6];
+        ctx->X[7] = X7 ^ w[7];
+        Skein_Show_Round(BLK_BITS,&ctx->h,SKEIN_RND_FEED_FWD,ctx->X);
+
+        ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+        }
+    while (--blkCnt);
+    ctx->h.T[0] = ts[0];
+    ctx->h.T[1] = ts[1];
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_512_Process_Block_CodeSize(void)
+    {
+    return ((u08b_t *) Skein_512_Process_Block_CodeSize) -
+           ((u08b_t *) Skein_512_Process_Block);
+    }
+uint_t Skein_512_Unroll_Cnt(void)
+    {
+    return SKEIN_UNROLL_512;
+    }
+#endif
+#endif
+
+/*****************************  Skein1024 ******************************/
+#if !(SKEIN_USE_ASM & 1024)
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd)
+    { /* do it in C, always looping (unrolled is bigger AND slower!) */
+    enum
+        {
+        WCNT = SKEIN1024_STATE_WORDS
+        };
+#undef  RCNT
+#define RCNT  (SKEIN1024_ROUNDS_TOTAL/8)
+
+#ifdef  SKEIN_LOOP                              /* configure how much to unroll the loop */
+#define SKEIN_UNROLL_1024 ((SKEIN_LOOP)%10)
+#else
+#define SKEIN_UNROLL_1024 (0)
+#endif
+
+#if (SKEIN_UNROLL_1024 != 0)
+#if (RCNT % SKEIN_UNROLL_1024)
+#error "Invalid SKEIN_UNROLL_1024"              /* sanity check on unroll count */
+#endif
+    size_t  r;
+    u64b_t  kw[WCNT+4+RCNT*2];                  /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+    u64b_t  kw[WCNT+4];                         /* key schedule words : chaining vars + tweak */
+#endif
+
+    u64b_t  X00,X01,X02,X03,X04,X05,X06,X07,    /* local copy of vars, for speed */
+            X08,X09,X10,X11,X12,X13,X14,X15;
+    u64b_t  w [WCNT];                           /* local copy of input block */
+#ifdef SKEIN_DEBUG
+    const u64b_t *Xptr[16];                     /* use for debugging (help compiler put Xn in registers) */
+    Xptr[ 0] = &X00;  Xptr[ 1] = &X01;  Xptr[ 2] = &X02;  Xptr[ 3] = &X03;
+    Xptr[ 4] = &X04;  Xptr[ 5] = &X05;  Xptr[ 6] = &X06;  Xptr[ 7] = &X07;
+    Xptr[ 8] = &X08;  Xptr[ 9] = &X09;  Xptr[10] = &X10;  Xptr[11] = &X11;
+    Xptr[12] = &X12;  Xptr[13] = &X13;  Xptr[14] = &X14;  Xptr[15] = &X15;
+#endif
+
+    Skein_assert(blkCnt != 0);                  /* never call with blkCnt == 0! */
+    ts[0] = ctx->h.T[0];
+    ts[1] = ctx->h.T[1];
+    do  {
+        /* this implementation only supports 2**64 input bytes (no carry out here) */
+        ts[0] += byteCntAdd;                    /* update processed length */
+
+        /* precompute the key schedule for this block */
+        ks[ 0] = ctx->X[ 0];
+        ks[ 1] = ctx->X[ 1];
+        ks[ 2] = ctx->X[ 2];
+        ks[ 3] = ctx->X[ 3];
+        ks[ 4] = ctx->X[ 4];
+        ks[ 5] = ctx->X[ 5];
+        ks[ 6] = ctx->X[ 6];
+        ks[ 7] = ctx->X[ 7];
+        ks[ 8] = ctx->X[ 8];
+        ks[ 9] = ctx->X[ 9];
+        ks[10] = ctx->X[10];
+        ks[11] = ctx->X[11];
+        ks[12] = ctx->X[12];
+        ks[13] = ctx->X[13];
+        ks[14] = ctx->X[14];
+        ks[15] = ctx->X[15];
+        ks[16] = ks[ 0] ^ ks[ 1] ^ ks[ 2] ^ ks[ 3] ^
+                 ks[ 4] ^ ks[ 5] ^ ks[ 6] ^ ks[ 7] ^
+                 ks[ 8] ^ ks[ 9] ^ ks[10] ^ ks[11] ^
+                 ks[12] ^ ks[13] ^ ks[14] ^ ks[15] ^ SKEIN_KS_PARITY;
+
+        ts[2]  = ts[0] ^ ts[1];
+
+        Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+        DebugSaveTweak(ctx);
+        Skein_Show_Block(BLK_BITS,&ctx->h,ctx->X,blkPtr,w,ks,ts);
+
+        X00    = w[ 0] + ks[ 0];                 /* do the first full key injection */
+        X01    = w[ 1] + ks[ 1];
+        X02    = w[ 2] + ks[ 2];
+        X03    = w[ 3] + ks[ 3];
+        X04    = w[ 4] + ks[ 4];
+        X05    = w[ 5] + ks[ 5];
+        X06    = w[ 6] + ks[ 6];
+        X07    = w[ 7] + ks[ 7];
+        X08    = w[ 8] + ks[ 8];
+        X09    = w[ 9] + ks[ 9];
+        X10    = w[10] + ks[10];
+        X11    = w[11] + ks[11];
+        X12    = w[12] + ks[12];
+        X13    = w[13] + ks[13] + ts[0];
+        X14    = w[14] + ks[14] + ts[1];
+        X15    = w[15] + ks[15];
+
+        Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INITIAL,Xptr);
+
+#define Round1024(p0,p1,p2,p3,p4,p5,p6,p7,p8,p9,pA,pB,pC,pD,pE,pF,ROT,rNum) \
+    X##p0 += X##p1; X##p1 = RotL_64(X##p1,ROT##_0); X##p1 ^= X##p0;   \
+    X##p2 += X##p3; X##p3 = RotL_64(X##p3,ROT##_1); X##p3 ^= X##p2;   \
+    X##p4 += X##p5; X##p5 = RotL_64(X##p5,ROT##_2); X##p5 ^= X##p4;   \
+    X##p6 += X##p7; X##p7 = RotL_64(X##p7,ROT##_3); X##p7 ^= X##p6;   \
+    X##p8 += X##p9; X##p9 = RotL_64(X##p9,ROT##_4); X##p9 ^= X##p8;   \
+    X##pA += X##pB; X##pB = RotL_64(X##pB,ROT##_5); X##pB ^= X##pA;   \
+    X##pC += X##pD; X##pD = RotL_64(X##pD,ROT##_6); X##pD ^= X##pC;   \
+    X##pE += X##pF; X##pF = RotL_64(X##pF,ROT##_7); X##pF ^= X##pE;   \
+
+#if SKEIN_UNROLL_1024 == 0                      
+#define R1024(p0,p1,p2,p3,p4,p5,p6,p7,p8,p9,pA,pB,pC,pD,pE,pF,ROT,rn) \
+    Round1024(p0,p1,p2,p3,p4,p5,p6,p7,p8,p9,pA,pB,pC,pD,pE,pF,ROT,rn) \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,rn,Xptr);
+
+#define I1024(R)                                                      \
+    X00   += ks[((R)+ 1) % 17]; /* inject the key schedule value */   \
+    X01   += ks[((R)+ 2) % 17];                                       \
+    X02   += ks[((R)+ 3) % 17];                                       \
+    X03   += ks[((R)+ 4) % 17];                                       \
+    X04   += ks[((R)+ 5) % 17];                                       \
+    X05   += ks[((R)+ 6) % 17];                                       \
+    X06   += ks[((R)+ 7) % 17];                                       \
+    X07   += ks[((R)+ 8) % 17];                                       \
+    X08   += ks[((R)+ 9) % 17];                                       \
+    X09   += ks[((R)+10) % 17];                                       \
+    X10   += ks[((R)+11) % 17];                                       \
+    X11   += ks[((R)+12) % 17];                                       \
+    X12   += ks[((R)+13) % 17];                                       \
+    X13   += ks[((R)+14) % 17] + ts[((R)+1) % 3];                     \
+    X14   += ks[((R)+15) % 17] + ts[((R)+2) % 3];                     \
+    X15   += ks[((R)+16) % 17] +     (R)+1;                           \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr); 
+#else                                       /* looping version */
+#define R1024(p0,p1,p2,p3,p4,p5,p6,p7,p8,p9,pA,pB,pC,pD,pE,pF,ROT,rn) \
+    Round1024(p0,p1,p2,p3,p4,p5,p6,p7,p8,p9,pA,pB,pC,pD,pE,pF,ROT,rn) \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,4*(r-1)+rn,Xptr);
+
+#define I1024(R)                                                      \
+    X00   += ks[r+(R)+ 0];    /* inject the key schedule value */     \
+    X01   += ks[r+(R)+ 1];                                            \
+    X02   += ks[r+(R)+ 2];                                            \
+    X03   += ks[r+(R)+ 3];                                            \
+    X04   += ks[r+(R)+ 4];                                            \
+    X05   += ks[r+(R)+ 5];                                            \
+    X06   += ks[r+(R)+ 6];                                            \
+    X07   += ks[r+(R)+ 7];                                            \
+    X08   += ks[r+(R)+ 8];                                            \
+    X09   += ks[r+(R)+ 9];                                            \
+    X10   += ks[r+(R)+10];                                            \
+    X11   += ks[r+(R)+11];                                            \
+    X12   += ks[r+(R)+12];                                            \
+    X13   += ks[r+(R)+13] + ts[r+(R)+0];                              \
+    X14   += ks[r+(R)+14] + ts[r+(R)+1];                              \
+    X15   += ks[r+(R)+15] +    r+(R)   ;                              \
+    ks[r  +       (R)+16] = ks[r+(R)-1];  /* rotate key schedule */   \
+    ts[r  +       (R)+ 2] = ts[r+(R)-1];                              \
+    Skein_Show_R_Ptr(BLK_BITS,&ctx->h,SKEIN_RND_KEY_INJECT,Xptr);
+
+    for (r=1;r <= 2*RCNT;r+=2*SKEIN_UNROLL_1024)    /* loop thru it */
+#endif  
+        {
+#define R1024_8_rounds(R)    /* do 8 full rounds */                               \
+        R1024(00,01,02,03,04,05,06,07,08,09,10,11,12,13,14,15,R1024_0,8*(R) + 1); \
+        R1024(00,09,02,13,06,11,04,15,10,07,12,03,14,05,08,01,R1024_1,8*(R) + 2); \
+        R1024(00,07,02,05,04,03,06,01,12,15,14,13,08,11,10,09,R1024_2,8*(R) + 3); \
+        R1024(00,15,02,11,06,13,04,09,14,01,08,05,10,03,12,07,R1024_3,8*(R) + 4); \
+        I1024(2*(R));                                                             \
+        R1024(00,01,02,03,04,05,06,07,08,09,10,11,12,13,14,15,R1024_4,8*(R) + 5); \
+        R1024(00,09,02,13,06,11,04,15,10,07,12,03,14,05,08,01,R1024_5,8*(R) + 6); \
+        R1024(00,07,02,05,04,03,06,01,12,15,14,13,08,11,10,09,R1024_6,8*(R) + 7); \
+        R1024(00,15,02,11,06,13,04,09,14,01,08,05,10,03,12,07,R1024_7,8*(R) + 8); \
+        I1024(2*(R)+1);
+
+        R1024_8_rounds( 0);
+
+#define R1024_Unroll_R(NN) ((SKEIN_UNROLL_1024 == 0 && SKEIN1024_ROUNDS_TOTAL/8 > (NN)) || (SKEIN_UNROLL_1024 > (NN)))
+
+  #if   R1024_Unroll_R( 1)
+        R1024_8_rounds( 1);
+  #endif
+  #if   R1024_Unroll_R( 2)
+        R1024_8_rounds( 2);
+  #endif
+  #if   R1024_Unroll_R( 3)
+        R1024_8_rounds( 3);
+  #endif
+  #if   R1024_Unroll_R( 4)
+        R1024_8_rounds( 4);
+  #endif
+  #if   R1024_Unroll_R( 5)
+        R1024_8_rounds( 5);
+  #endif
+  #if   R1024_Unroll_R( 6)
+        R1024_8_rounds( 6);
+  #endif
+  #if   R1024_Unroll_R( 7)
+        R1024_8_rounds( 7);
+  #endif
+  #if   R1024_Unroll_R( 8)
+        R1024_8_rounds( 8);
+  #endif
+  #if   R1024_Unroll_R( 9)
+        R1024_8_rounds( 9);
+  #endif
+  #if   R1024_Unroll_R(10)
+        R1024_8_rounds(10);
+  #endif
+  #if   R1024_Unroll_R(11)
+        R1024_8_rounds(11);
+  #endif
+  #if   R1024_Unroll_R(12)
+        R1024_8_rounds(12);
+  #endif
+  #if   R1024_Unroll_R(13)
+        R1024_8_rounds(13);
+  #endif
+  #if   R1024_Unroll_R(14)
+        R1024_8_rounds(14);
+  #endif
+  #if  (SKEIN_UNROLL_1024 > 14)
+#error  "need more unrolling in Skein_1024_Process_Block"
+  #endif
+        }
+        /* do the final "feedforward" xor, update context chaining vars */
+
+        ctx->X[ 0] = X00 ^ w[ 0];
+        ctx->X[ 1] = X01 ^ w[ 1];
+        ctx->X[ 2] = X02 ^ w[ 2];
+        ctx->X[ 3] = X03 ^ w[ 3];
+        ctx->X[ 4] = X04 ^ w[ 4];
+        ctx->X[ 5] = X05 ^ w[ 5];
+        ctx->X[ 6] = X06 ^ w[ 6];
+        ctx->X[ 7] = X07 ^ w[ 7];
+        ctx->X[ 8] = X08 ^ w[ 8];
+        ctx->X[ 9] = X09 ^ w[ 9];
+        ctx->X[10] = X10 ^ w[10];
+        ctx->X[11] = X11 ^ w[11];
+        ctx->X[12] = X12 ^ w[12];
+        ctx->X[13] = X13 ^ w[13];
+        ctx->X[14] = X14 ^ w[14];
+        ctx->X[15] = X15 ^ w[15];
+
+        Skein_Show_Round(BLK_BITS,&ctx->h,SKEIN_RND_FEED_FWD,ctx->X);
+        
+        ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+        blkPtr += SKEIN1024_BLOCK_BYTES;
+        }
+    while (--blkCnt);
+    ctx->h.T[0] = ts[0];
+    ctx->h.T[1] = ts[1];
+    }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein1024_Process_Block_CodeSize(void)
+    {
+    return ((u08b_t *) Skein1024_Process_Block_CodeSize) -
+           ((u08b_t *) Skein1024_Process_Block);
+    }
+uint_t Skein1024_Unroll_Cnt(void)
+    {
+    return SKEIN_UNROLL_1024;
+    }
+#endif
+#endif
diff --git a/Optimized_32bit/skein_debug.c b/Optimized_32bit/skein_debug.c
new file mode 100644
index 000000000000..fac5038598ea
--- /dev/null
+++ b/Optimized_32bit/skein_debug.c
@@ -0,0 +1,247 @@
+/***********************************************************************
+**
+** Debug output functions for Skein hashing.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+#include <stdio.h>
+
+#ifdef SKEIN_DEBUG  /* only instantiate this code if SKEIN_DEBUG is on */
+#include "skein.h"
+
+static const char INDENT[] =  "    ";  /* how much to indent on new line */
+
+uint_t skein_DebugFlag = 0;  /* off by default. Must be set externally */
+
+static void Show64_step(size_t cnt,const u64b_t *X,size_t step)
+    {
+    size_t i,j;
+    for (i=j=0;i < cnt;i++,j+=step)
+        {
+        if (i % 4 ==  0) printf(INDENT);
+        printf(" %08X.%08X ",(uint_32t)(X[j] >> 32),(uint_32t)X[j]);
+        if (i % 4 ==  3 || i==cnt-1) printf("\n");
+        fflush(stdout);
+        }
+    }
+
+#define Show64(cnt,X) Show64_step(cnt,X,1)
+
+static void Show64_flag(size_t cnt,const u64b_t *X)
+    {
+    size_t xptr = (size_t) X;
+    size_t step = (xptr & 1) ? 2 : 1;
+    if (step != 1)
+        {
+        X = (const u64b_t *) (xptr & ~1);
+        }
+    Show64_step(cnt,X,step);
+    }
+
+static void Show08(size_t cnt,const u08b_t *b)
+    {
+    size_t i;
+    for (i=0;i < cnt;i++)
+        {
+        if (i %16 ==  0) printf(INDENT);
+        else if (i % 4 == 0) printf(" ");
+        printf(" %02X",b[i]);
+        if (i %16 == 15 || i==cnt-1) printf("\n");
+        fflush(stdout);
+        }
+    }
+
+static const char *AlgoHeader(uint_t bits)
+    {
+    if (skein_DebugFlag & SKEIN_DEBUG_THREEFISH)
+        switch (bits)
+            {
+            case  256:  return ":Threefish-256: ";
+            case  512:  return ":Threefish-512: ";
+            case 1024:  return ":Threefish-1024:";
+            }
+    else
+        switch (bits)
+            {
+            case  256:  return ":Skein-256: ";
+            case  512:  return ":Skein-512: ";
+            case 1024:  return ":Skein-1024:";
+            }
+    return NULL;
+    }
+
+void Skein_Show_Final(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t cnt,const u08b_t *outPtr)
+    {
+    if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+    if (skein_DebugFlag & SKEIN_DEBUG_FINAL)
+        {
+        printf("\n%s Final output=\n",AlgoHeader(bits));
+        Show08(cnt,outPtr);
+        printf("    ++++++++++\n");
+        fflush(stdout);
+        }
+    }
+
+/* show state after a round (or "pseudo-round") */
+void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X)
+    {
+    static uint_t injectNum=0;  /* not multi-thread safe! */
+
+    if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+    if (skein_DebugFlag)
+        {
+        if (r >= SKEIN_RND_SPECIAL) 
+            {       /* a key injection (or feedforward) point */
+            injectNum = (r == SKEIN_RND_KEY_INITIAL) ? 0 : injectNum+1;
+            if (  skein_DebugFlag & SKEIN_DEBUG_INJECT ||
+                ((skein_DebugFlag & SKEIN_DEBUG_FINAL) && r == SKEIN_RND_FEED_FWD))
+                {
+                printf("\n%s",AlgoHeader(bits));
+                switch (r)
+                    {
+                    case SKEIN_RND_KEY_INITIAL:
+                        printf(" [state after initial key injection]");
+                        break;
+                    case SKEIN_RND_KEY_INJECT:
+                        printf(" [state after key injection #%02d]",injectNum);
+                        break;
+                    case SKEIN_RND_FEED_FWD:
+                        printf(" [state after plaintext feedforward]");
+                        injectNum = 0;
+                        break;
+                    }
+                printf("=\n");
+                Show64(bits/64,X);
+                if (r== SKEIN_RND_FEED_FWD)
+                    printf("    ----------\n");
+                }
+            }
+        else if (skein_DebugFlag & SKEIN_DEBUG_ROUNDS)
+            {
+            uint_t j;
+            u64b_t p[SKEIN_MAX_STATE_WORDS];
+            const u08b_t *perm;
+            const static u08b_t PERM_256 [4][ 4] = { { 0,1,2,3 }, { 0,3,2,1 }, { 0,1,2,3 }, { 0,3,2,1 } };
+            const static u08b_t PERM_512 [4][ 8] = { { 0,1,2,3,4,5,6,7 },
+                                                     { 2,1,4,7,6,5,0,3 },
+                                                     { 4,1,6,3,0,5,2,7 },
+                                                     { 6,1,0,7,2,5,4,3 }
+                                                   };
+            const static u08b_t PERM_1024[4][16] = { { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,10,11,12,13,14,15 },
+                                                     { 0, 9, 2,13, 6,11, 4,15,10, 7,12, 3,14, 5, 8, 1 },
+                                                     { 0, 7, 2, 5, 4, 3, 6, 1,12,15,14,13, 8,11,10, 9 },
+                                                     { 0,15, 2,11, 6,13, 4, 9,14, 1, 8, 5,10, 3,12, 7 }
+                                                   };
+                    
+            if ((skein_DebugFlag & SKEIN_DEBUG_PERMUTE) && (r & 3))
+                {
+                printf("\n%s [state after round %2d (permuted)]=\n",AlgoHeader(bits),(int)r);
+                switch (bits)
+                    {
+                    case  256: perm = PERM_256 [r&3];   break;
+                    case  512: perm = PERM_512 [r&3];   break;
+                    default:   perm = PERM_1024[r&3];   break;
+                    }
+                for (j=0;j<bits/64;j++)
+                    p[j] = X[perm[j]];
+                Show64(bits/64,p);
+                }
+            else
+                {
+                printf("\n%s [state after round %2d]=\n",AlgoHeader(bits),(int)r);
+                Show64(bits/64,X);
+                }
+            }
+        }
+    }
+
+/* show state after a round (or "pseudo-round"), given a list of pointers */
+void Skein_Show_R_Ptr(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X_ptr[])
+    {
+    uint_t i;
+    u64b_t X[SKEIN_MAX_STATE_WORDS];
+
+    for (i=0;i<bits/64;i++)     /* copy over the words */ 
+        X[i] = X_ptr[i][0];
+    Skein_Show_Round(bits,h,r,X);
+    }
+
+
+/* show the state at the start of a block */
+void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,const u08b_t *blkPtr,
+                      const u64b_t *wPtr, const u64b_t *ksPtr, const u64b_t *tsPtr)
+    {
+    uint_t n;
+    if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+    if (skein_DebugFlag)
+        {
+        if (skein_DebugFlag & SKEIN_DEBUG_HDR)
+            {
+            printf("\n%s Block: outBits=%4d. T0=%06X.",AlgoHeader(bits),(uint_t) h->hashBitLen,(uint_t)h->T[0]);
+            printf(" Type=");
+            n = (uint_t) ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) >> SKEIN_T1_POS_BLK_TYPE);
+            switch (n)
+                {
+                case SKEIN_BLK_TYPE_KEY:  printf("KEY. ");  break;
+                case SKEIN_BLK_TYPE_CFG:  printf("CFG. ");  break;
+                case SKEIN_BLK_TYPE_PERS: printf("PERS.");  break;
+                case SKEIN_BLK_TYPE_PK :  printf("PK.  ");  break;
+                case SKEIN_BLK_TYPE_KDF:  printf("KDF. ");  break;
+                case SKEIN_BLK_TYPE_MSG:  printf("MSG. ");  break;
+                case SKEIN_BLK_TYPE_OUT:  printf("OUT. ");  break;
+                default:    printf("0x%02X.",n); break;
+                }
+            printf(" Flags=");
+            printf((h->T[1] & SKEIN_T1_FLAG_FIRST)   ? " First":"      ");
+            printf((h->T[1] & SKEIN_T1_FLAG_FINAL)   ? " Final":"      ");
+            printf((h->T[1] & SKEIN_T1_FLAG_BIT_PAD) ? " Pad"  :"    ");
+            n = (uint_t) ((h->T[1] & SKEIN_T1_TREE_LVL_MASK) >> SKEIN_T1_POS_TREE_LVL);
+            if (n)
+                printf("  TreeLevel = %02X",n);
+            printf("\n");
+            fflush(stdout);
+            }
+        if (skein_DebugFlag & SKEIN_DEBUG_TWEAK)
+            {
+            printf("  Tweak:\n");
+            Show64(2,h->T);
+            }
+        if (skein_DebugFlag & SKEIN_DEBUG_STATE)
+            {
+            printf("  %s words:\n",(skein_DebugFlag & SKEIN_DEBUG_THREEFISH)?"Key":"State");
+            Show64(bits/64,X);
+            }
+        if (skein_DebugFlag & SKEIN_DEBUG_KEYSCHED)
+            {
+            printf("  Tweak schedule:\n");
+            Show64_flag(3,tsPtr);
+            printf("  Key   schedule:\n");
+            Show64_flag((bits/64)+1,ksPtr);
+            }
+        if (skein_DebugFlag & SKEIN_DEBUG_INPUT_64)
+            {
+            printf("  Input block (words):\n");
+            Show64(bits/64,wPtr);
+            }
+        if (skein_DebugFlag & SKEIN_DEBUG_INPUT_08)
+            {
+            printf("  Input block (bytes):\n");
+            Show08(bits/8,blkPtr);
+            }
+        }
+    }
+
+void Skein_Show_Key(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u08b_t *key,size_t keyBytes)
+    {
+    if (keyBytes)
+    if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+    if (skein_DebugFlag & SKEIN_DEBUG_KEY)
+        {
+        printf("\n%s MAC key = %4u bytes\n",AlgoHeader(bits),(unsigned) keyBytes);
+        Show08(keyBytes,key);
+        }
+    }
+#endif
diff --git a/Optimized_32bit/skein_debug.h b/Optimized_32bit/skein_debug.h
new file mode 100644
index 000000000000..7775c0165c0a
--- /dev/null
+++ b/Optimized_32bit/skein_debug.h
@@ -0,0 +1,48 @@
+#ifndef _SKEIN_DEBUG_H_
+#define _SKEIN_DEBUG_H_
+/***********************************************************************
+**
+** Interface definitions for Skein hashing debug output.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#ifdef  SKEIN_DEBUG
+/* callout functions used inside Skein code */
+void    Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,const u08b_t *blkPtr,
+                         const u64b_t *wPtr,const u64b_t *ksPtr,const u64b_t *tsPtr);
+void    Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X);
+void    Skein_Show_R_Ptr(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X_ptr[]);
+void    Skein_Show_Final(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t cnt,const u08b_t *outPtr);
+void    Skein_Show_Key  (uint_t bits,const Skein_Ctxt_Hdr_t *h,const u08b_t *key,size_t keyBytes);
+
+extern  uint_t skein_DebugFlag;            /* flags to control debug output (0 --> none) */
+
+#define SKEIN_RND_SPECIAL       (1000u)
+#define SKEIN_RND_KEY_INITIAL   (SKEIN_RND_SPECIAL+0u)
+#define SKEIN_RND_KEY_INJECT    (SKEIN_RND_SPECIAL+1u)
+#define SKEIN_RND_FEED_FWD      (SKEIN_RND_SPECIAL+2u)
+
+/* flag bits:  skein_DebugFlag */
+#define SKEIN_DEBUG_KEY         (1u << 1)  /* show MAC key */
+#define SKEIN_DEBUG_CONFIG      (1u << 2)  /* show config block processing */
+#define SKEIN_DEBUG_STATE       (1u << 3)  /* show input state during Show_Block() */
+#define SKEIN_DEBUG_TWEAK       (1u << 4)  /* show input state during Show_Block() */
+#define SKEIN_DEBUG_KEYSCHED    (1u << 5)  /* show expanded key schedule */
+#define SKEIN_DEBUG_INPUT_64    (1u << 6)  /* show input block as 64-bit words */
+#define SKEIN_DEBUG_INPUT_08    (1u << 7)  /* show input block as  8-bit bytes */
+#define SKEIN_DEBUG_INJECT      (1u << 8)  /* show state after key injection & feedforward points */
+#define SKEIN_DEBUG_ROUNDS      (1u << 9)  /* show state after all rounds */
+#define SKEIN_DEBUG_FINAL       (1u <<10)  /* show final output of Skein */
+#define SKEIN_DEBUG_HDR         (1u <<11)  /* show block header */
+#define SKEIN_DEBUG_THREEFISH   (1u <<12)  /* use Threefish name instead of Skein */
+#define SKEIN_DEBUG_PERMUTE     (1u <<13)  /* use word permutations */
+#define SKEIN_DEBUG_ALL         ((~0u) & ~(SKEIN_DEBUG_THREEFISH | SKEIN_DEBUG_PERMUTE))
+#define THREEFISH_DEBUG_ALL     (SKEIN_DEBUG_ALL | SKEIN_DEBUG_THREEFISH)
+
+#endif /*  SKEIN_DEBUG    */
+
+#endif /* _SKEIN_DEBUG_H_ */
diff --git a/Optimized_32bit/skein_iv.h b/Optimized_32bit/skein_iv.h
new file mode 100644
index 000000000000..a8f54a41d345
--- /dev/null
+++ b/Optimized_32bit/skein_iv.h
@@ -0,0 +1,199 @@
+#ifndef _SKEIN_IV_H_
+#define _SKEIN_IV_H_
+
+#include "skein.h"    /* get Skein macros and types */
+
+/*
+***************** Pre-computed Skein IVs *******************
+**
+** NOTE: these values are not "magic" constants, but
+** are generated using the Threefish block function.
+** They are pre-computed here only for speed; i.e., to
+** avoid the need for a Threefish call during Init().
+**
+** The IV for any fixed hash length may be pre-computed.
+** Only the most common values are included here.
+**
+************************************************************
+**/
+
+#define MK_64 SKEIN_MK_64
+
+/* blkSize =  256 bits. hashSize =  128 bits */
+const u64b_t SKEIN_256_IV_128[] =
+    {
+    MK_64(0xE1111906,0x964D7260),
+    MK_64(0x883DAAA7,0x7C8D811C),
+    MK_64(0x10080DF4,0x91960F7A),
+    MK_64(0xCCF7DDE5,0xB45BC1C2)
+    };
+
+/* blkSize =  256 bits. hashSize =  160 bits */
+const u64b_t SKEIN_256_IV_160[] =
+    {
+    MK_64(0x14202314,0x72825E98),
+    MK_64(0x2AC4E9A2,0x5A77E590),
+    MK_64(0xD47A5856,0x8838D63E),
+    MK_64(0x2DD2E496,0x8586AB7D)
+    };
+
+/* blkSize =  256 bits. hashSize =  224 bits */
+const u64b_t SKEIN_256_IV_224[] =
+    {
+    MK_64(0xC6098A8C,0x9AE5EA0B),
+    MK_64(0x876D5686,0x08C5191C),
+    MK_64(0x99CB88D7,0xD7F53884),
+    MK_64(0x384BDDB1,0xAEDDB5DE)
+    };
+
+/* blkSize =  256 bits. hashSize =  256 bits */
+const u64b_t SKEIN_256_IV_256[] =
+    {
+    MK_64(0xFC9DA860,0xD048B449),
+    MK_64(0x2FCA6647,0x9FA7D833),
+    MK_64(0xB33BC389,0x6656840F),
+    MK_64(0x6A54E920,0xFDE8DA69)
+    };
+
+/* blkSize =  512 bits. hashSize =  128 bits */
+const u64b_t SKEIN_512_IV_128[] =
+    {
+    MK_64(0xA8BC7BF3,0x6FBF9F52),
+    MK_64(0x1E9872CE,0xBD1AF0AA),
+    MK_64(0x309B1790,0xB32190D3),
+    MK_64(0xBCFBB854,0x3F94805C),
+    MK_64(0x0DA61BCD,0x6E31B11B),
+    MK_64(0x1A18EBEA,0xD46A32E3),
+    MK_64(0xA2CC5B18,0xCE84AA82),
+    MK_64(0x6982AB28,0x9D46982D)
+    };
+
+/* blkSize =  512 bits. hashSize =  160 bits */
+const u64b_t SKEIN_512_IV_160[] =
+    {
+    MK_64(0x28B81A2A,0xE013BD91),
+    MK_64(0xC2F11668,0xB5BDF78F),
+    MK_64(0x1760D8F3,0xF6A56F12),
+    MK_64(0x4FB74758,0x8239904F),
+    MK_64(0x21EDE07F,0x7EAF5056),
+    MK_64(0xD908922E,0x63ED70B8),
+    MK_64(0xB8EC76FF,0xECCB52FA),
+    MK_64(0x01A47BB8,0xA3F27A6E)
+    };
+
+/* blkSize =  512 bits. hashSize =  224 bits */
+const u64b_t SKEIN_512_IV_224[] =
+    {
+    MK_64(0xCCD06162,0x48677224),
+    MK_64(0xCBA65CF3,0xA92339EF),
+    MK_64(0x8CCD69D6,0x52FF4B64),
+    MK_64(0x398AED7B,0x3AB890B4),
+    MK_64(0x0F59D1B1,0x457D2BD0),
+    MK_64(0x6776FE65,0x75D4EB3D),
+    MK_64(0x99FBC70E,0x997413E9),
+    MK_64(0x9E2CFCCF,0xE1C41EF7)
+    };
+
+/* blkSize =  512 bits. hashSize =  256 bits */
+const u64b_t SKEIN_512_IV_256[] =
+    {
+    MK_64(0xCCD044A1,0x2FDB3E13),
+    MK_64(0xE8359030,0x1A79A9EB),
+    MK_64(0x55AEA061,0x4F816E6F),
+    MK_64(0x2A2767A4,0xAE9B94DB),
+    MK_64(0xEC06025E,0x74DD7683),
+    MK_64(0xE7A436CD,0xC4746251),
+    MK_64(0xC36FBAF9,0x393AD185),
+    MK_64(0x3EEDBA18,0x33EDFC13)
+    };
+
+/* blkSize =  512 bits. hashSize =  384 bits */
+const u64b_t SKEIN_512_IV_384[] =
+    {
+    MK_64(0xA3F6C6BF,0x3A75EF5F),
+    MK_64(0xB0FEF9CC,0xFD84FAA4),
+    MK_64(0x9D77DD66,0x3D770CFE),
+    MK_64(0xD798CBF3,0xB468FDDA),
+    MK_64(0x1BC4A666,0x8A0E4465),
+    MK_64(0x7ED7D434,0xE5807407),
+    MK_64(0x548FC1AC,0xD4EC44D6),
+    MK_64(0x266E1754,0x6AA18FF8)
+    };
+
+/* blkSize =  512 bits. hashSize =  512 bits */
+const u64b_t SKEIN_512_IV_512[] =
+    {
+    MK_64(0x4903ADFF,0x749C51CE),
+    MK_64(0x0D95DE39,0x9746DF03),
+    MK_64(0x8FD19341,0x27C79BCE),
+    MK_64(0x9A255629,0xFF352CB1),
+    MK_64(0x5DB62599,0xDF6CA7B0),
+    MK_64(0xEABE394C,0xA9D5C3F4),
+    MK_64(0x991112C7,0x1A75B523),
+    MK_64(0xAE18A40B,0x660FCC33)
+    };
+
+/* blkSize = 1024 bits. hashSize =  384 bits */
+const u64b_t SKEIN1024_IV_384[] =
+    {
+    MK_64(0x5102B6B8,0xC1894A35),
+    MK_64(0xFEEBC9E3,0xFE8AF11A),
+    MK_64(0x0C807F06,0xE32BED71),
+    MK_64(0x60C13A52,0xB41A91F6),
+    MK_64(0x9716D35D,0xD4917C38),
+    MK_64(0xE780DF12,0x6FD31D3A),
+    MK_64(0x797846B6,0xC898303A),
+    MK_64(0xB172C2A8,0xB3572A3B),
+    MK_64(0xC9BC8203,0xA6104A6C),
+    MK_64(0x65909338,0xD75624F4),
+    MK_64(0x94BCC568,0x4B3F81A0),
+    MK_64(0x3EBBF51E,0x10ECFD46),
+    MK_64(0x2DF50F0B,0xEEB08542),
+    MK_64(0x3B5A6530,0x0DBC6516),
+    MK_64(0x484B9CD2,0x167BBCE1),
+    MK_64(0x2D136947,0xD4CBAFEA)
+    };
+
+/* blkSize = 1024 bits. hashSize =  512 bits */
+const u64b_t SKEIN1024_IV_512[] =
+    {
+    MK_64(0xCAEC0E5D,0x7C1B1B18),
+    MK_64(0xA01B0E04,0x5F03E802),
+    MK_64(0x33840451,0xED912885),
+    MK_64(0x374AFB04,0xEAEC2E1C),
+    MK_64(0xDF25A0E2,0x813581F7),
+    MK_64(0xE4004093,0x8B12F9D2),
+    MK_64(0xA662D539,0xC2ED39B6),
+    MK_64(0xFA8B85CF,0x45D8C75A),
+    MK_64(0x8316ED8E,0x29EDE796),
+    MK_64(0x053289C0,0x2E9F91B8),
+    MK_64(0xC3F8EF1D,0x6D518B73),
+    MK_64(0xBDCEC3C4,0xD5EF332E),
+    MK_64(0x549A7E52,0x22974487),
+    MK_64(0x67070872,0x5B749816),
+    MK_64(0xB9CD28FB,0xF0581BD1),
+    MK_64(0x0E2940B8,0x15804974)
+    };
+
+/* blkSize = 1024 bits. hashSize = 1024 bits */
+const u64b_t SKEIN1024_IV_1024[] =
+    {
+    MK_64(0xD593DA07,0x41E72355),
+    MK_64(0x15B5E511,0xAC73E00C),
+    MK_64(0x5180E5AE,0xBAF2C4F0),
+    MK_64(0x03BD41D3,0xFCBCAFAF),
+    MK_64(0x1CAEC6FD,0x1983A898),
+    MK_64(0x6E510B8B,0xCDD0589F),
+    MK_64(0x77E2BDFD,0xC6394ADA),
+    MK_64(0xC11E1DB5,0x24DCB0A3),
+    MK_64(0xD6D14AF9,0xC6329AB5),
+    MK_64(0x6A9B0BFC,0x6EB67E0D),
+    MK_64(0x9243C60D,0xCCFF1332),
+    MK_64(0x1A1F1DDE,0x743F02D4),
+    MK_64(0x0996753C,0x10ED0BB8),
+    MK_64(0x6572DD22,0xF2B4969A),
+    MK_64(0x61FD3062,0xD00A579A),
+    MK_64(0x1DE0536E,0x8682E539)
+    };
+
+#endif /* _SKEIN_IV_H_ */
diff --git a/Optimized_32bit/skein_port.h b/Optimized_32bit/skein_port.h
new file mode 100644
index 000000000000..653302de7467
--- /dev/null
+++ b/Optimized_32bit/skein_port.h
@@ -0,0 +1,124 @@
+#ifndef _SKEIN_PORT_H_
+#define _SKEIN_PORT_H_
+/*******************************************************************
+**
+** Platform-specific definitions for Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+** Many thanks to Brian Gladman for his portable header files.
+**
+** To port Skein to an "unsupported" platform, change the definitions
+** in this file appropriately.
+** 
+********************************************************************/
+
+#include "brg_types.h"                      /* get integer type definitions */
+
+typedef unsigned int    uint_t;             /* native unsigned integer */
+typedef uint_8t         u08b_t;             /*  8-bit unsigned integer */
+typedef uint_64t        u64b_t;             /* 64-bit unsigned integer */
+
+#ifndef RotL_64
+#define RotL_64(x,N)    (((x) << (N)) | ((x) >> (64-(N))))
+#endif
+
+/*
+ * Skein is "natively" little-endian (unlike SHA-xxx), for optimal
+ * performance on x86 CPUs.  The Skein code requires the following
+ * definitions for dealing with endianness:
+ *
+ *    SKEIN_NEED_SWAP:  0 for little-endian, 1 for big-endian
+ *    Skein_Put64_LSB_First
+ *    Skein_Get64_LSB_First
+ *    Skein_Swap64
+ *
+ * If SKEIN_NEED_SWAP is defined at compile time, it is used here
+ * along with the portable versions of Put64/Get64/Swap64, which 
+ * are slow in general.
+ *
+ * Otherwise, an "auto-detect" of endianness is attempted below.
+ * If the default handling doesn't work well, the user may insert
+ * platform-specific code instead (e.g., for big-endian CPUs).
+ *
+ */
+#ifndef SKEIN_NEED_SWAP /* compile-time "override" for endianness? */
+
+#include "brg_endian.h"                     /* get endianness selection */
+#if   PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN
+    /* here for big-endian CPUs */
+#define SKEIN_NEED_SWAP   (1)
+#elif PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN
+    /* here for x86 and x86-64 CPUs (and other detected little-endian CPUs) */
+#define SKEIN_NEED_SWAP   (0)
+#if   PLATFORM_MUST_ALIGN == 0              /* ok to use "fast" versions? */
+#define Skein_Put64_LSB_First(dst08,src64,bCnt) memcpy(dst08,src64,bCnt)
+#define Skein_Get64_LSB_First(dst64,src08,wCnt) memcpy(dst64,src08,8*(wCnt))
+#endif
+#else
+#error "Skein needs endianness setting!"
+#endif
+
+#endif /* ifndef SKEIN_NEED_SWAP */
+
+/*
+ ******************************************************************
+ *      Provide any definitions still needed.
+ ******************************************************************
+ */
+#ifndef Skein_Swap64  /* swap for big-endian, nop for little-endian */
+#if     SKEIN_NEED_SWAP
+#define Skein_Swap64(w64)                       \
+  ( (( ((u64b_t)(w64))       & 0xFF) << 56) |   \
+    (((((u64b_t)(w64)) >> 8) & 0xFF) << 48) |   \
+    (((((u64b_t)(w64)) >>16) & 0xFF) << 40) |   \
+    (((((u64b_t)(w64)) >>24) & 0xFF) << 32) |   \
+    (((((u64b_t)(w64)) >>32) & 0xFF) << 24) |   \
+    (((((u64b_t)(w64)) >>40) & 0xFF) << 16) |   \
+    (((((u64b_t)(w64)) >>48) & 0xFF) <<  8) |   \
+    (((((u64b_t)(w64)) >>56) & 0xFF)      ) )
+#else
+#define Skein_Swap64(w64)  (w64)
+#endif
+#endif  /* ifndef Skein_Swap64 */
+
+
+#ifndef Skein_Put64_LSB_First
+void    Skein_Put64_LSB_First(u08b_t *dst,const u64b_t *src,size_t bCnt)
+#ifdef  SKEIN_PORT_CODE /* instantiate the function code here? */
+    { /* this version is fully portable (big-endian or little-endian), but slow */
+    size_t n;
+
+    for (n=0;n<bCnt;n++)
+        dst[n] = (u08b_t) (src[n>>3] >> (8*(n&7)));
+    }
+#else
+    ;    /* output only the function prototype */
+#endif
+#endif   /* ifndef Skein_Put64_LSB_First */
+
+
+#ifndef Skein_Get64_LSB_First
+void    Skein_Get64_LSB_First(u64b_t *dst,const u08b_t *src,size_t wCnt)
+#ifdef  SKEIN_PORT_CODE /* instantiate the function code here? */
+    { /* this version is fully portable (big-endian or little-endian), but slow */
+    size_t n;
+
+    f