aboutsummaryrefslogtreecommitdiff
path: root/test/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen')
-rw-r--r--test/CodeGen/AArch64/arm64-ccmp.ll40
-rw-r--r--test/CodeGen/AArch64/arm64-named-reg-alloc.ll2
-rw-r--r--test/CodeGen/AArch64/arm64-named-reg-notareg.ll2
-rw-r--r--test/CodeGen/AArch64/global-merge-ignore-single-use-minsize.ll74
-rw-r--r--test/CodeGen/AArch64/minmax.ll11
-rw-r--r--test/CodeGen/AArch64/special-reg.ll48
-rw-r--r--test/CodeGen/ARM/atomic-ops-v8.ll14
-rw-r--r--test/CodeGen/ARM/build-attributes.ll6
-rw-r--r--test/CodeGen/ARM/ifcvt-callback.ll22
-rw-r--r--test/CodeGen/ARM/jump-table-islands-split.ll52
-rw-r--r--test/CodeGen/ARM/jump-table-islands.ll40
-rw-r--r--test/CodeGen/ARM/jumptable-label.ll4
-rw-r--r--test/CodeGen/ARM/ldrd.ll16
-rw-r--r--test/CodeGen/ARM/named-reg-alloc.ll2
-rw-r--r--test/CodeGen/ARM/named-reg-notareg.ll2
-rw-r--r--test/CodeGen/ARM/special-reg-acore.ll78
-rw-r--r--test/CodeGen/ARM/special-reg-mcore.ll143
-rw-r--r--test/CodeGen/ARM/special-reg.ll78
-rw-r--r--test/CodeGen/BPF/alu8.ll3
-rw-r--r--test/CodeGen/BPF/atomics.ll3
-rw-r--r--test/CodeGen/BPF/basictest.ll2
-rw-r--r--test/CodeGen/BPF/cc_args.ll3
-rw-r--r--test/CodeGen/BPF/cc_args_be.ll96
-rw-r--r--test/CodeGen/BPF/cc_ret.ll2
-rw-r--r--test/CodeGen/BPF/ex1.ll2
-rw-r--r--test/CodeGen/BPF/intrinsics.ll2
-rw-r--r--test/CodeGen/BPF/load.ll2
-rw-r--r--test/CodeGen/BPF/loops.ll2
-rw-r--r--test/CodeGen/BPF/sanity.ll2
-rw-r--r--test/CodeGen/BPF/setcc.ll2
-rw-r--r--test/CodeGen/BPF/shifts.ll3
-rw-r--r--test/CodeGen/BPF/sockex2.ll3
-rw-r--r--test/CodeGen/Generic/stop-after.ll3
-rw-r--r--test/CodeGen/Hexagon/args.ll8
-rw-r--r--test/CodeGen/Hexagon/calling-conv.ll73
-rw-r--r--test/CodeGen/Hexagon/cext-valid-packet1.ll1
-rw-r--r--test/CodeGen/Hexagon/cext-valid-packet2.ll1
-rw-r--r--test/CodeGen/Hexagon/compound.ll17
-rw-r--r--test/CodeGen/Hexagon/dualstore.ll15
-rw-r--r--test/CodeGen/Hexagon/duplex.ll7
-rw-r--r--test/CodeGen/Hexagon/relax.ll14
-rw-r--r--test/CodeGen/Hexagon/sube.ll4
-rw-r--r--test/CodeGen/MIR/lit.local.cfg2
-rw-r--r--test/CodeGen/MIR/llvm-ir-error-reported.mir22
-rw-r--r--test/CodeGen/MIR/llvmIR.mir32
-rw-r--r--test/CodeGen/MIR/llvmIRMissing.mir5
-rw-r--r--test/CodeGen/MIR/machine-function-missing-name.mir22
-rw-r--r--test/CodeGen/MIR/machine-function.mir24
-rw-r--r--test/CodeGen/Mips/Fast-ISel/bswap1.ll58
-rw-r--r--test/CodeGen/Mips/Fast-ISel/div1.ll55
-rw-r--r--test/CodeGen/Mips/Fast-ISel/memtest1.ll74
-rw-r--r--test/CodeGen/Mips/Fast-ISel/mul1.ll18
-rw-r--r--test/CodeGen/Mips/Fast-ISel/rem1.ll56
-rw-r--r--test/CodeGen/Mips/Fast-ISel/sel1.ll91
-rw-r--r--test/CodeGen/Mips/dynamic-stack-realignment.ll299
-rw-r--r--test/CodeGen/Mips/ehframe-indirect.ll26
-rw-r--r--test/CodeGen/Mips/emergency-spill-slot-near-fp.ll4
-rw-r--r--test/CodeGen/NVPTX/access-non-generic.ll16
-rw-r--r--test/CodeGen/NVPTX/bug21465.ll12
-rw-r--r--test/CodeGen/NVPTX/call-with-alloca-buffer.ll7
-rw-r--r--test/CodeGen/NVPTX/globals_init.ll23
-rw-r--r--test/CodeGen/NVPTX/lower-kernel-ptr-arg.ll20
-rw-r--r--test/CodeGen/NVPTX/pr13291-i1-store.ll8
-rw-r--r--test/CodeGen/NVPTX/surf-read-cuda.ll8
-rw-r--r--test/CodeGen/NVPTX/tex-read-cuda.ll8
-rw-r--r--test/CodeGen/PowerPC/fma.ll83
-rw-r--r--test/CodeGen/PowerPC/vsx-fma-sp.ll167
-rw-r--r--test/CodeGen/R600/cgp-addressing-modes.ll242
-rw-r--r--test/CodeGen/R600/coalescer_remat.ll6
-rw-r--r--test/CodeGen/R600/codegen-prepare-addrmode-sext.ll10
-rw-r--r--test/CodeGen/R600/fmul.ll16
-rw-r--r--test/CodeGen/R600/half.ll543
-rw-r--r--test/CodeGen/R600/imm.ll2
-rw-r--r--test/CodeGen/R600/loop-address.ll5
-rw-r--r--test/CodeGen/R600/loop-idiom.ll4
-rw-r--r--test/CodeGen/R600/max.ll51
-rw-r--r--test/CodeGen/R600/min.ll51
-rw-r--r--test/CodeGen/R600/sext-in-reg.ll5
-rw-r--r--test/CodeGen/R600/si-vector-hang.ll3
-rw-r--r--test/CodeGen/R600/subreg-eliminate-dead.ll19
-rw-r--r--test/CodeGen/R600/trunc-store-f64-to-f16.ll56
-rw-r--r--test/CodeGen/R600/unroll.ll5
-rw-r--r--test/CodeGen/R600/wrong-transalu-pos-fix.ll7
-rw-r--r--test/CodeGen/Thumb2/constant-islands-jump-table.ll2
-rw-r--r--test/CodeGen/Thumb2/float-ops.ll6
-rw-r--r--test/CodeGen/Thumb2/thumb2-tbh.ll12
-rw-r--r--test/CodeGen/X86/asm-reject-reg-type-mismatch.ll10
-rw-r--r--test/CodeGen/X86/avx-vperm2x128.ll2
-rw-r--r--test/CodeGen/X86/avx2-vector-shifts.ll390
-rw-r--r--test/CodeGen/X86/avx512-build-vector.ll10
-rw-r--r--test/CodeGen/X86/avx512-intrinsics.ll45
-rw-r--r--test/CodeGen/X86/avx512-shuffle.ll336
-rw-r--r--test/CodeGen/X86/avx512-vec-cmp.ll7
-rw-r--r--test/CodeGen/X86/avx512vl-intrinsics.ll40
-rw-r--r--test/CodeGen/X86/buildvec-insertvec.ll55
-rw-r--r--test/CodeGen/X86/critical-anti-dep-breaker.ll3
-rw-r--r--test/CodeGen/X86/machine-cp.ll55
-rw-r--r--test/CodeGen/X86/pic.ll8
-rw-r--r--test/CodeGen/X86/pr23603.ll24
-rw-r--r--test/CodeGen/X86/pr23664.ll14
-rw-r--r--test/CodeGen/X86/recip-fastmath.ll4
-rw-r--r--test/CodeGen/X86/sibcall-win64.ll24
-rw-r--r--test/CodeGen/X86/sqrt-fastmath.ll2
-rw-r--r--test/CodeGen/X86/stack-folding-x86_64.ll51
-rw-r--r--test/CodeGen/X86/statepoint-far-call.ll22
-rw-r--r--test/CodeGen/X86/switch-or.ll21
-rw-r--r--test/CodeGen/X86/switch.ll15
-rw-r--r--test/CodeGen/X86/tail-call-got.ll18
-rw-r--r--test/CodeGen/X86/tailcallpic1.ll3
-rw-r--r--test/CodeGen/X86/tailcallpic3.ll73
-rw-r--r--test/CodeGen/X86/vec_fp_to_int.ll2
-rw-r--r--test/CodeGen/X86/vec_shift8.ll1016
-rw-r--r--test/CodeGen/X86/vector-ctpop.ll159
-rw-r--r--test/CodeGen/X86/vector-lzcnt-128.ll1915
-rw-r--r--test/CodeGen/X86/vector-lzcnt-256.ll1305
-rw-r--r--test/CodeGen/X86/vector-popcnt-128.ll462
-rw-r--r--test/CodeGen/X86/vector-popcnt-256.ll220
-rw-r--r--test/CodeGen/X86/vector-shuffle-256-v4.ll1
-rw-r--r--test/CodeGen/X86/vector-shuffle-512-v8.ll713
-rw-r--r--test/CodeGen/X86/vector-tzcnt-128.ll1788
-rw-r--r--test/CodeGen/X86/vector-tzcnt-256.ll1195
-rw-r--r--test/CodeGen/X86/vector-zext.ll10
-rw-r--r--test/CodeGen/X86/win32-eh-states.ll112
-rw-r--r--test/CodeGen/X86/win32-eh.ll59
124 files changed, 12311 insertions, 967 deletions
diff --git a/test/CodeGen/AArch64/arm64-ccmp.ll b/test/CodeGen/AArch64/arm64-ccmp.ll
index ff18f7364337..11228c7e8808 100644
--- a/test/CodeGen/AArch64/arm64-ccmp.ll
+++ b/test/CodeGen/AArch64/arm64-ccmp.ll
@@ -287,3 +287,43 @@ sw.bb.i.i:
%code1.i.i.phi.trans.insert = getelementptr inbounds %str1, %str1* %0, i64 0, i32 0, i32 0, i64 16
br label %sw.bb.i.i
}
+
+; CHECK-LABEL: select_and
+define i64 @select_and(i32 %v1, i32 %v2, i64 %a, i64 %b) {
+; CHECK: cmp
+; CHECK: ccmp{{.*}}, #0, ne
+; CHECK: csel{{.*}}, lt
+ %1 = icmp slt i32 %v1, %v2
+ %2 = icmp ne i32 5, %v2
+ %3 = and i1 %1, %2
+ %sel = select i1 %3, i64 %a, i64 %b
+ ret i64 %sel
+}
+
+; CHECK-LABEL: select_or
+define i64 @select_or(i32 %v1, i32 %v2, i64 %a, i64 %b) {
+; CHECK: cmp
+; CHECK: ccmp{{.*}}, #8, eq
+; CHECK: csel{{.*}}, lt
+ %1 = icmp slt i32 %v1, %v2
+ %2 = icmp ne i32 5, %v2
+ %3 = or i1 %1, %2
+ %sel = select i1 %3, i64 %a, i64 %b
+ ret i64 %sel
+}
+
+; CHECK-LABEL: select_complicated
+define i16 @select_complicated(double %v1, double %v2, i16 %a, i16 %b) {
+; CHECK: fcmp
+; CHECK: fccmp{{.*}}, #4, ne
+; CHECK: fccmp{{.*}}, #1, ne
+; CHECK: fccmp{{.*}}, #4, vc
+; CEHCK: csel{{.*}}, eq
+ %1 = fcmp one double %v1, %v2
+ %2 = fcmp oeq double %v2, 13.0
+ %3 = fcmp oeq double %v1, 42.0
+ %or0 = or i1 %2, %3
+ %or1 = or i1 %1, %or0
+ %sel = select i1 %or1, i16 %a, i16 %b
+ ret i16 %sel
+}
diff --git a/test/CodeGen/AArch64/arm64-named-reg-alloc.ll b/test/CodeGen/AArch64/arm64-named-reg-alloc.ll
index 0c564544a538..5d48c17e1286 100644
--- a/test/CodeGen/AArch64/arm64-named-reg-alloc.ll
+++ b/test/CodeGen/AArch64/arm64-named-reg-alloc.ll
@@ -4,7 +4,7 @@
define i32 @get_stack() nounwind {
entry:
; FIXME: Include an allocatable-specific error message
-; CHECK: Invalid register name global variable
+; CHECK: Invalid register name "x5".
%sp = call i32 @llvm.read_register.i32(metadata !0)
ret i32 %sp
}
diff --git a/test/CodeGen/AArch64/arm64-named-reg-notareg.ll b/test/CodeGen/AArch64/arm64-named-reg-notareg.ll
index 759bc15807b5..8a5fd6f1ac8b 100644
--- a/test/CodeGen/AArch64/arm64-named-reg-notareg.ll
+++ b/test/CodeGen/AArch64/arm64-named-reg-notareg.ll
@@ -3,7 +3,7 @@
define i32 @get_stack() nounwind {
entry:
-; CHECK: Invalid register name global variable
+; CHECK: Invalid register name "notareg".
%sp = call i32 @llvm.read_register.i32(metadata !0)
ret i32 %sp
}
diff --git a/test/CodeGen/AArch64/global-merge-ignore-single-use-minsize.ll b/test/CodeGen/AArch64/global-merge-ignore-single-use-minsize.ll
new file mode 100644
index 000000000000..e83cbab140a7
--- /dev/null
+++ b/test/CodeGen/AArch64/global-merge-ignore-single-use-minsize.ll
@@ -0,0 +1,74 @@
+; RUN: llc -mtriple=aarch64-apple-ios -asm-verbose=false -aarch64-collect-loh=false \
+; RUN: -O1 -global-merge-group-by-use -global-merge-ignore-single-use \
+; RUN: %s -o - | FileCheck %s
+
+; Check that, at -O1, we only merge globals used in minsize functions.
+; We assume that globals of the same size aren't reordered inside a set.
+; We use -global-merge-ignore-single-use, and thus only expect one merged set.
+
+@m1 = internal global i32 0, align 4
+@n1 = internal global i32 0, align 4
+
+; CHECK-LABEL: f1:
+define void @f1(i32 %a1, i32 %a2) minsize nounwind {
+; CHECK-NEXT: adrp x8, [[SET:__MergedGlobals]]@PAGE
+; CHECK-NEXT: add x8, x8, [[SET]]@PAGEOFF
+; CHECK-NEXT: stp w0, w1, [x8]
+; CHECK-NEXT: ret
+ store i32 %a1, i32* @m1, align 4
+ store i32 %a2, i32* @n1, align 4
+ ret void
+}
+
+@m2 = internal global i32 0, align 4
+@n2 = internal global i32 0, align 4
+
+; CHECK-LABEL: f2:
+define void @f2(i32 %a1, i32 %a2) nounwind {
+; CHECK-NEXT: adrp x8, _m2@PAGE
+; CHECK-NEXT: adrp x9, _n2@PAGE
+; CHECK-NEXT: str w0, [x8, _m2@PAGEOFF]
+; CHECK-NEXT: str w1, [x9, _n2@PAGEOFF]
+; CHECK-NEXT: ret
+ store i32 %a1, i32* @m2, align 4
+ store i32 %a2, i32* @n2, align 4
+ ret void
+}
+
+; If we have use sets partially overlapping between a minsize and a non-minsize
+; function, explicitly check that we only consider the globals used in the
+; minsize function for merging.
+
+@m3 = internal global i32 0, align 4
+@n3 = internal global i32 0, align 4
+
+; CHECK-LABEL: f3:
+define void @f3(i32 %a1, i32 %a2) minsize nounwind {
+; CHECK-NEXT: adrp x8, [[SET]]@PAGE
+; CHECK-NEXT: add x8, x8, [[SET]]@PAGEOFF
+; CHECK-NEXT: stp w0, w1, [x8, #8]
+; CHECK-NEXT: ret
+ store i32 %a1, i32* @m3, align 4
+ store i32 %a2, i32* @n3, align 4
+ ret void
+}
+
+@n4 = internal global i32 0, align 4
+
+; CHECK-LABEL: f4:
+define void @f4(i32 %a1, i32 %a2) nounwind {
+; CHECK-NEXT: adrp x8, [[SET]]@PAGE
+; CHECK-NEXT: add x8, x8, [[SET]]@PAGEOFF
+; CHECK-NEXT: adrp x9, _n4@PAGE
+; CHECK-NEXT: str w0, [x8, #8]
+; CHECK-NEXT: str w1, [x9, _n4@PAGEOFF]
+; CHECK-NEXT: ret
+ store i32 %a1, i32* @m3, align 4
+ store i32 %a2, i32* @n4, align 4
+ ret void
+}
+
+; CHECK-DAG: .zerofill __DATA,__bss,[[SET]],16,3
+; CHECK-DAG: .zerofill __DATA,__bss,_m2,4,2
+; CHECK-DAG: .zerofill __DATA,__bss,_n2,4,2
+; CHECK-DAG: .zerofill __DATA,__bss,_n4,4,2
diff --git a/test/CodeGen/AArch64/minmax.ll b/test/CodeGen/AArch64/minmax.ll
index a6b5adebe107..df4912ca1f7a 100644
--- a/test/CodeGen/AArch64/minmax.ll
+++ b/test/CodeGen/AArch64/minmax.ll
@@ -94,3 +94,14 @@ define <16 x i32> @t11(<16 x i32> %a, <16 x i32> %b) {
%t2 = select <16 x i1> %t1, <16 x i32> %a, <16 x i32> %b
ret <16 x i32> %t2
}
+
+; CHECK-LABEL: t12
+; CHECK-NOT: umin
+; The icmp is used by two instructions, so don't produce a umin node.
+define <16 x i8> @t12(<16 x i8> %a, <16 x i8> %b) {
+ %t1 = icmp ugt <16 x i8> %b, %a
+ %t2 = select <16 x i1> %t1, <16 x i8> %a, <16 x i8> %b
+ %t3 = zext <16 x i1> %t1 to <16 x i8>
+ %t4 = add <16 x i8> %t3, %t2
+ ret <16 x i8> %t4
+}
diff --git a/test/CodeGen/AArch64/special-reg.ll b/test/CodeGen/AArch64/special-reg.ll
new file mode 100644
index 000000000000..91c32158d420
--- /dev/null
+++ b/test/CodeGen/AArch64/special-reg.ll
@@ -0,0 +1,48 @@
+; RUN: llc < %s -mtriple=aarch64-none-eabi -mcpu=cortex-a57 2>&1 | FileCheck %s
+
+define i64 @read_encoded_register() nounwind {
+entry:
+; CHECK-LABEL: read_encoded_register:
+; CHECK: mrs x0, S1_2_C3_C4_5
+ %reg = call i64 @llvm.read_register.i64(metadata !0)
+ ret i64 %reg
+}
+
+define i64 @read_daif() nounwind {
+entry:
+; CHECK-LABEL: read_daif:
+; CHECK: mrs x0, DAIF
+ %reg = call i64 @llvm.read_register.i64(metadata !1)
+ ret i64 %reg
+}
+
+define void @write_encoded_register(i64 %x) nounwind {
+entry:
+; CHECK-LABEL: write_encoded_register:
+; CHECK: msr S1_2_C3_C4_5, x0
+ call void @llvm.write_register.i64(metadata !0, i64 %x)
+ ret void
+}
+
+define void @write_daif(i64 %x) nounwind {
+entry:
+; CHECK-LABEL: write_daif:
+; CHECK: msr DAIF, x0
+ call void @llvm.write_register.i64(metadata !1, i64 %x)
+ ret void
+}
+
+define void @write_daifset() nounwind {
+entry:
+; CHECK-LABEL: write_daifset:
+; CHECK: msr DAIFSET, #2
+ call void @llvm.write_register.i64(metadata !2, i64 2)
+ ret void
+}
+
+declare i64 @llvm.read_register.i64(metadata) nounwind
+declare void @llvm.write_register.i64(metadata, i64) nounwind
+
+!0 = !{!"1:2:3:4:5"}
+!1 = !{!"daif"}
+!2 = !{!"daifset"}
diff --git a/test/CodeGen/ARM/atomic-ops-v8.ll b/test/CodeGen/ARM/atomic-ops-v8.ll
index db5007b0758d..86287c1178db 100644
--- a/test/CodeGen/ARM/atomic-ops-v8.ll
+++ b/test/CodeGen/ARM/atomic-ops-v8.ll
@@ -664,7 +664,7 @@ define void @test_atomic_load_min_i64(i64 %offset) nounwind {
; CHECK: movt r[[ADDR]], :upper16:var64
; CHECK: .LBB{{[0-9]+}}_1:
-; CHECK: ldaexd [[OLD1:r[0-9]+]], [[OLD2:r[0-9]+]], [r[[ADDR]]]
+; CHECK: ldaexd [[OLD1:r[0-9]+|lr]], [[OLD2:r[0-9]+|lr]], [r[[ADDR]]]
; r0, r1 below is a reasonable guess but could change: it certainly comes into the
; function there.
; CHECK-ARM: mov [[LOCARRY:r[0-9]+|lr]], #0
@@ -782,7 +782,7 @@ define void @test_atomic_load_max_i64(i64 %offset) nounwind {
; CHECK: movt r[[ADDR]], :upper16:var64
; CHECK: .LBB{{[0-9]+}}_1:
-; CHECK: ldrexd [[OLD1:r[0-9]+]], [[OLD2:r[0-9]+]], [r[[ADDR]]]
+; CHECK: ldrexd [[OLD1:r[0-9]+]], [[OLD2:r[0-9]+|lr]], [r[[ADDR]]]
; r0, r1 below is a reasonable guess but could change: it certainly comes into the
; function there.
; CHECK-ARM: mov [[LOCARRY:r[0-9]+|lr]], #0
@@ -900,7 +900,7 @@ define void @test_atomic_load_umin_i64(i64 %offset) nounwind {
; CHECK: movt r[[ADDR]], :upper16:var64
; CHECK: .LBB{{[0-9]+}}_1:
-; CHECK: ldaexd [[OLD1:r[0-9]+]], [[OLD2:r[0-9]+]], [r[[ADDR]]]
+; CHECK: ldaexd [[OLD1:r[0-9]+|lr]], [[OLD2:r[0-9]+|lr]], [r[[ADDR]]]
; r0, r1 below is a reasonable guess but could change: it certainly comes into the
; function there.
; CHECK-ARM: mov [[LOCARRY:r[0-9]+|lr]], #0
@@ -1018,7 +1018,7 @@ define void @test_atomic_load_umax_i64(i64 %offset) nounwind {
; CHECK: movt r[[ADDR]], :upper16:var64
; CHECK: .LBB{{[0-9]+}}_1:
-; CHECK: ldaexd [[OLD1:r[0-9]+]], [[OLD2:r[0-9]+]], [r[[ADDR]]]
+; CHECK: ldaexd [[OLD1:r[0-9]+|lr]], [[OLD2:r[0-9]+|lr]], [r[[ADDR]]]
; r0, r1 below is a reasonable guess but could change: it certainly comes into the
; function there.
; CHECK-ARM: mov [[LOCARRY:r[0-9]+|lr]], #0
@@ -1146,10 +1146,12 @@ define void @test_atomic_cmpxchg_i64(i64 %wanted, i64 %new) nounwind {
; function there.
; CHECK-LE-DAG: eor{{(\.w)?}} [[MISMATCH_LO:r[0-9]+|lr]], [[OLD1]], r0
; CHECK-LE-DAG: eor{{(\.w)?}} [[MISMATCH_HI:r[0-9]+|lr]], [[OLD2]], r1
-; CHECK-LE: orrs{{(\.w)?}} {{r[0-9]+}}, [[MISMATCH_LO]], [[MISMATCH_HI]]
+; CHECK-ARM-LE: orrs{{(\.w)?}} {{r[0-9]+}}, [[MISMATCH_LO]], [[MISMATCH_HI]]
+; CHECK-THUMB-LE: orrs{{(\.w)?}} {{(r[0-9]+, )?}}[[MISMATCH_HI]], [[MISMATCH_LO]]
; CHECK-BE-DAG: eor{{(\.w)?}} [[MISMATCH_HI:r[0-9]+|lr]], [[OLD2]], r1
; CHECK-BE-DAG: eor{{(\.w)?}} [[MISMATCH_LO:r[0-9]+|lr]], [[OLD1]], r0
-; CHECK-BE: orrs{{(\.w)?}} {{r[0-9]+}}, [[MISMATCH_HI]], [[MISMATCH_LO]]
+; CHECK-ARM-BE: orrs{{(\.w)?}} {{r[0-9]+}}, [[MISMATCH_HI]], [[MISMATCH_LO]]
+; CHECK-THUMB-BE: orrs{{(\.w)?}} {{(r[0-9]+, )?}}[[MISMATCH_LO]], [[MISMATCH_HI]]
; CHECK-NEXT: bne .LBB{{[0-9]+}}_3
; CHECK-NEXT: BB#2:
; As above, r2, r3 is a reasonable guess.
diff --git a/test/CodeGen/ARM/build-attributes.ll b/test/CodeGen/ARM/build-attributes.ll
index 1982fa98ef41..e9de52a3e1a0 100644
--- a/test/CodeGen/ARM/build-attributes.ll
+++ b/test/CodeGen/ARM/build-attributes.ll
@@ -923,7 +923,7 @@
; CORTEX-M4-SOFT: .eabi_attribute 7, 77
; CORTEX-M4-SOFT: .eabi_attribute 8, 0
; CORTEX-M4-SOFT: .eabi_attribute 9, 2
-; CORTEX-M4-SOFT: .fpu vfpv4-d16
+; CORTEX-M4-SOFT: .fpu fpv4-sp-d16
; CORTEX-M4-SOFT-NOT: .eabi_attribute 19
;; We default to IEEE 754 compliance
; CORTEX-M4-SOFT: .eabi_attribute 20, 1
@@ -953,7 +953,7 @@
; CORTEX-M4-HARD: .eabi_attribute 7, 77
; CORTEX-M4-HARD: .eabi_attribute 8, 0
; CORTEX-M4-HARD: .eabi_attribute 9, 2
-; CORTEX-M4-HARD: .fpu vfpv4-d16
+; CORTEX-M4-HARD: .fpu fpv4-sp-d16
; CORTEX-M4-HARD-NOT: .eabi_attribute 19
;; We default to IEEE 754 compliance
; CORTEX-M4-HARD: .eabi_attribute 20, 1
@@ -984,7 +984,7 @@
; CORTEX-M7: .eabi_attribute 8, 0
; CORTEX-M7: .eabi_attribute 9, 2
; CORTEX-M7-SOFT-NOT: .fpu
-; CORTEX-M7-SINGLE: .fpu fpv5-d16
+; CORTEX-M7-SINGLE: .fpu fpv5-sp-d16
; CORTEX-M7-DOUBLE: .fpu fpv5-d16
; CORTEX-M7: .eabi_attribute 17, 1
; CORTEX-M7-NOT: .eabi_attribute 19
diff --git a/test/CodeGen/ARM/ifcvt-callback.ll b/test/CodeGen/ARM/ifcvt-callback.ll
new file mode 100644
index 000000000000..62a66e745b39
--- /dev/null
+++ b/test/CodeGen/ARM/ifcvt-callback.ll
@@ -0,0 +1,22 @@
+; RUN: llc -march thumb %s -o - | FileCheck %s
+
+; This test checks that if-conversion pass is unconditionally added to the pass
+; pipeline and is conditionally executed based on the per-function targert-cpu
+; attribute.
+
+; CHECK: ite eq
+
+define i32 @test_ifcvt(i32 %a, i32 %b) #0 {
+ %tmp2 = icmp eq i32 %a, 0
+ br i1 %tmp2, label %cond_false, label %cond_true
+
+cond_true:
+ %tmp5 = add i32 %b, 1
+ ret i32 %tmp5
+
+cond_false:
+ %tmp7 = add i32 %b, -1
+ ret i32 %tmp7
+}
+
+attributes #0 = { "target-cpu"="cortex-a8" }
diff --git a/test/CodeGen/ARM/jump-table-islands-split.ll b/test/CodeGen/ARM/jump-table-islands-split.ll
new file mode 100644
index 000000000000..deba21b4dbb1
--- /dev/null
+++ b/test/CodeGen/ARM/jump-table-islands-split.ll
@@ -0,0 +1,52 @@
+; RUN: llc -mtriple=thumbv7s-apple-ios8.0 -o - %s | FileCheck %s
+
+declare void @foo(double)
+declare i32 @llvm.arm.space(i32, i32)
+
+; The constpool entry used to call @foo should be directly between where we want
+; the tbb and its table. Fortunately, the flow is simple enough that we can
+; eliminate the entry calculation (ADD) and use the ADR as the base.
+;
+; I'm hoping this won't be fragile, but if it does break the most likely fix is
+; adjusting the @llvm.arm.space call slightly. If this happens too many times
+; the test should probably be removed.
+define i32 @test_jumptable_not_adjacent(i1 %tst, i32 %sw, i32 %l) {
+; CHECK-LABEL: test_jumptable_not_adjacent:
+; CHECK: vldr {{d[0-9]+}}, [[DBL_CONST:LCPI[0-9]+_[0-9]+]]
+; [...]
+; CHECK: adr.w r[[BASE:[0-9]+]], [[JUMP_TABLE:LJTI[0-9]+_[0-9]+]]
+; CHECK-NOT: r[[BASE]]
+
+; CHECK: [[TBB_KEY:LCPI[0-9]+_[0-9]+]]:
+; CHECK-NEXT: tbb [r[[BASE]], {{r[0-9]+}}]
+
+; CHECK: [[DBL_CONST]]:
+; CHECK: .long
+; CHECK: .long
+; CHECK: [[JUMP_TABLE]]:
+; CHECK: .byte (LBB{{[0-9]+}}_{{[0-9]+}}-([[TBB_KEY]]+4)
+
+ br label %complex
+
+complex:
+ call void @foo(double 12345.0)
+ call i32 @llvm.arm.space(i32 970, i32 undef)
+ switch i32 %sw, label %second [ i32 0, label %other
+ i32 1, label %third
+ i32 2, label %end
+ i32 3, label %other ]
+
+second:
+ ret i32 43
+third:
+ ret i32 0
+
+other:
+ call void @bar()
+ unreachable
+
+end:
+ ret i32 42
+}
+
+declare void @bar()
diff --git a/test/CodeGen/ARM/jump-table-islands.ll b/test/CodeGen/ARM/jump-table-islands.ll
new file mode 100644
index 000000000000..6b4f174c0928
--- /dev/null
+++ b/test/CodeGen/ARM/jump-table-islands.ll
@@ -0,0 +1,40 @@
+; RUN: llc -mtriple=armv7-apple-ios8.0 -o - %s | FileCheck %s
+
+%BigInt = type i5500
+
+define %BigInt @test_moved_jumptable(i1 %tst, i32 %sw, %BigInt %l) {
+; CHECK-LABEL: test_moved_jumptable:
+
+; CHECK: adr {{r[0-9]+}}, [[JUMP_TABLE:LJTI[0-9]+_[0-9]+]]
+; CHECK: b [[SKIP_TABLE:LBB[0-9]+_[0-9]+]]
+
+; CHECK: [[JUMP_TABLE]]:
+; CHECK: .data_region jt32
+; CHECK: .long LBB{{[0-9]+_[0-9]+}}-[[JUMP_TABLE]]
+
+; CHECK: [[SKIP_TABLE]]:
+; CHECK: add pc, {{r[0-9]+}}, {{r[0-9]+}}
+ br i1 %tst, label %simple, label %complex
+
+simple:
+ br label %end
+
+complex:
+ switch i32 %sw, label %simple [ i32 0, label %other
+ i32 1, label %third
+ i32 5, label %end
+ i32 6, label %other ]
+
+third:
+ ret %BigInt 0
+
+other:
+ call void @bar()
+ unreachable
+
+end:
+ %val = phi %BigInt [ %l, %complex ], [ -1, %simple ]
+ ret %BigInt %val
+}
+
+declare void @bar()
diff --git a/test/CodeGen/ARM/jumptable-label.ll b/test/CodeGen/ARM/jumptable-label.ll
index 49d698672f82..2ba90dc97365 100644
--- a/test/CodeGen/ARM/jumptable-label.ll
+++ b/test/CodeGen/ARM/jumptable-label.ll
@@ -2,8 +2,8 @@
; test that we print the label of a bb that is only used in a jump table.
-; CHECK: .long LBB0_2
-; CHECK: LBB0_2:
+; CHECK: .long [[JUMPTABLE_DEST:LBB[0-9]+_[0-9]+]]
+; CHECK: [[JUMPTABLE_DEST]]:
define i32 @calculate() {
entry:
diff --git a/test/CodeGen/ARM/ldrd.ll b/test/CodeGen/ARM/ldrd.ll
index 7ce846844e05..a8070ea68aa2 100644
--- a/test/CodeGen/ARM/ldrd.ll
+++ b/test/CodeGen/ARM/ldrd.ll
@@ -92,6 +92,22 @@ entry:
ret void
}
+declare void @extfunc(i32, i32, i32, i32)
+
+; CHECK-LABEL: Func2:
+; A8: ldrd
+; A8: blx
+; A8: pop
+define void @Func2(i32* %p) {
+entry:
+ %addr0 = getelementptr i32, i32* %p, i32 0
+ %addr1 = getelementptr i32, i32* %p, i32 1
+ %v0 = load i32, i32* %addr0
+ %v1 = load i32, i32* %addr1
+ ; try to force %v0/%v1 into non-adjacent registers
+ call void @extfunc(i32 %v0, i32 0, i32 0, i32 %v1)
+ ret void
+}
declare void @llvm.lifetime.start(i64, i8* nocapture) nounwind
declare void @llvm.lifetime.end(i64, i8* nocapture) nounwind
diff --git a/test/CodeGen/ARM/named-reg-alloc.ll b/test/CodeGen/ARM/named-reg-alloc.ll
index 380cf39734ff..d41fa64882c8 100644
--- a/test/CodeGen/ARM/named-reg-alloc.ll
+++ b/test/CodeGen/ARM/named-reg-alloc.ll
@@ -4,7 +4,7 @@
define i32 @get_stack() nounwind {
entry:
; FIXME: Include an allocatable-specific error message
-; CHECK: Invalid register name global variable
+; CHECK: Invalid register name "r5".
%sp = call i32 @llvm.read_register.i32(metadata !0)
ret i32 %sp
}
diff --git a/test/CodeGen/ARM/named-reg-notareg.ll b/test/CodeGen/ARM/named-reg-notareg.ll
index 3ac03f4fdaaa..45cb38f30f35 100644
--- a/test/CodeGen/ARM/named-reg-notareg.ll
+++ b/test/CodeGen/ARM/named-reg-notareg.ll
@@ -3,7 +3,7 @@
define i32 @get_stack() nounwind {
entry:
-; CHECK: Invalid register name global variable
+; CHECK: Invalid register name "notareg".
%sp = call i32 @llvm.read_register.i32(metadata !0)
ret i32 %sp
}
diff --git a/test/CodeGen/ARM/special-reg-acore.ll b/test/CodeGen/ARM/special-reg-acore.ll
new file mode 100644
index 000000000000..3d65ff44bfb0
--- /dev/null
+++ b/test/CodeGen/ARM/special-reg-acore.ll
@@ -0,0 +1,78 @@
+; RUN: llc < %s -mtriple=arm-none-eabi -mcpu=cortex-a8 2>&1 | FileCheck %s --check-prefix=ACORE
+; RUN: not llc < %s -mtriple=thumb-none-eabi -mcpu=cortex-m4 2>&1 | FileCheck %s --check-prefix=MCORE
+
+; MCORE: LLVM ERROR: Invalid register name "cpsr".
+
+define i32 @read_cpsr() nounwind {
+ ; ACORE-LABEL: read_cpsr:
+ ; ACORE: mrs r0, apsr
+ %reg = call i32 @llvm.read_register.i32(metadata !1)
+ ret i32 %reg
+}
+
+define i32 @read_aclass_registers() nounwind {
+entry:
+ ; ACORE-LABEL: read_aclass_registers:
+ ; ACORE: mrs r0, apsr
+ ; ACORE: mrs r1, spsr
+
+ %0 = call i32 @llvm.read_register.i32(metadata !0)
+ %1 = call i32 @llvm.read_register.i32(metadata !1)
+ %add1 = add i32 %1, %0
+ %2 = call i32 @llvm.read_register.i32(metadata !2)
+ %add2 = add i32 %add1, %2
+ ret i32 %add2
+}
+
+define void @write_aclass_registers(i32 %x) nounwind {
+entry:
+ ; ACORE-LABEL: write_aclass_registers:
+ ; ACORE: msr APSR_nzcvq, r0
+ ; ACORE: msr APSR_g, r0
+ ; ACORE: msr APSR_nzcvqg, r0
+ ; ACORE: msr CPSR_c, r0
+ ; ACORE: msr CPSR_x, r0
+ ; ACORE: msr APSR_g, r0
+ ; ACORE: msr APSR_nzcvq, r0
+ ; ACORE: msr CPSR_fsxc, r0
+ ; ACORE: msr SPSR_c, r0
+ ; ACORE: msr SPSR_x, r0
+ ; ACORE: msr SPSR_s, r0
+ ; ACORE: msr SPSR_f, r0
+ ; ACORE: msr SPSR_fsxc, r0
+
+ call void @llvm.write_register.i32(metadata !3, i32 %x)
+ call void @llvm.write_register.i32(metadata !4, i32 %x)
+ call void @llvm.write_register.i32(metadata !5, i32 %x)
+ call void @llvm.write_register.i32(metadata !6, i32 %x)
+ call void @llvm.write_register.i32(metadata !7, i32 %x)
+ call void @llvm.write_register.i32(metadata !8, i32 %x)
+ call void @llvm.write_register.i32(metadata !9, i32 %x)
+ call void @llvm.write_register.i32(metadata !10, i32 %x)
+ call void @llvm.write_register.i32(metadata !11, i32 %x)
+ call void @llvm.write_register.i32(metadata !12, i32 %x)
+ call void @llvm.write_register.i32(metadata !13, i32 %x)
+ call void @llvm.write_register.i32(metadata !14, i32 %x)
+ call void @llvm.write_register.i32(metadata !15, i32 %x)
+ ret void
+}
+
+declare i32 @llvm.read_register.i32(metadata) nounwind
+declare void @llvm.write_register.i32(metadata, i32) nounwind
+
+!0 = !{!"apsr"}
+!1 = !{!"cpsr"}
+!2 = !{!"spsr"}
+!3 = !{!"apsr_nzcvq"}
+!4 = !{!"apsr_g"}
+!5 = !{!"apsr_nzcvqg"}
+!6 = !{!"cpsr_c"}
+!7 = !{!"cpsr_x"}
+!8 = !{!"cpsr_s"}
+!9 = !{!"cpsr_f"}
+!10 = !{!"cpsr_cxsf"}
+!11 = !{!"spsr_c"}
+!12 = !{!"spsr_x"}
+!13 = !{!"spsr_s"}
+!14 = !{!"spsr_f"}
+!15 = !{!"spsr_cxsf"}
diff --git a/test/CodeGen/ARM/special-reg-mcore.ll b/test/CodeGen/ARM/special-reg-mcore.ll
new file mode 100644
index 000000000000..686da0f6b839
--- /dev/null
+++ b/test/CodeGen/ARM/special-reg-mcore.ll
@@ -0,0 +1,143 @@
+; RUN: llc < %s -mtriple=thumb-none-eabi -mcpu=cortex-m4 2>&1 | FileCheck %s --check-prefix=MCORE
+; RUN: not llc < %s -mtriple=thumb-none-eabi -mcpu=cortex-m3 2>&1 | FileCheck %s --check-prefix=M3CORE
+; RUN: not llc < %s -mtriple=arm-none-eabi -mcpu=cortex-a8 2>&1 | FileCheck %s --check-prefix=ACORE
+
+; ACORE: LLVM ERROR: Invalid register name "control".
+; M3CORE: LLVM ERROR: Invalid register name "control".
+
+define i32 @read_mclass_registers() nounwind {
+entry:
+ ; MCORE-LABEL: read_mclass_registers:
+ ; MCORE: mrs r0, apsr
+ ; MCORE: mrs r1, iapsr
+ ; MCORE: mrs r1, eapsr
+ ; MCORE: mrs r1, xpsr
+ ; MCORE: mrs r1, ipsr
+ ; MCORE: mrs r1, epsr
+ ; MCORE: mrs r1, iepsr
+ ; MCORE: mrs r1, msp
+ ; MCORE: mrs r1, psp
+ ; MCORE: mrs r1, primask
+ ; MCORE: mrs r1, basepri
+ ; MCORE: mrs r1, basepri_max
+ ; MCORE: mrs r1, faultmask
+ ; MCORE: mrs r1, control
+
+ %0 = call i32 @llvm.read_register.i32(metadata !0)
+ %1 = call i32 @llvm.read_register.i32(metadata !4)
+ %add1 = add i32 %1, %0
+ %2 = call i32 @llvm.read_register.i32(metadata !8)
+ %add2 = add i32 %add1, %2
+ %3 = call i32 @llvm.read_register.i32(metadata !12)
+ %add3 = add i32 %add2, %3
+ %4 = call i32 @llvm.read_register.i32(metadata !16)
+ %add4 = add i32 %add3, %4
+ %5 = call i32 @llvm.read_register.i32(metadata !17)
+ %add5 = add i32 %add4, %5
+ %6 = call i32 @llvm.read_register.i32(metadata !18)
+ %add6 = add i32 %add5, %6
+ %7 = call i32 @llvm.read_register.i32(metadata !19)
+ %add7 = add i32 %add6, %7
+ %8 = call i32 @llvm.read_register.i32(metadata !20)
+ %add8 = add i32 %add7, %8
+ %9 = call i32 @llvm.read_register.i32(metadata !21)
+ %add9 = add i32 %add8, %9
+ %10 = call i32 @llvm.read_register.i32(metadata !22)
+ %add10 = add i32 %add9, %10
+ %11 = call i32 @llvm.read_register.i32(metadata !23)
+ %add11 = add i32 %add10, %11
+ %12 = call i32 @llvm.read_register.i32(metadata !24)
+ %add12 = add i32 %add11, %12
+ %13 = call i32 @llvm.read_register.i32(metadata !25)
+ %add13 = add i32 %add12, %13
+ ret i32 %add13
+}
+
+define void @write_mclass_registers(i32 %x) nounwind {
+entry:
+ ; MCORE-LABEL: write_mclass_registers:
+ ; MCORE: msr apsr_nzcvqg, r0
+ ; MCORE: msr apsr_nzcvq, r0
+ ; MCORE: msr apsr_g, r0
+ ; MCORE: msr apsr_nzcvqg, r0
+ ; MCORE: msr iapsr_nzcvqg, r0
+ ; MCORE: msr iapsr_nzcvq, r0
+ ; MCORE: msr iapsr_g, r0
+ ; MCORE: msr iapsr_nzcvqg, r0
+ ; MCORE: msr eapsr_nzcvqg, r0
+ ; MCORE: msr eapsr_nzcvq, r0
+ ; MCORE: msr eapsr_g, r0
+ ; MCORE: msr eapsr_nzcvqg, r0
+ ; MCORE: msr xpsr_nzcvqg, r0
+ ; MCORE: msr xpsr_nzcvq, r0
+ ; MCORE: msr xpsr_g, r0
+ ; MCORE: msr xpsr_nzcvqg, r0
+ ; MCORE: msr ipsr, r0
+ ; MCORE: msr epsr, r0
+ ; MCORE: msr iepsr, r0
+ ; MCORE: msr msp, r0
+ ; MCORE: msr psp, r0
+ ; MCORE: msr primask, r0
+ ; MCORE: msr basepri, r0
+ ; MCORE: msr basepri_max, r0
+ ; MCORE: msr faultmask, r0
+ ; MCORE: msr control, r0
+
+ call void @llvm.write_register.i32(metadata !0, i32 %x)
+ call void @llvm.write_register.i32(metadata !1, i32 %x)
+ call void @llvm.write_register.i32(metadata !2, i32 %x)
+ call void @llvm.write_register.i32(metadata !3, i32 %x)
+ call void @llvm.write_register.i32(metadata !4, i32 %x)
+ call void @llvm.write_register.i32(metadata !5, i32 %x)
+ call void @llvm.write_register.i32(metadata !6, i32 %x)
+ call void @llvm.write_register.i32(metadata !7, i32 %x)
+ call void @llvm.write_register.i32(metadata !8, i32 %x)
+ call void @llvm.write_register.i32(metadata !9, i32 %x)
+ call void @llvm.write_register.i32(metadata !10, i32 %x)
+ call void @llvm.write_register.i32(metadata !11, i32 %x)
+ call void @llvm.write_register.i32(metadata !12, i32 %x)
+ call void @llvm.write_register.i32(metadata !13, i32 %x)
+ call void @llvm.write_register.i32(metadata !14, i32 %x)
+ call void @llvm.write_register.i32(metadata !15, i32 %x)
+ call void @llvm.write_register.i32(metadata !16, i32 %x)
+ call void @llvm.write_register.i32(metadata !17, i32 %x)
+ call void @llvm.write_register.i32(metadata !18, i32 %x)
+ call void @llvm.write_register.i32(metadata !19, i32 %x)
+ call void @llvm.write_register.i32(metadata !20, i32 %x)
+ call void @llvm.write_register.i32(metadata !21, i32 %x)
+ call void @llvm.write_register.i32(metadata !22, i32 %x)
+ call void @llvm.write_register.i32(metadata !23, i32 %x)
+ call void @llvm.write_register.i32(metadata !24, i32 %x)
+ call void @llvm.write_register.i32(metadata !25, i32 %x)
+ ret void
+}
+
+declare i32 @llvm.read_register.i32(metadata) nounwind
+declare void @llvm.write_register.i32(metadata, i32) nounwind
+
+!0 = !{!"apsr"}
+!1 = !{!"apsr_nzcvq"}
+!2 = !{!"apsr_g"}
+!3 = !{!"apsr_nzcvqg"}
+!4 = !{!"iapsr"}
+!5 = !{!"iapsr_nzcvq"}
+!6 = !{!"iapsr_g"}
+!7 = !{!"iapsr_nzcvqg"}
+!8 = !{!"eapsr"}
+!9 = !{!"eapsr_nzcvq"}
+!10 = !{!"eapsr_g"}
+!11 = !{!"eapsr_nzcvqg"}
+!12 = !{!"xpsr"}
+!13 = !{!"xpsr_nzcvq"}
+!14 = !{!"xpsr_g"}
+!15 = !{!"xpsr_nzcvqg"}
+!16 = !{!"ipsr"}
+!17 = !{!"epsr"}
+!18 = !{!"iepsr"}
+!19 = !{!"msp"}
+!20 = !{!"psp"}
+!21 = !{!"primask"}
+!22 = !{!"basepri"}
+!23 = !{!"basepri_max"}
+!24 = !{!"faultmask"}
+!25 = !{!"control"}
diff --git a/test/CodeGen/ARM/special-reg.ll b/test/CodeGen/ARM/special-reg.ll
new file mode 100644
index 000000000000..7ccb490f5d4a
--- /dev/null
+++ b/test/CodeGen/ARM/special-reg.ll
@@ -0,0 +1,78 @@
+; RUN: llc < %s -mtriple=arm-none-eabi -mcpu=cortex-a8 2>&1 | FileCheck %s --check-prefix=ARM --check-prefix=ACORE
+; RUN: llc < %s -mtriple=thumb-none-eabi -mcpu=cortex-m4 2>&1 | FileCheck %s --check-prefix=ARM --check-prefix=MCORE
+
+define i32 @read_i32_encoded_register() nounwind {
+entry:
+; ARM-LABEL: read_i32_encoded_register:
+; ARM: mrc p1, #2, r0, c3, c4, #5
+ %reg = call i32 @llvm.read_register.i32(metadata !0)
+ ret i32 %reg
+}
+
+define i64 @read_i64_encoded_register() nounwind {
+entry:
+; ARM-LABEL: read_i64_encoded_register:
+; ARM: mrrc p1, #2, r0, r1, c3
+ %reg = call i64 @llvm.read_register.i64(metadata !1)
+ ret i64 %reg
+}
+
+define i32 @read_apsr() nounwind {
+entry:
+; ARM-LABEL: read_apsr:
+; ARM: mrs r0, apsr
+ %reg = call i32 @llvm.read_register.i32(metadata !2)
+ ret i32 %reg
+}
+
+define i32 @read_fpscr() nounwind {
+entry:
+; ARM-LABEL: read_fpscr:
+; ARM: vmrs r0, fpscr
+ %reg = call i32 @llvm.read_register.i32(metadata !3)
+ ret i32 %reg
+}
+
+define void @write_i32_encoded_register(i32 %x) nounwind {
+entry:
+; ARM-LABEL: write_i32_encoded_register:
+; ARM: mcr p1, #2, r0, c3, c4, #5
+ call void @llvm.write_register.i32(metadata !0, i32 %x)
+ ret void
+}
+
+define void @write_i64_encoded_register(i64 %x) nounwind {
+entry:
+; ARM-LABEL: write_i64_encoded_register:
+; ARM: mcrr p1, #2, r0, r1, c3
+ call void @llvm.write_register.i64(metadata !1, i64 %x)
+ ret void
+}
+
+define void @write_apsr(i32 %x) nounwind {
+entry:
+; ARM-LABEL: write_apsr:
+; ACORE: msr APSR_nzcvq, r0
+; MCORE: msr apsr_nzcvq, r0
+ call void @llvm.write_register.i32(metadata !4, i32 %x)
+ ret void
+}
+
+define void @write_fpscr(i32 %x) nounwind {
+entry:
+; ARM-LABEL: write_fpscr:
+; ARM: vmsr fpscr, r0
+ call void @llvm.write_register.i32(metadata !3, i32 %x)
+ ret void
+}
+
+declare i32 @llvm.read_register.i32(metadata) nounwind
+declare i64 @llvm.read_register.i64(metadata) nounwind
+declare void @llvm.write_register.i32(metadata, i32) nounwind
+declare void @llvm.write_register.i64(metadata, i64) nounwind
+
+!0 = !{!"cp1:2:c3:c4:5"}
+!1 = !{!"cp1:2:c3"}
+!2 = !{!"apsr"}
+!3 = !{!"fpscr"}
+!4 = !{!"apsr_nzcvq"}
diff --git a/test/CodeGen/BPF/alu8.ll b/test/CodeGen/BPF/alu8.ll
index 0233225f81b5..c1c2bd29f247 100644
--- a/test/CodeGen/BPF/alu8.ll
+++ b/test/CodeGen/BPF/alu8.ll
@@ -1,5 +1,4 @@
-; RUN: llc -march=bpf -show-mc-encoding < %s | FileCheck %s
-; test little endian only for now
+; RUN: llc -march=bpfel -show-mc-encoding < %s | FileCheck %s
define i8 @mov(i8 %a, i8 %b) nounwind {
; CHECK-LABEL: mov:
diff --git a/test/CodeGen/BPF/atomics.ll b/test/CodeGen/BPF/atomics.ll
index 2f9730ddddef..a2e17d291968 100644
--- a/test/CodeGen/BPF/atomics.ll
+++ b/test/CodeGen/BPF/atomics.ll
@@ -1,5 +1,4 @@
-; RUN: llc < %s -march=bpf -verify-machineinstrs -show-mc-encoding | FileCheck %s
-; test little endian only for now
+; RUN: llc < %s -march=bpfel -verify-machineinstrs -show-mc-encoding | FileCheck %s
; CHECK-LABEL: test_load_add_32
; CHECK: xadd32
diff --git a/test/CodeGen/BPF/basictest.ll b/test/CodeGen/BPF/basictest.ll
index 2a2d49878a63..82feb43d005c 100644
--- a/test/CodeGen/BPF/basictest.ll
+++ b/test/CodeGen/BPF/basictest.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
define i32 @test0(i32 %X) {
%tmp.1 = add i32 %X, 1
diff --git a/test/CodeGen/BPF/cc_args.ll b/test/CodeGen/BPF/cc_args.ll
index 5085fe5684eb..8e3f8604ac88 100644
--- a/test/CodeGen/BPF/cc_args.ll
+++ b/test/CodeGen/BPF/cc_args.ll
@@ -1,5 +1,4 @@
-; RUN: llc < %s -march=bpf -show-mc-encoding | FileCheck %s
-; test little endian only for now
+; RUN: llc < %s -march=bpfel -show-mc-encoding | FileCheck %s
define void @test() #0 {
entry:
diff --git a/test/CodeGen/BPF/cc_args_be.ll b/test/CodeGen/BPF/cc_args_be.ll
new file mode 100644
index 000000000000..59a7439728ba
--- /dev/null
+++ b/test/CodeGen/BPF/cc_args_be.ll
@@ -0,0 +1,96 @@
+; RUN: llc < %s -march=bpfeb -show-mc-encoding | FileCheck %s
+; test big endian
+
+define void @test() #0 {
+entry:
+; CHECK: test:
+
+; CHECK: mov r1, 123 # encoding: [0xb7,0x10,0x00,0x00,0x00,0x00,0x00,0x7b]
+; CHECK: call f_i16
+ call void @f_i16(i16 123)
+
+; CHECK: mov r1, 12345678 # encoding: [0xb7,0x10,0x00,0x00,0x00,0xbc,0x61,0x4e]
+; CHECK: call f_i32
+ call void @f_i32(i32 12345678)
+
+; CHECK: ld_64 r1, 72623859790382856 # encoding: [0x18,0x10,0x00,0x00,0x05,0x06,0x07,0x08,0x00,0x00,0x00,0x00,0x01,0x02,0x03,0x04]
+; CHECK: call f_i64
+ call void @f_i64(i64 72623859790382856)
+
+; CHECK: mov r1, 1234
+; CHECK: mov r2, 5678
+; CHECK: call f_i32_i32
+ call void @f_i32_i32(i32 1234, i32 5678)
+
+; CHECK: mov r1, 2
+; CHECK: mov r2, 3
+; CHECK: mov r3, 4
+; CHECK: call f_i16_i32_i16
+ call void @f_i16_i32_i16(i16 2, i32 3, i16 4)
+
+; CHECK: mov r1, 5
+; CHECK: ld_64 r2, 7262385979038285
+; CHECK: mov r3, 6
+; CHECK: call f_i16_i64_i16
+ call void @f_i16_i64_i16(i16 5, i64 7262385979038285, i16 6)
+
+ ret void
+}
+
+@g_i16 = common global i16 0, align 2
+@g_i32 = common global i32 0, align 2
+@g_i64 = common global i64 0, align 4
+
+define void @f_i16(i16 %a) #0 {
+; CHECK: f_i16:
+; CHECK: sth 0(r2), r1 # encoding: [0x6b,0x21,0x00,0x00,0x00,0x00,0x00,0x00]
+ store volatile i16 %a, i16* @g_i16, align 2
+ ret void
+}
+
+define void @f_i32(i32 %a) #0 {
+; CHECK: f_i32:
+; CHECK: sth 2(r2), r1 # encoding: [0x6b,0x21,0x00,0x02,0x00,0x00,0x00,0x00]
+; CHECK: sth 0(r2), r1 # encoding: [0x6b,0x21,0x00,0x00,0x00,0x00,0x00,0x00]
+ store volatile i32 %a, i32* @g_i32, align 2
+ ret void
+}
+
+define void @f_i64(i64 %a) #0 {
+; CHECK: f_i64:
+; CHECK: stw 4(r2), r1 # encoding: [0x63,0x21,0x00,0x04,0x00,0x00,0x00,0x00]
+; CHECK: stw 0(r2), r1
+ store volatile i64 %a, i64* @g_i64, align 2
+ ret void
+}
+
+define void @f_i32_i32(i32 %a, i32 %b) #0 {
+; CHECK: f_i32_i32:
+; CHECK: stw 0(r3), r1
+ store volatile i32 %a, i32* @g_i32, align 4
+; CHECK: stw 0(r3), r2
+ store volatile i32 %b, i32* @g_i32, align 4
+ ret void
+}
+
+define void @f_i16_i32_i16(i16 %a, i32 %b, i16 %c) #0 {
+; CHECK: f_i16_i32_i16:
+; CHECK: sth 0(r4), r1
+ store volatile i16 %a, i16* @g_i16, align 2
+; CHECK: stw 0(r1), r2
+ store volatile i32 %b, i32* @g_i32, align 4
+; CHECK: sth 0(r4), r3
+ store volatile i16 %c, i16* @g_i16, align 2
+ ret void
+}
+
+define void @f_i16_i64_i16(i16 %a, i64 %b, i16 %c) #0 {
+; CHECK: f_i16_i64_i16:
+; CHECK: sth 0(r4), r1
+ store volatile i16 %a, i16* @g_i16, align 2
+; CHECK: std 0(r1), r2 # encoding: [0x7b,0x12,0x00,0x00,0x00,0x00,0x00,0x00]
+ store volatile i64 %b, i64* @g_i64, align 8
+; CHECK: sth 0(r4), r3
+ store volatile i16 %c, i16* @g_i16, align 2
+ ret void
+}
diff --git a/test/CodeGen/BPF/cc_ret.ll b/test/CodeGen/BPF/cc_ret.ll
index e32b17bcc61c..09574922f325 100644
--- a/test/CodeGen/BPF/cc_ret.ll
+++ b/test/CodeGen/BPF/cc_ret.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
define void @test() #0 {
entry:
diff --git a/test/CodeGen/BPF/ex1.ll b/test/CodeGen/BPF/ex1.ll
index be038e9a3d8c..546e5d49da69 100644
--- a/test/CodeGen/BPF/ex1.ll
+++ b/test/CodeGen/BPF/ex1.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
%struct.bpf_context = type { i64, i64, i64, i64, i64, i64, i64 }
%struct.sk_buff = type { i64, i64, i64, i64, i64, i64, i64 }
diff --git a/test/CodeGen/BPF/intrinsics.ll b/test/CodeGen/BPF/intrinsics.ll
index 98b57deb7c8d..483473e922fc 100644
--- a/test/CodeGen/BPF/intrinsics.ll
+++ b/test/CodeGen/BPF/intrinsics.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf -show-mc-encoding | FileCheck %s
+; RUN: llc < %s -march=bpfel -show-mc-encoding | FileCheck %s
; Function Attrs: nounwind uwtable
define i32 @ld_b(i64 %foo, i64* nocapture %bar, i8* %ctx, i8* %ctx2) #0 {
diff --git a/test/CodeGen/BPF/load.ll b/test/CodeGen/BPF/load.ll
index 03fb17c965b5..d4ba315b5f18 100644
--- a/test/CodeGen/BPF/load.ll
+++ b/test/CodeGen/BPF/load.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
define i16 @am1(i16* %a) nounwind {
%1 = load i16, i16* %a
diff --git a/test/CodeGen/BPF/loops.ll b/test/CodeGen/BPF/loops.ll
index 4798d78842ca..00be54b3bac5 100644
--- a/test/CodeGen/BPF/loops.ll
+++ b/test/CodeGen/BPF/loops.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
define zeroext i16 @add(i16* nocapture %a, i16 zeroext %n) nounwind readonly {
entry:
diff --git a/test/CodeGen/BPF/sanity.ll b/test/CodeGen/BPF/sanity.ll
index 09a6b65d0854..7f0ef889ff99 100644
--- a/test/CodeGen/BPF/sanity.ll
+++ b/test/CodeGen/BPF/sanity.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=bpf | FileCheck %s
+; RUN: llc < %s -march=bpfel | FileCheck %s
@foo_printf.fmt = private unnamed_addr constant [9 x i8] c"hello \0A\00", align 1
diff --git a/test/CodeGen/BPF/setcc.ll b/test/CodeGen/BPF/setcc.ll
index eabb6c9bf2d6..f6c6db6c6836 100644
--- a/test/CodeGen/BPF/setcc.ll
+++ b/test/CodeGen/BPF/setcc.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=bpf < %s | FileCheck %s
+; RUN: llc -march=bpfel < %s | FileCheck %s
define i16 @sccweqand(i16 %a, i16 %b) nounwind {
%t1 = and i16 %a, %b
diff --git a/test/CodeGen/BPF/shifts.ll b/test/CodeGen/BPF/shifts.ll
index 898ae2d46123..cb000b92fcd9 100644
--- a/test/CodeGen/BPF/shifts.ll
+++ b/test/CodeGen/BPF/shifts.ll
@@ -1,5 +1,4 @@
-; RUN: llc < %s -march=bpf -show-mc-encoding | FileCheck %s
-; test little endian only for now
+; RUN: llc < %s -march=bpfel -show-mc-encoding | FileCheck %s
define zeroext i8 @lshr8(i8 zeroext %a, i8 zeroext %cnt) nounwind readnone {
entry:
diff --git a/test/CodeGen/BPF/sockex2.ll b/test/CodeGen/BPF/sockex2.ll
index 6ae5e1c8d6bf..d372a5982f68 100644
--- a/test/CodeGen/BPF/sockex2.ll
+++ b/test/CodeGen/BPF/sockex2.ll
@@ -1,5 +1,4 @@
-; RUN: llc < %s -march=bpf -show-mc-encoding | FileCheck %s
-; test little endian only for now
+; RUN: llc < %s -march=bpfel -show-mc-encoding | FileCheck %s
%struct.bpf_map_def = type { i32, i32, i32, i32 }
%struct.sk_buff = type opaque
diff --git a/test/CodeGen/Generic/stop-after.ll b/test/CodeGen/Generic/stop-after.ll
index 557e097840af..791378c3737d 100644
--- a/test/CodeGen/Generic/stop-after.ll
+++ b/test/CodeGen/Generic/stop-after.ll
@@ -1,9 +1,10 @@
; RUN: llc < %s -debug-pass=Structure -stop-after=loop-reduce -o /dev/null 2>&1 | FileCheck %s -check-prefix=STOP
; RUN: llc < %s -debug-pass=Structure -start-after=loop-reduce -o /dev/null 2>&1 | FileCheck %s -check-prefix=START
-; STOP: -loop-reduce -print-module
+; STOP: -loop-reduce
; STOP: Loop Strength Reduction
; STOP-NEXT: Machine Function Analysis
+; STOP-NEXT: MIR Printing Pass
; START: -machine-branch-prob -gc-lowering
; START: FunctionPass Manager
diff --git a/test/CodeGen/Hexagon/args.ll b/test/CodeGen/Hexagon/args.ll
index aea4ffe2eee5..1c470f68aa27 100644
--- a/test/CodeGen/Hexagon/args.ll
+++ b/test/CodeGen/Hexagon/args.ll
@@ -1,8 +1,8 @@
-; RUN: llc -march=hexagon -mcpu=hexagonv4 -disable-dfa-sched -disable-hexagon-misched < %s | FileCheck %s
-; CHECK: memw(r29{{ *}}+{{ *}}#0){{ *}}={{ *}}#7
-; CHECK: r1:0 = combine(#2, #1)
-; CHECK: r3:2 = combine(#4, #3)
+; RUN: llc -march=hexagon < %s | FileCheck %s
; CHECK: r5:4 = combine(#6, #5)
+; CHECK: r3:2 = combine(#4, #3)
+; CHECK: r1:0 = combine(#2, #1)
+; CHECK: memw(r29{{ *}}+{{ *}}#0){{ *}}={{ *}}#7
define void @foo() nounwind {
diff --git a/test/CodeGen/Hexagon/calling-conv.ll b/test/CodeGen/Hexagon/calling-conv.ll
deleted file mode 100644
index 7133c1ae7aad..000000000000
--- a/test/CodeGen/Hexagon/calling-conv.ll
+++ /dev/null
@@ -1,73 +0,0 @@
-; RUN: llc -march=hexagon -mcpu=hexagonv5 <%s | \
-; RUN: FileCheck %s --check-prefix=CHECK-ONE
-; RUN: llc -march=hexagon -mcpu=hexagonv5 <%s | \
-; RUN: FileCheck %s --check-prefix=CHECK-TWO
-; RUN: llc -march=hexagon -mcpu=hexagonv5 <%s | \
-; RUN: FileCheck %s --check-prefix=CHECK-THREE
-
-%struct.test_struct = type { i32, i8, i64 }
-%struct.test_struct_long = type { i8, i64 }
-
-@mystruct = external global %struct.test_struct*, align 4
-
-; CHECK-ONE: memw(r29+#48) = r2
-; CHECK-TWO: memw(r29+#52) = r2
-; CHECK-THREE: memw(r29+#56) = r2
-; Function Attrs: nounwind
-define void @foo(%struct.test_struct* noalias sret %agg.result, i32 %a, i8 zeroext %c, %struct.test_struct* byval %s, %struct.test_struct_long* byval %t) #0 {
-entry:
- %a.addr = alloca i32, align 4
- %c.addr = alloca i8, align 1
- %z = alloca i32, align 4
- %ret = alloca %struct.test_struct, align 8
- store i32 %a, i32* %a.addr, align 4
- store i8 %c, i8* %c.addr, align 1
- %0 = bitcast i32* %z to i8*
- call void @llvm.lifetime.start(i64 4, i8* %0) #1
- store i32 45, i32* %z, align 4
- %1 = bitcast %struct.test_struct* %ret to i8*
- call void @llvm.lifetime.start(i64 16, i8* %1) #1
- %2 = load i32, i32* %z, align 4
- %3 = load %struct.test_struct*, %struct.test_struct** @mystruct, align 4
- %4 = load %struct.test_struct*, %struct.test_struct** @mystruct, align 4
- %5 = load i8, i8* %c.addr, align 1
- %6 = load i32, i32* %a.addr, align 4
- %conv = sext i32 %6 to i64
- %add = add nsw i64 %conv, 1
- %7 = load i32, i32* %a.addr, align 4
- %add1 = add nsw i32 %7, 2
- %8 = load i32, i32* %a.addr, align 4
- %conv2 = sext i32 %8 to i64
- %add3 = add nsw i64 %conv2, 3
- %9 = load i8, i8* %c.addr, align 1
- %10 = load i8, i8* %c.addr, align 1
- %11 = load i8, i8* %c.addr, align 1
- %12 = load i32, i32* %z, align 4
- call void @bar(%struct.test_struct* sret %ret, i32 %2, %struct.test_struct* byval %3, %struct.test_struct* byval %4, i8 zeroext %5, i64 %add, i32 %add1, i64 %add3, i8 zeroext %9, i8 zeroext %10, i8 zeroext %11, i32 %12)
- %x = getelementptr inbounds %struct.test_struct, %struct.test_struct* %ret, i32 0, i32 0
- store i32 20, i32* %x, align 4
- %13 = bitcast %struct.test_struct* %agg.result to i8*
- %14 = bitcast %struct.test_struct* %ret to i8*
- call void @llvm.memcpy.p0i8.p0i8.i32(i8* %13, i8* %14, i32 16, i32 8, i1 false)
- %15 = bitcast %struct.test_struct* %ret to i8*
- call void @llvm.lifetime.end(i64 16, i8* %15) #1
- %16 = bitcast i32* %z to i8*
- call void @llvm.lifetime.end(i64 4, i8* %16) #1
- ret void
-}
-
-; Function Attrs: nounwind
-declare void @llvm.lifetime.start(i64, i8* nocapture) #1
-
-declare void @bar(%struct.test_struct* sret, i32, %struct.test_struct* byval, %struct.test_struct* byval, i8 zeroext, i64, i32, i64, i8 zeroext, i8 zeroext, i8 zeroext, i32) #2
-
-; Function Attrs: nounwind
-declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i32, i1) #1
-
-; Function Attrs: nounwind
-declare void @llvm.lifetime.end(i64, i8* nocapture) #1
-
-attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="hexagonv4" "unsafe-fp-math"="false" "use-soft-float"="false" }
-attributes #1 = { nounwind }
-attributes #2 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="hexagonv4" "unsafe-fp-math"="false" "use-soft-float"="false" }
-
diff --git a/test/CodeGen/Hexagon/cext-valid-packet1.ll b/test/CodeGen/Hexagon/cext-valid-packet1.ll
index a479d37e4ae5..35e7b364b508 100644
--- a/test/CodeGen/Hexagon/cext-valid-packet1.ll
+++ b/test/CodeGen/Hexagon/cext-valid-packet1.ll
@@ -1,4 +1,5 @@
; RUN: llc -march=hexagon -mcpu=hexagonv4 < %s | FileCheck %s
+; XFAIL:
; Check that the packetizer generates valid packets with constant
; extended instructions.
diff --git a/test/CodeGen/Hexagon/cext-valid-packet2.ll b/test/CodeGen/Hexagon/cext-valid-packet2.ll
index 2eba74329960..c3a4915ec2e0 100644
--- a/test/CodeGen/Hexagon/cext-valid-packet2.ll
+++ b/test/CodeGen/Hexagon/cext-valid-packet2.ll
@@ -1,4 +1,5 @@
; RUN: llc -march=hexagon -mcpu=hexagonv4 < %s | FileCheck %s
+; XFAIL:
; Check that the packetizer generates valid packets with constant
; extended add and base+offset store instructions.
diff --git a/test/CodeGen/Hexagon/compound.ll b/test/CodeGen/Hexagon/compound.ll
new file mode 100644
index 000000000000..f8d36b8b77d9
--- /dev/null
+++ b/test/CodeGen/Hexagon/compound.ll
@@ -0,0 +1,17 @@
+; RUN: llc -march=hexagon -filetype=obj -o - %s | llvm-objdump -d - | FileCheck %s
+
+; CHECK: p0 = cmp.gt(r0,#-1); if (!p0.new) jump:nt
+
+declare void @a()
+declare void @b()
+
+define void @foo(i32 %a) {
+%b = icmp sgt i32 %a, -1
+br i1 %b, label %x, label %y
+x:
+call void @a()
+ret void
+y:
+call void @b()
+ret void
+} \ No newline at end of file
diff --git a/test/CodeGen/Hexagon/dualstore.ll b/test/CodeGen/Hexagon/dualstore.ll
index 33d9ce9b9351..9f4569d6459c 100644
--- a/test/CodeGen/Hexagon/dualstore.ll
+++ b/test/CodeGen/Hexagon/dualstore.ll
@@ -1,12 +1,11 @@
-; RUN: llc -march=hexagon -disable-hexagon-misched < %s | FileCheck %s
+; RUN: llc -march=hexagon -filetype=obj %s -o - | llvm-objdump -d - | FileCheck %s
; Check that we generate dual stores in one packet in V4
-; CHECK: memw(r{{[0-9]+}}{{ *}}+{{ *}}#{{[0-9]+}}){{ *}}=
-; CHECK-NEXT: memw(r{{[0-9]+}}{{ *}}+{{ *}}#{{[0-9]+}}){{ *}}=
+; CHECK: 00 40 9f 52 529f4000
+; CHECK: 10 10 00 f0 f0001010
-define i32 @main(i32 %v, i32* %p1, i32* %p2) nounwind {
-entry:
- store i32 %v, i32* %p1, align 4
- store i32 %v, i32* %p2, align 4
- ret i32 0
+define void @foo(i32* %a, i32* %b) {
+ store i32 0, i32* %a
+ store i32 0, i32* %b
+ ret void
}
diff --git a/test/CodeGen/Hexagon/duplex.ll b/test/CodeGen/Hexagon/duplex.ll
new file mode 100644
index 000000000000..80fe61ceccca
--- /dev/null
+++ b/test/CodeGen/Hexagon/duplex.ll
@@ -0,0 +1,7 @@
+; RUN: llc -march=hexagon -filetype=obj -o - %s | llvm-objdump -d - | FileCheck %s
+
+; CHECK: c0 3f 00 48 48003fc0
+
+define i32 @foo() {
+ret i32 0
+} \ No newline at end of file
diff --git a/test/CodeGen/Hexagon/relax.ll b/test/CodeGen/Hexagon/relax.ll
new file mode 100644
index 000000000000..9823d4d1cd9c
--- /dev/null
+++ b/test/CodeGen/Hexagon/relax.ll
@@ -0,0 +1,14 @@
+; RUN: llc -march=hexagon -filetype=obj < %s | llvm-objdump -d -r - | FileCheck %s
+
+declare void @bar()
+
+define void @foo() {
+call void @bar()
+ret void
+}
+
+
+; CHECK: { allocframe(#0) }
+; CHECK: { call 0 }
+; CHECK: 00000004: R_HEX_B22_PCREL
+; CHECK: { dealloc_return } \ No newline at end of file
diff --git a/test/CodeGen/Hexagon/sube.ll b/test/CodeGen/Hexagon/sube.ll
index 873f52b2d5df..9735894c419e 100644
--- a/test/CodeGen/Hexagon/sube.ll
+++ b/test/CodeGen/Hexagon/sube.ll
@@ -3,10 +3,10 @@
; CHECK: r{{[0-9]+:[0-9]+}} = #1
; CHECK: r{{[0-9]+:[0-9]+}} = #0
; CHECK: p{{[0-9]+}} = cmp.gtu(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
-; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
-; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
+; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
+; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = combine(r{{[0-9]+}}, r{{[0-9]+}})
define void @check_sube_subc(i64 %AL, i64 %AH, i64 %BL, i64 %BH, i64* %RL, i64* %RH) {
diff --git a/test/CodeGen/MIR/lit.local.cfg b/test/CodeGen/MIR/lit.local.cfg
new file mode 100644
index 000000000000..e69aa5765356
--- /dev/null
+++ b/test/CodeGen/MIR/lit.local.cfg
@@ -0,0 +1,2 @@
+config.suffixes = ['.mir']
+
diff --git a/test/CodeGen/MIR/llvm-ir-error-reported.mir b/test/CodeGen/MIR/llvm-ir-error-reported.mir
new file mode 100644
index 000000000000..013b28cd7890
--- /dev/null
+++ b/test/CodeGen/MIR/llvm-ir-error-reported.mir
@@ -0,0 +1,22 @@
+# RUN: not llc -start-after branch-folder -stop-after branch-folder -o /dev/null %s 2>&1 | FileCheck %s
+# This test ensures an error is reported if the embedded LLVM IR contains an
+# error.
+
+--- |
+
+ ; CHECK: [[@LINE+3]]:15: error: use of undefined value '%a'
+ define i32 @foo(i32 %x, i32 %y) {
+ %z = alloca i32, align 4
+ store i32 %a, i32* %z, align 4
+ br label %Test
+ Test:
+ %m = load i32, i32* %z, align 4
+ %cond = icmp eq i32 %y, %m
+ br i1 %cond, label %IfEqual, label %IfUnequal
+ IfEqual:
+ ret i32 1
+ IfUnequal:
+ ret i32 0
+ }
+
+...
diff --git a/test/CodeGen/MIR/llvmIR.mir b/test/CodeGen/MIR/llvmIR.mir
new file mode 100644
index 000000000000..7a7b46b62638
--- /dev/null
+++ b/test/CodeGen/MIR/llvmIR.mir
@@ -0,0 +1,32 @@
+# RUN: llc -start-after branch-folder -stop-after branch-folder -o /dev/null %s | FileCheck %s
+# This test ensures that the LLVM IR that's embedded with MIR is parsed
+# correctly.
+
+--- |
+ ; CHECK: define i32 @foo(i32 %x, i32 %y)
+ ; CHECK: %z = alloca i32, align 4
+ ; CHECK: store i32 %x, i32* %z, align 4
+ ; CHECK: br label %Test
+ ; CHECK: Test:
+ ; CHECK: %m = load i32, i32* %z, align 4
+ ; CHECK: %cond = icmp eq i32 %y, %m
+ ; CHECK: br i1 %cond, label %IfEqual, label %IfUnequal
+ ; CHECK: IfEqual:
+ ; CHECK: ret i32 1
+ ; CHECK: IfUnequal:
+ ; CHECK: ret i32 0
+ define i32 @foo(i32 %x, i32 %y) {
+ %z = alloca i32, align 4
+ store i32 %x, i32* %z, align 4
+ br label %Test
+ Test:
+ %m = load i32, i32* %z, align 4
+ %cond = icmp eq i32 %y, %m
+ br i1 %cond, label %IfEqual, label %IfUnequal
+ IfEqual:
+ ret i32 1
+ IfUnequal:
+ ret i32 0
+ }
+
+...
diff --git a/test/CodeGen/MIR/llvmIRMissing.mir b/test/CodeGen/MIR/llvmIRMissing.mir
new file mode 100644
index 000000000000..2acbcd1f9884
--- /dev/null
+++ b/test/CodeGen/MIR/llvmIRMissing.mir
@@ -0,0 +1,5 @@
+# RUN: llc -start-after branch-folder -stop-after branch-folder -o /dev/null %s
+# This test ensures that the MIR parser accepts files without the LLVM IR.
+
+---
+...
diff --git a/test/CodeGen/MIR/machine-function-missing-name.mir b/test/CodeGen/MIR/machine-function-missing-name.mir
new file mode 100644
index 000000000000..54668f1a5efe
--- /dev/null
+++ b/test/CodeGen/MIR/machine-function-missing-name.mir
@@ -0,0 +1,22 @@
+# RUN: not llc -start-after branch-folder -stop-after branch-folder -o /dev/null %s 2>&1 | FileCheck %s
+# This test ensures that an error is reported when a machine function doesn't
+# have a name attribute.
+
+--- |
+
+ define i32 @foo() {
+ ret i32 0
+ }
+
+ define i32 @bar() {
+ ret i32 0
+ }
+
+...
+---
+# CHECK: [[@LINE+1]]:1: error: missing required key 'name'
+nme: foo
+...
+---
+name: bar
+...
diff --git a/test/CodeGen/MIR/machine-function.mir b/test/CodeGen/MIR/machine-function.mir
new file mode 100644
index 000000000000..679bfd2d1620
--- /dev/null
+++ b/test/CodeGen/MIR/machine-function.mir
@@ -0,0 +1,24 @@
+# RUN: llc -start-after branch-folder -stop-after branch-folder -o /dev/null %s | FileCheck %s
+# This test ensures that the MIR parser parses machine functions correctly.
+
+--- |
+
+ define i32 @foo() {
+ ret i32 0
+ }
+
+ define i32 @bar() {
+ ret i32 0
+ }
+
+...
+---
+# CHECK: name: foo
+# CHECK-NEXT: ...
+name: foo
+...
+---
+# CHECK: name: bar
+# CHECK-NEXT: ...
+name: bar
+...
diff --git a/test/CodeGen/Mips/Fast-ISel/bswap1.ll b/test/CodeGen/Mips/Fast-ISel/bswap1.ll
new file mode 100644
index 000000000000..8ac9753fa463
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/bswap1.ll
@@ -0,0 +1,58 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s \
+; RUN: -check-prefix=ALL -check-prefix=32R1
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s \
+; RUN: -check-prefix=ALL -check-prefix=32R2
+
+@a = global i16 -21829, align 2
+@b = global i32 -1430532899, align 4
+@a1 = common global i16 0, align 2
+@b1 = common global i32 0, align 4
+
+declare i16 @llvm.bswap.i16(i16)
+declare i32 @llvm.bswap.i32(i32)
+
+define void @b16() {
+ ; ALL-LABEL: b16:
+
+ ; ALL: lw $[[A_ADDR:[0-9]+]], %got(a)($[[GOT_ADDR:[0-9]+]])
+ ; ALL: lhu $[[A_VAL:[0-9]+]], 0($[[A_ADDR]])
+
+ ; 32R1: sll $[[TMP1:[0-9]+]], $[[A_VAL]], 8
+ ; 32R1: srl $[[TMP2:[0-9]+]], $[[A_VAL]], 8
+ ; 32R1: or $[[TMP3:[0-9]+]], $[[TMP1]], $[[TMP2]]
+ ; 32R1: andi $[[TMP4:[0-9]+]], $[[TMP3]], 65535
+
+ ; 32R2: wsbh $[[RESULT:[0-9]+]], $[[A_VAL]]
+
+ %1 = load i16, i16* @a, align 2
+ %2 = call i16 @llvm.bswap.i16(i16 %1)
+ store i16 %2, i16* @a1, align 2
+ ret void
+}
+
+define void @b32() {
+ ; ALL-LABEL: b32:
+
+ ; ALL: lw $[[B_ADDR:[0-9]+]], %got(b)($[[GOT_ADDR:[0-9]+]])
+ ; ALL: lw $[[B_VAL:[0-9]+]], 0($[[B_ADDR]])
+
+ ; 32R1: srl $[[TMP1:[0-9]+]], $[[B_VAL]], 8
+ ; 32R1: srl $[[TMP2:[0-9]+]], $[[B_VAL]], 24
+ ; 32R1: andi $[[TMP3:[0-9]+]], $[[TMP1]], 65280
+ ; 32R1: or $[[TMP4:[0-9]+]], $[[TMP2]], $[[TMP3]]
+ ; 32R1: andi $[[TMP5:[0-9]+]], $[[B_VAL]], 65280
+ ; 32R1: sll $[[TMP6:[0-9]+]], $[[TMP5]], 8
+ ; 32R1: sll $[[TMP7:[0-9]+]], $[[B_VAL]], 24
+ ; 32R1: or $[[TMP8:[0-9]+]], $[[TMP4]], $[[TMP6]]
+ ; 32R1: or $[[RESULT:[0-9]+]], $[[TMP7]], $[[TMP8]]
+
+ ; 32R2: wsbh $[[TMP:[0-9]+]], $[[B_VAL]]
+ ; 32R2: rotr $[[RESULT:[0-9]+]], $[[TMP]], 16
+
+ %1 = load i32, i32* @b, align 4
+ %2 = call i32 @llvm.bswap.i32(i32 %1)
+ store i32 %2, i32* @b1, align 4
+ ret void
+}
diff --git a/test/CodeGen/Mips/Fast-ISel/div1.ll b/test/CodeGen/Mips/Fast-ISel/div1.ll
new file mode 100644
index 000000000000..89e7f211251f
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/div1.ll
@@ -0,0 +1,55 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s
+
+@sj = global i32 200000, align 4
+@sk = global i32 -47, align 4
+@uj = global i32 200000, align 4
+@uk = global i32 43, align 4
+@si = common global i32 0, align 4
+@ui = common global i32 0, align 4
+
+define void @divs() {
+ ; CHECK-LABEL: divs:
+
+ ; CHECK: lui $[[GOT1:[0-9]+]], %hi(_gp_disp)
+ ; CHECK: addiu $[[GOT2:[0-9]+]], $[[GOT1]], %lo(_gp_disp)
+ ; CHECK: addu $[[GOT:[0-9]+]], $[[GOT2:[0-9]+]], $25
+ ; CHECK-DAG: lw $[[I_ADDR:[0-9]+]], %got(si)($[[GOT]])
+ ; CHECK-DAG: lw $[[K_ADDR:[0-9]+]], %got(sk)($[[GOT]])
+ ; CHECK-DAG: lw $[[J_ADDR:[0-9]+]], %got(sj)($[[GOT]])
+ ; CHECK-DAG: lw $[[J:[0-9]+]], 0($[[J_ADDR]])
+ ; CHECK-DAG: lw $[[K:[0-9]+]], 0($[[K_ADDR]])
+ ; CHECK-DAG: div $zero, $[[J]], $[[K]]
+ ; CHECK_DAG: teq $[[K]], $zero, 7
+ ; CHECK-DAG: mflo $[[RESULT:[0-9]+]]
+ ; CHECK: sw $[[RESULT]], 0($[[I_ADDR]])
+ %1 = load i32, i32* @sj, align 4
+ %2 = load i32, i32* @sk, align 4
+ %div = sdiv i32 %1, %2
+ store i32 %div, i32* @si, align 4
+ ret void
+}
+
+define void @divu() {
+ ; CHECK-LABEL: divu:
+
+ ; CHECK: lui $[[GOT1:[0-9]+]], %hi(_gp_disp)
+ ; CHECK: addiu $[[GOT2:[0-9]+]], $[[GOT1]], %lo(_gp_disp)
+ ; CHECK: addu $[[GOT:[0-9]+]], $[[GOT2:[0-9]+]], $25
+ ; CHECK-DAG: lw $[[I_ADDR:[0-9]+]], %got(ui)($[[GOT]])
+ ; CHECK-DAG: lw $[[K_ADDR:[0-9]+]], %got(uk)($[[GOT]])
+ ; CHECK-DAG: lw $[[J_ADDR:[0-9]+]], %got(uj)($[[GOT]])
+ ; CHECK-DAG: lw $[[J:[0-9]+]], 0($[[J_ADDR]])
+ ; CHECK-DAG: lw $[[K:[0-9]+]], 0($[[K_ADDR]])
+ ; CHECK-DAG: divu $zero, $[[J]], $[[K]]
+ ; CHECK_DAG: teq $[[K]], $zero, 7
+ ; CHECK-DAG: mflo $[[RESULT:[0-9]+]]
+ ; CHECK: sw $[[RESULT]], 0($[[I_ADDR]])
+ %1 = load i32, i32* @uj, align 4
+ %2 = load i32, i32* @uk, align 4
+ %div = udiv i32 %1, %2
+ store i32 %div, i32* @ui, align 4
+ ret void
+}
diff --git a/test/CodeGen/Mips/Fast-ISel/memtest1.ll b/test/CodeGen/Mips/Fast-ISel/memtest1.ll
new file mode 100644
index 000000000000..a3fc4a32981c
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/memtest1.ll
@@ -0,0 +1,74 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s \
+; RUN: -check-prefix=ALL -check-prefix=32R1
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s \
+; RUN: -check-prefix=ALL -check-prefix=32R2
+
+@str = private unnamed_addr constant [12 x i8] c"hello there\00", align 1
+@src = global i8* getelementptr inbounds ([12 x i8], [12 x i8]* @str, i32 0, i32 0), align 4
+@i = global i32 12, align 4
+@dest = common global [50 x i8] zeroinitializer, align 1
+
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i32, i1)
+declare void @llvm.memmove.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i32, i1)
+declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1)
+
+define void @cpy(i8* %src, i32 %i) {
+ ; ALL-LABEL: cpy:
+
+ ; ALL-DAG: lw $[[T0:[0-9]+]], %got(dest)(${{[0-9]+}})
+ ; ALL-DAG: sw $4, 24($sp)
+ ; ALL-DAG: move $4, $[[T0]]
+ ; ALL-DAG: sw $5, 20($sp)
+ ; ALL-DAG: lw $[[T1:[0-9]+]], 24($sp)
+ ; ALL-DAG: move $5, $[[T1]]
+ ; ALL-DAG: lw $6, 20($sp)
+ ; ALL-DAG: lw $[[T2:[0-9]+]], %got(memcpy)(${{[0-9]+}})
+ ; ALL: jalr $[[T2]]
+ ; ALL-NEXT: nop
+ ; ALL-NOT: {{.*}}$2{{.*}}
+ call void @llvm.memcpy.p0i8.p0i8.i32(i8* getelementptr inbounds ([50 x i8], [50 x i8]* @dest, i32 0, i32 0),
+ i8* %src, i32 %i, i32 1, i1 false)
+ ret void
+}
+
+define void @mov(i8* %src, i32 %i) {
+ ; ALL-LABEL: mov:
+
+
+ ; ALL-DAG: lw $[[T0:[0-9]+]], %got(dest)(${{[0-9]+}})
+ ; ALL-DAG: sw $4, 24($sp)
+ ; ALL-DAG: move $4, $[[T0]]
+ ; ALL-DAG: sw $5, 20($sp)
+ ; ALL-DAG: lw $[[T1:[0-9]+]], 24($sp)
+ ; ALL-DAG: move $5, $[[T1]]
+ ; ALL-DAG: lw $6, 20($sp)
+ ; ALL-DAG: lw $[[T2:[0-9]+]], %got(memmove)(${{[0-9]+}})
+ ; ALL: jalr $[[T2]]
+ ; ALL-NEXT: nop
+ ; ALL-NOT: {{.*}}$2{{.*}}
+ call void @llvm.memmove.p0i8.p0i8.i32(i8* getelementptr inbounds ([50 x i8], [50 x i8]* @dest, i32 0, i32 0),
+ i8* %src, i32 %i, i32 1, i1 false)
+ ret void
+}
+
+define void @clear(i32 %i) {
+ ; ALL-LABEL: clear:
+
+ ; ALL-DAG: lw $[[T0:[0-9]+]], %got(dest)(${{[0-9]+}})
+ ; ALL-DAG: sw $4, 16($sp)
+ ; ALL-DAG: move $4, $[[T0]]
+ ; ALL-DAG: addiu $[[T1:[0-9]+]], $zero, 42
+ ; 32R1-DAG: sll $[[T2:[0-9]+]], $[[T1]], 24
+ ; 32R1-DAG: sra $5, $[[T2]], 24
+ ; 32R2-DAG: seb $5, $[[T1]]
+ ; ALL-DAG: lw $6, 16($sp)
+ ; ALL-DAG: lw $[[T2:[0-9]+]], %got(memset)(${{[0-9]+}})
+ ; ALL: jalr $[[T2]]
+ ; ALL-NEXT: nop
+ ; ALL-NOT: {{.*}}$2{{.*}}
+ call void @llvm.memset.p0i8.i32(i8* getelementptr inbounds ([50 x i8], [50 x i8]* @dest, i32 0, i32 0),
+ i8 42, i32 %i, i32 1, i1 false)
+ ret void
+}
diff --git a/test/CodeGen/Mips/Fast-ISel/mul1.ll b/test/CodeGen/Mips/Fast-ISel/mul1.ll
new file mode 100644
index 000000000000..0ee044bea0a7
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/mul1.ll
@@ -0,0 +1,18 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32 -O0 \
+; RUN: -fast-isel -mips-fast-isel -relocation-model=pic
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O0 \
+; RUN: -fast-isel -mips-fast-isel -relocation-model=pic
+
+; The test is just to make sure it is able to allocate
+; registers for this example. There was an issue with allocating AC0
+; after a mul instruction.
+
+declare { i32, i1 } @llvm.smul.with.overflow.i32(i32, i32)
+
+define i32 @foo(i32 %a, i32 %b) {
+entry:
+ %0 = mul i32 %a, %b
+ %1 = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %0, i32 %b)
+ %2 = extractvalue { i32, i1 } %1, 0
+ ret i32 %2
+}
diff --git a/test/CodeGen/Mips/Fast-ISel/rem1.ll b/test/CodeGen/Mips/Fast-ISel/rem1.ll
new file mode 100644
index 000000000000..9b5e440d0eaa
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/rem1.ll
@@ -0,0 +1,56 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O0 -relocation-model=pic \
+; RUN: -fast-isel=true -mips-fast-isel -fast-isel-abort=1 | FileCheck %s
+
+@sj = global i32 200, align 4
+@sk = global i32 -47, align 4
+@uj = global i32 200, align 4
+@uk = global i32 43, align 4
+@si = common global i32 0, align 4
+@ui = common global i32 0, align 4
+
+define void @rems() {
+ ; CHECK-LABEL: rems:
+
+ ; CHECK: lui $[[GOT1:[0-9]+]], %hi(_gp_disp)
+ ; CHECK: addiu $[[GOT2:[0-9]+]], $[[GOT1]], %lo(_gp_disp)
+ ; CHECK: addu $[[GOT:[0-9]+]], $[[GOT2:[0-9]+]], $25
+ ; CHECK-DAG: lw $[[I_ADDR:[0-9]+]], %got(si)($[[GOT]])
+ ; CHECK-DAG: lw $[[K_ADDR:[0-9]+]], %got(sk)($[[GOT]])
+ ; CHECK-DAG: lw $[[J_ADDR:[0-9]+]], %got(sj)($[[GOT]])
+ ; CHECK-DAG: lw $[[J:[0-9]+]], 0($[[J_ADDR]])
+ ; CHECK-DAG: lw $[[K:[0-9]+]], 0($[[K_ADDR]])
+ ; CHECK-DAG: div $zero, $[[J]], $[[K]]
+ ; CHECK_DAG: teq $[[K]], $zero, 7
+ ; CHECK-DAG: mfhi $[[RESULT:[0-9]+]]
+ ; CHECK: sw $[[RESULT]], 0($[[I_ADDR]])
+ %1 = load i32, i32* @sj, align 4
+ %2 = load i32, i32* @sk, align 4
+ %rem = srem i32 %1, %2
+ store i32 %rem, i32* @si, align 4
+ ret void
+}
+
+; Function Attrs: noinline nounwind
+define void @remu() {
+ ; CHECK-LABEL: remu:
+
+ ; CHECK: lui $[[GOT1:[0-9]+]], %hi(_gp_disp)
+ ; CHECK: addiu $[[GOT2:[0-9]+]], $[[GOT1]], %lo(_gp_disp)
+ ; CHECK: addu $[[GOT:[0-9]+]], $[[GOT2:[0-9]+]], $25
+ ; CHECK-DAG: lw $[[I_ADDR:[0-9]+]], %got(ui)($[[GOT]])
+ ; CHECK-DAG: lw $[[K_ADDR:[0-9]+]], %got(uk)($[[GOT]])
+ ; CHECK-DAG: lw $[[J_ADDR:[0-9]+]], %got(uj)($[[GOT]])
+ ; CHECK-DAG: lw $[[J:[0-9]+]], 0($[[J_ADDR]])
+ ; CHECK-DAG: lw $[[K:[0-9]+]], 0($[[K_ADDR]])
+ ; CHECK-DAG: divu $zero, $[[J]], $[[K]]
+ ; CHECK_DAG: teq $[[K]], $zero, 7
+ ; CHECK-DAG: mfhi $[[RESULT:[0-9]+]]
+ ; CHECK: sw $[[RESULT]], 0($[[I_ADDR]])
+ %1 = load i32, i32* @uj, align 4
+ %2 = load i32, i32* @uk, align 4
+ %rem = urem i32 %1, %2
+ store i32 %rem, i32* @ui, align 4
+ ret void
+}
diff --git a/test/CodeGen/Mips/Fast-ISel/sel1.ll b/test/CodeGen/Mips/Fast-ISel/sel1.ll
new file mode 100644
index 000000000000..47b6a895cde8
--- /dev/null
+++ b/test/CodeGen/Mips/Fast-ISel/sel1.ll
@@ -0,0 +1,91 @@
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 -O2 -relocation-model=pic \
+; RUN: -fast-isel -mips-fast-isel -fast-isel-abort=1 | FileCheck %s
+
+define i1 @sel_i1(i1 %j, i1 %k, i1 %l) {
+entry:
+ ; CHECK-LABEL: sel_i1:
+
+ ; FIXME: The following instruction is redundant.
+ ; CHECK: xor $[[T0:[0-9]+]], $4, $zero
+ ; CHECK-NEXT: sltu $[[T1:[0-9]+]], $zero, $[[T0]]
+ ; CHECK-NEXT: movn $6, $5, $[[T1]]
+ ; CHECK: move $2, $6
+ %cond = icmp ne i1 %j, 0
+ %res = select i1 %cond, i1 %k, i1 %l
+ ret i1 %res
+}
+
+define i8 @sel_i8(i8 %j, i8 %k, i8 %l) {
+entry:
+ ; CHECK-LABEL: sel_i8:
+
+ ; CHECK-DAG: seb $[[T0:[0-9]+]], $4
+ ; FIXME: The following 2 instructions are redundant.
+ ; CHECK-DAG: seb $[[T1:[0-9]+]], $zero
+ ; CHECK: xor $[[T2:[0-9]+]], $[[T0]], $[[T1]]
+ ; CHECK-NEXT: sltu $[[T3:[0-9]+]], $zero, $[[T2]]
+ ; CHECK-NEXT: movn $6, $5, $[[T3]]
+ ; CHECK: move $2, $6
+ %cond = icmp ne i8 %j, 0
+ %res = select i1 %cond, i8 %k, i8 %l
+ ret i8 %res
+}
+
+define i16 @sel_i16(i16 %j, i16 %k, i16 %l) {
+entry:
+ ; CHECK-LABEL: sel_i16:
+
+ ; CHECK-DAG: seh $[[T0:[0-9]+]], $4
+ ; FIXME: The following 2 instructions are redundant.
+ ; CHECK-DAG: seh $[[T1:[0-9]+]], $zero
+ ; CHECK: xor $[[T2:[0-9]+]], $[[T0]], $[[T1]]
+ ; CHECK-NEXT: sltu $[[T3:[0-9]+]], $zero, $[[T2]]
+ ; CHECK-NEXT: movn $6, $5, $[[T3]]
+ ; CHECK: move $2, $6
+ %cond = icmp ne i16 %j, 0
+ %res = select i1 %cond, i16 %k, i16 %l
+ ret i16 %res
+}
+
+define i32 @sel_i32(i32 %j, i32 %k, i32 %l) {
+entry:
+ ; CHECK-LABEL: sel_i32:
+
+ ; FIXME: The following instruction is redundant.
+ ; CHECK: xor $[[T0:[0-9]+]], $4, $zero
+ ; CHECK-NEXT: sltu $[[T1:[0-9]+]], $zero, $[[T0]]
+ ; CHECK-NEXT: movn $6, $5, $[[T1]]
+ ; CHECK: move $2, $6
+ %cond = icmp ne i32 %j, 0
+ %res = select i1 %cond, i32 %k, i32 %l
+ ret i32 %res
+}
+
+define float @sel_float(i32 %j, float %k, float %l) {
+entry:
+ ; CHECK-LABEL: sel_float:
+
+ ; CHECK-DAG: mtc1 $6, $f0
+ ; CHECK-DAG: mtc1 $5, $f1
+ ; CHECK-DAG: xor $[[T0:[0-9]+]], $4, $zero
+ ; CHECK: sltu $[[T1:[0-9]+]], $zero, $[[T0]]
+ ; CHECK: movn.s $f0, $f1, $[[T1]]
+ %cond = icmp ne i32 %j, 0
+ %res = select i1 %cond, float %k, float %l
+ ret float %res
+}
+
+define double @sel_double(i32 %j, double %k, double %l) {
+entry:
+ ; CHECK-LABEL: sel_double:
+
+ ; CHECK-DAG: mtc1 $6, $f2
+ ; CHECK-DAG: mthc1 $7, $f2
+ ; CHECK-DAG: ldc1 $f0, 16($sp)
+ ; CHECK-DAG: xor $[[T0:[0-9]+]], $4, $zero
+ ; CHECK: sltu $[[T1:[0-9]+]], $zero, $[[T0]]
+ ; CHECK: movn.d $f0, $f2, $[[T1]]
+ %cond = icmp ne i32 %j, 0
+ %res = select i1 %cond, double %k, double %l
+ ret double %res
+}
diff --git a/test/CodeGen/Mips/dynamic-stack-realignment.ll b/test/CodeGen/Mips/dynamic-stack-realignment.ll
new file mode 100644
index 000000000000..777930a37ad5
--- /dev/null
+++ b/test/CodeGen/Mips/dynamic-stack-realignment.ll
@@ -0,0 +1,299 @@
+; RUN: llc < %s -march=mips -mcpu=mips2 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP32
+; RUN: llc < %s -march=mips -mcpu=mips32 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP32
+; RUN: llc < %s -march=mips -mcpu=mips32r6 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP32
+; RUN: llc < %s -march=mips64 -mcpu=mips3 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N64
+; RUN: llc < %s -march=mips64 -mcpu=mips64 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N64
+; RUN: llc < %s -march=mips64 -mcpu=mips64r6 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N64
+; RUN: llc < %s -march=mips64 -mcpu=mips3 -target-abi n32 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N32
+; RUN: llc < %s -march=mips64 -mcpu=mips64 -target-abi n32 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N32
+; RUN: llc < %s -march=mips64 -mcpu=mips64r6 -target-abi n32 | FileCheck %s \
+; RUN: --check-prefix=ALL --check-prefix=GP64 -check-prefix=N32
+
+; Check dynamic stack realignment in functions without variable-sized objects.
+
+declare void @helper_01(i32, i32, i32, i32, i32*)
+
+; O32 ABI
+define void @func_01() {
+entry:
+; GP32-LABEL: func_01:
+
+ ; prologue
+ ; FIXME: We are currently over-allocating stack space. This particular case
+ ; needs a frame of up to between 16 and 512-bytes but currently
+ ; allocates between 1024 and 1536 bytes
+ ; GP32: addiu $sp, $sp, -1024
+ ; GP32: sw $ra, 1020($sp)
+ ; GP32: sw $fp, 1016($sp)
+ ;
+ ; GP32: move $fp, $sp
+ ; GP32: addiu $[[T0:[0-9]+|ra|gp]], $zero, -512
+ ; GP32-NEXT: and $sp, $sp, $[[T0]]
+
+ ; body
+ ; GP32: addiu $[[T1:[0-9]+]], $sp, 512
+ ; GP32: sw $[[T1]], 16($sp)
+
+ ; epilogue
+ ; GP32: move $sp, $fp
+ ; GP32: lw $fp, 1016($sp)
+ ; GP32: lw $ra, 1020($sp)
+ ; GP32: addiu $sp, $sp, 1024
+
+ %a = alloca i32, align 512
+ call void @helper_01(i32 0, i32 0, i32 0, i32 0, i32* %a)
+ ret void
+}
+
+declare void @helper_02(i32, i32, i32, i32,
+ i32, i32, i32, i32, i32*)
+
+; N32/N64 ABIs
+define void @func_02() {
+entry:
+; GP64-LABEL: func_02:
+
+ ; prologue
+ ; FIXME: We are currently over-allocating stack space. This particular case
+ ; needs a frame of up to between 16 and 512-bytes but currently
+ ; allocates between 1024 and 1536 bytes
+ ; N32: addiu $sp, $sp, -1024
+ ; N64: daddiu $sp, $sp, -1024
+ ; GP64: sd $ra, 1016($sp)
+ ; GP64: sd $fp, 1008($sp)
+ ; N32: sd $gp, 1000($sp)
+ ;
+ ; GP64: move $fp, $sp
+ ; N32: addiu $[[T0:[0-9]+|ra]], $zero, -512
+ ; N64: daddiu $[[T0:[0-9]+|ra]], $zero, -512
+ ; GP64-NEXT: and $sp, $sp, $[[T0]]
+
+ ; body
+ ; N32: addiu $[[T1:[0-9]+]], $sp, 512
+ ; N64: daddiu $[[T1:[0-9]+]], $sp, 512
+ ; GP64: sd $[[T1]], 0($sp)
+
+ ; epilogue
+ ; GP64: move $sp, $fp
+ ; N32: ld $gp, 1000($sp)
+ ; GP64: ld $fp, 1008($sp)
+ ; GP64: ld $ra, 1016($sp)
+ ; N32: addiu $sp, $sp, 1024
+ ; N64: daddiu $sp, $sp, 1024
+
+ %a = alloca i32, align 512
+ call void @helper_02(i32 0, i32 0, i32 0, i32 0,
+ i32 0, i32 0, i32 0, i32 0, i32* %a)
+ ret void
+}
+
+; Verify that we use $fp for referencing incoming arguments.
+
+declare void @helper_03(i32, i32, i32, i32, i32*, i32*)
+
+; O32 ABI
+define void @func_03(i32 %p0, i32 %p1, i32 %p2, i32 %p3, i32* %b) {
+entry:
+; GP32-LABEL: func_03:
+
+ ; body
+ ; FIXME: We are currently over-allocating stack space.
+ ; GP32-DAG: addiu $[[T0:[0-9]+]], $sp, 512
+ ; GP32-DAG: sw $[[T0]], 16($sp)
+ ; GP32-DAG: lw $[[T1:[0-9]+]], 1040($fp)
+ ; GP32-DAG: sw $[[T1]], 20($sp)
+
+ %a = alloca i32, align 512
+ call void @helper_03(i32 0, i32 0, i32 0, i32 0, i32* %a, i32* %b)
+ ret void
+}
+
+declare void @helper_04(i32, i32, i32, i32,
+ i32, i32, i32, i32, i32*, i32*)
+
+; N32/N64 ABIs
+define void @func_04(i32 %p0, i32 %p1, i32 %p2, i32 %p3,
+ i32 %p4, i32 %p5, i32 %p6, i32 %p7,
+ i32* %b) {
+entry:
+; GP64-LABEL: func_04:
+
+ ; body
+ ; FIXME: We are currently over-allocating stack space.
+ ; N32-DAG: addiu $[[T0:[0-9]+]], $sp, 512
+ ; N64-DAG: daddiu $[[T0:[0-9]+]], $sp, 512
+ ; GP64-DAG: sd $[[T0]], 0($sp)
+ ; GP64-DAG: ld $[[T1:[0-9]+]], 1024($fp)
+ ; GP64-DAG: sd $[[T1]], 8($sp)
+
+ %a = alloca i32, align 512
+ call void @helper_04(i32 0, i32 0, i32 0, i32 0,
+ i32 0, i32 0, i32 0, i32 0, i32* %a, i32* %b)
+ ret void
+}
+
+; Check dynamic stack realignment in functions with variable-sized objects.
+
+; O32 ABI
+define void @func_05(i32 %sz) {
+entry:
+; GP32-LABEL: func_05:
+
+ ; prologue
+ ; FIXME: We are currently over-allocating stack space.
+ ; GP32: addiu $sp, $sp, -1024
+ ; GP32: sw $fp, 1020($sp)
+ ; GP32: sw $23, 1016($sp)
+ ;
+ ; GP32: move $fp, $sp
+ ; GP32: addiu $[[T0:[0-9]+|gp]], $zero, -512
+ ; GP32-NEXT: and $sp, $sp, $[[T0]]
+ ; GP32-NEXT: move $23, $sp
+
+ ; body
+ ; GP32: addiu $[[T1:[0-9]+]], $zero, 222
+ ; GP32: sw $[[T1]], 508($23)
+
+ ; epilogue
+ ; GP32: move $sp, $fp
+ ; GP32: lw $23, 1016($sp)
+ ; GP32: lw $fp, 1020($sp)
+ ; GP32: addiu $sp, $sp, 1024
+
+ %a0 = alloca i32, i32 %sz, align 512
+ %a1 = alloca i32, align 4
+
+ store volatile i32 111, i32* %a0, align 512
+ store volatile i32 222, i32* %a1, align 4
+
+ ret void
+}
+
+; N32/N64 ABIs
+define void @func_06(i32 %sz) {
+entry:
+; GP64-LABEL: func_06:
+
+ ; prologue
+ ; FIXME: We are currently over-allocating stack space.
+ ; N32: addiu $sp, $sp, -1024
+ ; N64: daddiu $sp, $sp, -1024
+ ; GP64: sd $fp, 1016($sp)
+ ; GP64: sd $23, 1008($sp)
+ ;
+ ; GP64: move $fp, $sp
+ ; GP64: addiu $[[T0:[0-9]+|gp]], $zero, -512
+ ; GP64-NEXT: and $sp, $sp, $[[T0]]
+ ; GP64-NEXT: move $23, $sp
+
+ ; body
+ ; GP64: addiu $[[T1:[0-9]+]], $zero, 222
+ ; GP64: sw $[[T1]], 508($23)
+
+ ; epilogue
+ ; GP64: move $sp, $fp
+ ; GP64: ld $23, 1008($sp)
+ ; GP64: ld $fp, 1016($sp)
+ ; N32: addiu $sp, $sp, 1024
+ ; N64: daddiu $sp, $sp, 1024
+
+ %a0 = alloca i32, i32 %sz, align 512
+ %a1 = alloca i32, align 4
+
+ store volatile i32 111, i32* %a0, align 512
+ store volatile i32 222, i32* %a1, align 4
+
+ ret void
+}
+
+; Verify that we use $fp for referencing incoming arguments and $sp for
+; building outbound arguments for nested function calls.
+
+; O32 ABI
+define void @func_07(i32 %p0, i32 %p1, i32 %p2, i32 %p3, i32 %sz) {
+entry:
+; GP32-LABEL: func_07:
+
+ ; body
+ ; FIXME: We are currently over-allocating stack space.
+ ; GP32-DAG: lw $[[T0:[0-9]+]], 1040($fp)
+ ;
+ ; GP32-DAG: addiu $[[T1:[0-9]+]], $zero, 222
+ ; GP32-DAG: sw $[[T1]], 508($23)
+ ;
+ ; GP32-DAG: sw $[[T2:[0-9]+]], 16($sp)
+
+ %a0 = alloca i32, i32 %sz, align 512
+ %a1 = alloca i32, align 4
+
+ store volatile i32 111, i32* %a0, align 512
+ store volatile i32 222, i32* %a1, align 4
+
+ call void @helper_01(i32 0, i32 0, i32 0, i32 0, i32* %a1)
+
+ ret void
+}
+
+; N32/N64 ABIs
+define void @func_08(i32 %p0, i32 %p1, i32 %p2, i32 %p3,
+ i32 %p4, i32 %p5, i32 %p6, i32 %p7,
+ i32 %sz) {
+entry:
+; GP64-LABEL: func_08:
+
+ ; body
+ ; FIXME: We are currently over-allocating stack space.
+ ; N32-DAG: lw $[[T0:[0-9]+]], 1028($fp)
+ ; N64-DAG: lwu $[[T0:[0-9]+]], 1028($fp)
+ ;
+ ; GP64-DAG: addiu $[[T1:[0-9]+]], $zero, 222
+ ; GP64-DAG: sw $[[T1]], 508($23)
+ ;
+ ; GP64-DAG: sd $[[T2:[0-9]+]], 0($sp)
+
+ %a0 = alloca i32, i32 %sz, align 512
+ %a1 = alloca i32, align 4
+
+ store volatile i32 111, i32* %a0, align 512
+ store volatile i32 222, i32* %a1, align 4
+
+ call void @helper_02(i32 0, i32 0, i32 0, i32 0,
+ i32 0, i32 0, i32 0, i32 0, i32* %a1)
+ ret void
+}
+
+; Check that we do not perform dynamic stack realignment in the presence of
+; the "no-realign-stack" function attribute.
+define void @func_09() "no-realign-stack" {
+entry:
+; ALL-LABEL: func_09:
+
+ ; ALL-NOT: and $sp, $sp, $[[T0:[0-9]+|ra|gp]]
+
+ %a = alloca i32, align 512
+ call void @helper_01(i32 0, i32 0, i32 0, i32 0, i32* %a)
+ ret void
+}
+
+define void @func_10(i32 %sz) "no-realign-stack" {
+entry:
+; ALL-LABEL: func_10:
+
+ ; ALL-NOT: and $sp, $sp, $[[T0:[0-9]+|ra|gp]]
+
+ %a0 = alloca i32, i32 %sz, align 512
+ %a1 = alloca i32, align 4
+
+ store volatile i32 111, i32* %a0, align 512
+ store volatile i32 222, i32* %a1, align 4
+
+ ret void
+}
diff --git a/test/CodeGen/Mips/ehframe-indirect.ll b/test/CodeGen/Mips/ehframe-indirect.ll
index f124881a472f..dc06ef7840ff 100644
--- a/test/CodeGen/Mips/ehframe-indirect.ll
+++ b/test/CodeGen/Mips/ehframe-indirect.ll
@@ -1,9 +1,11 @@
-; RUN: llc -mtriple=mipsel-linux-gnu < %s | FileCheck -check-prefix=ALL -check-prefix=O32 %s
-; RUN: llc -mtriple=mipsel-linux-android < %s | FileCheck -check-prefix=ALL -check-prefix=O32 %s
-; RUN: llc -mtriple=mips64el-linux-gnu -target-abi=n32 < %s | FileCheck -check-prefix=ALL -check-prefix=N32 %s
-; RUN: llc -mtriple=mips64el-linux-android -target-abi=n32 < %s | FileCheck -check-prefix=ALL -check-prefix=N32 %s
-; RUN: llc -mtriple=mips64el-linux-gnu < %s | FileCheck -check-prefix=ALL -check-prefix=N64 %s
-; RUN: llc -mtriple=mips64el-linux-android < %s | FileCheck -check-prefix=ALL -check-prefix=N64 %s
+; RUN: llc -mtriple=mipsel-linux-gnu < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=O32 %s
+; RUN: llc -mtriple=mipsel-linux-android < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=O32 %s
+; RUN: llc -mtriple=mips64el-linux-gnu -target-abi=n32 < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N32 %s
+; RUN: llc -mtriple=mips64el-linux-android -target-abi=n32 < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N32 %s
+; RUN: llc -mtriple=mips64el-linux-gnu < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N64 %s
+; RUN: llc -mtriple=mips64el-linux-android < %s -asm-verbose | FileCheck -check-prefix=ALL -check-prefix=N64 %s
+
+@_ZTISt9exception = external constant i8*
define i32 @main() {
; ALL: .cfi_startproc
@@ -16,7 +18,9 @@ entry:
lpad:
%0 = landingpad { i8*, i32 } personality i8*
- bitcast (i32 (...)* @__gxx_personality_v0 to i8*) catch i8* null
+ bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
+ catch i8* null
+ catch i8* bitcast (i8** @_ZTISt9exception to i8*)
ret i32 0
cont:
@@ -28,6 +32,14 @@ declare i32 @__gxx_personality_v0(...)
declare void @foo()
+; ALL: GCC_except_table{{[0-9]+}}:
+; ALL: .byte 155 # @TType Encoding = indirect pcrel sdata4
+; ALL: $[[PC_LABEL:tmp[0-9]+]]:
+; ALL: .4byte ($_ZTISt9exception.DW.stub)-($[[PC_LABEL]])
+; ALL: $_ZTISt9exception.DW.stub:
+; O32: .4byte _ZTISt9exception
+; N32: .4byte _ZTISt9exception
+; N64: .8byte _ZTISt9exception
; ALL: .hidden DW.ref.__gxx_personality_v0
; ALL: .weak DW.ref.__gxx_personality_v0
; ALL: .section .data.DW.ref.__gxx_personality_v0,"aGw",@progbits,DW.ref.__gxx_personality_v0,comdat
diff --git a/test/CodeGen/Mips/emergency-spill-slot-near-fp.ll b/test/CodeGen/Mips/emergency-spill-slot-near-fp.ll
index 3dc1cde77095..779620e10128 100644
--- a/test/CodeGen/Mips/emergency-spill-slot-near-fp.ll
+++ b/test/CodeGen/Mips/emergency-spill-slot-near-fp.ll
@@ -1,8 +1,8 @@
; Check that register scavenging spill slot is close to $fp.
; RUN: llc -march=mipsel -O0 < %s | FileCheck %s
-; CHECK: sw ${{.*}}, 4($fp)
-; CHECK: lw ${{.*}}, 4($fp)
+; CHECK: sw ${{.*}}, 4($sp)
+; CHECK: lw ${{.*}}, 4($sp)
define i32 @main(i32 signext %argc, i8** %argv) "no-frame-pointer-elim"="true" {
entry:
diff --git a/test/CodeGen/NVPTX/access-non-generic.ll b/test/CodeGen/NVPTX/access-non-generic.ll
index e709302918f5..5deefe881e3f 100644
--- a/test/CodeGen/NVPTX/access-non-generic.ll
+++ b/test/CodeGen/NVPTX/access-non-generic.ll
@@ -85,6 +85,22 @@ define i32 @ld_int_from_float() {
ret i32 %1
}
+define i32 @ld_int_from_global_float(float addrspace(1)* %input, i32 %i, i32 %j) {
+; IR-LABEL: @ld_int_from_global_float(
+; PTX-LABEL: ld_int_from_global_float(
+ %1 = addrspacecast float addrspace(1)* %input to float*
+ %2 = getelementptr float, float* %1, i32 %i
+; IR-NEXT: getelementptr float, float addrspace(1)* %input, i32 %i
+ %3 = getelementptr float, float* %2, i32 %j
+; IR-NEXT: getelementptr float, float addrspace(1)* {{%[^,]+}}, i32 %j
+ %4 = bitcast float* %3 to i32*
+; IR-NEXT: bitcast float addrspace(1)* {{%[^ ]+}} to i32 addrspace(1)*
+ %5 = load i32, i32* %4
+; IR-NEXT: load i32, i32 addrspace(1)* {{%.+}}
+; PTX-LABEL: ld.global
+ ret i32 %5
+}
+
declare void @llvm.cuda.syncthreads() #3
attributes #3 = { noduplicate nounwind }
diff --git a/test/CodeGen/NVPTX/bug21465.ll b/test/CodeGen/NVPTX/bug21465.ll
index 76af386c6516..2eae41f73a0c 100644
--- a/test/CodeGen/NVPTX/bug21465.ll
+++ b/test/CodeGen/NVPTX/bug21465.ll
@@ -1,4 +1,5 @@
-; RUN: opt < %s -nvptx-lower-struct-args -S | FileCheck %s
+; RUN: opt < %s -nvptx-lower-kernel-args -S | FileCheck %s
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_35 | FileCheck %s --check-prefix PTX
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"
target triple = "nvptx64-unknown-unknown"
@@ -8,12 +9,15 @@ target triple = "nvptx64-unknown-unknown"
; Function Attrs: nounwind
define void @_Z11TakesStruct1SPi(%struct.S* byval nocapture readonly %input, i32* nocapture %output) #0 {
entry:
-; CHECK-LABEL @_Z22TakesStruct1SPi
-; CHECK: bitcast %struct.S* %input to i8*
-; CHECK: call i8 addrspace(101)* @llvm.nvvm.ptr.gen.to.param.p101i8.p0i8
+; CHECK-LABEL: @_Z11TakesStruct1SPi
+; PTX-LABEL: .visible .entry _Z11TakesStruct1SPi(
+; CHECK: addrspacecast %struct.S* %input to %struct.S addrspace(101)*
%b = getelementptr inbounds %struct.S, %struct.S* %input, i64 0, i32 1
%0 = load i32, i32* %b, align 4
+; PTX-NOT: ld.param.u32 {{%r[0-9]+}}, [{{%rd[0-9]+}}]
+; PTX: ld.param.u32 [[value:%r[0-9]+]], [{{%rd[0-9]+}}+4]
store i32 %0, i32* %output, align 4
+; PTX-NEXT: st.global.u32 [{{%rd[0-9]+}}], [[value]]
ret void
}
diff --git a/test/CodeGen/NVPTX/call-with-alloca-buffer.ll b/test/CodeGen/NVPTX/call-with-alloca-buffer.ll
index 58b191129917..c70670da13d6 100644
--- a/test/CodeGen/NVPTX/call-with-alloca-buffer.ll
+++ b/test/CodeGen/NVPTX/call-with-alloca-buffer.ll
@@ -24,7 +24,10 @@ entry:
; CHECK: cvta.local.u64 %SP, %rd[[BUF_REG]]
; CHECK: ld.param.u64 %rd[[A_REG:[0-9]+]], [kernel_func_param_0]
-; CHECK: ld.f32 %f[[A0_REG:[0-9]+]], [%rd[[A_REG]]]
+; CHECK: cvta.to.global.u64 %rd[[A1_REG:[0-9]+]], %rd[[A_REG]]
+; FIXME: casting A1_REG to A2_REG is unnecessary; A2_REG is essentially A_REG
+; CHECK: cvta.global.u64 %rd[[A2_REG:[0-9]+]], %rd[[A1_REG]]
+; CHECK: ld.global.f32 %f[[A0_REG:[0-9]+]], [%rd[[A1_REG]]]
; CHECK: st.f32 [%SP+0], %f[[A0_REG]]
%0 = load float, float* %a, align 4
@@ -48,7 +51,7 @@ entry:
; CHECK: add.u64 %rd[[SP_REG:[0-9]+]], %SP, 0
; CHECK: .param .b64 param0;
-; CHECK-NEXT: st.param.b64 [param0+0], %rd[[A_REG]]
+; CHECK-NEXT: st.param.b64 [param0+0], %rd[[A2_REG]]
; CHECK-NEXT: .param .b64 param1;
; CHECK-NEXT: st.param.b64 [param1+0], %rd[[SP_REG]]
; CHECK-NEXT: call.uni
diff --git a/test/CodeGen/NVPTX/globals_init.ll b/test/CodeGen/NVPTX/globals_init.ll
new file mode 100644
index 000000000000..5b45f410156c
--- /dev/null
+++ b/test/CodeGen/NVPTX/globals_init.ll
@@ -0,0 +1,23 @@
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
+
+; Make sure the globals constant initializers are not prone to host endianess
+; issues.
+
+; CHECK-DAG: .b8 Gbli08[2] = {171, 205};
+@Gbli08 = global [2 x i8] [i8 171, i8 205]
+
+; CHECK-DAG: .b8 Gbli16[4] = {205, 171, 1, 239};
+@Gbli16 = global [2 x i16] [i16 43981, i16 61185]
+
+; CHECK-DAG: .b8 Gbli32[8] = {1, 239, 205, 171, 137, 103, 69, 35};
+@Gbli32 = global [2 x i32] [i32 2882400001, i32 591751049]
+
+; CHECK-DAG: .b8 Gbli64[16] = {137, 103, 69, 35, 1, 239, 205, 171, 239, 205, 171, 137, 103, 69, 35, 1};
+@Gbli64 = global [2 x i64] [i64 12379813738877118345, i64 81985529216486895]
+
+; CHECK-DAG: .b8 Gblf32[8] = {192, 225, 100, 75, 0, 96, 106, 69};
+@Gblf32 = global [2 x float] [float 1.5e+7, float 3.75e+3]
+
+; CHECK-DAG: .b8 Gblf64[16] = {116, 10, 181, 48, 134, 62, 230, 58, 106, 222, 138, 98, 204, 250, 200, 75};
+@Gblf64 = global [2 x double] [double 5.75e-25, double 12.25e+56]
+
diff --git a/test/CodeGen/NVPTX/lower-kernel-ptr-arg.ll b/test/CodeGen/NVPTX/lower-kernel-ptr-arg.ll
new file mode 100644
index 000000000000..53220bd905bd
--- /dev/null
+++ b/test/CodeGen/NVPTX/lower-kernel-ptr-arg.ll
@@ -0,0 +1,20 @@
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"
+target triple = "nvptx64-unknown-unknown"
+
+; Verify that both %input and %output are converted to global pointers and then
+; addrspacecast'ed back to the original type.
+define void @kernel(float* %input, float* %output) {
+; CHECK-LABEL: .visible .entry kernel(
+; CHECK: cvta.to.global.u64
+; CHECK: cvta.to.global.u64
+ %1 = load float, float* %input, align 4
+; CHECK: ld.global.f32
+ store float %1, float* %output, align 4
+; CHECK: st.global.f32
+ ret void
+}
+
+!nvvm.annotations = !{!0}
+!0 = !{void (float*, float*)* @kernel, !"kernel", i32 1}
diff --git a/test/CodeGen/NVPTX/pr13291-i1-store.ll b/test/CodeGen/NVPTX/pr13291-i1-store.ll
index d4f7c3bd210a..934df30a3a7d 100644
--- a/test/CodeGen/NVPTX/pr13291-i1-store.ll
+++ b/test/CodeGen/NVPTX/pr13291-i1-store.ll
@@ -3,19 +3,19 @@
define ptx_kernel void @t1(i1* %a) {
; PTX32: mov.u16 %rs{{[0-9]+}}, 0;
-; PTX32-NEXT: st.u8 [%r{{[0-9]+}}], %rs{{[0-9]+}};
+; PTX32-NEXT: st.global.u8 [%r{{[0-9]+}}], %rs{{[0-9]+}};
; PTX64: mov.u16 %rs{{[0-9]+}}, 0;
-; PTX64-NEXT: st.u8 [%rd{{[0-9]+}}], %rs{{[0-9]+}};
+; PTX64-NEXT: st.global.u8 [%rd{{[0-9]+}}], %rs{{[0-9]+}};
store i1 false, i1* %a
ret void
}
define ptx_kernel void @t2(i1* %a, i8* %b) {
-; PTX32: ld.u8 %rs{{[0-9]+}}, [%r{{[0-9]+}}]
+; PTX32: ld.global.u8 %rs{{[0-9]+}}, [%r{{[0-9]+}}]
; PTX32: and.b16 %rs{{[0-9]+}}, %rs{{[0-9]+}}, 1;
; PTX32: setp.eq.b16 %p{{[0-9]+}}, %rs{{[0-9]+}}, 1;
-; PTX64: ld.u8 %rs{{[0-9]+}}, [%rd{{[0-9]+}}]
+; PTX64: ld.global.u8 %rs{{[0-9]+}}, [%rd{{[0-9]+}}]
; PTX64: and.b16 %rs{{[0-9]+}}, %rs{{[0-9]+}}, 1;
; PTX64: setp.eq.b16 %p{{[0-9]+}}, %rs{{[0-9]+}}, 1;
diff --git a/test/CodeGen/NVPTX/surf-read-cuda.ll b/test/CodeGen/NVPTX/surf-read-cuda.ll
index ed021346c0f9..c17c71e01d3e 100644
--- a/test/CodeGen/NVPTX/surf-read-cuda.ll
+++ b/test/CodeGen/NVPTX/surf-read-cuda.ll
@@ -18,8 +18,8 @@ define void @foo(i64 %img, float* %red, i32 %idx) {
; SM20: cvt.rn.f32.s32 %f[[REDF:[0-9]+]], %r[[RED]]
; SM30: cvt.rn.f32.s32 %f[[REDF:[0-9]+]], %r[[RED]]
%ret = sitofp i32 %val to float
-; SM20: st.f32 [%r{{[0-9]+}}], %f[[REDF]]
-; SM30: st.f32 [%r{{[0-9]+}}], %f[[REDF]]
+; SM20: st.global.f32 [%r{{[0-9]+}}], %f[[REDF]]
+; SM30: st.global.f32 [%r{{[0-9]+}}], %f[[REDF]]
store float %ret, float* %red
ret void
}
@@ -37,8 +37,8 @@ define void @bar(float* %red, i32 %idx) {
; SM20: cvt.rn.f32.s32 %f[[REDF:[0-9]+]], %r[[RED]]
; SM30: cvt.rn.f32.s32 %f[[REDF:[0-9]+]], %r[[RED]]
%ret = sitofp i32 %val to float
-; SM20: st.f32 [%r{{[0-9]+}}], %f[[REDF]]
-; SM30: st.f32 [%r{{[0-9]+}}], %f[[REDF]]
+; SM20: st.global.f32 [%r{{[0-9]+}}], %f[[REDF]]
+; SM30: st.global.f32 [%r{{[0-9]+}}], %f[[REDF]]
store float %ret, float* %red
ret void
}
diff --git a/test/CodeGen/NVPTX/tex-read-cuda.ll b/test/CodeGen/NVPTX/tex-read-cuda.ll
index c5b5600de874..d5f7c1667f17 100644
--- a/test/CodeGen/NVPTX/tex-read-cuda.ll
+++ b/test/CodeGen/NVPTX/tex-read-cuda.ll
@@ -16,8 +16,8 @@ define void @foo(i64 %img, float* %red, i32 %idx) {
; SM30: tex.1d.v4.f32.s32 {%f[[RED:[0-9]+]], %f[[GREEN:[0-9]+]], %f[[BLUE:[0-9]+]], %f[[ALPHA:[0-9]+]]}, [%rd[[TEXREG]], {%r{{[0-9]+}}}]
%val = tail call { float, float, float, float } @llvm.nvvm.tex.unified.1d.v4f32.s32(i64 %img, i32 %idx)
%ret = extractvalue { float, float, float, float } %val, 0
-; SM20: st.f32 [%r{{[0-9]+}}], %f[[RED]]
-; SM30: st.f32 [%r{{[0-9]+}}], %f[[RED]]
+; SM20: st.global.f32 [%r{{[0-9]+}}], %f[[RED]]
+; SM30: st.global.f32 [%r{{[0-9]+}}], %f[[RED]]
store float %ret, float* %red
ret void
}
@@ -34,8 +34,8 @@ define void @bar(float* %red, i32 %idx) {
; SM30: tex.1d.v4.f32.s32 {%f[[RED:[0-9]+]], %f[[GREEN:[0-9]+]], %f[[BLUE:[0-9]+]], %f[[ALPHA:[0-9]+]]}, [%rd[[TEXHANDLE]], {%r{{[0-9]+}}}]
%val = tail call { float, float, float, float } @llvm.nvvm.tex.unified.1d.v4f32.s32(i64 %texHandle, i32 %idx)
%ret = extractvalue { float, float, float, float } %val, 0
-; SM20: st.f32 [%r{{[0-9]+}}], %f[[RED]]
-; SM30: st.f32 [%r{{[0-9]+}}], %f[[RED]]
+; SM20: st.global.f32 [%r{{[0-9]+}}], %f[[RED]]
+; SM30: st.global.f32 [%r{{[0-9]+}}], %f[[RED]]
store float %ret, float* %red
ret void
}
diff --git a/test/CodeGen/PowerPC/fma.ll b/test/CodeGen/PowerPC/fma.ll
index ab5251b2a554..9cfef398edfd 100644
--- a/test/CodeGen/PowerPC/fma.ll
+++ b/test/CodeGen/PowerPC/fma.ll
@@ -1,9 +1,12 @@
; RUN: llc < %s -march=ppc32 -fp-contract=fast -mattr=-vsx | FileCheck %s
; RUN: llc < %s -mtriple=powerpc64-unknown-linux-gnu -fp-contract=fast -mattr=+vsx -mcpu=pwr7 | FileCheck -check-prefix=CHECK-VSX %s
+; RUN: llc < %s -mtriple=powerpc64-unknown-linux-gnu -fp-contract=fast -mcpu=pwr8 | FileCheck -check-prefix=CHECK-P8 %s
+; RUN: llc < %s -mtriple=powerpc64le-unknown-linux-gnu -fp-contract=fast -mcpu=pwr8 | FileCheck -check-prefix=CHECK-P8 %s
declare double @dummy1(double) #0
declare double @dummy2(double, double) #0
declare double @dummy3(double, double, double) #0
+declare float @dummy4(float, float) #0
define double @test_FMADD1(double %A, double %B, double %C) {
%D = fmul double %A, %B ; <double> [#uses=1]
@@ -126,3 +129,83 @@ define float @test_FNMSUBS(float %A, float %B, float %C) {
; CHECK-VSX: fnmsubs
; CHECK-VSX-NEXT: blr
}
+
+define float @test_XSMADDMSP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fadd float %C, %D ; <float> [#uses=1]
+ ret float %E
+; CHECK-P8-LABEL: test_XSMADDMSP:
+; CHECK-P8: xsmaddmsp
+; CHECK-P8-NEXT: blr
+}
+
+define float @test_XSMSUBMSP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fsub float %D, %C ; <float> [#uses=1]
+ ret float %E
+; CHECK-P8-LABEL: test_XSMSUBMSP:
+; CHECK-P8: xsmsubmsp
+; CHECK-P8-NEXT: blr
+}
+
+define float @test_XSMADDASP(float %A, float %B, float %C, float %D) {
+ %E = fmul float %A, %B ; <float> [#uses=2]
+ %F = fadd float %E, %C ; <float> [#uses=1]
+ %G = fsub float %E, %D ; <float> [#uses=1]
+ %H = call float @dummy4(float %F, float %G) ; <float> [#uses=1]
+ ret float %H
+; CHECK-P8-LABEL: test_XSMADDASP:
+; CHECK-P8: xsmaddasp
+; CHECK-P8-NEXT: xsmsubmsp
+}
+
+define float @test_XSMSUBASP(float %A, float %B, float %C, float %D) {
+ %E = fmul float %A, %B ; <float> [#uses=2]
+ %F = fsub float %E, %C ; <float> [#uses=1]
+ %G = fsub float %E, %D ; <float> [#uses=1]
+ %H = call float @dummy4(float %F, float %G) ; <float> [#uses=1]
+ ret float %H
+; CHECK-P8-LABEL: test_XSMSUBASP:
+; CHECK-P8: xsmsubasp
+; CHECK-P8-NEXT: xsmsubmsp
+}
+
+define float @test_XSNMADDMSP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fadd float %D, %C ; <float> [#uses=1]
+ %F = fsub float -0.000000e+00, %E ; <float> [#uses=1]
+ ret float %F
+; CHECK-P8-LABEL: test_XSNMADDMSP:
+; CHECK-P8: xsnmaddmsp
+; CHECK-P8-NEXT: blr
+}
+
+define float @test_XSNMSUBMSP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fsub float %D, %C ; <float> [#uses=1]
+ %F = fsub float -0.000000e+00, %E ; <float> [#uses=1]
+ ret float %F
+; CHECK-P8-LABEL: test_XSNMSUBMSP:
+; CHECK-P8: xsnmsubmsp
+; CHECK-P8-NEXT: blr
+}
+
+define float @test_XSNMADDASP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fadd float %D, %C ; <float> [#uses=1]
+ %F = fsub float -0.000000e+00, %E ; <float> [#uses=1]
+ %H = call float @dummy4(float %E, float %F) ; <float> [#uses=1]
+ ret float %F
+; CHECK-P8-LABEL: test_XSNMADDASP:
+; CHECK-P8: xsnmaddasp
+}
+
+define float @test_XSNMSUBASP(float %A, float %B, float %C) {
+ %D = fmul float %A, %B ; <float> [#uses=1]
+ %E = fsub float %D, %C ; <float> [#uses=1]
+ %F = fsub float -0.000000e+00, %E ; <float> [#uses=1]
+ %H = call float @dummy4(float %E, float %F) ; <float> [#uses=1]
+ ret float %F
+; CHECK-P8-LABEL: test_XSNMSUBASP:
+; CHECK-P8: xsnmsubasp
+}
diff --git a/test/CodeGen/PowerPC/vsx-fma-sp.ll b/test/CodeGen/PowerPC/vsx-fma-sp.ll
new file mode 100644
index 000000000000..1c3e457f92cb
--- /dev/null
+++ b/test/CodeGen/PowerPC/vsx-fma-sp.ll
@@ -0,0 +1,167 @@
+; RUN: llc < %s -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 -mattr=+vsx | FileCheck %s
+; RUN: llc < %s -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 -mattr=+vsx -fast-isel -O0 | FileCheck -check-prefix=CHECK-FISL %s
+define void @test1sp(float %a, float %b, float %c, float %e, float* nocapture %d) #0 {
+entry:
+ %0 = tail call float @llvm.fma.f32(float %b, float %c, float %a)
+ store float %0, float* %d, align 4
+ %1 = tail call float @llvm.fma.f32(float %b, float %e, float %a)
+ %arrayidx1 = getelementptr inbounds float, float* %d, i64 1
+ store float %1, float* %arrayidx1, align 4
+ ret void
+
+; CHECK-LABEL: @test1sp
+; CHECK-DAG: li [[C1:[0-9]+]], 4
+; CHECK-DAG: xsmaddmsp 3, 2, 1
+; CHECK-DAG: xsmaddasp 1, 2, 4
+; CHECK-DAG: stxsspx 3, 0, 7
+; CHECK-DAG: stxsspx 1, 7, [[C1]]
+; CHECK: blr
+
+; CHECK-FISL-LABEL: @test1sp
+; CHECK-FISL-DAG: fmr 0, 1
+; CHECK-FISL-DAG: xsmaddasp 0, 2, 3
+; CHECK-FISL-DAG: stxsspx 0, 0, 7
+; CHECK-FISL-DAG: xsmaddasp 1, 2, 4
+; CHECK-FISL-DAG: li [[C1:[0-9]+]], 4
+; CHECK-FISL-DAG: stxsspx 1, 7, [[C1]]
+; CHECK-FISL: blr
+}
+
+define void @test2sp(float %a, float %b, float %c, float %e, float %f, float* nocapture %d) #0 {
+entry:
+ %0 = tail call float @llvm.fma.f32(float %b, float %c, float %a)
+ store float %0, float* %d, align 4
+ %1 = tail call float @llvm.fma.f32(float %b, float %e, float %a)
+ %arrayidx1 = getelementptr inbounds float, float* %d, i64 1
+ store float %1, float* %arrayidx1, align 4
+ %2 = tail call float @llvm.fma.f32(float %b, float %f, float %a)
+ %arrayidx2 = getelementptr inbounds float, float* %d, i64 2
+ store float %2, float* %arrayidx2, align 4
+ ret void
+
+; CHECK-LABEL: @test2sp
+; CHECK-DAG: li [[C1:[0-9]+]], 4
+; CHECK-DAG: li [[C2:[0-9]+]], 8
+; CHECK-DAG: xsmaddmsp 3, 2, 1
+; CHECK-DAG: xsmaddmsp 4, 2, 1
+; CHECK-DAG: xsmaddasp 1, 2, 5
+; CHECK-DAG: stxsspx 3, 0, 8
+; CHECK-DAG: stxsspx 4, 8, [[C1]]
+; CHECK-DAG: stxsspx 1, 8, [[C2]]
+; CHECK: blr
+
+; CHECK-FISL-LABEL: @test2sp
+; CHECK-FISL-DAG: fmr 0, 1
+; CHECK-FISL-DAG: xsmaddasp 0, 2, 3
+; CHECK-FISL-DAG: stxsspx 0, 0, 8
+; CHECK-FISL-DAG: fmr 0, 1
+; CHECK-FISL-DAG: xsmaddasp 0, 2, 4
+; CHECK-FISL-DAG: li [[C1:[0-9]+]], 4
+; CHECK-FISL-DAG: stxsspx 0, 8, [[C1]]
+; CHECK-FISL-DAG: xsmaddasp 1, 2, 5
+; CHECK-FISL-DAG: li [[C2:[0-9]+]], 8
+; CHECK-FISL-DAG: stxsspx 1, 8, [[C2]]
+; CHECK-FISL: blr
+}
+
+define void @test3sp(float %a, float %b, float %c, float %e, float %f, float* nocapture %d) #0 {
+entry:
+ %0 = tail call float @llvm.fma.f32(float %b, float %c, float %a)
+ store float %0, float* %d, align 4
+ %1 = tail call float @llvm.fma.f32(float %b, float %e, float %a)
+ %2 = tail call float @llvm.fma.f32(float %b, float %c, float %1)
+ %arrayidx1 = getelementptr inbounds float, float* %d, i64 3
+ store float %2, float* %arrayidx1, align 4
+ %3 = tail call float @llvm.fma.f32(float %b, float %f, float %a)
+ %arrayidx2 = getelementptr inbounds float, float* %d, i64 2
+ store float %3, float* %arrayidx2, align 4
+ %arrayidx3 = getelementptr inbounds float, float* %d, i64 1
+ store float %1, float* %arrayidx3, align 4
+ ret void
+
+; CHECK-LABEL: @test3sp
+; CHECK-DAG: fmr [[F1:[0-9]+]], 1
+; CHECK-DAG: li [[C1:[0-9]+]], 12
+; CHECK-DAG: li [[C2:[0-9]+]], 8
+; CHECK-DAG: li [[C3:[0-9]+]], 4
+; CHECK-DAG: xsmaddmsp 4, 2, 1
+; CHECK-DAG: xsmaddasp 1, 2, 5
+
+; Note: We could convert this next FMA to M-type as well, but it would require
+; re-ordering the instructions.
+; CHECK-DAG: xsmaddasp [[F1]], 2, 3
+
+; CHECK-DAG: xsmaddmsp 3, 2, 4
+; CHECK-DAG: stxsspx [[F1]], 0, 8
+; CHECK-DAG: stxsspx 3, 8, [[C1]]
+; CHECK-DAG: stxsspx 1, 8, [[C2]]
+; CHECK-DAG: stxsspx 4, 8, [[C3]]
+; CHECK: blr
+
+; CHECK-FISL-LABEL: @test3sp
+; CHECK-FISL-DAG: fmr [[F1:[0-9]+]], 1
+; CHECK-FISL-DAG: xsmaddasp [[F1]], 2, 4
+; CHECK-FISL-DAG: fmr 4, [[F1]]
+; CHECK-FISL-DAG: xsmaddasp 4, 2, 3
+; CHECK-FISL-DAG: li [[C1:[0-9]+]], 12
+; CHECK-FISL-DAG: stxsspx 4, 8, [[C1]]
+; CHECK-FISL-DAG: xsmaddasp 1, 2, 5
+; CHECK-FISL-DAG: li [[C2:[0-9]+]], 8
+; CHECK-FISL-DAG: stxsspx 1, 8, [[C2]]
+; CHECK-FISL-DAG: li [[C3:[0-9]+]], 4
+; CHECK-FISL-DAG: stxsspx 0, 8, [[C3]]
+; CHECK-FISL: blr
+}
+
+define void @test4sp(float %a, float %b, float %c, float %e, float %f, float* nocapture %d) #0 {
+entry:
+ %0 = tail call float @llvm.fma.f32(float %b, float %c, float %a)
+ store float %0, float* %d, align 4
+ %1 = tail call float @llvm.fma.f32(float %b, float %e, float %a)
+ %arrayidx1 = getelementptr inbounds float, float* %d, i64 1
+ store float %1, float* %arrayidx1, align 4
+ %2 = tail call float @llvm.fma.f32(float %b, float %c, float %1)
+ %arrayidx3 = getelementptr inbounds float, float* %d, i64 3
+ store float %2, float* %arrayidx3, align 4
+ %3 = tail call float @llvm.fma.f32(float %b, float %f, float %a)
+ %arrayidx4 = getelementptr inbounds float, float* %d, i64 2
+ store float %3, float* %arrayidx4, align 4
+ ret void
+
+; CHECK-LABEL: @test4sp
+; CHECK-DAG: fmr [[F1:[0-9]+]], 1
+; CHECK-DAG: li [[C1:[0-9]+]], 4
+; CHECK-DAG: li [[C2:[0-9]+]], 8
+; CHECK-DAG: xsmaddmsp 4, 2, 1
+
+; Note: We could convert this next FMA to M-type as well, but it would require
+; re-ordering the instructions.
+; CHECK-DAG: xsmaddasp 1, 2, 5
+
+; CHECK-DAG: xsmaddasp [[F1]], 2, 3
+; CHECK-DAG: stxsspx [[F1]], 0, 8
+; CHECK-DAG: stxsspx 4, 8, [[C1]]
+; CHECK-DAG: li [[C3:[0-9]+]], 12
+; CHECK-DAG: xsmaddasp 4, 2, 3
+; CHECK-DAG: stxsspx 4, 8, [[C3]]
+; CHECK-DAG: stxsspx 1, 8, [[C2]]
+; CHECK: blr
+
+; CHECK-FISL-LABEL: @test4sp
+; CHECK-FISL-DAG: fmr [[F1:[0-9]+]], 1
+; CHECK-FISL-DAG: xsmaddasp [[F1]], 2, 3
+; CHECK-FISL-DAG: stxsspx 0, 0, 8
+; CHECK-FISL-DAG: fmr [[F1]], 1
+; CHECK-FISL-DAG: xsmaddasp [[F1]], 2, 4
+; CHECK-FISL-DAG: li [[C3:[0-9]+]], 4
+; CHECK-FISL-DAG: stxsspx 0, 8, [[C3]]
+; CHECK-FISL-DAG: xsmaddasp 0, 2, 3
+; CHECK-FISL-DAG: li [[C1:[0-9]+]], 12
+; CHECK-FISL-DAG: stxsspx 0, 8, [[C1]]
+; CHECK-FISL-DAG: xsmaddasp 1, 2, 5
+; CHECK-FISL-DAG: li [[C2:[0-9]+]], 8
+; CHECK-FISL-DAG: stxsspx 1, 8, [[C2]]
+; CHECK-FISL: blr
+}
+
+declare float @llvm.fma.f32(float, float, float) #0
diff --git a/test/CodeGen/R600/cgp-addressing-modes.ll b/test/CodeGen/R600/cgp-addressing-modes.ll
new file mode 100644
index 000000000000..3d36bd19937e
--- /dev/null
+++ b/test/CodeGen/R600/cgp-addressing-modes.ll
@@ -0,0 +1,242 @@
+; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown < %s | FileCheck -check-prefix=OPT %s
+; RUN: llc -march=amdgcn -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN %s
+
+declare i32 @llvm.r600.read.tidig.x() #0
+
+; OPT-LABEL: @test_sink_global_small_offset_i32(
+; OPT-NOT: getelementptr i32, i32 addrspace(1)* %in
+; OPT: br i1
+; OPT: ptrtoint
+
+; GCN-LABEL: {{^}}test_sink_global_small_offset_i32:
+; GCN: {{^}}BB0_2:
+define void @test_sink_global_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond) {
+entry:
+ %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i64 7
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i32, i32 addrspace(1)* %in.gep
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+; OPT-LABEL: @test_sink_global_small_max_i32_ds_offset(
+; OPT: %in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 65535
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_global_small_max_i32_ds_offset:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, s{{[0-9]+$}}
+; GCN: {{^}}BB1_2:
+; GCN: s_or_b64 exec
+define void @test_sink_global_small_max_i32_ds_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
+entry:
+ %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 99999
+ %in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 65535
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i8, i8 addrspace(1)* %in.gep
+ %tmp2 = sext i8 %tmp1 to i32
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+; GCN-LABEL: {{^}}test_sink_global_small_max_mubuf_offset:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:4095{{$}}
+; GCN: {{^}}BB2_2:
+; GCN: s_or_b64 exec
+define void @test_sink_global_small_max_mubuf_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
+entry:
+ %out.gep = getelementptr i32, i32 addrspace(1)* %out, i32 1024
+ %in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 4095
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i8, i8 addrspace(1)* %in.gep
+ %tmp2 = sext i8 %tmp1 to i32
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+; GCN-LABEL: {{^}}test_sink_global_small_max_plus_1_mubuf_offset:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, s{{[0-9]+$}}
+; GCN: {{^}}BB3_2:
+; GCN: s_or_b64 exec
+define void @test_sink_global_small_max_plus_1_mubuf_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
+entry:
+ %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 99999
+ %in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 4096
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i8, i8 addrspace(1)* %in.gep
+ %tmp2 = sext i8 %tmp1 to i32
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+; OPT-LABEL: @test_no_sink_flat_small_offset_i32(
+; OPT: getelementptr i32, i32 addrspace(4)* %in
+; OPT: br i1
+; OPT-NOT: ptrtoint
+
+; GCN-LABEL: {{^}}test_no_sink_flat_small_offset_i32:
+; GCN: flat_load_dword
+; GCN: {{^}}BB4_2:
+
+define void @test_no_sink_flat_small_offset_i32(i32 addrspace(4)* %out, i32 addrspace(4)* %in, i32 %cond) {
+entry:
+ %out.gep = getelementptr i32, i32 addrspace(4)* %out, i64 999999
+ %in.gep = getelementptr i32, i32 addrspace(4)* %in, i64 7
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i32, i32 addrspace(4)* %in.gep
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(4)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+; OPT-LABEL: @test_sink_scratch_small_offset_i32(
+; OPT-NOT: getelementptr [512 x i32]
+; OPT: br i1
+; OPT: ptrtoint
+
+; GCN-LABEL: {{^}}test_sink_scratch_small_offset_i32:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
+; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
+; GCN: {{^}}BB5_2:
+define void @test_sink_scratch_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
+entry:
+ %alloca = alloca [512 x i32], align 4
+ %out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i64 999998
+ %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+ %add.arg = add i32 %arg, 8
+ %alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1023
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ store volatile i32 123, i32* %alloca.gep
+ %tmp1 = load volatile i32, i32* %alloca.gep
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep.0
+ %load = load volatile i32, i32* %alloca.gep
+ store i32 %load, i32 addrspace(1)* %out.gep.1
+ br label %done
+
+done:
+ ret void
+}
+
+; OPT-LABEL: @test_no_sink_scratch_large_offset_i32(
+; OPT: %alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1024
+; OPT: br i1
+; OPT-NOT: ptrtoint
+
+; GCN-LABEL: {{^}}test_no_sink_scratch_large_offset_i32:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
+; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
+; GCN: {{^}}BB6_2:
+define void @test_no_sink_scratch_large_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
+entry:
+ %alloca = alloca [512 x i32], align 4
+ %out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i64 999998
+ %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+ %add.arg = add i32 %arg, 8
+ %alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1024
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ store volatile i32 123, i32* %alloca.gep
+ %tmp1 = load volatile i32, i32* %alloca.gep
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep.0
+ %load = load volatile i32, i32* %alloca.gep
+ store i32 %load, i32 addrspace(1)* %out.gep.1
+ br label %done
+
+done:
+ ret void
+}
+
+; GCN-LABEL: {{^}}test_sink_global_vreg_sreg_i32:
+; GCN: s_and_saveexec_b64
+; GCN: buffer_load_dword {{v[0-9]+}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: {{^}}BB7_2:
+define void @test_sink_global_vreg_sreg_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %offset, i32 %cond) {
+entry:
+ %offset.ext = zext i32 %offset to i64
+ %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i64 %offset.ext
+ %tmp0 = icmp eq i32 %cond, 0
+ br i1 %tmp0, label %endif, label %if
+
+if:
+ %tmp1 = load i32, i32 addrspace(1)* %in.gep
+ br label %endif
+
+endif:
+ %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+ store i32 %x, i32 addrspace(1)* %out.gep
+ br label %done
+
+done:
+ ret void
+}
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }
diff --git a/test/CodeGen/R600/coalescer_remat.ll b/test/CodeGen/R600/coalescer_remat.ll
index f78a77b36154..96730bcf2e8f 100644
--- a/test/CodeGen/R600/coalescer_remat.ll
+++ b/test/CodeGen/R600/coalescer_remat.ll
@@ -1,5 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -o - %s | FileCheck %s
-target triple="amdgcn--"
+; RUN: llc -march=amdgcn -verify-machineinstrs -mtriple=amdgcn-- -o - %s | FileCheck %s
declare float @llvm.fma.f32(float, float, float)
@@ -12,7 +11,8 @@ declare float @llvm.fma.f32(float, float, float)
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
-; CHECK: ; NumVgprs: 12
+; It's probably OK if this is slightly higher:
+; CHECK: ; NumVgprs: 9
define void @foobar(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in, i32 %flag) {
entry:
%cmpflag = icmp eq i32 %flag, 1
diff --git a/test/CodeGen/R600/codegen-prepare-addrmode-sext.ll b/test/CodeGen/R600/codegen-prepare-addrmode-sext.ll
index 0aecc189e0bf..585172092676 100644
--- a/test/CodeGen/R600/codegen-prepare-addrmode-sext.ll
+++ b/test/CodeGen/R600/codegen-prepare-addrmode-sext.ll
@@ -1,12 +1,10 @@
-; RUN: opt -codegenprepare -S -o - %s | FileCheck --check-prefix=OPT %s
-; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=SI-LLC %s
+; RUN: opt -mtriple=amdgcn-- -codegenprepare -S < %s | FileCheck -check-prefix=OPT %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI-LLC %s
-target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:32:32-p5:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
-target triple = "r600--"
-
-; OPT-LABEL: @test
+; OPT-LABEL: @test(
; OPT: mul nsw i32
; OPT-NEXT: sext
+
; SI-LLC-LABEL: {{^}}test:
; SI-LLC: s_mul_i32
; SI-LLC-NOT: mul
diff --git a/test/CodeGen/R600/fmul.ll b/test/CodeGen/R600/fmul.ll
index 68ebc4dedfe0..addc409c9eb1 100644
--- a/test/CodeGen/R600/fmul.ll
+++ b/test/CodeGen/R600/fmul.ll
@@ -73,4 +73,20 @@ define void @test_mul_2_k_inv(float addrspace(1)* %out, float %x) #0 {
ret void
}
+; There should be three multiplies here; %a should be used twice (once
+; negated), not duplicated into mul x, 5.0 and mul x, -5.0.
+; FUNC-LABEL: {{^}}test_mul_twouse:
+; SI: v_mul_f32
+; SI: v_mul_f32
+; SI: v_mul_f32
+; SI-NOT: v_mul_f32
+define void @test_mul_twouse(float addrspace(1)* %out, float %x, float %y) #0 {
+ %a = fmul float %x, 5.0
+ %b = fsub float -0.0, %a
+ %c = fmul float %b, %y
+ %d = fmul float %c, %a
+ store float %d, float addrspace(1)* %out
+ ret void
+}
+
attributes #0 = { "less-precise-fpmad"="true" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "unsafe-fp-math"="true" }
diff --git a/test/CodeGen/R600/half.ll b/test/CodeGen/R600/half.ll
index 42ee788e88d5..bf8f11860b50 100644
--- a/test/CodeGen/R600/half.ll
+++ b/test/CodeGen/R600/half.ll
@@ -1,62 +1,525 @@
-; RUN: llc < %s -march=amdgcn -mcpu=SI | FileCheck %s
-; RUN: llc < %s -march=amdgcn -mcpu=tonga | FileCheck %s
+; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
-define void @test_load_store(half addrspace(1)* %in, half addrspace(1)* %out) {
-; CHECK-LABEL: {{^}}test_load_store:
-; CHECK: buffer_load_ushort [[TMP:v[0-9]+]]
-; CHECK: buffer_store_short [[TMP]]
+; half args should be promoted to float
+
+; GCN-LABEL: {{^}}load_f16_arg:
+; GCN: s_load_dword [[ARG:s[0-9]+]]
+; GCN: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[ARG]]
+; GCN: buffer_store_short [[CVT]]
+define void @load_f16_arg(half addrspace(1)* %out, half %arg) #0 {
+ store half %arg, half addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}load_v2f16_arg:
+; GCN-DAG: buffer_load_ushort [[V0:v[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0 offset:44
+; GCN-DAG: buffer_load_ushort [[V1:v[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0 offset:46
+; GCN-DAG: buffer_store_short [[V0]], s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
+; GCN-DAG: buffer_store_short [[V1]], s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
+; GCN: s_endpgm
+define void @load_v2f16_arg(<2 x half> addrspace(1)* %out, <2 x half> %arg) #0 {
+ store <2 x half> %arg, <2 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}load_v3f16_arg:
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN-NOT: buffer_load
+; GCN-DAG: buffer_store_dword
+; GCN-DAG: buffer_store_short
+; GCN-NOT: buffer_store
+; GCN: s_endpgm
+define void @load_v3f16_arg(<3 x half> addrspace(1)* %out, <3 x half> %arg) #0 {
+ store <3 x half> %arg, <3 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}load_v4f16_arg:
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: s_endpgm
+define void @load_v4f16_arg(<4 x half> addrspace(1)* %out, <4 x half> %arg) #0 {
+ store <4 x half> %arg, <4 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}load_v8f16_arg:
+define void @load_v8f16_arg(<8 x half> addrspace(1)* %out, <8 x half> %arg) #0 {
+ store <8 x half> %arg, <8 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v2f16_arg:
+define void @extload_v2f16_arg(<2 x float> addrspace(1)* %out, <2 x half> %in) #0 {
+ %fpext = fpext <2 x half> %in to <2 x float>
+ store <2 x float> %fpext, <2 x float> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_f16_to_f32_arg:
+define void @extload_f16_to_f32_arg(float addrspace(1)* %out, half %arg) #0 {
+ %ext = fpext half %arg to float
+ store float %ext, float addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v2f16_to_v2f32_arg:
+define void @extload_v2f16_to_v2f32_arg(<2 x float> addrspace(1)* %out, <2 x half> %arg) #0 {
+ %ext = fpext <2 x half> %arg to <2 x float>
+ store <2 x float> %ext, <2 x float> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v3f16_to_v3f32_arg:
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN: buffer_load_ushort
+; GCN-NOT: buffer_load
+; GCN: v_cvt_f32_f16_e32
+; GCN: v_cvt_f32_f16_e32
+; GCN: v_cvt_f32_f16_e32
+; GCN-NOT: v_cvt_f32_f16
+; GCN-DAG: buffer_store_dword
+; GCN-DAG: buffer_store_dwordx2
+; GCN: s_endpgm
+define void @extload_v3f16_to_v3f32_arg(<3 x float> addrspace(1)* %out, <3 x half> %arg) #0 {
+ %ext = fpext <3 x half> %arg to <3 x float>
+ store <3 x float> %ext, <3 x float> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v4f16_to_v4f32_arg:
+define void @extload_v4f16_to_v4f32_arg(<4 x float> addrspace(1)* %out, <4 x half> %arg) #0 {
+ %ext = fpext <4 x half> %arg to <4 x float>
+ store <4 x float> %ext, <4 x float> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v8f16_to_v8f32_arg:
+define void @extload_v8f16_to_v8f32_arg(<8 x float> addrspace(1)* %out, <8 x half> %arg) #0 {
+ %ext = fpext <8 x half> %arg to <8 x float>
+ store <8 x float> %ext, <8 x float> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_f16_to_f64_arg:
+define void @extload_f16_to_f64_arg(double addrspace(1)* %out, half %arg) #0 {
+ %ext = fpext half %arg to double
+ store double %ext, double addrspace(1)* %out
+ ret void
+}
+; GCN-LABEL: {{^}}extload_v2f16_to_v2f64_arg:
+define void @extload_v2f16_to_v2f64_arg(<2 x double> addrspace(1)* %out, <2 x half> %arg) #0 {
+ %ext = fpext <2 x half> %arg to <2 x double>
+ store <2 x double> %ext, <2 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v3f16_to_v3f64_arg:
+define void @extload_v3f16_to_v3f64_arg(<3 x double> addrspace(1)* %out, <3 x half> %arg) #0 {
+ %ext = fpext <3 x half> %arg to <3 x double>
+ store <3 x double> %ext, <3 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v4f16_to_v4f64_arg:
+define void @extload_v4f16_to_v4f64_arg(<4 x double> addrspace(1)* %out, <4 x half> %arg) #0 {
+ %ext = fpext <4 x half> %arg to <4 x double>
+ store <4 x double> %ext, <4 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}extload_v8f16_to_v8f64_arg:
+define void @extload_v8f16_to_v8f64_arg(<8 x double> addrspace(1)* %out, <8 x half> %arg) #0 {
+ %ext = fpext <8 x half> %arg to <8 x double>
+ store <8 x double> %ext, <8 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_load_store_f16:
+; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
+; GCN: buffer_store_short [[TMP]]
+define void @global_load_store_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 {
%val = load half, half addrspace(1)* %in
- store half %val, half addrspace(1) * %out
+ store half %val, half addrspace(1)* %out
ret void
}
-define void @test_bitcast_from_half(half addrspace(1)* %in, i16 addrspace(1)* %out) {
-; CHECK-LABEL: {{^}}test_bitcast_from_half:
-; CHECK: buffer_load_ushort [[TMP:v[0-9]+]]
-; CHECK: buffer_store_short [[TMP]]
- %val = load half, half addrspace(1) * %in
- %val_int = bitcast half %val to i16
- store i16 %val_int, i16 addrspace(1)* %out
+; GCN-LABEL: {{^}}global_load_store_v2f16:
+; GCN: buffer_load_dword [[TMP:v[0-9]+]]
+; GCN: buffer_store_dword [[TMP]]
+define void @global_load_store_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+ %val = load <2 x half>, <2 x half> addrspace(1)* %in
+ store <2 x half> %val, <2 x half> addrspace(1)* %out
ret void
}
-define void @test_bitcast_to_half(half addrspace(1)* %out, i16 addrspace(1)* %in) {
-; CHECK-LABEL: {{^}}test_bitcast_to_half:
-; CHECK: buffer_load_ushort [[TMP:v[0-9]+]]
-; CHECK: buffer_store_short [[TMP]]
- %val = load i16, i16 addrspace(1)* %in
- %val_fp = bitcast i16 %val to half
- store half %val_fp, half addrspace(1)* %out
+; GCN-LABEL: {{^}}global_load_store_v4f16:
+; GCN: buffer_load_dwordx2 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: buffer_store_dwordx2 [[TMP]]
+define void @global_load_store_v4f16(<4 x half> addrspace(1)* %in, <4 x half> addrspace(1)* %out) #0 {
+ %val = load <4 x half>, <4 x half> addrspace(1)* %in
+ store <4 x half> %val, <4 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_load_store_v8f16:
+; GCN: buffer_load_dwordx4 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: buffer_store_dwordx4 [[TMP:v\[[0-9]+:[0-9]+\]]]
+; GCN: s_endpgm
+define void @global_load_store_v8f16(<8 x half> addrspace(1)* %out, <8 x half> addrspace(1)* %in) #0 {
+ %val = load <8 x half>, <8 x half> addrspace(1)* %in
+ store <8 x half> %val, <8 x half> addrspace(1)* %out
ret void
}
-define void @test_extend32(half addrspace(1)* %in, float addrspace(1)* %out) {
-; CHECK-LABEL: {{^}}test_extend32:
-; CHECK: v_cvt_f32_f16_e32
+; GCN-LABEL: {{^}}global_extload_f16_to_f32:
+; GCN: buffer_load_ushort [[LOAD:v[0-9]+]]
+; GCN: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], [[LOAD]]
+; GCN: buffer_store_dword [[CVT]]
+define void @global_extload_f16_to_f32(float addrspace(1)* %out, half addrspace(1)* %in) #0 {
+ %val = load half, half addrspace(1)* %in
+ %cvt = fpext half %val to float
+ store float %cvt, float addrspace(1)* %out
+ ret void
+}
- %val16 = load half, half addrspace(1)* %in
- %val32 = fpext half %val16 to float
- store float %val32, float addrspace(1)* %out
+; GCN-LABEL: {{^}}global_extload_v2f16_to_v2f32:
+define void @global_extload_v2f16_to_v2f32(<2 x float> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+ %val = load <2 x half>, <2 x half> addrspace(1)* %in
+ %cvt = fpext <2 x half> %val to <2 x float>
+ store <2 x float> %cvt, <2 x float> addrspace(1)* %out
ret void
}
-define void @test_extend64(half addrspace(1)* %in, double addrspace(1)* %out) {
-; CHECK-LABEL: {{^}}test_extend64:
-; CHECK: v_cvt_f32_f16_e32
-; CHECK: v_cvt_f64_f32_e32
+; GCN-LABEL: {{^}}global_extload_v3f16_to_v3f32:
+define void @global_extload_v3f16_to_v3f32(<3 x float> addrspace(1)* %out, <3 x half> addrspace(1)* %in) #0 {
+ %val = load <3 x half>, <3 x half> addrspace(1)* %in
+ %cvt = fpext <3 x half> %val to <3 x float>
+ store <3 x float> %cvt, <3 x float> addrspace(1)* %out
+ ret void
+}
- %val16 = load half, half addrspace(1)* %in
- %val64 = fpext half %val16 to double
- store double %val64, double addrspace(1)* %out
+; GCN-LABEL: {{^}}global_extload_v4f16_to_v4f32:
+define void @global_extload_v4f16_to_v4f32(<4 x float> addrspace(1)* %out, <4 x half> addrspace(1)* %in) #0 {
+ %val = load <4 x half>, <4 x half> addrspace(1)* %in
+ %cvt = fpext <4 x half> %val to <4 x float>
+ store <4 x float> %cvt, <4 x float> addrspace(1)* %out
ret void
}
-define void @test_trunc32(float addrspace(1)* %in, half addrspace(1)* %out) {
-; CHECK-LABEL: {{^}}test_trunc32:
-; CHECK: v_cvt_f16_f32_e32
+; GCN-LABEL: {{^}}global_extload_v8f16_to_v8f32:
+define void @global_extload_v8f16_to_v8f32(<8 x float> addrspace(1)* %out, <8 x half> addrspace(1)* %in) #0 {
+ %val = load <8 x half>, <8 x half> addrspace(1)* %in
+ %cvt = fpext <8 x half> %val to <8 x float>
+ store <8 x float> %cvt, <8 x float> addrspace(1)* %out
+ ret void
+}
- %val32 = load float, float addrspace(1)* %in
- %val16 = fptrunc float %val32 to half
- store half %val16, half addrspace(1)* %out
+; GCN-LABEL: {{^}}global_extload_v16f16_to_v16f32:
+define void @global_extload_v16f16_to_v16f32(<16 x float> addrspace(1)* %out, <16 x half> addrspace(1)* %in) #0 {
+ %val = load <16 x half>, <16 x half> addrspace(1)* %in
+ %cvt = fpext <16 x half> %val to <16 x float>
+ store <16 x float> %cvt, <16 x float> addrspace(1)* %out
ret void
}
+
+; GCN-LABEL: {{^}}global_extload_f16_to_f64:
+; GCN: buffer_load_ushort [[LOAD:v[0-9]+]]
+; GCN: v_cvt_f32_f16_e32 [[CVT0:v[0-9]+]], [[LOAD]]
+; GCN: v_cvt_f64_f32_e32 [[CVT1:v\[[0-9]+:[0-9]+\]]], [[CVT0]]
+; GCN: buffer_store_dwordx2 [[CVT1]]
+define void @global_extload_f16_to_f64(double addrspace(1)* %out, half addrspace(1)* %in) #0 {
+ %val = load half, half addrspace(1)* %in
+ %cvt = fpext half %val to double
+ store double %cvt, double addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_extload_v2f16_to_v2f64:
+define void @global_extload_v2f16_to_v2f64(<2 x double> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
+ %val = load <2 x half>, <2 x half> addrspace(1)* %in
+ %cvt = fpext <2 x half> %val to <2 x double>
+ store <2 x double> %cvt, <2 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_extload_v3f16_to_v3f64:
+define void @global_extload_v3f16_to_v3f64(<3 x double> addrspace(1)* %out, <3 x half> addrspace(1)* %in) #0 {
+ %val = load <3 x half>, <3 x half> addrspace(1)* %in
+ %cvt = fpext <3 x half> %val to <3 x double>
+ store <3 x double> %cvt, <3 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_extload_v4f16_to_v4f64:
+define void @global_extload_v4f16_to_v4f64(<4 x double> addrspace(1)* %out, <4 x half> addrspace(1)* %in) #0 {
+ %val = load <4 x half>, <4 x half> addrspace(1)* %in
+ %cvt = fpext <4 x half> %val to <4 x double>
+ store <4 x double> %cvt, <4 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_extload_v8f16_to_v8f64:
+define void @global_extload_v8f16_to_v8f64(<8 x double> addrspace(1)* %out, <8 x half> addrspace(1)* %in) #0 {
+ %val = load <8 x half>, <8 x half> addrspace(1)* %in
+ %cvt = fpext <8 x half> %val to <8 x double>
+ store <8 x double> %cvt, <8 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_extload_v16f16_to_v16f64:
+define void @global_extload_v16f16_to_v16f64(<16 x double> addrspace(1)* %out, <16 x half> addrspace(1)* %in) #0 {
+ %val = load <16 x half>, <16 x half> addrspace(1)* %in
+ %cvt = fpext <16 x half> %val to <16 x double>
+ store <16 x double> %cvt, <16 x double> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_f32_to_f16:
+; GCN: buffer_load_dword [[LOAD:v[0-9]+]]
+; GCN: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[LOAD]]
+; GCN: buffer_store_short [[CVT]]
+define void @global_truncstore_f32_to_f16(half addrspace(1)* %out, float addrspace(1)* %in) #0 {
+ %val = load float, float addrspace(1)* %in
+ %cvt = fptrunc float %val to half
+ store half %cvt, half addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v2f32_to_v2f16:
+; GCN: buffer_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
+; GCN-DAG: v_cvt_f16_f32_e32 [[CVT0:v[0-9]+]], v[[LO]]
+; GCN-DAG: v_cvt_f16_f32_e32 [[CVT1:v[0-9]+]], v[[HI]]
+; GCN-DAG: buffer_store_short [[CVT0]]
+; GCN-DAG: buffer_store_short [[CVT1]]
+; GCN: s_endpgm
+define void @global_truncstore_v2f32_to_v2f16(<2 x half> addrspace(1)* %out, <2 x float> addrspace(1)* %in) #0 {
+ %val = load <2 x float>, <2 x float> addrspace(1)* %in
+ %cvt = fptrunc <2 x float> %val to <2 x half>
+ store <2 x half> %cvt, <2 x half> addrspace(1)* %out
+ ret void
+}
+
+; FIXME: Shouldn't do 4th conversion
+; GCN-LABEL: {{^}}global_truncstore_v3f32_to_v3f16:
+; GCN: buffer_load_dwordx4
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: buffer_store_short
+; GCN: buffer_store_dword
+; GCN: s_endpgm
+define void @global_truncstore_v3f32_to_v3f16(<3 x half> addrspace(1)* %out, <3 x float> addrspace(1)* %in) #0 {
+ %val = load <3 x float>, <3 x float> addrspace(1)* %in
+ %cvt = fptrunc <3 x float> %val to <3 x half>
+ store <3 x half> %cvt, <3 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v4f32_to_v4f16:
+; GCN: buffer_load_dwordx4
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: s_endpgm
+define void @global_truncstore_v4f32_to_v4f16(<4 x half> addrspace(1)* %out, <4 x float> addrspace(1)* %in) #0 {
+ %val = load <4 x float>, <4 x float> addrspace(1)* %in
+ %cvt = fptrunc <4 x float> %val to <4 x half>
+ store <4 x half> %cvt, <4 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v8f32_to_v8f16:
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: s_endpgm
+define void @global_truncstore_v8f32_to_v8f16(<8 x half> addrspace(1)* %out, <8 x float> addrspace(1)* %in) #0 {
+ %val = load <8 x float>, <8 x float> addrspace(1)* %in
+ %cvt = fptrunc <8 x float> %val to <8 x half>
+ store <8 x half> %cvt, <8 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v16f32_to_v16f16:
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: buffer_load_dword
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: v_cvt_f16_f32_e32
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: buffer_store_short
+; GCN: s_endpgm
+define void @global_truncstore_v16f32_to_v16f16(<16 x half> addrspace(1)* %out, <16 x float> addrspace(1)* %in) #0 {
+ %val = load <16 x float>, <16 x float> addrspace(1)* %in
+ %cvt = fptrunc <16 x float> %val to <16 x half>
+ store <16 x half> %cvt, <16 x half> addrspace(1)* %out
+ ret void
+}
+
+; FIXME: Unsafe math should fold conversions away
+; GCN-LABEL: {{^}}fadd_f16:
+; SI-DAG: v_cvt_f32_f16_e32 v{{[0-9]+}},
+; SI-DAG: v_cvt_f32_f16_e32 v{{[0-9]+}},
+; SI-DAG: v_cvt_f32_f16_e32 v{{[0-9]+}},
+; SI-DAG: v_cvt_f32_f16_e32 v{{[0-9]+}},
+; SI: v_add_f32
+; GCN: s_endpgm
+define void @fadd_f16(half addrspace(1)* %out, half %a, half %b) #0 {
+ %add = fadd half %a, %b
+ store half %add, half addrspace(1)* %out, align 4
+ ret void
+}
+
+; GCN-LABEL: {{^}}fadd_v2f16:
+; SI: v_add_f32
+; SI: v_add_f32
+; GCN: s_endpgm
+define void @fadd_v2f16(<2 x half> addrspace(1)* %out, <2 x half> %a, <2 x half> %b) #0 {
+ %add = fadd <2 x half> %a, %b
+ store <2 x half> %add, <2 x half> addrspace(1)* %out, align 8
+ ret void
+}
+
+; GCN-LABEL: {{^}}fadd_v4f16:
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; GCN: s_endpgm
+define void @fadd_v4f16(<4 x half> addrspace(1)* %out, <4 x half> addrspace(1)* %in) #0 {
+ %b_ptr = getelementptr <4 x half>, <4 x half> addrspace(1)* %in, i32 1
+ %a = load <4 x half>, <4 x half> addrspace(1)* %in, align 16
+ %b = load <4 x half>, <4 x half> addrspace(1)* %b_ptr, align 16
+ %result = fadd <4 x half> %a, %b
+ store <4 x half> %result, <4 x half> addrspace(1)* %out, align 16
+ ret void
+}
+
+; GCN-LABEL: {{^}}fadd_v8f16:
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; SI: v_add_f32
+; GCN: s_endpgm
+define void @fadd_v8f16(<8 x half> addrspace(1)* %out, <8 x half> %a, <8 x half> %b) #0 {
+ %add = fadd <8 x half> %a, %b
+ store <8 x half> %add, <8 x half> addrspace(1)* %out, align 32
+ ret void
+}
+
+; GCN-LABEL: {{^}}fsub_f16:
+; GCN: v_subrev_f32_e32
+; GCN: s_endpgm
+define void @fsub_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 {
+ %b_ptr = getelementptr half, half addrspace(1)* %in, i32 1
+ %a = load half, half addrspace(1)* %in
+ %b = load half, half addrspace(1)* %b_ptr
+ %sub = fsub half %a, %b
+ store half %sub, half addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}test_bitcast_from_half:
+; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
+; GCN: buffer_store_short [[TMP]]
+define void @test_bitcast_from_half(half addrspace(1)* %in, i16 addrspace(1)* %out) #0 {
+ %val = load half, half addrspace(1)* %in
+ %val_int = bitcast half %val to i16
+ store i16 %val_int, i16 addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}test_bitcast_to_half:
+; GCN: buffer_load_ushort [[TMP:v[0-9]+]]
+; GCN: buffer_store_short [[TMP]]
+define void @test_bitcast_to_half(half addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
+ %val = load i16, i16 addrspace(1)* %in
+ %val_fp = bitcast i16 %val to half
+ store half %val_fp, half addrspace(1)* %out
+ ret void
+}
+
+attributes #0 = { nounwind }
diff --git a/test/CodeGen/R600/imm.ll b/test/CodeGen/R600/imm.ll
index 8917cd6dba33..12eed550eb1f 100644
--- a/test/CodeGen/R600/imm.ll
+++ b/test/CodeGen/R600/imm.ll
@@ -36,7 +36,7 @@ define void @store_imm_neg_0.0_i64(i64 addrspace(1) *%out) {
; CHECK-LABEL: {{^}}store_inline_imm_neg_0.0_i32:
; CHECK: v_mov_b32_e32 [[REG:v[0-9]+]], 0x80000000
-; CHECK-NEXT: buffer_store_dword [[REG]]
+; CHECK: buffer_store_dword [[REG]]
define void @store_inline_imm_neg_0.0_i32(i32 addrspace(1)* %out) {
store i32 -2147483648, i32 addrspace(1)* %out
ret void
diff --git a/test/CodeGen/R600/loop-address.ll b/test/CodeGen/R600/loop-address.ll
index 7fadb8dba7b8..f60d574497de 100644
--- a/test/CodeGen/R600/loop-address.ll
+++ b/test/CodeGen/R600/loop-address.ll
@@ -1,13 +1,10 @@
-;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+;RUN: llc < %s -march=r600 -mcpu=redwood < %s | FileCheck %s
;CHECK: ALU_PUSH
;CHECK: LOOP_START_DX10 @11
;CHECK: LOOP_BREAK @10
;CHECK: POP @10
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
-target triple = "r600--"
-
define void @loop_ge(i32 addrspace(1)* nocapture %out, i32 %iterations) #0 {
entry:
%cmp5 = icmp sgt i32 %iterations, 0
diff --git a/test/CodeGen/R600/loop-idiom.ll b/test/CodeGen/R600/loop-idiom.ll
index 810b34fed865..5fd9806813cd 100644
--- a/test/CodeGen/R600/loop-idiom.ll
+++ b/test/CodeGen/R600/loop-idiom.ll
@@ -2,10 +2,6 @@
; RUN: opt -basicaa -loop-idiom -S < %s -march=amdgcn -mcpu=SI -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s
; RUN: opt -basicaa -loop-idiom -S < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s
-target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:32:32-p5:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
-target triple = "r600--"
-
-
; Make sure loop-idiom doesn't create memcpy or memset. There are no library
; implementations of these for R600.
diff --git a/test/CodeGen/R600/max.ll b/test/CodeGen/R600/max.ll
index 1aa9e6883011..fef3e2f0a21c 100644
--- a/test/CodeGen/R600/max.ll
+++ b/test/CodeGen/R600/max.ll
@@ -115,3 +115,54 @@ define void @s_test_umax_ugt_i32(i32 addrspace(1)* %out, i32 %a, i32 %b) nounwin
store i32 %val, i32 addrspace(1)* %out, align 4
ret void
}
+
+; Make sure redundant and removed
+; FUNC-LABEL: {{^}}simplify_demanded_bits_test_umax_ugt_i16:
+; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xb
+; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xc
+; SI: s_max_u32 [[MIN:s[0-9]+]], [[A]], [[B]]
+; SI-NEXT: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
+; SI-NEXT: buffer_store_dword [[VMIN]]
+define void @simplify_demanded_bits_test_umax_ugt_i16(i32 addrspace(1)* %out, i16 zeroext %a, i16 zeroext %b) nounwind {
+ %a.ext = zext i16 %a to i32
+ %b.ext = zext i16 %b to i32
+ %cmp = icmp ugt i32 %a.ext, %b.ext
+ %val = select i1 %cmp, i32 %a.ext, i32 %b.ext
+ %mask = and i32 %val, 65535
+ store i32 %mask, i32 addrspace(1)* %out
+ ret void
+}
+
+; Make sure redundant sign_extend_inreg removed.
+
+; FUNC-LABEL: {{^}}simplify_demanded_bits_test_min_slt_i16:
+; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xb
+; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xc
+; SI: s_max_i32 [[MIN:s[0-9]+]], [[A]], [[B]]
+; SI-NEXT: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
+; SI-NEXT: buffer_store_dword [[VMIN]]
+define void @simplify_demanded_bits_test_min_slt_i16(i32 addrspace(1)* %out, i16 signext %a, i16 signext %b) nounwind {
+ %a.ext = sext i16 %a to i32
+ %b.ext = sext i16 %b to i32
+ %cmp = icmp sgt i32 %a.ext, %b.ext
+ %val = select i1 %cmp, i32 %a.ext, i32 %b.ext
+ %shl = shl i32 %val, 16
+ %sextinreg = ashr i32 %shl, 16
+ store i32 %sextinreg, i32 addrspace(1)* %out
+ ret void
+}
+
+; FIXME: Should get match min/max through extends inserted by
+; legalization.
+
+; FUNC-LABEL: {{^}}s_test_imin_sge_i16:
+; SI: s_sext_i32_i16
+; SI: s_sext_i32_i16
+; SI: v_cmp_ge_i32_e32
+; SI: v_cndmask_b32
+define void @s_test_imin_sge_i16(i16 addrspace(1)* %out, i16 %a, i16 %b) nounwind {
+ %cmp = icmp sge i16 %a, %b
+ %val = select i1 %cmp, i16 %a, i16 %b
+ store i16 %val, i16 addrspace(1)* %out
+ ret void
+}
diff --git a/test/CodeGen/R600/min.ll b/test/CodeGen/R600/min.ll
index 275e9a7d899b..0332d1a8e407 100644
--- a/test/CodeGen/R600/min.ll
+++ b/test/CodeGen/R600/min.ll
@@ -136,3 +136,54 @@ define void @v_test_umin_ult_i32_multi_use(i32 addrspace(1)* %out0, i1 addrspace
store i1 %cmp, i1 addrspace(1)* %outgep1
ret void
}
+
+; Make sure redundant and removed
+; FUNC-LABEL: {{^}}simplify_demanded_bits_test_umin_ult_i16:
+; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xb
+; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xc
+; SI: s_min_u32 [[MIN:s[0-9]+]], [[A]], [[B]]
+; SI-NEXT: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
+; SI-NEXT: buffer_store_dword [[VMIN]]
+define void @simplify_demanded_bits_test_umin_ult_i16(i32 addrspace(1)* %out, i16 zeroext %a, i16 zeroext %b) nounwind {
+ %a.ext = zext i16 %a to i32
+ %b.ext = zext i16 %b to i32
+ %cmp = icmp ult i32 %a.ext, %b.ext
+ %val = select i1 %cmp, i32 %a.ext, i32 %b.ext
+ %mask = and i32 %val, 65535
+ store i32 %mask, i32 addrspace(1)* %out
+ ret void
+}
+
+; Make sure redundant sign_extend_inreg removed.
+
+; FUNC-LABEL: {{^}}simplify_demanded_bits_test_min_slt_i16:
+; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xb
+; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0xc
+; SI: s_min_i32 [[MIN:s[0-9]+]], [[A]], [[B]]
+; SI-NEXT: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
+; SI-NEXT: buffer_store_dword [[VMIN]]
+define void @simplify_demanded_bits_test_min_slt_i16(i32 addrspace(1)* %out, i16 signext %a, i16 signext %b) nounwind {
+ %a.ext = sext i16 %a to i32
+ %b.ext = sext i16 %b to i32
+ %cmp = icmp slt i32 %a.ext, %b.ext
+ %val = select i1 %cmp, i32 %a.ext, i32 %b.ext
+ %shl = shl i32 %val, 16
+ %sextinreg = ashr i32 %shl, 16
+ store i32 %sextinreg, i32 addrspace(1)* %out
+ ret void
+}
+
+; FIXME: Should get match min/max through extends inserted by
+; legalization.
+
+; FUNC-LABEL: {{^}}s_test_imin_sle_i16:
+; SI: s_sext_i32_i16
+; SI: s_sext_i32_i16
+; SI: v_cmp_le_i32_e32
+; SI: v_cndmask_b32
+define void @s_test_imin_sle_i16(i16 addrspace(1)* %out, i16 %a, i16 %b) nounwind {
+ %cmp = icmp sle i16 %a, %b
+ %val = select i1 %cmp, i16 %a, i16 %b
+ store i16 %val, i16 addrspace(1)* %out
+ ret void
+}
diff --git a/test/CodeGen/R600/sext-in-reg.ll b/test/CodeGen/R600/sext-in-reg.ll
index d9ad4935968d..5aedda2ce1a9 100644
--- a/test/CodeGen/R600/sext-in-reg.ll
+++ b/test/CodeGen/R600/sext-in-reg.ll
@@ -450,13 +450,10 @@ define void @vgpr_sext_in_reg_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x
ret void
}
-; FIXME: The BFE should really be eliminated. I think it should happen
-; when computeKnownBitsForTargetNode is implemented for imax.
-
; FUNC-LABEL: {{^}}sext_in_reg_to_illegal_type:
; SI: buffer_load_sbyte
; SI: v_max_i32
-; SI: v_bfe_i32
+; SI-NOT: bfe
; SI: buffer_store_short
define void @sext_in_reg_to_illegal_type(i16 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %src) nounwind {
%tmp5 = load i8, i8 addrspace(1)* %src, align 1
diff --git a/test/CodeGen/R600/si-vector-hang.ll b/test/CodeGen/R600/si-vector-hang.ll
index 94c47fe3c600..bd427dd3ed46 100644
--- a/test/CodeGen/R600/si-vector-hang.ll
+++ b/test/CodeGen/R600/si-vector-hang.ll
@@ -11,10 +11,7 @@
; CHECK: buffer_store_byte
; CHECK: buffer_store_byte
; ModuleID = 'radeon'
-target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
-target triple = "r600--"
-; Function Attrs: nounwind
define void @test_8_min_char(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture readonly %in0, i8 addrspace(1)* nocapture readonly %in1) #0 {
entry:
%0 = load i8, i8 addrspace(1)* %in0, align 1
diff --git a/test/CodeGen/R600/subreg-eliminate-dead.ll b/test/CodeGen/R600/subreg-eliminate-dead.ll
new file mode 100644
index 000000000000..8bd995a8ecbb
--- /dev/null
+++ b/test/CodeGen/R600/subreg-eliminate-dead.ll
@@ -0,0 +1,19 @@
+; RUN: llc -mtriple=amdgcn-- -verify-machineinstrs -o - %s | FileCheck %s
+; LiveRangeEdit::eliminateDeadDef did not update LiveInterval sub ranges
+; properly.
+
+; Just make sure this test doesn't crash.
+; CHECK-LABEL: foobar:
+; CHECK: s_endpgm
+define void @foobar() {
+ %v0 = icmp eq <4 x i32> undef, <i32 0, i32 1, i32 2, i32 3>
+ %v3 = sext <4 x i1> %v0 to <4 x i32>
+ %v4 = extractelement <4 x i32> %v3, i32 1
+ %v5 = icmp ne i32 %v4, 0
+ %v6 = select i1 %v5, i32 undef, i32 0
+ %v15 = insertelement <2 x i32> undef, i32 %v6, i32 1
+ store <2 x i32> %v15, <2 x i32> addrspace(1)* undef, align 8
+ ret void
+}
+
+declare double @llvm.fma.f64(double, double, double)
diff --git a/test/CodeGen/R600/trunc-store-f64-to-f16.ll b/test/CodeGen/R600/trunc-store-f64-to-f16.ll
new file mode 100644
index 000000000000..c29872beef86
--- /dev/null
+++ b/test/CodeGen/R600/trunc-store-f64-to-f16.ll
@@ -0,0 +1,56 @@
+; XFAIL: *
+; RUN: llc -march=amdgcn -mcpu=SI < %s
+
+; GCN-LABEL: {{^}}global_truncstore_f64_to_f16:
+; GCN: s_endpgm
+define void @global_truncstore_f64_to_f16(half addrspace(1)* %out, double addrspace(1)* %in) #0 {
+ %val = load double, double addrspace(1)* %in
+ %cvt = fptrunc double %val to half
+ store half %cvt, half addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v2f64_to_v2f16:
+; GCN: s_endpgm
+define void @global_truncstore_v2f64_to_v2f16(<2 x half> addrspace(1)* %out, <2 x double> addrspace(1)* %in) #0 {
+ %val = load <2 x double>, <2 x double> addrspace(1)* %in
+ %cvt = fptrunc <2 x double> %val to <2 x half>
+ store <2 x half> %cvt, <2 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v3f64_to_v3f16:
+; GCN: s_endpgm
+define void @global_truncstore_v3f64_to_v3f16(<3 x half> addrspace(1)* %out, <3 x double> addrspace(1)* %in) #0 {
+ %val = load <3 x double>, <3 x double> addrspace(1)* %in
+ %cvt = fptrunc <3 x double> %val to <3 x half>
+ store <3 x half> %cvt, <3 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v4f64_to_v4f16:
+; GCN: s_endpgm
+define void @global_truncstore_v4f64_to_v4f16(<4 x half> addrspace(1)* %out, <4 x double> addrspace(1)* %in) #0 {
+ %val = load <4 x double>, <4 x double> addrspace(1)* %in
+ %cvt = fptrunc <4 x double> %val to <4 x half>
+ store <4 x half> %cvt, <4 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v8f64_to_v8f16:
+; GCN: s_endpgm
+define void @global_truncstore_v8f64_to_v8f16(<8 x half> addrspace(1)* %out, <8 x double> addrspace(1)* %in) #0 {
+ %val = load <8 x double>, <8 x double> addrspace(1)* %in
+ %cvt = fptrunc <8 x double> %val to <8 x half>
+ store <8 x half> %cvt, <8 x half> addrspace(1)* %out
+ ret void
+}
+
+; GCN-LABEL: {{^}}global_truncstore_v16f64_to_v16f16:
+; GCN: s_endpgm
+define void @global_truncstore_v16f64_to_v16f16(<16 x half> addrspace(1)* %out, <16 x double> addrspace(1)* %in) #0 {
+ %val = load <16 x double>, <16 x double> addrspace(1)* %in
+ %cvt = fptrunc <16 x double> %val to <16 x half>
+ store <16 x half> %cvt, <16 x half> addrspace(1)* %out
+ ret void
+}
diff --git a/test/CodeGen/R600/unroll.ll b/test/CodeGen/R600/unroll.ll
index ca8d822ec7ed..411a15a4b839 100644
--- a/test/CodeGen/R600/unroll.ll
+++ b/test/CodeGen/R600/unroll.ll
@@ -1,7 +1,6 @@
-; RUN: opt -loop-unroll -simplifycfg -sroa %s -S -o - | FileCheck %s
+; RUN: opt -mtriple=amdgcn-- -loop-unroll -simplifycfg -sroa %s -S -o - | FileCheck %s
+; RUN: opt -mtriple=r600-- -loop-unroll -simplifycfg -sroa %s -S -o - | FileCheck %s
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
-target triple = "r600--"
; This test contains a simple loop that initializes an array declared in
; private memory. We want to make sure these kinds of loops are always
diff --git a/test/CodeGen/R600/wrong-transalu-pos-fix.ll b/test/CodeGen/R600/wrong-transalu-pos-fix.ll
index 5ab465338e15..8b383e4c393d 100644
--- a/test/CodeGen/R600/wrong-transalu-pos-fix.ll
+++ b/test/CodeGen/R600/wrong-transalu-pos-fix.ll
@@ -1,14 +1,9 @@
-; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
+; RUN: llc -march=r600 -mcpu=redwood -mtriple=r600-- < %s | FileCheck %s
; We want all MULLO_INT inst to be last in their instruction group
;CHECK: {{^}}fill3d:
;CHECK-NOT: MULLO_INT T[0-9]+
-; ModuleID = 'radeon'
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
-target triple = "r600--"
-
-; Function Attrs: nounwind
define void @fill3d(i32 addrspace(1)* nocapture %out) #0 {
entry:
%x.i = tail call i32 @llvm.r600.read.global.size.x() #1
diff --git a/test/CodeGen/Thumb2/constant-islands-jump-table.ll b/test/CodeGen/Thumb2/constant-islands-jump-table.ll
index 0dd7092291ba..5ffe1f9b09f6 100644
--- a/test/CodeGen/Thumb2/constant-islands-jump-table.ll
+++ b/test/CodeGen/Thumb2/constant-islands-jump-table.ll
@@ -1,7 +1,7 @@
; RUN: llc < %s -mtriple=thumbv7-linux-gnueabihf -O1 %s -o - | FileCheck %s
; CHECK-LABEL: test_jump_table:
-; CHECK: b .LBB
+; CHECK: b{{.*}} .LBB
; CHECK-NOT: tbh
define i32 @test_jump_table(i32 %x, float %in) {
diff --git a/test/CodeGen/Thumb2/float-ops.ll b/test/CodeGen/Thumb2/float-ops.ll
index 7ec08f866655..4c42908ce13b 100644
--- a/test/CodeGen/Thumb2/float-ops.ll
+++ b/test/CodeGen/Thumb2/float-ops.ll
@@ -109,7 +109,7 @@ entry:
define double @load_d(double* %a) {
entry:
; CHECK-LABEL: load_d:
-; NONE: ldm r0, {r0, r1}
+; NONE: ldrd r0, r1, [r0]
; HARD: vldr d0, [r0]
%0 = load double, double* %a, align 8
ret double %0
@@ -127,9 +127,7 @@ entry:
define void @store_d(double* %a, double %b) {
entry:
; CHECK-LABEL: store_d:
-; NONE: mov r1, r3
-; NONE: str r2, [r0]
-; NONE: str r1, [r0, #4]
+; NONE: strd r2, r3, [r0]
; HARD: vstr d0, [r0]
store double %b, double* %a, align 8
ret void
diff --git a/test/CodeGen/Thumb2/thumb2-tbh.ll b/test/CodeGen/Thumb2/thumb2-tbh.ll
index a5a5ed0c8da2..0761ed589a26 100644
--- a/test/CodeGen/Thumb2/thumb2-tbh.ll
+++ b/test/CodeGen/Thumb2/thumb2-tbh.ll
@@ -14,9 +14,19 @@ declare void @Z_fatal(i8*) noreturn nounwind
declare noalias i8* @calloc(i32, i32) nounwind
+; Jump tables are not anchored next to the TBB/TBH any more. Make sure the
+; correct address is still calculated (i.e. via a PC-relative symbol *at* the
+; TBB/TBH).
define i32 @main(i32 %argc, i8** nocapture %argv) nounwind {
; CHECK-LABEL: main:
-; CHECK: tbb
+; CHECK-NOT: adr {{r[0-9]+}}, LJTI
+; CHECK: [[PCREL_ANCHOR:LCPI[0-9]+_[0-9]+]]:
+; CHECK-NEXT: tbb [pc, {{r[0-9]+}}]
+
+; CHECK: LJTI0_0:
+; CHECK-NEXT: .data_region jt8
+; CHECK-NEXT: .byte (LBB{{[0-9]+_[0-9]+}}-([[PCREL_ANCHOR]]+4))/2
+
entry:
br label %bb42.i
diff --git a/test/CodeGen/X86/asm-reject-reg-type-mismatch.ll b/test/CodeGen/X86/asm-reject-reg-type-mismatch.ll
new file mode 100644
index 000000000000..016e2d261eef
--- /dev/null
+++ b/test/CodeGen/X86/asm-reject-reg-type-mismatch.ll
@@ -0,0 +1,10 @@
+; RUN: not llc -no-integrated-as %s -o - 2> %t1
+; RUN: FileCheck %s < %t1
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--"
+
+; CHECK: error: couldn't allocate output register for constraint '{ax}'
+define i128 @blup() {
+ %v = tail call i128 asm "", "={ax},0,~{dirflag},~{fpsr},~{flags}"(i128 0)
+ ret i128 %v
+}
diff --git a/test/CodeGen/X86/avx-vperm2x128.ll b/test/CodeGen/X86/avx-vperm2x128.ll
index 10ed079a264e..74d20f348b52 100644
--- a/test/CodeGen/X86/avx-vperm2x128.ll
+++ b/test/CodeGen/X86/avx-vperm2x128.ll
@@ -147,8 +147,8 @@ define <16 x i16> @E5i(<16 x i16>* %a, <16 x i16>* %b) nounwind uwtable readnone
; AVX1-LABEL: E5i:
; AVX1: ## BB#0: ## %entry
; AVX1-NEXT: vmovdqa (%rdi), %ymm0
-; AVX1-NEXT: vpaddw {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vmovaps (%rsi), %ymm1
+; AVX1-NEXT: vpaddw {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT: retq
;
diff --git a/test/CodeGen/X86/avx2-vector-shifts.ll b/test/CodeGen/X86/avx2-vector-shifts.ll
index 8aae90c3c03d..5d99269ae1dc 100644
--- a/test/CodeGen/X86/avx2-vector-shifts.ll
+++ b/test/CodeGen/X86/avx2-vector-shifts.ll
@@ -300,6 +300,56 @@ define <16 x i16> @shl_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
ret <16 x i16> %shl
}
+define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: shl_32i8
+; CHECK: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpsllw $4, %xmm3, %xmm2
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm8 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
+; CHECK-NEXT: vpand %xmm8, %xmm2, %xmm5
+; CHECK-NEXT: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpsllw $5, %xmm2, %xmm2
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm9 = [224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224]
+; CHECK-NEXT: vpand %xmm9, %xmm2, %xmm7
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; CHECK-NEXT: vpand %xmm7, %xmm2, %xmm4
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm4, %xmm4
+; CHECK-NEXT: vpblendvb %xmm4, %xmm5, %xmm3, %xmm3
+; CHECK-NEXT: vpsllw $2, %xmm3, %xmm4
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm5 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
+; CHECK-NEXT: vpand %xmm5, %xmm4, %xmm4
+; CHECK-NEXT: vpaddb %xmm7, %xmm7, %xmm7
+; CHECK-NEXT: vpand %xmm7, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpaddb %xmm3, %xmm3, %xmm4
+; CHECK-NEXT: vpaddb %xmm7, %xmm7, %xmm6
+; CHECK-NEXT: vpand %xmm6, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsllw $4, %xmm0, %xmm4
+; CHECK-NEXT: vpand %xmm8, %xmm4, %xmm4
+; CHECK-NEXT: vpsllw $5, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm9, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm6
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm6, %xmm6
+; CHECK-NEXT: vpblendvb %xmm6, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vpsllw $2, %xmm0, %xmm4
+; CHECK-NEXT: vpand %xmm5, %xmm4, %xmm4
+; CHECK-NEXT: vpaddb %xmm1, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm5
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm5, %xmm5
+; CHECK-NEXT: vpblendvb %xmm5, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vpaddb %xmm0, %xmm0, %xmm4
+; CHECK-NEXT: vpaddb %xmm1, %xmm1, %xmm1
+; CHECK-NEXT: vpand %xmm1, %xmm2, %xmm1
+; CHECK-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
+; CHECK-NEXT: vpblendvb %xmm1, %xmm4, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %shl = shl <32 x i8> %r, %a
+ ret <32 x i8> %shl
+}
+
define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; CHECK-LABEL: ashr_8i16
; CHECK: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
@@ -329,6 +379,176 @@ define <16 x i16> @ashr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
ret <16 x i16> %ashr
}
+define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: ashr_32i8
+; CHECK: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpextrb $1, %xmm2, %ecx
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpextrb $1, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $0, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: movzbl %dl, %edx
+; CHECK-NEXT: vpextrb $2, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm3, %esi
+; CHECK-NEXT: sarb %cl, %sil
+; CHECK-NEXT: vmovd %edx, %xmm4
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $4, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $5, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $8, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $9, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $12, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $13, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm3, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $15, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm3, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $1, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $1, %xmm0, %esi
+; CHECK-NEXT: sarb %cl, %sil
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vpinsrb $14, %ecx, %xmm4, %xmm2
+; CHECK-NEXT: vpextrb $0, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpextrb $2, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm0, %edi
+; CHECK-NEXT: sarb %cl, %dil
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vmovd %ecx, %xmm3
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $4, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $5, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $8, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $9, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $12, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $13, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm0, %eax
+; CHECK-NEXT: sarb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $15, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm0, %edx
+; CHECK-NEXT: sarb %cl, %dl
+; CHECK-NEXT: vpinsrb $14, %eax, %xmm3, %xmm0
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %ashr = ashr <32 x i8> %r, %a
+ ret <32 x i8> %ashr
+}
+
define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; CHECK-LABEL: lshr_8i16
; CHECK: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
@@ -357,3 +577,173 @@ define <16 x i16> @lshr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind {
%lshr = lshr <16 x i16> %r, %a
ret <16 x i16> %lshr
}
+
+define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind {
+; CHECK-LABEL: lshr_32i8
+; CHECK: vextracti128 $1, %ymm1, %xmm2
+; CHECK-NEXT: vpextrb $1, %xmm2, %ecx
+; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm3
+; CHECK-NEXT: vpextrb $1, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $0, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: movzbl %dl, %edx
+; CHECK-NEXT: vpextrb $2, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm3, %esi
+; CHECK-NEXT: shrb %cl, %sil
+; CHECK-NEXT: vmovd %edx, %xmm4
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $4, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $5, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $8, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $9, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $12, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $13, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm3, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4
+; CHECK-NEXT: vpextrb $15, %xmm2, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm3, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $1, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $1, %xmm0, %esi
+; CHECK-NEXT: shrb %cl, %sil
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vpinsrb $14, %ecx, %xmm4, %xmm2
+; CHECK-NEXT: vpextrb $0, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $0, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpextrb $2, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $2, %xmm0, %edi
+; CHECK-NEXT: shrb %cl, %dil
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2
+; CHECK-NEXT: movzbl %sil, %eax
+; CHECK-NEXT: movzbl %dl, %ecx
+; CHECK-NEXT: vmovd %ecx, %xmm3
+; CHECK-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: vpextrb $3, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $3, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $4, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $4, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $5, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $5, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $6, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $6, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $7, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $7, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $8, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $8, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $9, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $9, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $10, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $10, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $11, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $11, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $12, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $12, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3
+; CHECK-NEXT: vpextrb $13, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $13, %xmm0, %eax
+; CHECK-NEXT: shrb %cl, %al
+; CHECK-NEXT: vpextrb $14, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $14, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpextrb $15, %xmm1, %ecx
+; CHECK-NEXT: vpextrb $15, %xmm0, %edx
+; CHECK-NEXT: shrb %cl, %dl
+; CHECK-NEXT: vpinsrb $14, %eax, %xmm3, %xmm0
+; CHECK-NEXT: movzbl %dl, %eax
+; CHECK-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0
+; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; CHECK-NEXT: retq
+ %lshr = lshr <32 x i8> %r, %a
+ ret <32 x i8> %lshr
+}
diff --git a/test/CodeGen/X86/avx512-build-vector.ll b/test/CodeGen/X86/avx512-build-vector.ll
index 8373c6da2619..e70d9f3ad521 100644
--- a/test/CodeGen/X86/avx512-build-vector.ll
+++ b/test/CodeGen/X86/avx512-build-vector.ll
@@ -2,13 +2,9 @@
define <16 x i32> @test1(i32* %x) {
; CHECK-LABEL: test1:
-; CHECK: ## BB#0:
-; CHECK-NEXT: vmovd (%rdi), %xmm0
-; CHECK-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; CHECK-NEXT: vpxor %ymm1, %ymm1, %ymm1
-; CHECK-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4],ymm1[5,6,7]
-; CHECK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; CHECK-NEXT: retq
+; CHECK: vmovd (%rdi), %xmm
+; CHECK: vmovdqa32
+; CHECK: vpermt2d %zmm
%y = load i32, i32* %x, align 4
%res = insertelement <16 x i32>zeroinitializer, i32 %y, i32 4
ret <16 x i32>%res
diff --git a/test/CodeGen/X86/avx512-intrinsics.ll b/test/CodeGen/X86/avx512-intrinsics.ll
index 471e34cdedce..9387192f8aa4 100644
--- a/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/test/CodeGen/X86/avx512-intrinsics.ll
@@ -98,18 +98,55 @@ define <4 x float> @test_rcp14_ss(<4 x float> %a0) {
declare <4 x float> @llvm.x86.avx512.rcp14.ss(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone
define <8 x double> @test_sqrt_pd_512(<8 x double> %a0) {
+ ; CHECK-LABEL: test_sqrt_pd_512
; CHECK: vsqrtpd
- %res = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 4) ; <<8 x double>> [#uses=1]
+ %res = call <8 x double> @llvm.x86.avx512.mask.sqrt.pd.512(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 4)
ret <8 x double> %res
}
-declare <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
+declare <8 x double> @llvm.x86.avx512.mask.sqrt.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
define <16 x float> @test_sqrt_ps_512(<16 x float> %a0) {
+ ; CHECK-LABEL: test_sqrt_ps_512
; CHECK: vsqrtps
- %res = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4) ; <<16 x float>> [#uses=1]
+ %res = call <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4)
ret <16 x float> %res
}
-declare <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
+define <16 x float> @test_sqrt_round_ps_512(<16 x float> %a0) {
+ ; CHECK-LABEL: test_sqrt_round_ps_512
+ ; CHECK: vsqrtps {rz-sae}
+ %res = call <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 3)
+ ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
+
+define <8 x double> @test_getexp_pd_512(<8 x double> %a0) {
+ ; CHECK-LABEL: test_getexp_pd_512
+ ; CHECK: vgetexppd
+ %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 4)
+ ret <8 x double> %res
+}
+define <8 x double> @test_getexp_round_pd_512(<8 x double> %a0) {
+ ; CHECK-LABEL: test_getexp_round_pd_512
+ ; CHECK: vgetexppd {sae}
+ %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1, i32 8)
+ ret <8 x double> %res
+}
+declare <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
+
+define <16 x float> @test_getexp_ps_512(<16 x float> %a0) {
+ ; CHECK-LABEL: test_getexp_ps_512
+ ; CHECK: vgetexpps
+ %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4)
+ ret <16 x float> %res
+}
+
+define <16 x float> @test_getexp_round_ps_512(<16 x float> %a0) {
+ ; CHECK-LABEL: test_getexp_round_ps_512
+ ; CHECK: vgetexpps {sae}
+ %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
+ ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
define <4 x float> @test_sqrt_ss(<4 x float> %a0, <4 x float> %a1) {
; CHECK: vsqrtss {{.*}}encoding: [0x62
diff --git a/test/CodeGen/X86/avx512-shuffle.ll b/test/CodeGen/X86/avx512-shuffle.ll
new file mode 100644
index 000000000000..2683d6fe238c
--- /dev/null
+++ b/test/CodeGen/X86/avx512-shuffle.ll
@@ -0,0 +1,336 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=CHECK-SKX
+
+; CHECK-LABEL: test1:
+; CHECK: vpermps
+; CHECK: ret
+define <16 x float> @test1(<16 x float> %a) nounwind {
+ %c = shufflevector <16 x float> %a, <16 x float> undef, <16 x i32> <i32 2, i32 5, i32 undef, i32 undef, i32 7, i32 undef, i32 10, i32 1, i32 0, i32 5, i32 undef, i32 4, i32 7, i32 undef, i32 10, i32 1>
+ ret <16 x float> %c
+}
+
+; CHECK-LABEL: test2:
+; CHECK: vpermd
+; CHECK: ret
+define <16 x i32> @test2(<16 x i32> %a) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> undef, <16 x i32> <i32 2, i32 5, i32 undef, i32 undef, i32 7, i32 undef, i32 10, i32 1, i32 0, i32 5, i32 undef, i32 4, i32 7, i32 undef, i32 10, i32 1>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: test3:
+; CHECK: vpermq
+; CHECK: ret
+define <8 x i64> @test3(<8 x i64> %a) nounwind {
+ %c = shufflevector <8 x i64> %a, <8 x i64> undef, <8 x i32> <i32 2, i32 5, i32 1, i32 undef, i32 7, i32 undef, i32 3, i32 1>
+ ret <8 x i64> %c
+}
+
+; CHECK-LABEL: test4:
+; CHECK: vpermpd
+; CHECK: ret
+define <8 x double> @test4(<8 x double> %a) nounwind {
+ %c = shufflevector <8 x double> %a, <8 x double> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ ret <8 x double> %c
+}
+
+; CHECK-LABEL: test5:
+; CHECK: vpermt2pd
+; CHECK: ret
+define <8 x double> @test5(<8 x double> %a, <8 x double> %b) nounwind {
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 2, i32 8, i32 0, i32 1, i32 6, i32 10, i32 4, i32 5>
+ ret <8 x double> %c
+}
+
+; CHECK-LABEL: test6:
+; CHECK: vpermq $30
+; CHECK: ret
+define <8 x i64> @test6(<8 x i64> %a) nounwind {
+ %c = shufflevector <8 x i64> %a, <8 x i64> undef, <8 x i32> <i32 2, i32 3, i32 1, i32 0, i32 6, i32 7, i32 5, i32 4>
+ ret <8 x i64> %c
+}
+
+; CHECK-LABEL: test7:
+; CHECK: vpermt2q
+; CHECK: ret
+define <8 x i64> @test7(<8 x i64> %a, <8 x i64> %b) nounwind {
+ %c = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 2, i32 8, i32 0, i32 1, i32 6, i32 10, i32 4, i32 5>
+ ret <8 x i64> %c
+}
+
+; CHECK-LABEL: test8:
+; CHECK: vpermt2d
+; CHECK: ret
+define <16 x i32> @test8(<16 x i32> %a, <16 x i32> %b) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 15, i32 31, i32 14, i32 22, i32 13, i32 29, i32 4, i32 28, i32 11, i32 27, i32 10, i32 26, i32 9, i32 25, i32 8, i32 24>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: test9:
+; CHECK: vpermt2ps
+; CHECK: ret
+define <16 x float> @test9(<16 x float> %a, <16 x float> %b) nounwind {
+ %c = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 15, i32 31, i32 14, i32 22, i32 13, i32 29, i32 4, i32 28, i32 11, i32 27, i32 10, i32 26, i32 9, i32 25, i32 8, i32 24>
+ ret <16 x float> %c
+}
+
+; CHECK-LABEL: test10:
+; CHECK: vpermt2ps (
+; CHECK: ret
+define <16 x float> @test10(<16 x float> %a, <16 x float>* %b) nounwind {
+ %c = load <16 x float>, <16 x float>* %b
+ %d = shufflevector <16 x float> %a, <16 x float> %c, <16 x i32> <i32 15, i32 31, i32 14, i32 22, i32 13, i32 29, i32 4, i32 28, i32 11, i32 27, i32 10, i32 26, i32 9, i32 25, i32 8, i32 24>
+ ret <16 x float> %d
+}
+
+; CHECK-LABEL: test11:
+; CHECK: vpermt2d
+; CHECK: ret
+define <16 x i32> @test11(<16 x i32> %a, <16 x i32>* %b) nounwind {
+ %c = load <16 x i32>, <16 x i32>* %b
+ %d = shufflevector <16 x i32> %a, <16 x i32> %c, <16 x i32> <i32 15, i32 31, i32 14, i32 22, i32 13, i32 29, i32 4, i32 28, i32 11, i32 27, i32 10, i32 26, i32 9, i32 25, i32 8, i32 24>
+ ret <16 x i32> %d
+}
+
+; CHECK-LABEL: test13
+; CHECK: vpermilps $177, %zmm
+; CHECK: ret
+define <16 x float> @test13(<16 x float> %a) {
+ %b = shufflevector <16 x float> %a, <16 x float> undef, <16 x i32><i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+ ret <16 x float> %b
+}
+
+; CHECK-LABEL: test14
+; CHECK: vpermilpd $203, %zmm
+; CHECK: ret
+define <8 x double> @test14(<8 x double> %a) {
+ %b = shufflevector <8 x double> %a, <8 x double> undef, <8 x i32><i32 1, i32 1, i32 2, i32 3, i32 4, i32 4, i32 7, i32 7>
+ ret <8 x double> %b
+}
+
+; CHECK-LABEL: test15
+; CHECK: vpshufd $177, %zmm
+; CHECK: ret
+define <16 x i32> @test15(<16 x i32> %a) {
+; mask 1-0-3-2 = 10110001 = 0xb1 = 177
+ %b = shufflevector <16 x i32> %a, <16 x i32> undef, <16 x i32><i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+ ret <16 x i32> %b
+}
+; CHECK-LABEL: test16
+; CHECK: valignq $2, %zmm0, %zmm1
+; CHECK: ret
+define <8 x double> @test16(<8 x double> %a, <8 x double> %b) nounwind {
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9>
+ ret <8 x double> %c
+}
+
+; CHECK-LABEL: test17
+; CHECK: vshufpd $19, %zmm1, %zmm0
+; CHECK: ret
+define <8 x double> @test17(<8 x double> %a, <8 x double> %b) nounwind {
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 9, i32 2, i32 10, i32 5, i32 undef, i32 undef, i32 undef>
+ ret <8 x double> %c
+}
+
+; CHECK-LABEL: test18
+; CHECK: vpunpckhdq %zmm
+; CHECK: ret
+define <16 x i32> @test18(<16 x i32> %a, <16 x i32> %c) {
+ %b = shufflevector <16 x i32> %a, <16 x i32> %c, <16 x i32><i32 2, i32 18, i32 3, i32 19, i32 6, i32 22, i32 7, i32 23, i32 10, i32 26, i32 11, i32 27, i32 14, i32 30, i32 15, i32 31>
+ ret <16 x i32> %b
+}
+
+; CHECK-LABEL: test19
+; CHECK: vpunpckldq %zmm
+; CHECK: ret
+define <16 x i32> @test19(<16 x i32> %a, <16 x i32> %c) {
+ %b = shufflevector <16 x i32> %a, <16 x i32> %c, <16 x i32><i32 0, i32 16, i32 1, i32 17, i32 4, i32 20, i32 5, i32 21, i32 8, i32 24, i32 9, i32 25, i32 12, i32 28, i32 13, i32 29>
+ ret <16 x i32> %b
+}
+
+; CHECK-LABEL: test20
+; CHECK: vpunpckhqdq %zmm
+; CHECK: ret
+define <8 x i64> @test20(<8 x i64> %a, <8 x i64> %c) {
+ %b = shufflevector <8 x i64> %a, <8 x i64> %c, <8 x i32><i32 1, i32 9, i32 3, i32 11, i32 5, i32 13, i32 7, i32 15>
+ ret <8 x i64> %b
+}
+
+; CHECK-LABEL: test21
+; CHECK: vbroadcastsd %xmm0, %zmm
+; CHECK: ret
+define <8 x double> @test21(<8 x double> %a, <8 x double> %b) {
+ %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+ ret <8 x double> %shuffle
+}
+
+; CHECK-LABEL: test22
+; CHECK: vpbroadcastq %xmm0, %zmm
+; CHECK: ret
+define <8 x i64> @test22(<8 x i64> %a, <8 x i64> %b) {
+ %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+ ret <8 x i64> %shuffle
+}
+
+; CHECK-LABEL: @test23
+; CHECK: vshufps
+; CHECK: vshufps
+; CHECK: ret
+define <16 x i32> @test23(<16 x i32> %a, <16 x i32> %b) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: @test24
+; CHECK: vpermt2d
+; CHECK: ret
+define <16 x i32> @test24(<16 x i32> %a, <16 x i32> %b) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 25, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: @test25
+; CHECK: vshufps $52
+; CHECK: ret
+define <16 x i32> @test25(<16 x i32> %a, <16 x i32> %b) nounwind {
+; mask - 0-1-3-0 00110100 = 0x34 = 52
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 19, i32 16, i32 4, i32 5, i32 23, i32 undef, i32 8, i32 9, i32 27, i32 undef, i32 12, i32 13, i32 undef, i32 undef>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: @test26
+; CHECK: vmovshdup
+; CHECK: ret
+define <16 x i32> @test26(<16 x i32> %a) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> undef, <16 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 undef, i32 9, i32 9, i32 undef, i32 11, i32 13, i32 undef, i32 undef, i32 undef>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: @test27
+; CHECK: ret
+define <16 x i32> @test27(<4 x i32>%a) {
+ %res = shufflevector <4 x i32> %a, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ ret <16 x i32> %res
+}
+
+; CHECK-LABEL: test28
+; CHECK: vpshufhw $177, %ymm
+; CHECK: ret
+define <16 x i16> @test28(<16 x i16> %a) {
+ %b = shufflevector <16 x i16> %a, <16 x i16> undef, <16 x i32><i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 7, i32 6, i32 8, i32 9, i32 10, i32 11, i32 13, i32 12, i32 15, i32 14>
+ ret <16 x i16> %b
+}
+
+; CHECK-LABEL: test29
+; CHECK: vunpcklps %zmm
+; CHECK: ret
+define <16 x float> @test29(<16 x float> %a, <16 x float> %c) {
+ %b = shufflevector <16 x float> %a, <16 x float> %c, <16 x i32><i32 0, i32 16, i32 1, i32 17, i32 4, i32 20, i32 5, i32 21, i32 8, i32 24, i32 9, i32 25, i32 12, i32 28, i32 13, i32 29>
+ ret <16 x float> %b
+}
+
+; CHECK-LABEL: @test30
+; CHECK: vshufps $144, %zmm
+; CHECK: ret
+define <16 x float> @test30(<16 x float> %a, <16 x float> %c) {
+ %b = shufflevector <16 x float> %a, <16 x float> %c, <16 x i32><i32 0, i32 0, i32 17, i32 18, i32 4, i32 4, i32 21, i32 22, i32 8, i32 8, i32 25, i32 26, i32 12, i32 12, i32 29, i32 30>
+ ret <16 x float> %b
+}
+
+; CHECK-LABEL: test31
+; CHECK: valignd $3, %zmm0, %zmm1
+; CHECK: ret
+define <16 x i32> @test31(<16 x i32> %a, <16 x i32> %b) nounwind {
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 3, i32 4, i32 5, i32 undef, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+ ret <16 x i32> %c
+}
+
+; CHECK-LABEL: test32
+; CHECK: vshufpd $99, %zmm0, %zmm1
+; CHECK: ret
+define <8 x double> @test32(<8 x double> %a, <8 x double> %b) nounwind {
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 1, i32 10, i32 2, i32 undef, i32 5, i32 15, i32 undef>
+ ret <8 x double> %c
+}
+
+define <16 x i32> @test_align_v16i32_rr(<16 x i32> %a, <16 x i32> %b) nounwind {
+; CHECK-LABEL: test_align_v16i32_rr:
+; CHECK: ## BB#0:
+; CHECK-NEXT: valignd $3, %zmm0, %zmm1, %zmm0
+; CHECK-NEXT: retq
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 3, i32 4, i32 5, i32 undef, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+ ret <16 x i32> %c
+}
+
+define <16 x i32> @test_align_v16i32_rm(<16 x i32>* %a.ptr, <16 x i32> %b) nounwind {
+; CHECK-LABEL: test_align_v16i32_rm:
+; CHECK: ## BB#0:
+; CHECK-NEXT: valignd $3, (%rdi), %zmm0, %zmm0
+; CHECK-NEXT: retq
+ %a = load <16 x i32>, <16 x i32>* %a.ptr
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 3, i32 4, i32 5, i32 undef, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+ ret <16 x i32> %c
+}
+
+define <16 x i32> @test_align_v16i32_rm_mask(<16 x i32>* %a.ptr, <16 x i32> %b, <16 x i1> %mask) nounwind {
+; CHECK-LABEL: test_align_v16i32_rm_mask:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpmovsxbd %xmm1, %zmm1
+; CHECK-NEXT: vpandd {{.*}}(%rip){1to16}, %zmm1, %zmm1
+; CHECK-NEXT: vptestmd %zmm1, %zmm1, %k1
+; CHECK-NEXT: vmovdqa32 (%rdi), %zmm1
+; CHECK-NEXT: valignd $3, %zmm1, %zmm0, %zmm1 {%k1}
+; CHECK-NEXT: vmovaps %zmm1, %zmm0
+; CHECK-NEXT: retq
+;
+; CHECK-SKX-LABEL: test_align_v16i32_rm_mask:
+; CHECK-SKX: ## BB#0:
+; CHECK-SKX-NEXT: vpmovb2m %xmm1, %k1
+; CHECK-SKX-NEXT: vmovdqa32 (%rdi), %zmm1
+; CHECK-SKX-NEXT: valignd $3, %zmm1, %zmm0, %zmm1 {%k1}
+; CHECK-SKX-NEXT: vmovaps %zmm1, %zmm0
+; CHECK-SKX-NEXT: retq
+ %a = load <16 x i32>, <16 x i32>* %a.ptr
+ %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 3, i32 4, i32 5, i32 undef, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+ %res = select <16 x i1> %mask,<16 x i32> %c, <16 x i32> %a
+ ret <16 x i32> %res
+}
+
+define <8 x double> @test_align_v8f64_rr(<8 x double> %a, <8 x double> %b) nounwind {
+; CHECK-LABEL: test_align_v8f64_rr:
+; CHECK: ## BB#0:
+; CHECK-NEXT: valignq $3, %zmm0, %zmm1, %zmm0
+; CHECK-NEXT: retq
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
+ ret <8 x double> %c
+}
+
+define <8 x double> @test_align_v18f64_rm(<8 x double>* %a.ptr, <8 x double> %b) nounwind {
+; CHECK-LABEL: test_align_v18f64_rm:
+; CHECK: ## BB#0:
+; CHECK-NEXT: valignq $3, (%rdi), %zmm0, %zmm0
+; CHECK-NEXT: retq
+ %a = load <8 x double>, <8 x double>* %a.ptr
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
+ ret <8 x double> %c
+}
+
+define <8 x double> @test_align_v18f64_rm_mask(<8 x double>* %a.ptr, <8 x double> %b, <8 x i1> %mask) nounwind {
+; CHECK-LABEL: test_align_v18f64_rm_mask:
+; CHECK: ## BB#0:
+; CHECK-NEXT: vpmovsxwq %xmm1, %zmm1
+; CHECK-NEXT: vpandq {{.*}}(%rip){1to8}, %zmm1, %zmm1
+; CHECK-NEXT: vptestmq %zmm1, %zmm1, %k1
+; CHECK-NEXT: valignq $3, (%rdi), %zmm0, %zmm0 {%k1} {z}
+; CHECK-NEXT: retq
+;
+; CHECK-SKX-LABEL: test_align_v18f64_rm_mask:
+; CHECK-SKX: ## BB#0:
+; CHECK-SKX-NEXT: vpmovw2m %xmm1, %k1
+; CHECK-SKX-NEXT: valignq $3, (%rdi), %zmm0, %zmm0 {%k1} {z}
+; CHECK-SKX-NEXT: retq
+ %a = load <8 x double>, <8 x double>* %a.ptr
+ %c = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
+ %res = select <8 x i1> %mask,<8 x double> %c, <8 x double> zeroinitializer
+ ret <8 x double> %res
+}
+
diff --git a/test/CodeGen/X86/avx512-vec-cmp.ll b/test/CodeGen/X86/avx512-vec-cmp.ll
index e1f6276c6ef4..04028a1da510 100644
--- a/test/CodeGen/X86/avx512-vec-cmp.ll
+++ b/test/CodeGen/X86/avx512-vec-cmp.ll
@@ -116,11 +116,8 @@ define <2 x double> @test8(<2 x double> %a, <2 x double> %b) {
define <8 x i32> @test9(<8 x i32> %x, <8 x i32> %y) nounwind {
; KNL-LABEL: test9:
; KNL: ## BB#0:
-; KNL-NEXT: ## kill: YMM1<def> YMM1<kill> ZMM1<def>
-; KNL-NEXT: ## kill: YMM0<def> YMM0<kill> ZMM0<def>
; KNL-NEXT: vpcmpeqd %zmm1, %zmm0, %k1
; KNL-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
-; KNL-NEXT: ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
; KNL-NEXT: retq
%mask = icmp eq <8 x i32> %x, %y
%max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
@@ -130,11 +127,8 @@ define <8 x i32> @test9(<8 x i32> %x, <8 x i32> %y) nounwind {
define <8 x float> @test10(<8 x float> %x, <8 x float> %y) nounwind {
; KNL-LABEL: test10:
; KNL: ## BB#0:
-; KNL-NEXT: ## kill: YMM1<def> YMM1<kill> ZMM1<def>
-; KNL-NEXT: ## kill: YMM0<def> YMM0<kill> ZMM0<def>
; KNL-NEXT: vcmpeqps %zmm1, %zmm0, %k1
; KNL-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1}
-; KNL-NEXT: ## kill: YMM0<def> YMM0<kill> ZMM0<kill>
; KNL-NEXT: retq
; SKX-LABEL: test10:
; SKX: ## BB#0:
@@ -166,7 +160,6 @@ define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind {
; KNL-NEXT: vpcmpeqq %zmm3, %zmm1, %k1
; KNL-NEXT: kunpckbw %k0, %k1, %k0
; KNL-NEXT: kmovw %k0, %eax
-; KNL-NEXT: ## kill: AX<def> AX<kill> EAX<kill>
; KNL-NEXT: retq
%res = icmp eq <16 x i64> %a, %b
%res1 = bitcast <16 x i1> %res to i16
diff --git a/test/CodeGen/X86/avx512vl-intrinsics.ll b/test/CodeGen/X86/avx512vl-intrinsics.ll
index fd76ed5d0dbd..9d96c272f355 100644
--- a/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -668,7 +668,7 @@ declare <4 x float> @llvm.x86.avx512.mask.compress.ps.128(<4 x float> %data, <4
; CHECK-LABEL: compr7
; CHECK-NOT: vcompress
-; CHECK: vmovapd
+; CHECK: vmovupd
define void @compr7(i8* %addr, <8 x double> %data) {
call void @llvm.x86.avx512.mask.compress.store.pd.512(i8* %addr, <8 x double> %data, i8 -1)
ret void
@@ -757,7 +757,7 @@ declare <4 x float> @llvm.x86.avx512.mask.expand.ps.128(<4 x float> %data, <4 x
; CHECK-LABEL: expand7
; CHECK-NOT: vexpand
-; CHECK: vmovapd
+; CHECK: vmovupd
define <8 x double> @expand7(i8* %addr, <8 x double> %data) {
%res = call <8 x double> @llvm.x86.avx512.mask.expand.load.pd.512(i8* %addr, <8 x double> %data, i8 -1)
ret <8 x double> %res
@@ -2552,4 +2552,38 @@ define <4 x float> @test_mm512_min_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %
%res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
ret <4 x float> %res
}
-declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8) \ No newline at end of file
+declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)
+
+define <4 x double> @test_sqrt_pd_256(<4 x double> %a0, i8 %mask) {
+ ; CHECK-LABEL: test_sqrt_pd_256
+ ; CHECK: vsqrtpd
+ %res = call <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 %mask)
+ ret <4 x double> %res
+}
+declare <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
+
+define <8 x float> @test_sqrt_ps_256(<8 x float> %a0, i8 %mask) {
+ ; CHECK-LABEL: test_sqrt_ps_256
+ ; CHECK: vsqrtps
+ %res = call <8 x float> @llvm.x86.avx512.mask.sqrt.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
+ ret <8 x float> %res
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.sqrt.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
+
+define <4 x double> @test_getexp_pd_256(<4 x double> %a0) {
+ ; CHECK-LABEL: test_getexp_pd_256
+ ; CHECK: vgetexppd
+ %res = call <4 x double> @llvm.x86.avx512.mask.getexp.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 -1)
+ ret <4 x double> %res
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.getexp.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
+
+define <8 x float> @test_getexp_ps_256(<8 x float> %a0) {
+ ; CHECK-LABEL: test_getexp_ps_256
+ ; CHECK: vgetexpps
+ %res = call <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 -1)
+ ret <8 x float> %res
+}
+declare <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone \ No newline at end of file
diff --git a/test/CodeGen/X86/buildvec-insertvec.ll b/test/CodeGen/X86/buildvec-insertvec.ll
index 3fb69a48b3c7..73dbe1f650a1 100644
--- a/test/CodeGen/X86/buildvec-insertvec.ll
+++ b/test/CodeGen/X86/buildvec-insertvec.ll
@@ -1,15 +1,56 @@
-; RUN: llc < %s -mcpu=corei7 -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s
define void @foo(<3 x float> %in, <4 x i8>* nocapture %out) nounwind {
+; CHECK-LABEL: foo:
+; CHECK: # BB#0:
+; CHECK-NEXT: cvttps2dq %xmm0, %xmm0
+; CHECK-NEXT: movl $255, %eax
+; CHECK-NEXT: pinsrd $3, %eax, %xmm0
+; CHECK-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT: movd %xmm0, (%rdi)
+; CHECK-NEXT: retq
%t0 = fptoui <3 x float> %in to <3 x i8>
%t1 = shufflevector <3 x i8> %t0, <3 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 undef>
%t2 = insertelement <4 x i8> %t1, i8 -1, i32 3
store <4 x i8> %t2, <4 x i8>* %out, align 4
ret void
-; CHECK: foo
-; CHECK: cvttps2dq
-; CHECK-NOT: pextrd
-; CHECK: pinsrd
-; CHECK-NEXT: pshufb
-; CHECK: ret
+}
+
+; Verify that the DAGCombiner doesn't wrongly fold a build_vector into a
+; blend with a zero vector if the build_vector contains negative zero.
+;
+; TODO: the codegen for function 'test_negative_zero_1' is sub-optimal.
+; Ideally, we should generate a single shuffle blend operation.
+
+define <4 x float> @test_negative_zero_1(<4 x float> %A) {
+; CHECK-LABEL: test_negative_zero_1:
+; CHECK: # BB#0: # %entry
+; CHECK-NEXT: movapd %xmm0, %xmm1
+; CHECK-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1,0]
+; CHECK-NEXT: xorps %xmm2, %xmm2
+; CHECK-NEXT: blendps {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; CHECK-NEXT: retq
+entry:
+ %0 = extractelement <4 x float> %A, i32 0
+ %1 = insertelement <4 x float> undef, float %0, i32 0
+ %2 = insertelement <4 x float> %1, float -0.0, i32 1
+ %3 = extractelement <4 x float> %A, i32 2
+ %4 = insertelement <4 x float> %2, float %3, i32 2
+ %5 = insertelement <4 x float> %4, float 0.0, i32 3
+ ret <4 x float> %5
+}
+
+define <2 x double> @test_negative_zero_2(<2 x double> %A) {
+; CHECK-LABEL: test_negative_zero_2:
+; CHECK: # BB#0: # %entry
+; CHECK-NEXT: movhpd {{.*}}(%rip), %xmm0
+; CHECK-NEXT: retq
+entry:
+ %0 = extractelement <2 x double> %A, i32 0
+ %1 = insertelement <2 x double> undef, double %0, i32 0
+ %2 = insertelement <2 x double> %1, double -0.0, i32 1
+ ret <2 x double> %2
}
diff --git a/test/CodeGen/X86/critical-anti-dep-breaker.ll b/test/CodeGen/X86/critical-anti-dep-breaker.ll
index 86afc1f245ad..de5744d3a88f 100644
--- a/test/CodeGen/X86/critical-anti-dep-breaker.ll
+++ b/test/CodeGen/X86/critical-anti-dep-breaker.ll
@@ -11,8 +11,7 @@
@NullToken = external global i64
; CHECK-LABEL: Part_Create:
-; CHECK-DAG: # kill: RDI<def>
-; CHECK-DAG: movq PartClass@GOTPCREL(%rip), %r10
+; CHECK: movq PartClass@GOTPCREL(%rip), %r10
define i32 @Part_Create(i64* %Anchor, i32 %TypeNum, i32 %F, i32 %Z, i32* %Status, i64* %PartTkn) {
%PartObj = alloca i64*, align 8
%Vchunk = alloca i64, align 8
diff --git a/test/CodeGen/X86/machine-cp.ll b/test/CodeGen/X86/machine-cp.ll
index 0006b6ea7133..aaed0f0a23dc 100644
--- a/test/CodeGen/X86/machine-cp.ll
+++ b/test/CodeGen/X86/machine-cp.ll
@@ -58,3 +58,58 @@ while.end: ; preds = %while.body, %entry
%t = trunc i64 %a.addr.0.lcssa to i32
ret i32 %t
}
+
+; Check that copy propagation does not kill thing like:
+; dst = copy src <-- do not kill that.
+; ... = op1 dst<undef>
+; ... = op2 dst <-- this is used here.
+;
+; CHECK-LABEL: foo:
+; CHECK: psllw $7,
+; CHECK: psllw $7,
+; CHECK-NEXT: pand
+; CHECK-NEXT: pcmpgtb
+; CHECK-NEXT: pand %xmm{{[0-9]+}}, [[SRC:%xmm[0-9]+]]
+; Machine propagation used to delete the first copy as the
+; first few uses were <undef>.
+; CHECK-NEXT: movdqa [[SRC]], [[CPY1:%xmm[0-9]+]]
+; CHECK-NEXT: movdqa [[SRC]], [[CPY2:%xmm[0-9]+]]
+; CHECK-NEXT: punpckhbw [[SRC]],
+; Check that CPY1 is not redefined.
+; CHECK-NOT: , [[CPY1]]
+; undef use, we do not care.
+; CHECK: punpcklwd [[CPY1]],
+; Check that CPY1 is not redefined.
+; CHECK-NOT: , [[CPY1]]
+; CHECK: punpcklbw [[CPY2]], [[CPY2]]
+; CHECK-NEXT: punpckhwd [[CPY2]], [[CPY2]]
+; CHECK-NEXT pslld $31, [[CPY2]]
+; Check that CPY1 is not redefined.
+; CHECK-NOT: , [[CPY1]]
+; CHECK: punpcklbw [[CPY1]], [[CPY1]]
+; CHECK-NEXT: punpcklwd [[CPY1]], [[CPY1]]
+; CHECK-NEXT pslld $31, [[CPY1]]
+define <16 x float> @foo(<16 x float> %x) {
+bb:
+ %v3 = icmp slt <16 x i32> undef, zeroinitializer
+ %v14 = zext <16 x i1> %v3 to <16 x i32>
+ %v16 = fcmp olt <16 x float> %x, zeroinitializer
+ %v17 = sext <16 x i1> %v16 to <16 x i32>
+ %v18 = zext <16 x i1> %v16 to <16 x i32>
+ %v19 = xor <16 x i32> %v14, %v18
+ %v20 = or <16 x i32> %v17, undef
+ %v21 = fptosi <16 x float> %x to <16 x i32>
+ %v22 = sitofp <16 x i32> %v21 to <16 x float>
+ %v69 = fcmp ogt <16 x float> %v22, zeroinitializer
+ %v75 = and <16 x i1> %v69, %v3
+ %v77 = bitcast <16 x float> %v22 to <16 x i32>
+ %v79 = sext <16 x i1> %v75 to <16 x i32>
+ %v80 = and <16 x i32> undef, %v79
+ %v81 = xor <16 x i32> %v77, %v80
+ %v82 = and <16 x i32> undef, %v81
+ %v83 = xor <16 x i32> %v19, %v82
+ %v84 = and <16 x i32> %v83, %v20
+ %v85 = xor <16 x i32> %v19, %v84
+ %v86 = bitcast <16 x i32> %v85 to <16 x float>
+ ret <16 x float> %v86
+}
diff --git a/test/CodeGen/X86/pic.ll b/test/CodeGen/X86/pic.ll
index d543deb804d1..73be234db81c 100644
--- a/test/CodeGen/X86/pic.ll
+++ b/test/CodeGen/X86/pic.ll
@@ -196,9 +196,11 @@ bb12:
; LINUX-NEXT: .LJTI7_0:
; LINUX: .long .LBB7_2@GOTOFF
; LINUX: .long .LBB7_8@GOTOFF
-; LINUX: .long .LBB7_14@GOTOFF
-; LINUX: .long .LBB7_9@GOTOFF
-; LINUX: .long .LBB7_10@GOTOFF
+; LINUX: .long .LBB7_4@GOTOFF
+; LINUX: .long .LBB7_6@GOTOFF
+; LINUX: .long .LBB7_5@GOTOFF
+; LINUX: .long .LBB7_8@GOTOFF
+; LINUX: .long .LBB7_7@GOTOFF
}
declare void @foo1(...)
diff --git a/test/CodeGen/X86/pr23603.ll b/test/CodeGen/X86/pr23603.ll
new file mode 100644
index 000000000000..6f856aedb8d5
--- /dev/null
+++ b/test/CodeGen/X86/pr23603.ll
@@ -0,0 +1,24 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
+
+declare void @free_v()
+
+define void @f(i32* %x, i32 %c32, i32* %y) {
+; CHECK-LABEL: f
+ entry:
+ %v = load i32, i32* %x, !invariant.load !0
+; CHECK: movl (%rdi), %ebx
+; CHECK: free_v
+; CHECK-NOT: movl (%rdi), %ebx
+ call void @free_v()
+ %c = icmp ne i32 %c32, 0
+ br i1 %c, label %left, label %merge
+
+ left:
+ store i32 %v, i32* %y
+ br label %merge
+
+ merge:
+ ret void
+}
+
+!0 = !{}
diff --git a/test/CodeGen/X86/pr23664.ll b/test/CodeGen/X86/pr23664.ll
new file mode 100644
index 000000000000..a501c0db837e
--- /dev/null
+++ b/test/CodeGen/X86/pr23664.ll
@@ -0,0 +1,14 @@
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+
+define i2 @f(i32 %arg) {
+ %trunc = trunc i32 %arg to i1
+ %sext = sext i1 %trunc to i2
+ %or = or i2 %sext, 1
+ ret i2 %or
+
+; CHECK-LABEL: f:
+; CHECK: addb %dil, %dil
+; CHECK-NEXT: orb $1, %dil
+; CHECK-NEXT: movb %dil, %al
+; CHECK-NEXT: retq
+}
diff --git a/test/CodeGen/X86/recip-fastmath.ll b/test/CodeGen/X86/recip-fastmath.ll
index fcd077092dab..7f1521a83bcf 100644
--- a/test/CodeGen/X86/recip-fastmath.ll
+++ b/test/CodeGen/X86/recip-fastmath.ll
@@ -1,6 +1,6 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=sse2 | FileCheck %s
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx,use-recip-est | FileCheck %s --check-prefix=RECIP
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx,use-recip-est -x86-recip-refinement-steps=2 | FileCheck %s --check-prefix=REFINE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx -recip=divf,vec-divf | FileCheck %s --check-prefix=RECIP
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx -recip=divf:2,vec-divf:2 | FileCheck %s --check-prefix=REFINE
; If the target's divss/divps instructions are substantially
; slower than rcpss/rcpps with a Newton-Raphson refinement,
diff --git a/test/CodeGen/X86/sibcall-win64.ll b/test/CodeGen/X86/sibcall-win64.ll
index f7038726f9ca..204e1f8b050b 100644
--- a/test/CodeGen/X86/sibcall-win64.ll
+++ b/test/CodeGen/X86/sibcall-win64.ll
@@ -1,7 +1,11 @@
; RUN: llc < %s -mtriple=x86_64-pc-linux | FileCheck %s
declare x86_64_win64cc void @win64_callee(i32)
+declare x86_64_win64cc void (i32)* @win64_indirect()
+declare x86_64_win64cc void @win64_other(i32)
declare void @sysv_callee(i32)
+declare void (i32)* @sysv_indirect()
+declare void @sysv_other(i32)
define void @sysv_caller(i32 %p1) {
entry:
@@ -40,3 +44,23 @@ define x86_64_win64cc void @win64_matched(i32 %p1) {
; CHECK-LABEL: win64_matched:
; CHECK: jmp win64_callee # TAILCALL
+
+define x86_64_win64cc void @win64_indirect_caller(i32 %p1) {
+ %1 = call x86_64_win64cc void (i32)* @win64_indirect()
+ call x86_64_win64cc void @win64_other(i32 0)
+ tail call x86_64_win64cc void %1(i32 %p1)
+ ret void
+}
+
+; CHECK-LABEL: win64_indirect_caller:
+; CHECK: jmpq *%{{rax|rcx|rdx|r8|r9|r11}} # TAILCALL
+
+define void @sysv_indirect_caller(i32 %p1) {
+ %1 = call void (i32)* @sysv_indirect()
+ call void @sysv_other(i32 0)
+ tail call void %1(i32 %p1)
+ ret void
+}
+
+; CHECK-LABEL: sysv_indirect_caller:
+; CHECK: jmpq *%{{rax|rcx|rdx|rsi|rdi|r8|r9|r11}} # TAILCALL
diff --git a/test/CodeGen/X86/sqrt-fastmath.ll b/test/CodeGen/X86/sqrt-fastmath.ll
index 4c6b521156e0..373fa53c970f 100644
--- a/test/CodeGen/X86/sqrt-fastmath.ll
+++ b/test/CodeGen/X86/sqrt-fastmath.ll
@@ -1,5 +1,5 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=sse2 | FileCheck %s
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx,use-sqrt-est | FileCheck %s --check-prefix=ESTIMATE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx -recip=sqrtf,vec-sqrtf | FileCheck %s --check-prefix=ESTIMATE
declare double @__sqrt_finite(double) #0
declare float @__sqrtf_finite(float) #0
diff --git a/test/CodeGen/X86/stack-folding-x86_64.ll b/test/CodeGen/X86/stack-folding-x86_64.ll
new file mode 100644
index 000000000000..211227916a09
--- /dev/null
+++ b/test/CodeGen/X86/stack-folding-x86_64.ll
@@ -0,0 +1,51 @@
+; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mcpu=x86-64 < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-unknown"
+
+; Stack reload folding tests.
+;
+; By including a nop call with sideeffects we can force a partial register spill of the
+; relevant registers and check that the reload is correctly folded into the instruction.
+
+;TODO stack_fold_bsf_i16
+declare i16 @llvm.cttz.i16(i16, i1)
+
+define i32 @stack_fold_bsf_i32(i32 %a0) {
+ ;CHECK-LABEL: stack_fold_bsf_i32
+ ;CHECK: bsfl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
+ %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+ %2 = call i32 @llvm.cttz.i32(i32 %a0, i1 -1)
+ ret i32 %2
+}
+declare i32 @llvm.cttz.i32(i32, i1)
+
+define i64 @stack_fold_bsf_i64(i64 %a0) {
+ ;CHECK-LABEL: stack_fold_bsf_i64
+ ;CHECK: bsfq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
+ %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+ %2 = call i64 @llvm.cttz.i64(i64 %a0, i1 -1)
+ ret i64 %2
+}
+declare i64 @llvm.cttz.i64(i64, i1)
+
+;TODO stack_fold_bsr_i16
+declare i16 @llvm.ctlz.i16(i16, i1)
+
+define i32 @stack_fold_bsr_i32(i32 %a0) {
+ ;CHECK-LABEL: stack_fold_bsr_i32
+ ;CHECK: bsrl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
+ %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+ %2 = call i32 @llvm.ctlz.i32(i32 %a0, i1 -1)
+ ret i32 %2
+}
+declare i32 @llvm.ctlz.i32(i32, i1)
+
+define i64 @stack_fold_bsr_i64(i64 %a0) {
+ ;CHECK-LABEL: stack_fold_bsr_i64
+ ;CHECK: bsrq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
+ %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+ %2 = call i64 @llvm.ctlz.i64(i64 %a0, i1 -1)
+ ret i64 %2
+}
+declare i64 @llvm.ctlz.i64(i64, i1)
diff --git a/test/CodeGen/X86/statepoint-far-call.ll b/test/CodeGen/X86/statepoint-far-call.ll
new file mode 100644
index 000000000000..cd8dd0f35a20
--- /dev/null
+++ b/test/CodeGen/X86/statepoint-far-call.ll
@@ -0,0 +1,22 @@
+; RUN: llc < %s | FileCheck %s
+; Test to check that Statepoints with X64 far-immediate targets
+; are lowered correctly to an indirect call via a scratch register.
+
+target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-pc-win64"
+
+define void @test_far_call() gc "statepoint-example" {
+; CHECK-LABEL: test_far_call
+; CHECK: pushq %rax
+; CHECK: movabsq $140727162896504, %rax
+; CHECK: callq *%rax
+; CHECK: popq %rax
+; CHECK: retq
+
+entry:
+ %safepoint_token = call i32 (i64, i32, void ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_isVoidf(i64 0, i32 0, void ()* inttoptr (i64 140727162896504 to void ()*), i32 0, i32 0, i32 0, i32 0)
+ ret void
+}
+
+declare i32 @llvm.experimental.gc.statepoint.p0f_isVoidf(i64, i32, void ()*, i32, i32, ...)
+
diff --git a/test/CodeGen/X86/switch-or.ll b/test/CodeGen/X86/switch-or.ll
index 6e6b013d9fa8..4642accfff8d 100644
--- a/test/CodeGen/X86/switch-or.ll
+++ b/test/CodeGen/X86/switch-or.ll
@@ -1,10 +1,11 @@
; RUN: llc -march=x86 -asm-verbose=false < %s | FileCheck %s
; Check that merging switch cases that differ in one bit works.
+; CHECK-LABEL: test1
; CHECK: orl $2
; CHECK-NEXT: cmpl $6
-define void @foo(i32 %variable) nounwind {
+define void @test1(i32 %variable) nounwind {
entry:
switch i32 %variable, label %if.end [
i32 4, label %if.then
@@ -19,4 +20,22 @@ if.end:
ret void
}
+; CHECK-LABEL: test2
+; CHECK: orl $-2147483648
+; CHECK-NEXT: cmpl $-2147483648
+define void @test2(i32 %variable) nounwind {
+entry:
+ switch i32 %variable, label %if.end [
+ i32 0, label %if.then
+ i32 -2147483648, label %if.then
+ ]
+
+if.then:
+ %call = tail call i32 (...) @bar() nounwind
+ ret void
+
+if.end:
+ ret void
+}
+
declare i32 @bar(...) nounwind
diff --git a/test/CodeGen/X86/switch.ll b/test/CodeGen/X86/switch.ll
index 66a739c8470c..a4dece65479c 100644
--- a/test/CodeGen/X86/switch.ll
+++ b/test/CodeGen/X86/switch.ll
@@ -534,3 +534,18 @@ return: ret void
; CHECK-NOT: cmpl
; CHECK: cmpl $99
}
+
+
+define void @pr23738(i4 %x) {
+entry:
+ switch i4 %x, label %bb0 [
+ i4 0, label %bb1
+ i4 1, label %bb1
+ i4 -5, label %bb1
+ ]
+bb0: tail call void @g(i32 0) br label %return
+bb1: tail call void @g(i32 1) br label %return
+return: ret void
+; Don't assert due to truncating the bitwidth (64) to i4 when checking
+; that the bit-test range fits in a word.
+}
diff --git a/test/CodeGen/X86/tail-call-got.ll b/test/CodeGen/X86/tail-call-got.ll
index 84d561dcd8c3..20d1a87b626a 100644
--- a/test/CodeGen/X86/tail-call-got.ll
+++ b/test/CodeGen/X86/tail-call-got.ll
@@ -1,12 +1,14 @@
; RUN: llc < %s -relocation-model=pic -mattr=+sse2 | FileCheck %s
+; We used to do tail calls through the GOT for these symbols, but it was
+; disabled due to PR15086.
+
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32-n8:16:32"
target triple = "i386-unknown-freebsd9.0"
define double @test1(double %x) nounwind readnone {
; CHECK-LABEL: test1:
-; CHECK: movl foo@GOT
-; CHECK-NEXT: jmpl
+; CHECK: calll foo@PLT
%1 = tail call double @foo(double %x) nounwind readnone
ret double %1
}
@@ -15,10 +17,18 @@ declare double @foo(double) readnone
define double @test2(double %x) nounwind readnone {
; CHECK-LABEL: test2:
-; CHECK: movl sin@GOT
-; CHECK-NEXT: jmpl
+; CHECK: calll sin@PLT
%1 = tail call double @sin(double %x) nounwind readnone
ret double %1
}
declare double @sin(double) readnone
+
+define double @test3(double %x) nounwind readnone {
+; CHECK-LABEL: test3:
+; CHECK: calll sin2@PLT
+ %1 = tail call double @sin2(double %x) nounwind readnone
+ ret double %1
+}
+
+declare double @sin2(double) readnone
diff --git a/test/CodeGen/X86/tailcallpic1.ll b/test/CodeGen/X86/tailcallpic1.ll
index ff590a1fd3e9..ed101fcccd2d 100644
--- a/test/CodeGen/X86/tailcallpic1.ll
+++ b/test/CodeGen/X86/tailcallpic1.ll
@@ -1,5 +1,8 @@
; RUN: llc < %s -tailcallopt -mtriple=i686-pc-linux-gnu -relocation-model=pic | FileCheck %s
+; This test uses guaranteed TCO so these will be tail calls, despite the early
+; binding issues.
+
define protected fastcc i32 @tailcallee(i32 %a1, i32 %a2, i32 %a3, i32 %a4) {
entry:
ret i32 %a3
diff --git a/test/CodeGen/X86/tailcallpic3.ll b/test/CodeGen/X86/tailcallpic3.ll
new file mode 100644
index 000000000000..edc58052d82f
--- /dev/null
+++ b/test/CodeGen/X86/tailcallpic3.ll
@@ -0,0 +1,73 @@
+; RUN: llc < %s -mtriple=i686-pc-linux-gnu -relocation-model=pic | FileCheck %s
+
+; While many of these could be tail called, we don't do it because it forces
+; early binding.
+
+declare void @external()
+
+define hidden void @tailcallee_hidden() {
+entry:
+ ret void
+}
+
+define void @tailcall_hidden() {
+entry:
+ tail call void @tailcallee_hidden()
+ ret void
+}
+; CHECK: tailcall_hidden:
+; CHECK: jmp tailcallee_hidden
+
+define internal void @tailcallee_internal() {
+entry:
+ ret void
+}
+
+define void @tailcall_internal() {
+entry:
+ tail call void @tailcallee_internal()
+ ret void
+}
+; CHECK: tailcall_internal:
+; CHECK: jmp tailcallee_internal
+
+define default void @tailcallee_default() {
+entry:
+ ret void
+}
+
+define void @tailcall_default() {
+entry:
+ tail call void @tailcallee_default()
+ ret void
+}
+; CHECK: tailcall_default:
+; CHECK: calll tailcallee_default@PLT
+
+define void @tailcallee_default_implicit() {
+entry:
+ ret void
+}
+
+define void @tailcall_default_implicit() {
+entry:
+ tail call void @tailcallee_default_implicit()
+ ret void
+}
+; CHECK: tailcall_default_implicit:
+; CHECK: calll tailcallee_default_implicit@PLT
+
+define void @tailcall_external() {
+ tail call void @external()
+ ret void
+}
+; CHECK: tailcall_external:
+; CHECK: calll external@PLT
+
+define void @musttail_external() {
+ musttail call void @external()
+ ret void
+}
+; CHECK: musttail_external:
+; CHECK: movl external@GOT
+; CHECK: jmpl
diff --git a/test/CodeGen/X86/vec_fp_to_int.ll b/test/CodeGen/X86/vec_fp_to_int.ll
index 9f1c7afa295b..3e72212d85d3 100644
--- a/test/CodeGen/X86/vec_fp_to_int.ll
+++ b/test/CodeGen/X86/vec_fp_to_int.ll
@@ -239,7 +239,6 @@ define <4 x i64> @fptoui_4vf64(<4 x double> %a) {
; SSE2: # BB#0:
; SSE2-NEXT: movapd %xmm0, %xmm2
; SSE2-NEXT: movsd {{.*#+}} xmm3 = mem[0],zero
-; SSE2-NEXT: {{.*#+}} kill: XMM0<def> XMM2<kill>
; SSE2-NEXT: subsd %xmm3, %xmm0
; SSE2-NEXT: cvttsd2si %xmm0, %rcx
; SSE2-NEXT: movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
@@ -589,7 +588,6 @@ define <8 x i32> @fptoui_8vf32(<8 x float> %a) {
; SSE2-LABEL: fptoui_8vf32:
; SSE2: # BB#0:
; SSE2-NEXT: movaps %xmm0, %xmm2
-; SSE2-NEXT: {{.*#+}} kill: XMM0<def> XMM2<kill>
; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
; SSE2-NEXT: cvttss2si %xmm0, %rax
; SSE2-NEXT: movd %eax, %xmm0
diff --git a/test/CodeGen/X86/vec_shift8.ll b/test/CodeGen/X86/vec_shift8.ll
new file mode 100644
index 000000000000..a32cb30b0b26
--- /dev/null
+++ b/test/CodeGen/X86/vec_shift8.ll
@@ -0,0 +1,1016 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
+
+;
+; Vectorized integer shifts
+;
+
+define <2 x i64> @shl_8i16(<8 x i16> %r, <8 x i16> %a) nounwind readnone ssp {
+entry:
+; SSE2: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: pextrw $7, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: pextrw $3, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %eax
+; SSE2-NEXT: pextrw $5, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: pextrw $1, %xmm0, %eax
+; SSE2-NEXT: pextrw $1, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %eax
+; SSE2-NEXT: pextrw $6, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: pextrw $2, %xmm0, %eax
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: pextrw $4, %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: movd %xmm1, %ecx
+; SSE2-NEXT: shll %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: retq
+;
+; SSE41: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: pextrw $1, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: movd %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %edx
+; SSE41-NEXT: movd %edx, %xmm2
+; SSE41-NEXT: pinsrw $1, %eax, %xmm2
+; SSE41-NEXT: pextrw $2, %xmm0, %eax
+; SSE41-NEXT: pextrw $2, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $2, %eax, %xmm2
+; SSE41-NEXT: pextrw $3, %xmm0, %eax
+; SSE41-NEXT: pextrw $3, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $3, %eax, %xmm2
+; SSE41-NEXT: pextrw $4, %xmm0, %eax
+; SSE41-NEXT: pextrw $4, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $4, %eax, %xmm2
+; SSE41-NEXT: pextrw $5, %xmm0, %eax
+; SSE41-NEXT: pextrw $5, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $5, %eax, %xmm2
+; SSE41-NEXT: pextrw $6, %xmm0, %eax
+; SSE41-NEXT: pextrw $6, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $6, %eax, %xmm2
+; SSE41-NEXT: pextrw $7, %xmm0, %eax
+; SSE41-NEXT: pextrw $7, %xmm1, %ecx
+; SSE41-NEXT: shll %cl, %eax
+; SSE41-NEXT: pinsrw $7, %eax, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: vpextrw $1, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: vmovd %xmm1, %ecx
+; AVX-NEXT: shll %cl, %edx
+; AVX-NEXT: vmovd %edx, %xmm2
+; AVX-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $2, %xmm0, %eax
+; AVX-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $3, %xmm0, %eax
+; AVX-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $4, %xmm0, %eax
+; AVX-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $5, %xmm0, %eax
+; AVX-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $6, %xmm0, %eax
+; AVX-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $7, %xmm0, %eax
+; AVX-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX-NEXT: shll %cl, %eax
+; AVX-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX-NEXT: retq
+ %shl = shl <8 x i16> %r, %a
+ %tmp2 = bitcast <8 x i16> %shl to <2 x i64>
+ ret <2 x i64> %tmp2
+}
+
+define <2 x i64> @shl_16i8(<16 x i8> %r, <16 x i8> %a) nounwind readnone ssp {
+entry:
+; SSE2: psllw $5, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE2-NEXT: movdqa %xmm2, %xmm3
+; SSE2-NEXT: pand %xmm1, %xmm3
+; SSE2-NEXT: pcmpeqb %xmm2, %xmm3
+; SSE2-NEXT: movdqa %xmm3, %xmm4
+; SSE2-NEXT: pandn %xmm0, %xmm4
+; SSE2-NEXT: psllw $4, %xmm0
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: por %xmm4, %xmm0
+; SSE2-NEXT: paddb %xmm1, %xmm1
+; SSE2-NEXT: movdqa %xmm2, %xmm3
+; SSE2-NEXT: pand %xmm1, %xmm3
+; SSE2-NEXT: pcmpeqb %xmm2, %xmm3
+; SSE2-NEXT: movdqa %xmm3, %xmm4
+; SSE2-NEXT: pandn %xmm0, %xmm4
+; SSE2-NEXT: psllw $2, %xmm0
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: por %xmm4, %xmm0
+; SSE2-NEXT: paddb %xmm1, %xmm1
+; SSE2-NEXT: pand %xmm2, %xmm1
+; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm2
+; SSE2-NEXT: pandn %xmm0, %xmm2
+; SSE2-NEXT: paddb %xmm0, %xmm0
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: por %xmm2, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE41: movdqa %xmm0, %xmm2
+; SSE41-NEXT: psllw $5, %xmm1
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm5
+; SSE41-NEXT: paddb %xmm5, %xmm5
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE41-NEXT: movdqa %xmm3, %xmm4
+; SSE41-NEXT: pand %xmm5, %xmm4
+; SSE41-NEXT: pcmpeqb %xmm3, %xmm4
+; SSE41-NEXT: pand %xmm3, %xmm1
+; SSE41-NEXT: pcmpeqb %xmm3, %xmm1
+; SSE41-NEXT: movdqa %xmm2, %xmm6
+; SSE41-NEXT: psllw $4, %xmm6
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm6
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: pblendvb %xmm6, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm1
+; SSE41-NEXT: psllw $2, %xmm1
+; SSE41-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE41-NEXT: movdqa %xmm4, %xmm0
+; SSE41-NEXT: pblendvb %xmm1, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm1
+; SSE41-NEXT: paddb %xmm1, %xmm1
+; SSE41-NEXT: paddb %xmm5, %xmm5
+; SSE41-NEXT: pand %xmm3, %xmm5
+; SSE41-NEXT: pcmpeqb %xmm5, %xmm3
+; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: pblendvb %xmm1, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpsllw $5, %xmm1, %xmm1
+; AVX-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX-NEXT: vpaddb %xmm1, %xmm1, %xmm2
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX-NEXT: vpand %xmm2, %xmm3, %xmm4
+; AVX-NEXT: vpcmpeqb %xmm3, %xmm4, %xmm4
+; AVX-NEXT: vpand %xmm1, %xmm3, %xmm1
+; AVX-NEXT: vpcmpeqb %xmm3, %xmm1, %xmm1
+; AVX-NEXT: vpsllw $4, %xmm0, %xmm5
+; AVX-NEXT: vpand {{.*}}(%rip), %xmm5, %xmm5
+; AVX-NEXT: vpblendvb %xmm1, %xmm5, %xmm0, %xmm0
+; AVX-NEXT: vpsllw $2, %xmm0, %xmm1
+; AVX-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX-NEXT: vpblendvb %xmm4, %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpaddb %xmm0, %xmm0, %xmm1
+; AVX-NEXT: vpaddb %xmm2, %xmm2, %xmm2
+; AVX-NEXT: vpand %xmm2, %xmm3, %xmm2
+; AVX-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm2
+; AVX-NEXT: vpblendvb %xmm2, %xmm1, %xmm0, %xmm0
+; AVX-NEXT: retq
+ %shl = shl <16 x i8> %r, %a
+ %tmp2 = bitcast <16 x i8> %shl to <2 x i64>
+ ret <2 x i64> %tmp2
+}
+
+define <2 x i64> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind readnone ssp {
+entry:
+; SSE2: pextrw $7, %xmm1, %ecx
+; SSE2-NEXT: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: pextrw $3, %xmm1, %ecx
+; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $5, %xmm1, %ecx
+; SSE2-NEXT: pextrw $5, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: pextrw $1, %xmm1, %ecx
+; SSE2-NEXT: pextrw $1, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE2-NEXT: pextrw $6, %xmm1, %ecx
+; SSE2-NEXT: pextrw $6, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
+; SSE2-NEXT: pextrw $2, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE2-NEXT: pextrw $4, %xmm1, %ecx
+; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: movd %xmm1, %ecx
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: sarw %cl, %ax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: retq
+;
+; SSE41: pextrw $1, %xmm1, %ecx
+; SSE41-NEXT: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: movd %xmm1, %ecx
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: sarw %cl, %dx
+; SSE41-NEXT: movd %edx, %xmm2
+; SSE41-NEXT: pinsrw $1, %eax, %xmm2
+; SSE41-NEXT: pextrw $2, %xmm1, %ecx
+; SSE41-NEXT: pextrw $2, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $2, %eax, %xmm2
+; SSE41-NEXT: pextrw $3, %xmm1, %ecx
+; SSE41-NEXT: pextrw $3, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $3, %eax, %xmm2
+; SSE41-NEXT: pextrw $4, %xmm1, %ecx
+; SSE41-NEXT: pextrw $4, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $4, %eax, %xmm2
+; SSE41-NEXT: pextrw $5, %xmm1, %ecx
+; SSE41-NEXT: pextrw $5, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $5, %eax, %xmm2
+; SSE41-NEXT: pextrw $6, %xmm1, %ecx
+; SSE41-NEXT: pextrw $6, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $6, %eax, %xmm2
+; SSE41-NEXT: pextrw $7, %xmm1, %ecx
+; SSE41-NEXT: pextrw $7, %xmm0, %eax
+; SSE41-NEXT: sarw %cl, %ax
+; SSE41-NEXT: pinsrw $7, %eax, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpextrw $1, %xmm1, %ecx
+; AVX-NEXT: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vmovd %xmm1, %ecx
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: sarw %cl, %dx
+; AVX-NEXT: vmovd %edx, %xmm2
+; AVX-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX-NEXT: vpextrw $2, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX-NEXT: vpextrw $3, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX-NEXT: vpextrw $4, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX-NEXT: vpextrw $5, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX-NEXT: vpextrw $6, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX-NEXT: vpextrw $7, %xmm0, %eax
+; AVX-NEXT: sarw %cl, %ax
+; AVX-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX-NEXT: retq
+ %ashr = ashr <8 x i16> %r, %a
+ %tmp2 = bitcast <8 x i16> %ashr to <2 x i64>
+ ret <2 x i64> %tmp2
+}
+
+define <2 x i64> @ashr_16i8(<16 x i8> %r, <16 x i8> %a) nounwind readnone ssp {
+entry:
+;
+; SSE2: pushq %rbp
+; SSE2-NEXT: pushq %r15
+; SSE2-NEXT: pushq %r14
+; SSE2-NEXT: pushq %r13
+; SSE2-NEXT: pushq %r12
+; SSE2-NEXT: pushq %rbx
+; SSE2-NEXT: movaps %xmm1, -24(%rsp)
+; SSE2-NEXT: movaps %xmm0, -40(%rsp)
+; SSE2-NEXT: movb -9(%rsp), %cl
+; SSE2-NEXT: movb -25(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movb -17(%rsp), %cl
+; SSE2-NEXT: movb -33(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -13(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movl %eax, -44(%rsp)
+; SSE2-NEXT: movb -29(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movzbl %al, %r9d
+; SSE2-NEXT: movb -21(%rsp), %cl
+; SSE2-NEXT: movb -37(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -11(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r10d
+; SSE2-NEXT: movb -27(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -19(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r11d
+; SSE2-NEXT: movb -35(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -15(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r14d
+; SSE2-NEXT: movb -31(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movzbl %al, %r15d
+; SSE2-NEXT: movb -23(%rsp), %cl
+; SSE2-NEXT: movb -39(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -10(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r12d
+; SSE2-NEXT: movb -26(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -18(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r13d
+; SSE2-NEXT: movb -34(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -14(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r8d
+; SSE2-NEXT: movb -30(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -22(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %ebp
+; SSE2-NEXT: movb -38(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movb -12(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %edi
+; SSE2-NEXT: movb -28(%rsp), %dl
+; SSE2-NEXT: sarb %cl, %dl
+; SSE2-NEXT: movb -20(%rsp), %cl
+; SSE2-NEXT: movzbl %dl, %esi
+; SSE2-NEXT: movb -36(%rsp), %bl
+; SSE2-NEXT: sarb %cl, %bl
+; SSE2-NEXT: movb -16(%rsp), %cl
+; SSE2-NEXT: movzbl %bl, %ebx
+; SSE2-NEXT: movb -32(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movzbl %al, %edx
+; SSE2-NEXT: movb -24(%rsp), %cl
+; SSE2-NEXT: movb -40(%rsp), %al
+; SSE2-NEXT: sarb %cl, %al
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movd -44(%rsp), %xmm1
+; SSE2: movd %r9d, %xmm2
+; SSE2-NEXT: movd %r10d, %xmm3
+; SSE2-NEXT: movd %r11d, %xmm4
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %r14d, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: movd %r15d, %xmm1
+; SSE2-NEXT: movd %r12d, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE2-NEXT: movd %r13d, %xmm0
+; SSE2-NEXT: movd %r8d, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %ebp, %xmm0
+; SSE2-NEXT: movd %edi, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE2-NEXT: movd %esi, %xmm0
+; SSE2-NEXT: movd %ebx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %edx, %xmm4
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: popq %r12
+; SSE2-NEXT: popq %r13
+; SSE2-NEXT: popq %r14
+; SSE2-NEXT: popq %r15
+; SSE2-NEXT: popq %rbp
+; SSE2-NEXT: retq
+;
+; SSE41: pextrb $1, %xmm1, %ecx
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pextrb $0, %xmm1, %ecx
+; SSE41-NEXT: pextrb $0, %xmm0, %edx
+; SSE41-NEXT: sarb %cl, %dl
+; SSE41-NEXT: movzbl %dl, %ecx
+; SSE41-NEXT: movd %ecx, %xmm2
+; SSE41-NEXT: pinsrb $1, %eax, %xmm2
+; SSE41-NEXT: pextrb $2, %xmm1, %ecx
+; SSE41-NEXT: pextrb $2, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $2, %eax, %xmm2
+; SSE41-NEXT: pextrb $3, %xmm1, %ecx
+; SSE41-NEXT: pextrb $3, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $3, %eax, %xmm2
+; SSE41-NEXT: pextrb $4, %xmm1, %ecx
+; SSE41-NEXT: pextrb $4, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $4, %eax, %xmm2
+; SSE41-NEXT: pextrb $5, %xmm1, %ecx
+; SSE41-NEXT: pextrb $5, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $5, %eax, %xmm2
+; SSE41-NEXT: pextrb $6, %xmm1, %ecx
+; SSE41-NEXT: pextrb $6, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $6, %eax, %xmm2
+; SSE41-NEXT: pextrb $7, %xmm1, %ecx
+; SSE41-NEXT: pextrb $7, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $7, %eax, %xmm2
+; SSE41-NEXT: pextrb $8, %xmm1, %ecx
+; SSE41-NEXT: pextrb $8, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $8, %eax, %xmm2
+; SSE41-NEXT: pextrb $9, %xmm1, %ecx
+; SSE41-NEXT: pextrb $9, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $9, %eax, %xmm2
+; SSE41-NEXT: pextrb $10, %xmm1, %ecx
+; SSE41-NEXT: pextrb $10, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $10, %eax, %xmm2
+; SSE41-NEXT: pextrb $11, %xmm1, %ecx
+; SSE41-NEXT: pextrb $11, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $11, %eax, %xmm2
+; SSE41-NEXT: pextrb $12, %xmm1, %ecx
+; SSE41-NEXT: pextrb $12, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $12, %eax, %xmm2
+; SSE41-NEXT: pextrb $13, %xmm1, %ecx
+; SSE41-NEXT: pextrb $13, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $13, %eax, %xmm2
+; SSE41-NEXT: pextrb $14, %xmm1, %ecx
+; SSE41-NEXT: pextrb $14, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $14, %eax, %xmm2
+; SSE41-NEXT: pextrb $15, %xmm1, %ecx
+; SSE41-NEXT: pextrb $15, %xmm0, %eax
+; SSE41-NEXT: sarb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $15, %eax, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpextrb $1, %xmm1, %ecx
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX-NEXT: vpextrb $0, %xmm0, %edx
+; AVX-NEXT: sarb %cl, %dl
+; AVX-NEXT: movzbl %dl, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm2
+; AVX-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $2, %xmm1, %ecx
+; AVX-NEXT: vpextrb $2, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $3, %xmm1, %ecx
+; AVX-NEXT: vpextrb $3, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $4, %xmm1, %ecx
+; AVX-NEXT: vpextrb $4, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $5, %xmm1, %ecx
+; AVX-NEXT: vpextrb $5, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $6, %xmm1, %ecx
+; AVX-NEXT: vpextrb $6, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $7, %xmm1, %ecx
+; AVX-NEXT: vpextrb $7, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $8, %xmm1, %ecx
+; AVX-NEXT: vpextrb $8, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $9, %xmm1, %ecx
+; AVX-NEXT: vpextrb $9, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $10, %xmm1, %ecx
+; AVX-NEXT: vpextrb $10, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $11, %xmm1, %ecx
+; AVX-NEXT: vpextrb $11, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $12, %xmm1, %ecx
+; AVX-NEXT: vpextrb $12, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $13, %xmm1, %ecx
+; AVX-NEXT: vpextrb $13, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $14, %xmm1, %ecx
+; AVX-NEXT: vpextrb $14, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $15, %xmm1, %ecx
+; AVX-NEXT: vpextrb $15, %xmm0, %eax
+; AVX-NEXT: sarb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX-NEXT: retq
+ %ashr = ashr <16 x i8> %r, %a
+ %tmp2 = bitcast <16 x i8> %ashr to <2 x i64>
+ ret <2 x i64> %tmp2
+}
+
+define <2 x i64> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind readnone ssp {
+entry:
+
+; SSE2: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: pextrw $7, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: pextrw $3, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %eax
+; SSE2-NEXT: pextrw $5, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: pextrw $1, %xmm0, %eax
+; SSE2-NEXT: pextrw $1, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %eax
+; SSE2-NEXT: pextrw $6, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: pextrw $2, %xmm0, %eax
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: pextrw $4, %xmm1, %ecx
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: movd %xmm1, %ecx
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: movzwl %ax, %eax
+; SSE2-NEXT: shrl %cl, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: retq
+;
+; SSE41: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: pextrw $1, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: movd %xmm1, %ecx
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: movzwl %dx, %edx
+; SSE41-NEXT: shrl %cl, %edx
+; SSE41-NEXT: movd %edx, %xmm2
+; SSE41-NEXT: pinsrw $1, %eax, %xmm2
+; SSE41-NEXT: pextrw $2, %xmm0, %eax
+; SSE41-NEXT: pextrw $2, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $2, %eax, %xmm2
+; SSE41-NEXT: pextrw $3, %xmm0, %eax
+; SSE41-NEXT: pextrw $3, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $3, %eax, %xmm2
+; SSE41-NEXT: pextrw $4, %xmm0, %eax
+; SSE41-NEXT: pextrw $4, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $4, %eax, %xmm2
+; SSE41-NEXT: pextrw $5, %xmm0, %eax
+; SSE41-NEXT: pextrw $5, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $5, %eax, %xmm2
+; SSE41-NEXT: pextrw $6, %xmm0, %eax
+; SSE41-NEXT: pextrw $6, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $6, %eax, %xmm2
+; SSE41-NEXT: pextrw $7, %xmm0, %eax
+; SSE41-NEXT: pextrw $7, %xmm1, %ecx
+; SSE41-NEXT: shrl %cl, %eax
+; SSE41-NEXT: pinsrw $7, %eax, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: vpextrw $1, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vmovd %xmm1, %ecx
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: movzwl %dx, %edx
+; AVX-NEXT: shrl %cl, %edx
+; AVX-NEXT: vmovd %edx, %xmm2
+; AVX-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $2, %xmm0, %eax
+; AVX-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $3, %xmm0, %eax
+; AVX-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $4, %xmm0, %eax
+; AVX-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $5, %xmm0, %eax
+; AVX-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $6, %xmm0, %eax
+; AVX-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrw $7, %xmm0, %eax
+; AVX-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX-NEXT: shrl %cl, %eax
+; AVX-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX-NEXT: retq
+ %lshr = lshr <8 x i16> %r, %a
+ %tmp2 = bitcast <8 x i16> %lshr to <2 x i64>
+ ret <2 x i64> %tmp2
+}
+
+define <2 x i64> @lshr_16i8(<16 x i8> %r, <16 x i8> %a) nounwind readnone ssp {
+entry:
+; SSE2: pushq %rbp
+; SSE2-NEXT: pushq %r15
+; SSE2-NEXT: pushq %r14
+; SSE2-NEXT: pushq %r13
+; SSE2-NEXT: pushq %r12
+; SSE2-NEXT: pushq %rbx
+; SSE2-NEXT: movaps %xmm1, -24(%rsp)
+; SSE2-NEXT: movaps %xmm0, -40(%rsp)
+; SSE2-NEXT: movb -9(%rsp), %cl
+; SSE2-NEXT: movb -25(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movb -17(%rsp), %cl
+; SSE2-NEXT: movb -33(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -13(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movl %eax, -44(%rsp)
+; SSE2-NEXT: movb -29(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movzbl %al, %r9d
+; SSE2-NEXT: movb -21(%rsp), %cl
+; SSE2-NEXT: movb -37(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -11(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r10d
+; SSE2-NEXT: movb -27(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -19(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r11d
+; SSE2-NEXT: movb -35(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -15(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r14d
+; SSE2-NEXT: movb -31(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movzbl %al, %r15d
+; SSE2-NEXT: movb -23(%rsp), %cl
+; SSE2-NEXT: movb -39(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -10(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r12d
+; SSE2-NEXT: movb -26(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -18(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r13d
+; SSE2-NEXT: movb -34(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -14(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %r8d
+; SSE2-NEXT: movb -30(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -22(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %ebp
+; SSE2-NEXT: movb -38(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movb -12(%rsp), %cl
+; SSE2-NEXT: movzbl %al, %edi
+; SSE2-NEXT: movb -28(%rsp), %dl
+; SSE2-NEXT: shrb %cl, %dl
+; SSE2-NEXT: movb -20(%rsp), %cl
+; SSE2-NEXT: movzbl %dl, %esi
+; SSE2-NEXT: movb -36(%rsp), %bl
+; SSE2-NEXT: shrb %cl, %bl
+; SSE2-NEXT: movb -16(%rsp), %cl
+; SSE2-NEXT: movzbl %bl, %ebx
+; SSE2-NEXT: movb -32(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movzbl %al, %edx
+; SSE2-NEXT: movb -24(%rsp), %cl
+; SSE2-NEXT: movb -40(%rsp), %al
+; SSE2-NEXT: shrb %cl, %al
+; SSE2-NEXT: movzbl %al, %eax
+; SSE2-NEXT: movd -44(%rsp), %xmm1
+; SSE2: movd %r9d, %xmm2
+; SSE2-NEXT: movd %r10d, %xmm3
+; SSE2-NEXT: movd %r11d, %xmm4
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %r14d, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: movd %r15d, %xmm1
+; SSE2-NEXT: movd %r12d, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE2-NEXT: movd %r13d, %xmm0
+; SSE2-NEXT: movd %r8d, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %ebp, %xmm0
+; SSE2-NEXT: movd %edi, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE2-NEXT: movd %esi, %xmm0
+; SSE2-NEXT: movd %ebx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: movd %edx, %xmm4
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: popq %r12
+; SSE2-NEXT: popq %r13
+; SSE2-NEXT: popq %r14
+; SSE2-NEXT: popq %r15
+; SSE2-NEXT: popq %rbp
+; SSE2-NEXT: retq
+;
+; SSE41: pextrb $1, %xmm1, %ecx
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pextrb $0, %xmm1, %ecx
+; SSE41-NEXT: pextrb $0, %xmm0, %edx
+; SSE41-NEXT: shrb %cl, %dl
+; SSE41-NEXT: movzbl %dl, %ecx
+; SSE41-NEXT: movd %ecx, %xmm2
+; SSE41-NEXT: pinsrb $1, %eax, %xmm2
+; SSE41-NEXT: pextrb $2, %xmm1, %ecx
+; SSE41-NEXT: pextrb $2, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $2, %eax, %xmm2
+; SSE41-NEXT: pextrb $3, %xmm1, %ecx
+; SSE41-NEXT: pextrb $3, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $3, %eax, %xmm2
+; SSE41-NEXT: pextrb $4, %xmm1, %ecx
+; SSE41-NEXT: pextrb $4, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $4, %eax, %xmm2
+; SSE41-NEXT: pextrb $5, %xmm1, %ecx
+; SSE41-NEXT: pextrb $5, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $5, %eax, %xmm2
+; SSE41-NEXT: pextrb $6, %xmm1, %ecx
+; SSE41-NEXT: pextrb $6, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $6, %eax, %xmm2
+; SSE41-NEXT: pextrb $7, %xmm1, %ecx
+; SSE41-NEXT: pextrb $7, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $7, %eax, %xmm2
+; SSE41-NEXT: pextrb $8, %xmm1, %ecx
+; SSE41-NEXT: pextrb $8, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $8, %eax, %xmm2
+; SSE41-NEXT: pextrb $9, %xmm1, %ecx
+; SSE41-NEXT: pextrb $9, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $9, %eax, %xmm2
+; SSE41-NEXT: pextrb $10, %xmm1, %ecx
+; SSE41-NEXT: pextrb $10, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $10, %eax, %xmm2
+; SSE41-NEXT: pextrb $11, %xmm1, %ecx
+; SSE41-NEXT: pextrb $11, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $11, %eax, %xmm2
+; SSE41-NEXT: pextrb $12, %xmm1, %ecx
+; SSE41-NEXT: pextrb $12, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $12, %eax, %xmm2
+; SSE41-NEXT: pextrb $13, %xmm1, %ecx
+; SSE41-NEXT: pextrb $13, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $13, %eax, %xmm2
+; SSE41-NEXT: pextrb $14, %xmm1, %ecx
+; SSE41-NEXT: pextrb $14, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $14, %eax, %xmm2
+; SSE41-NEXT: pextrb $15, %xmm1, %ecx
+; SSE41-NEXT: pextrb $15, %xmm0, %eax
+; SSE41-NEXT: shrb %cl, %al
+; SSE41-NEXT: movzbl %al, %eax
+; SSE41-NEXT: pinsrb $15, %eax, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX: vpextrb $1, %xmm1, %ecx
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX-NEXT: vpextrb $0, %xmm0, %edx
+; AVX-NEXT: shrb %cl, %dl
+; AVX-NEXT: movzbl %dl, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm2
+; AVX-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $2, %xmm1, %ecx
+; AVX-NEXT: vpextrb $2, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $3, %xmm1, %ecx
+; AVX-NEXT: vpextrb $3, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $4, %xmm1, %ecx
+; AVX-NEXT: vpextrb $4, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $5, %xmm1, %ecx
+; AVX-NEXT: vpextrb $5, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $6, %xmm1, %ecx
+; AVX-NEXT: vpextrb $6, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $7, %xmm1, %ecx
+; AVX-NEXT: vpextrb $7, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $8, %xmm1, %ecx
+; AVX-NEXT: vpextrb $8, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $9, %xmm1, %ecx
+; AVX-NEXT: vpextrb $9, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $10, %xmm1, %ecx
+; AVX-NEXT: vpextrb $10, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $11, %xmm1, %ecx
+; AVX-NEXT: vpextrb $11, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $12, %xmm1, %ecx
+; AVX-NEXT: vpextrb $12, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $13, %xmm1, %ecx
+; AVX-NEXT: vpextrb $13, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $14, %xmm1, %ecx
+; AVX-NEXT: vpextrb $14, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX-NEXT: vpextrb $15, %xmm1, %ecx
+; AVX-NEXT: vpextrb $15, %xmm0, %eax
+; AVX-NEXT: shrb %cl, %al
+; AVX-NEXT: movzbl %al, %eax
+; AVX-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX-NEXT: retq
+ %lshr = lshr <16 x i8> %r, %a
+ %tmp2 = bitcast <16 x i8> %lshr to <2 x i64>
+ ret <2 x i64> %tmp2
+}
diff --git a/test/CodeGen/X86/vector-ctpop.ll b/test/CodeGen/X86/vector-ctpop.ll
deleted file mode 100644
index 59d67928c6fa..000000000000
--- a/test/CodeGen/X86/vector-ctpop.ll
+++ /dev/null
@@ -1,159 +0,0 @@
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=avx2 | FileCheck -check-prefix=AVX2 %s
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=avx -mattr=-popcnt | FileCheck -check-prefix=AVX1-NOPOPCNT %s
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=avx2 -mattr=-popcnt | FileCheck -check-prefix=AVX2-NOPOPCNT %s
-
-; Vector version of:
-; v = v - ((v >> 1) & 0x55555555)
-; v = (v & 0x33333333) + ((v >> 2) & 0x33333333)
-; v = (v + (v >> 4) & 0xF0F0F0F)
-; v = v + (v >> 8)
-; v = v + (v >> 16)
-; v = v + (v >> 32) ; i64 only
-
-define <8 x i32> @test0(<8 x i32> %x) {
-; AVX2-LABEL: @test0
-entry:
-; AVX2: vpsrld $1, %ymm
-; AVX2-NEXT: vpbroadcastd
-; AVX2-NEXT: vpand
-; AVX2-NEXT: vpsubd
-; AVX2-NEXT: vpbroadcastd
-; AVX2-NEXT: vpand
-; AVX2-NEXT: vpsrld $2
-; AVX2-NEXT: vpand
-; AVX2-NEXT: vpaddd
-; AVX2-NEXT: vpsrld $4
-; AVX2-NEXT: vpaddd
-; AVX2-NEXT: vpbroadcastd
-; AVX2-NEXT: vpand
-; AVX2-NEXT: vpsrld $8
-; AVX2-NEXT: vpaddd
-; AVX2-NEXT: vpsrld $16
-; AVX2-NEXT: vpaddd
-; AVX2-NEXT: vpbroadcastd
-; AVX2-NEXT: vpand
- %y = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %x)
- ret <8 x i32> %y
-}
-
-define <4 x i64> @test1(<4 x i64> %x) {
-; AVX2-NOPOPCNT-LABEL: @test1
-entry:
-; AVX2-NOPOPCNT: vpsrlq $1, %ymm
-; AVX2-NOPOPCNT-NEXT: vpbroadcastq
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsubq
-; AVX2-NOPOPCNT-NEXT: vpbroadcastq
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrlq $2
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $4
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpbroadcastq
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrlq $8
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $16
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $32
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpbroadcastq
-; AVX2-NOPOPCNT-NEXT: vpand
- %y = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %x)
- ret <4 x i64> %y
-}
-
-define <4 x i32> @test2(<4 x i32> %x) {
-; AVX2-NOPOPCNT-LABEL: @test2
-; AVX1-NOPOPCNT-LABEL: @test2
-entry:
-; AVX2-NOPOPCNT: vpsrld $1, %xmm
-; AVX2-NOPOPCNT-NEXT: vpbroadcastd
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsubd
-; AVX2-NOPOPCNT-NEXT: vpbroadcastd
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrld $2
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpaddd
-; AVX2-NOPOPCNT-NEXT: vpsrld $4
-; AVX2-NOPOPCNT-NEXT: vpaddd
-; AVX2-NOPOPCNT-NEXT: vpbroadcastd
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrld $8
-; AVX2-NOPOPCNT-NEXT: vpaddd
-; AVX2-NOPOPCNT-NEXT: vpsrld $16
-; AVX2-NOPOPCNT-NEXT: vpaddd
-; AVX2-NOPOPCNT-NEXT: vpbroadcastd
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT: vpsrld $1, %xmm
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsubd
-; AVX1-NOPOPCNT-NEXT: vmovdqa
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsrld $2
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpaddd
-; AVX1-NOPOPCNT-NEXT: vpsrld $4
-; AVX1-NOPOPCNT-NEXT: vpaddd
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsrld $8
-; AVX1-NOPOPCNT-NEXT: vpaddd
-; AVX1-NOPOPCNT-NEXT: vpsrld $16
-; AVX1-NOPOPCNT-NEXT: vpaddd
-; AVX1-NOPOPCNT-NEXT: vpand
- %y = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %x)
- ret <4 x i32> %y
-}
-
-define <2 x i64> @test3(<2 x i64> %x) {
-; AVX2-NOPOPCNT-LABEL: @test3
-; AVX1-NOPOPCNT-LABEL: @test3
-entry:
-; AVX2-NOPOPCNT: vpsrlq $1, %xmm
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsubq
-; AVX2-NOPOPCNT-NEXT: vmovdqa
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrlq $2
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $4
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX2-NOPOPCNT-NEXT: vpsrlq $8
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $16
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpsrlq $32
-; AVX2-NOPOPCNT-NEXT: vpaddq
-; AVX2-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT: vpsrlq $1, %xmm
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsubq
-; AVX1-NOPOPCNT-NEXT: vmovdqa
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsrlq $2
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpaddq
-; AVX1-NOPOPCNT-NEXT: vpsrlq $4
-; AVX1-NOPOPCNT-NEXT: vpaddq
-; AVX1-NOPOPCNT-NEXT: vpand
-; AVX1-NOPOPCNT-NEXT: vpsrlq $8
-; AVX1-NOPOPCNT-NEXT: vpaddq
-; AVX1-NOPOPCNT-NEXT: vpsrlq $16
-; AVX1-NOPOPCNT-NEXT: vpaddq
-; AVX1-NOPOPCNT-NEXT: vpsrlq $32
-; AVX1-NOPOPCNT-NEXT: vpaddq
-; AVX1-NOPOPCNT-NEXT: vpand
- %y = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %x)
- ret <2 x i64> %y
-}
-
-declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>)
-declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>)
-
-declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>)
-declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>)
-
diff --git a/test/CodeGen/X86/vector-lzcnt-128.ll b/test/CodeGen/X86/vector-lzcnt-128.ll
new file mode 100644
index 000000000000..b43188b7c6ea
--- /dev/null
+++ b/test/CodeGen/X86/vector-lzcnt-128.ll
@@ -0,0 +1,1915 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <2 x i64> @testv2i64(<2 x i64> %in) {
+; SSE2-LABEL: testv2i64:
+; SSE2: # BB#0:
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsrq %rax, %rax
+; SSE2-NEXT: movl $127, %ecx
+; SSE2-NEXT: cmoveq %rcx, %rax
+; SSE2-NEXT: xorq $63, %rax
+; SSE2-NEXT: movd %rax, %xmm1
+; SSE2-NEXT: pshufd $78, %xmm0, %xmm0 # xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsrq %rax, %rax
+; SSE2-NEXT: cmoveq %rcx, %rax
+; SSE2-NEXT: xorq $63, %rax
+; SSE2-NEXT: movd %rax, %xmm0
+; SSE2-NEXT: punpcklqdq %xmm0, %xmm1 # xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv2i64:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsrq %rax, %rax
+; SSE3-NEXT: movl $127, %ecx
+; SSE3-NEXT: cmoveq %rcx, %rax
+; SSE3-NEXT: xorq $63, %rax
+; SSE3-NEXT: movd %rax, %xmm1
+; SSE3-NEXT: pshufd $78, %xmm0, %xmm0 # xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsrq %rax, %rax
+; SSE3-NEXT: cmoveq %rcx, %rax
+; SSE3-NEXT: xorq $63, %rax
+; SSE3-NEXT: movd %rax, %xmm0
+; SSE3-NEXT: punpcklqdq %xmm0, %xmm1 # xmm1 = xmm1[0],xmm0[0]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv2i64:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsrq %rax, %rax
+; SSSE3-NEXT: movl $127, %ecx
+; SSSE3-NEXT: cmoveq %rcx, %rax
+; SSSE3-NEXT: xorq $63, %rax
+; SSSE3-NEXT: movd %rax, %xmm1
+; SSSE3-NEXT: pshufd $78, %xmm0, %xmm0 # xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsrq %rax, %rax
+; SSSE3-NEXT: cmoveq %rcx, %rax
+; SSSE3-NEXT: xorq $63, %rax
+; SSSE3-NEXT: movd %rax, %xmm0
+; SSSE3-NEXT: punpcklqdq %xmm0, %xmm1 # xmm1 = xmm1[0],xmm0[0]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+
+;
+; SSE41-LABEL: testv2i64:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrq $1, %xmm0, %rax
+; SSE41-NEXT: bsrq %rax, %rax
+; SSE41-NEXT: movl $127, %ecx
+; SSE41-NEXT: cmoveq %rcx, %rax
+; SSE41-NEXT: xorq $63, %rax
+; SSE41-NEXT: movd %rax, %xmm1
+; SSE41-NEXT: movd %xmm0, %rax
+; SSE41-NEXT: bsrq %rax, %rax
+; SSE41-NEXT: cmoveq %rcx, %rax
+; SSE41-NEXT: xorq $63, %rax
+; SSE41-NEXT: movd %rax, %xmm0
+; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrq $1, %xmm0, %rax
+; AVX-NEXT: bsrq %rax, %rax
+; AVX-NEXT: movl $127, %ecx
+; AVX-NEXT: cmoveq %rcx, %rax
+; AVX-NEXT: xorq $63, %rax
+; AVX-NEXT: vmovq %rax, %xmm1
+; AVX-NEXT: vmovq %xmm0, %rax
+; AVX-NEXT: bsrq %rax, %rax
+; AVX-NEXT: cmoveq %rcx, %rax
+; AVX-NEXT: xorq $63, %rax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %in, i1 0)
+ ret <2 x i64> %out
+}
+
+define <2 x i64> @testv2i64u(<2 x i64> %in) {
+; SSE2-LABEL: testv2i64u:
+; SSE2: # BB#0:
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsrq %rax, %rax
+; SSE2-NEXT: xorq $63, %rax
+; SSE2-NEXT: movd %rax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsrq %rax, %rax
+; SSE2-NEXT: xorq $63, %rax
+; SSE2-NEXT: movd %rax, %xmm0
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv2i64u:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsrq %rax, %rax
+; SSE3-NEXT: xorq $63, %rax
+; SSE3-NEXT: movd %rax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsrq %rax, %rax
+; SSE3-NEXT: xorq $63, %rax
+; SSE3-NEXT: movd %rax, %xmm0
+; SSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv2i64u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsrq %rax, %rax
+; SSSE3-NEXT: xorq $63, %rax
+; SSSE3-NEXT: movd %rax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsrq %rax, %rax
+; SSSE3-NEXT: xorq $63, %rax
+; SSSE3-NEXT: movd %rax, %xmm0
+; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv2i64u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrq $1, %xmm0, %rax
+; SSE41-NEXT: bsrq %rax, %rax
+; SSE41-NEXT: xorq $63, %rax
+; SSE41-NEXT: movd %rax, %xmm1
+; SSE41-NEXT: movd %xmm0, %rax
+; SSE41-NEXT: bsrq %rax, %rax
+; SSE41-NEXT: xorq $63, %rax
+; SSE41-NEXT: movd %rax, %xmm0
+; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv2i64u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrq $1, %xmm0, %rax
+; AVX-NEXT: bsrq %rax, %rax
+; AVX-NEXT: xorq $63, %rax
+; AVX-NEXT: vmovq %rax, %xmm1
+; AVX-NEXT: vmovq %xmm0, %rax
+; AVX-NEXT: bsrq %rax, %rax
+; AVX-NEXT: xorq $63, %rax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %in, i1 -1)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @testv4i32(<4 x i32> %in) {
+; SSE2-LABEL: testv4i32:
+; SSE2: # BB#0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: movl $63, %ecx
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv4i32:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: movl $63, %ecx
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv4i32:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSSE3-NEXT: movd %xmm1, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: movl $63, %ecx
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSSE3-NEXT: movd %xmm2, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv4i32:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrd $1, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: movl $63, %ecx
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: bsrl %edx, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: xorl $31, %edx
+; SSE41-NEXT: movd %edx, %xmm1
+; SSE41-NEXT: pinsrd $1, %eax, %xmm1
+; SSE41-NEXT: pextrd $2, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: pinsrd $2, %eax, %xmm1
+; SSE41-NEXT: pextrd $3, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: pinsrd $3, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrd $1, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: movl $63, %ecx
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: bsrl %edx, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: xorl $31, %edx
+; AVX-NEXT: vmovd %edx, %xmm1
+; AVX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $2, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $3, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %in, i1 0)
+ ret <4 x i32> %out
+}
+
+define <4 x i32> @testv4i32u(<4 x i32> %in) {
+; SSE2-LABEL: testv4i32u:
+; SSE2: # BB#0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $31, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv4i32u:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $31, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv4i32u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSSE3-NEXT: movd %xmm1, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSSE3-NEXT: movd %xmm2, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $31, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv4i32u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrd $1, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: movd %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: xorl $31, %ecx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrd $1, %eax, %xmm1
+; SSE41-NEXT: pextrd $2, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: pinsrd $2, %eax, %xmm1
+; SSE41-NEXT: pextrd $3, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $31, %eax
+; SSE41-NEXT: pinsrd $3, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv4i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrd $1, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vmovd %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: xorl $31, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $2, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $3, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $31, %eax
+; AVX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %in, i1 -1)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @testv8i16(<8 x i16> %in) {
+; SSE2-LABEL: testv8i16:
+; SSE2: # BB#0:
+; SSE2-NEXT: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %cx
+; SSE2-NEXT: movw $31, %ax
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: pextrw $3, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: pextrw $1, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: pextrw $2, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: bsrw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: xorl $15, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv8i16:
+; SSE3: # BB#0:
+; SSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %cx
+; SSE3-NEXT: movw $31, %ax
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: pextrw $3, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE3-NEXT: pextrw $5, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE3-NEXT: pextrw $6, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: pextrw $2, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE3-NEXT: pextrw $4, %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: bsrw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: xorl $15, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv8i16:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %cx
+; SSSE3-NEXT: movw $31, %ax
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: pextrw $3, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSSE3-NEXT: pextrw $5, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSSE3-NEXT: pextrw $6, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: pextrw $2, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSSE3-NEXT: pextrw $4, %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: movd %xmm0, %ecx
+; SSSE3-NEXT: bsrw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: xorl $15, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv8i16:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %cx
+; SSE41-NEXT: movw $31, %ax
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: bsrw %dx, %dx
+; SSE41-NEXT: cmovew %ax, %dx
+; SSE41-NEXT: xorl $15, %edx
+; SSE41-NEXT: movd %edx, %xmm1
+; SSE41-NEXT: pinsrw $1, %ecx, %xmm1
+; SSE41-NEXT: pextrw $2, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $2, %ecx, %xmm1
+; SSE41-NEXT: pextrw $3, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $3, %ecx, %xmm1
+; SSE41-NEXT: pextrw $4, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $4, %ecx, %xmm1
+; SSE41-NEXT: pextrw $5, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $5, %ecx, %xmm1
+; SSE41-NEXT: pextrw $6, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $6, %ecx, %xmm1
+; SSE41-NEXT: pextrw $7, %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: pinsrw $7, %ecx, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %cx
+; AVX-NEXT: movw $31, %ax
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: bsrw %dx, %dx
+; AVX-NEXT: cmovew %ax, %dx
+; AVX-NEXT: xorl $15, %edx
+; AVX-NEXT: vmovd %edx, %xmm1
+; AVX-NEXT: vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %in, i1 0)
+ ret <8 x i16> %out
+}
+
+define <8 x i16> @testv8i16u(<8 x i16> %in) {
+; SSE2-LABEL: testv8i16u:
+; SSE2: # BB#0:
+; SSE2-NEXT: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: pextrw $1, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: pextrw $2, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsrw %ax, %ax
+; SSE2-NEXT: xorl $15, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv8i16u:
+; SSE3: # BB#0:
+; SSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pextrw $3, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE3-NEXT: pextrw $5, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: pextrw $1, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsrw %ax, %ax
+; SSE3-NEXT: xorl $15, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv8i16u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pextrw $3, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSSE3-NEXT: pextrw $5, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: pextrw $1, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsrw %ax, %ax
+; SSSE3-NEXT: xorl $15, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv8i16u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: movd %xmm0, %ecx
+; SSE41-NEXT: bsrw %cx, %cx
+; SSE41-NEXT: xorl $15, %ecx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrw $1, %eax, %xmm1
+; SSE41-NEXT: pextrw $2, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $2, %eax, %xmm1
+; SSE41-NEXT: pextrw $3, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $3, %eax, %xmm1
+; SSE41-NEXT: pextrw $4, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $4, %eax, %xmm1
+; SSE41-NEXT: pextrw $5, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $5, %eax, %xmm1
+; SSE41-NEXT: pextrw $6, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $6, %eax, %xmm1
+; SSE41-NEXT: pextrw $7, %xmm0, %eax
+; SSE41-NEXT: bsrw %ax, %ax
+; SSE41-NEXT: xorl $15, %eax
+; SSE41-NEXT: pinsrw $7, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv8i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vmovd %xmm0, %ecx
+; AVX-NEXT: bsrw %cx, %cx
+; AVX-NEXT: xorl $15, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrw $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $2, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $3, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $3, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $4, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $4, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $5, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $5, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $6, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $7, %xmm0, %eax
+; AVX-NEXT: bsrw %ax, %ax
+; AVX-NEXT: xorl $15, %eax
+; AVX-NEXT: vpinsrw $7, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %in, i1 -1)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @testv16i8(<16 x i8> %in) {
+; SSE2-LABEL: testv16i8:
+; SSE2: # BB#0:
+; SSE2: pushq %rbp
+; SSE2: movaps %xmm0, -24(%rsp)
+; SSE2-NEXT: movzbl -9(%rsp), %eax
+; SSE2-NEXT: bsrl %eax, %ecx
+; SSE2-NEXT: movl $15, %eax
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: movzbl -10(%rsp), %ebx
+; SSE2-NEXT: movzbl -11(%rsp), %edi
+; SSE2-NEXT: movzbl -12(%rsp), %r9d
+; SSE2-NEXT: movzbl -13(%rsp), %edx
+; SSE2-NEXT: movzbl -14(%rsp), %r11d
+; SSE2-NEXT: movzbl -15(%rsp), %esi
+; SSE2-NEXT: movzbl -16(%rsp), %r8d
+; SSE2-NEXT: movzbl -17(%rsp), %ecx
+; SSE2-NEXT: bsrl %ecx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsrl %edx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: movzbl -18(%rsp), %edx
+; SSE2-NEXT: movzbl -19(%rsp), %ecx
+; SSE2-NEXT: movzbl -20(%rsp), %r10d
+; SSE2-NEXT: movzbl -21(%rsp), %ebp
+; SSE2-NEXT: bsrl %ebp, %ebp
+; SSE2-NEXT: cmovel %eax, %ebp
+; SSE2-NEXT: xorl $7, %ebp
+; SSE2-NEXT: movd %ebp, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: bsrl %edi, %edi
+; SSE2-NEXT: cmovel %eax, %edi
+; SSE2-NEXT: xorl $7, %edi
+; SSE2-NEXT: movd %edi, %xmm1
+; SSE2-NEXT: bsrl %ecx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: bsrl %esi, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: movzbl -22(%rsp), %esi
+; SSE2-NEXT: movzbl -23(%rsp), %ecx
+; SSE2-NEXT: bsrl %ecx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsrl %ebx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: bsrl %edx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsrl %r11d, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: bsrl %esi, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE2-NEXT: bsrl %r9d, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: bsrl %r10d, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsrl %r8d, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm4
+; SSE2-NEXT: movzbl -24(%rsp), %ecx
+; SSE2-NEXT: bsrl %ecx, %ecx
+; SSE2-NEXT: cmovel %eax, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: popq %rbp
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv16i8:
+; SSE3: # BB#0:
+; SSE3: pushq %rbp
+; SSE3: movaps %xmm0, -24(%rsp)
+; SSE3-NEXT: movzbl -9(%rsp), %eax
+; SSE3-NEXT: bsrl %eax, %ecx
+; SSE3-NEXT: movl $15, %eax
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: movzbl -10(%rsp), %ebx
+; SSE3-NEXT: movzbl -11(%rsp), %edi
+; SSE3-NEXT: movzbl -12(%rsp), %r9d
+; SSE3-NEXT: movzbl -13(%rsp), %edx
+; SSE3-NEXT: movzbl -14(%rsp), %r11d
+; SSE3-NEXT: movzbl -15(%rsp), %esi
+; SSE3-NEXT: movzbl -16(%rsp), %r8d
+; SSE3-NEXT: movzbl -17(%rsp), %ecx
+; SSE3-NEXT: bsrl %ecx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsrl %edx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: movzbl -18(%rsp), %edx
+; SSE3-NEXT: movzbl -19(%rsp), %ecx
+; SSE3-NEXT: movzbl -20(%rsp), %r10d
+; SSE3-NEXT: movzbl -21(%rsp), %ebp
+; SSE3-NEXT: bsrl %ebp, %ebp
+; SSE3-NEXT: cmovel %eax, %ebp
+; SSE3-NEXT: xorl $7, %ebp
+; SSE3-NEXT: movd %ebp, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: bsrl %edi, %edi
+; SSE3-NEXT: cmovel %eax, %edi
+; SSE3-NEXT: xorl $7, %edi
+; SSE3-NEXT: movd %edi, %xmm1
+; SSE3-NEXT: bsrl %ecx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE3-NEXT: bsrl %esi, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: movzbl -22(%rsp), %esi
+; SSE3-NEXT: movzbl -23(%rsp), %ecx
+; SSE3-NEXT: bsrl %ecx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsrl %ebx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: bsrl %edx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsrl %r11d, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: bsrl %esi, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE3-NEXT: bsrl %r9d, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: bsrl %r10d, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsrl %r8d, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm4
+; SSE3-NEXT: movzbl -24(%rsp), %ecx
+; SSE3-NEXT: bsrl %ecx, %ecx
+; SSE3-NEXT: cmovel %eax, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: popq %rbp
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv16i8:
+; SSSE3: # BB#0:
+; SSSE3: pushq %rbp
+; SSSE3: movaps %xmm0, -24(%rsp)
+; SSSE3-NEXT: movzbl -9(%rsp), %eax
+; SSSE3-NEXT: bsrl %eax, %ecx
+; SSSE3-NEXT: movl $15, %eax
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: movzbl -10(%rsp), %ebx
+; SSSE3-NEXT: movzbl -11(%rsp), %edi
+; SSSE3-NEXT: movzbl -12(%rsp), %r9d
+; SSSE3-NEXT: movzbl -13(%rsp), %edx
+; SSSE3-NEXT: movzbl -14(%rsp), %r11d
+; SSSE3-NEXT: movzbl -15(%rsp), %esi
+; SSSE3-NEXT: movzbl -16(%rsp), %r8d
+; SSSE3-NEXT: movzbl -17(%rsp), %ecx
+; SSSE3-NEXT: bsrl %ecx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsrl %edx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: movzbl -18(%rsp), %edx
+; SSSE3-NEXT: movzbl -19(%rsp), %ecx
+; SSSE3-NEXT: movzbl -20(%rsp), %r10d
+; SSSE3-NEXT: movzbl -21(%rsp), %ebp
+; SSSE3-NEXT: bsrl %ebp, %ebp
+; SSSE3-NEXT: cmovel %eax, %ebp
+; SSSE3-NEXT: xorl $7, %ebp
+; SSSE3-NEXT: movd %ebp, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: bsrl %edi, %edi
+; SSSE3-NEXT: cmovel %eax, %edi
+; SSSE3-NEXT: xorl $7, %edi
+; SSSE3-NEXT: movd %edi, %xmm1
+; SSSE3-NEXT: bsrl %ecx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSSE3-NEXT: bsrl %esi, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: movzbl -22(%rsp), %esi
+; SSSE3-NEXT: movzbl -23(%rsp), %ecx
+; SSSE3-NEXT: bsrl %ecx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsrl %ebx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: bsrl %edx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsrl %r11d, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: bsrl %esi, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSSE3-NEXT: bsrl %r9d, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: bsrl %r10d, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsrl %r8d, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm4
+; SSSE3-NEXT: movzbl -24(%rsp), %ecx
+; SSSE3-NEXT: bsrl %ecx, %ecx
+; SSSE3-NEXT: cmovel %eax, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: popq %rbx
+; SSSE3-NEXT: popq %rbp
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv16i8:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %ecx
+; SSE41-NEXT: movl $15, %eax
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pextrb $0, %xmm0, %edx
+; SSE41-NEXT: bsrl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: xorl $7, %edx
+; SSE41-NEXT: movd %edx, %xmm1
+; SSE41-NEXT: pinsrb $1, %ecx, %xmm1
+; SSE41-NEXT: pextrb $2, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $2, %ecx, %xmm1
+; SSE41-NEXT: pextrb $3, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $3, %ecx, %xmm1
+; SSE41-NEXT: pextrb $4, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $4, %ecx, %xmm1
+; SSE41-NEXT: pextrb $5, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $5, %ecx, %xmm1
+; SSE41-NEXT: pextrb $6, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $6, %ecx, %xmm1
+; SSE41-NEXT: pextrb $7, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $7, %ecx, %xmm1
+; SSE41-NEXT: pextrb $8, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $8, %ecx, %xmm1
+; SSE41-NEXT: pextrb $9, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $9, %ecx, %xmm1
+; SSE41-NEXT: pextrb $10, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $10, %ecx, %xmm1
+; SSE41-NEXT: pextrb $11, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $11, %ecx, %xmm1
+; SSE41-NEXT: pextrb $12, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $12, %ecx, %xmm1
+; SSE41-NEXT: pextrb $13, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $13, %ecx, %xmm1
+; SSE41-NEXT: pextrb $14, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $14, %ecx, %xmm1
+; SSE41-NEXT: pextrb $15, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: cmovel %eax, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: pinsrb $15, %ecx, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %ecx
+; AVX-NEXT: movl $15, %eax
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpextrb $0, %xmm0, %edx
+; AVX-NEXT: bsrl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: xorl $7, %edx
+; AVX-NEXT: vmovd %edx, %xmm1
+; AVX-NEXT: vpinsrb $1, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $2, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $2, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $3, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $3, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $4, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $4, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $5, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $5, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $6, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $6, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $7, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $7, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $8, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $8, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $9, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $9, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $10, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $10, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $11, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $11, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $12, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $12, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $13, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $13, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $14, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $14, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $15, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: cmovel %eax, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vpinsrb $15, %ecx, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %in, i1 0)
+ ret <16 x i8> %out
+}
+
+define <16 x i8> @testv16i8u(<16 x i8> %in) {
+; SSE2-LABEL: testv16i8u:
+; SSE2: # BB#0:
+; SSE2: pushq %rbx
+; SSE2: movaps %xmm0, -16(%rsp)
+; SSE2-NEXT: movzbl -1(%rsp), %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movzbl -2(%rsp), %edi
+; SSE2-NEXT: movzbl -3(%rsp), %edx
+; SSE2-NEXT: movzbl -4(%rsp), %r9d
+; SSE2-NEXT: movzbl -5(%rsp), %eax
+; SSE2-NEXT: movzbl -6(%rsp), %r10d
+; SSE2-NEXT: movzbl -7(%rsp), %ecx
+; SSE2-NEXT: movzbl -8(%rsp), %r8d
+; SSE2-NEXT: movzbl -9(%rsp), %esi
+; SSE2-NEXT: bsrl %esi, %esi
+; SSE2-NEXT: xorl $7, %esi
+; SSE2-NEXT: movd %esi, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movzbl -10(%rsp), %eax
+; SSE2-NEXT: movzbl -11(%rsp), %esi
+; SSE2-NEXT: movzbl -12(%rsp), %r11d
+; SSE2-NEXT: movzbl -13(%rsp), %ebx
+; SSE2-NEXT: bsrl %ebx, %ebx
+; SSE2-NEXT: xorl $7, %ebx
+; SSE2-NEXT: movd %ebx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: bsrl %edx, %edx
+; SSE2-NEXT: xorl $7, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: bsrl %esi, %edx
+; SSE2-NEXT: xorl $7, %edx
+; SSE2-NEXT: movd %edx, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsrl %ecx, %ecx
+; SSE2-NEXT: xorl $7, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: movzbl -14(%rsp), %ecx
+; SSE2-NEXT: movzbl -15(%rsp), %edx
+; SSE2-NEXT: bsrl %edx, %edx
+; SSE2-NEXT: xorl $7, %edx
+; SSE2-NEXT: movd %edx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT: bsrl %edi, %edx
+; SSE2-NEXT: xorl $7, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: bsrl %r10d, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: bsrl %ecx, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT: bsrl %r9d, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: bsrl %r11d, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: bsrl %r8d, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: movzbl -16(%rsp), %eax
+; SSE2-NEXT: bsrl %eax, %eax
+; SSE2-NEXT: xorl $7, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv16i8u:
+; SSE3: # BB#0:
+; SSE3: pushq %rbx
+; SSE3: movaps %xmm0, -16(%rsp)
+; SSE3-NEXT: movzbl -1(%rsp), %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: movzbl -2(%rsp), %edi
+; SSE3-NEXT: movzbl -3(%rsp), %edx
+; SSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSE3-NEXT: movzbl -5(%rsp), %eax
+; SSE3-NEXT: movzbl -6(%rsp), %r10d
+; SSE3-NEXT: movzbl -7(%rsp), %ecx
+; SSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSE3-NEXT: movzbl -9(%rsp), %esi
+; SSE3-NEXT: bsrl %esi, %esi
+; SSE3-NEXT: xorl $7, %esi
+; SSE3-NEXT: movd %esi, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: movzbl -10(%rsp), %eax
+; SSE3-NEXT: movzbl -11(%rsp), %esi
+; SSE3-NEXT: movzbl -12(%rsp), %r11d
+; SSE3-NEXT: movzbl -13(%rsp), %ebx
+; SSE3-NEXT: bsrl %ebx, %ebx
+; SSE3-NEXT: xorl $7, %ebx
+; SSE3-NEXT: movd %ebx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE3-NEXT: bsrl %edx, %edx
+; SSE3-NEXT: xorl $7, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: bsrl %esi, %edx
+; SSE3-NEXT: xorl $7, %edx
+; SSE3-NEXT: movd %edx, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsrl %ecx, %ecx
+; SSE3-NEXT: xorl $7, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: movzbl -14(%rsp), %ecx
+; SSE3-NEXT: movzbl -15(%rsp), %edx
+; SSE3-NEXT: bsrl %edx, %edx
+; SSE3-NEXT: xorl $7, %edx
+; SSE3-NEXT: movd %edx, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE3-NEXT: bsrl %edi, %edx
+; SSE3-NEXT: xorl $7, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: bsrl %r10d, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: bsrl %ecx, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE3-NEXT: bsrl %r9d, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: bsrl %r11d, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: bsrl %r8d, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm4
+; SSE3-NEXT: movzbl -16(%rsp), %eax
+; SSE3-NEXT: bsrl %eax, %eax
+; SSE3-NEXT: xorl $7, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv16i8u:
+; SSSE3: # BB#0:
+; SSSE3: pushq %rbx
+; SSSE3: movaps %xmm0, -16(%rsp)
+; SSSE3-NEXT: movzbl -1(%rsp), %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: movzbl -2(%rsp), %edi
+; SSSE3-NEXT: movzbl -3(%rsp), %edx
+; SSSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSSE3-NEXT: movzbl -5(%rsp), %eax
+; SSSE3-NEXT: movzbl -6(%rsp), %r10d
+; SSSE3-NEXT: movzbl -7(%rsp), %ecx
+; SSSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSSE3-NEXT: movzbl -9(%rsp), %esi
+; SSSE3-NEXT: bsrl %esi, %esi
+; SSSE3-NEXT: xorl $7, %esi
+; SSSE3-NEXT: movd %esi, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: movzbl -10(%rsp), %eax
+; SSSE3-NEXT: movzbl -11(%rsp), %esi
+; SSSE3-NEXT: movzbl -12(%rsp), %r11d
+; SSSE3-NEXT: movzbl -13(%rsp), %ebx
+; SSSE3-NEXT: bsrl %ebx, %ebx
+; SSSE3-NEXT: xorl $7, %ebx
+; SSSE3-NEXT: movd %ebx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSSE3-NEXT: bsrl %edx, %edx
+; SSSE3-NEXT: xorl $7, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: bsrl %esi, %edx
+; SSSE3-NEXT: xorl $7, %edx
+; SSSE3-NEXT: movd %edx, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsrl %ecx, %ecx
+; SSSE3-NEXT: xorl $7, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: movzbl -14(%rsp), %ecx
+; SSSE3-NEXT: movzbl -15(%rsp), %edx
+; SSSE3-NEXT: bsrl %edx, %edx
+; SSSE3-NEXT: xorl $7, %edx
+; SSSE3-NEXT: movd %edx, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSSE3-NEXT: bsrl %edi, %edx
+; SSSE3-NEXT: xorl $7, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: bsrl %r10d, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: bsrl %ecx, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSSE3-NEXT: bsrl %r9d, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: bsrl %r11d, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: bsrl %r8d, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm4
+; SSSE3-NEXT: movzbl -16(%rsp), %eax
+; SSSE3-NEXT: bsrl %eax, %eax
+; SSSE3-NEXT: xorl $7, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: popq %rbx
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv16i8u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pextrb $0, %xmm0, %ecx
+; SSE41-NEXT: bsrl %ecx, %ecx
+; SSE41-NEXT: xorl $7, %ecx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrb $1, %eax, %xmm1
+; SSE41-NEXT: pextrb $2, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $2, %eax, %xmm1
+; SSE41-NEXT: pextrb $3, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $3, %eax, %xmm1
+; SSE41-NEXT: pextrb $4, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $4, %eax, %xmm1
+; SSE41-NEXT: pextrb $5, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $5, %eax, %xmm1
+; SSE41-NEXT: pextrb $6, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $6, %eax, %xmm1
+; SSE41-NEXT: pextrb $7, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $7, %eax, %xmm1
+; SSE41-NEXT: pextrb $8, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $8, %eax, %xmm1
+; SSE41-NEXT: pextrb $9, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $9, %eax, %xmm1
+; SSE41-NEXT: pextrb $10, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $10, %eax, %xmm1
+; SSE41-NEXT: pextrb $11, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $11, %eax, %xmm1
+; SSE41-NEXT: pextrb $12, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $12, %eax, %xmm1
+; SSE41-NEXT: pextrb $13, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $13, %eax, %xmm1
+; SSE41-NEXT: pextrb $14, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $14, %eax, %xmm1
+; SSE41-NEXT: pextrb $15, %xmm0, %eax
+; SSE41-NEXT: bsrl %eax, %eax
+; SSE41-NEXT: xorl $7, %eax
+; SSE41-NEXT: pinsrb $15, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv16i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX-NEXT: bsrl %ecx, %ecx
+; AVX-NEXT: xorl $7, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrb $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $2, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $3, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $3, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $4, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $4, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $5, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $5, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $6, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $6, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $7, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $7, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $8, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $8, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $9, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $9, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $10, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $10, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $11, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $11, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $12, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $13, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $13, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $14, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $15, %xmm0, %eax
+; AVX-NEXT: bsrl %eax, %eax
+; AVX-NEXT: xorl $7, %eax
+; AVX-NEXT: vpinsrb $15, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %in, i1 -1)
+ ret <16 x i8> %out
+}
+
+define <2 x i64> @foldv2i64() {
+; SSE-LABEL: foldv2i64:
+; SSE: # BB#0:
+; SSE-NEXT: movl $55, %eax
+; SSE-NEXT: movd %rax, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: movl $55, %eax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> <i64 256, i64 -1>, i1 0)
+ ret <2 x i64> %out
+}
+
+define <2 x i64> @foldv2i64u() {
+; SSE-LABEL: foldv2i64u:
+; SSE: # BB#0:
+; SSE-NEXT: movl $55, %eax
+; SSE-NEXT: movd %rax, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv2i64u:
+; AVX: # BB#0:
+; AVX-NEXT: movl $55, %eax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> <i64 256, i64 -1>, i1 -1)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @foldv4i32() {
+; SSE-LABEL: foldv4i32:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [23,0,32,24]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [23,0,32,24]
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> <i32 256, i32 -1, i32 0, i32 255>, i1 0)
+ ret <4 x i32> %out
+}
+
+define <4 x i32> @foldv4i32u() {
+; SSE-LABEL: foldv4i32u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [23,0,32,24]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv4i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [23,0,32,24]
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> <i32 256, i32 -1, i32 0, i32 255>, i1 -1)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @foldv8i16() {
+; SSE-LABEL: foldv8i16:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [7,0,16,8,16,13,11,9]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [7,0,16,8,16,13,11,9]
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88>, i1 0)
+ ret <8 x i16> %out
+}
+
+define <8 x i16> @foldv8i16u() {
+; SSE-LABEL: foldv8i16u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [7,0,16,8,16,13,11,9]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv8i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [7,0,16,8,16,13,11,9]
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88>, i1 -1)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @foldv16i8() {
+; SSE-LABEL: foldv16i8:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2]
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32>, i1 0)
+ ret <16 x i8> %out
+}
+
+define <16 x i8> @foldv16i8u() {
+; SSE-LABEL: foldv16i8u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv16i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2]
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32>, i1 -1)
+ ret <16 x i8> %out
+}
+
+declare <2 x i64> @llvm.ctlz.v2i64(<2 x i64>, i1)
+declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1)
+declare <8 x i16> @llvm.ctlz.v8i16(<8 x i16>, i1)
+declare <16 x i8> @llvm.ctlz.v16i8(<16 x i8>, i1)
diff --git a/test/CodeGen/X86/vector-lzcnt-256.ll b/test/CodeGen/X86/vector-lzcnt-256.ll
new file mode 100644
index 000000000000..48abe1290528
--- /dev/null
+++ b/test/CodeGen/X86/vector-lzcnt-256.ll
@@ -0,0 +1,1305 @@
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <4 x i64> @testv4i64(<4 x i64> %in) {
+; AVX1-LABEL: testv4i64:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrq $1, %xmm1, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: movl $127, %ecx
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm1, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm1
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT: vpextrq $1, %xmm0, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm0, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm0
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv4i64:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrq $1, %xmm1, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: movl $127, %ecx
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm1, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm1
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-NEXT: vpextrq $1, %xmm0, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm0, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm0
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %in, i1 0)
+ ret <4 x i64> %out
+}
+
+define <4 x i64> @testv4i64u(<4 x i64> %in) {
+; AVX1-LABEL: testv4i64u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrq $1, %xmm1, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm1, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm1
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT: vpextrq $1, %xmm0, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm0, %rax
+; AVX1-NEXT: bsrq %rax, %rax
+; AVX1-NEXT: xorq $63, %rax
+; AVX1-NEXT: vmovq %rax, %xmm0
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv4i64u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrq $1, %xmm1, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm1, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm1
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-NEXT: vpextrq $1, %xmm0, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm0, %rax
+; AVX2-NEXT: bsrq %rax, %rax
+; AVX2-NEXT: xorq $63, %rax
+; AVX2-NEXT: vmovq %rax, %xmm0
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %in, i1 -1)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @testv8i32(<8 x i32> %in) {
+; AVX1-LABEL: testv8i32:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %ecx
+; AVX1-NEXT: movl $63, %eax
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vmovd %xmm1, %edx
+; AVX1-NEXT: bsrl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: xorl $31, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vmovd %xmm0, %edx
+; AVX1-NEXT: bsrl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: xorl $31, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv8i32:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %ecx
+; AVX2-NEXT: movl $63, %eax
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vmovd %xmm1, %edx
+; AVX2-NEXT: bsrl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: xorl $31, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vmovd %xmm0, %edx
+; AVX2-NEXT: bsrl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: xorl $31, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %in, i1 0)
+ ret <8 x i32> %out
+}
+
+define <8 x i32> @testv8i32u(<8 x i32> %in) {
+; AVX1-LABEL: testv8i32u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vmovd %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vmovd %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: xorl $31, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $31, %eax
+; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv8i32u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vmovd %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vmovd %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: xorl $31, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $31, %eax
+; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %in, i1 -1)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @testv16i16(<16 x i16> %in) {
+; AVX1-LABEL: testv16i16:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %cx
+; AVX1-NEXT: movw $31, %ax
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vmovd %xmm1, %edx
+; AVX1-NEXT: bsrw %dx, %dx
+; AVX1-NEXT: cmovew %ax, %dx
+; AVX1-NEXT: xorl $15, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vmovd %xmm0, %edx
+; AVX1-NEXT: bsrw %dx, %dx
+; AVX1-NEXT: cmovew %ax, %dx
+; AVX1-NEXT: xorl $15, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv16i16:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %cx
+; AVX2-NEXT: movw $31, %ax
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vmovd %xmm1, %edx
+; AVX2-NEXT: bsrw %dx, %dx
+; AVX2-NEXT: cmovew %ax, %dx
+; AVX2-NEXT: xorl $15, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vmovd %xmm0, %edx
+; AVX2-NEXT: bsrw %dx, %dx
+; AVX2-NEXT: cmovew %ax, %dx
+; AVX2-NEXT: xorl $15, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %in, i1 0)
+ ret <16 x i16> %out
+}
+
+define <16 x i16> @testv16i16u(<16 x i16> %in) {
+; AVX1-LABEL: testv16i16u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vmovd %xmm1, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm1, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $7, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vmovd %xmm0, %ecx
+; AVX1-NEXT: bsrw %cx, %cx
+; AVX1-NEXT: xorl $15, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm0, %eax
+; AVX1-NEXT: bsrw %ax, %ax
+; AVX1-NEXT: xorl $15, %eax
+; AVX1-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv16i16u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vmovd %xmm1, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm1, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $7, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vmovd %xmm0, %ecx
+; AVX2-NEXT: bsrw %cx, %cx
+; AVX2-NEXT: xorl $15, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm0, %eax
+; AVX2-NEXT: bsrw %ax, %ax
+; AVX2-NEXT: xorl $15, %eax
+; AVX2-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %in, i1 -1)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @testv32i8(<32 x i8> %in) {
+; AVX1-LABEL: testv32i8:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %ecx
+; AVX1-NEXT: movl $15, %eax
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpextrb $0, %xmm1, %edx
+; AVX1-NEXT: bsrl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: xorl $7, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $7, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $8, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $9, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $10, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $11, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $12, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $13, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $14, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $15, %ecx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpextrb $0, %xmm0, %edx
+; AVX1-NEXT: bsrl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: xorl $7, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $7, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $8, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $9, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $10, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $11, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $12, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $13, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $14, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vpinsrb $15, %ecx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv32i8:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %ecx
+; AVX2-NEXT: movl $15, %eax
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpextrb $0, %xmm1, %edx
+; AVX2-NEXT: bsrl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: xorl $7, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $7, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $8, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $9, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $10, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $11, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $12, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $13, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $14, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $15, %ecx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpextrb $0, %xmm0, %edx
+; AVX2-NEXT: bsrl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: xorl $7, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $7, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $8, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $9, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $10, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $11, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $12, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $13, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $14, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vpinsrb $15, %ecx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %in, i1 0)
+ ret <32 x i8> %out
+}
+
+define <32 x i8> @testv32i8u(<32 x i8> %in) {
+; AVX1-LABEL: testv32i8u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm1, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $15, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX1-NEXT: bsrl %ecx, %ecx
+; AVX1-NEXT: xorl $7, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm0, %eax
+; AVX1-NEXT: bsrl %eax, %eax
+; AVX1-NEXT: xorl $7, %eax
+; AVX1-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv32i8u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm1, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $15, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX2-NEXT: bsrl %ecx, %ecx
+; AVX2-NEXT: xorl $7, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm0, %eax
+; AVX2-NEXT: bsrl %eax, %eax
+; AVX2-NEXT: xorl $7, %eax
+; AVX2-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %in, i1 -1)
+ ret <32 x i8> %out
+}
+
+define <4 x i64> @foldv4i64() {
+; AVX-LABEL: foldv4i64:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [55,0,64,56]
+; AVX-NEXT: retq
+ %out = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>, i1 0)
+ ret <4 x i64> %out
+}
+
+define <4 x i64> @foldv4i64u() {
+; AVX-LABEL: foldv4i64u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [55,0,64,56]
+; AVX-NEXT: retq
+ %out = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>, i1 -1)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @foldv8i32() {
+; AVX-LABEL: foldv8i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [23,0,32,24,0,29,27,25]
+; AVX-NEXT: retq
+ %out = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>, i1 0)
+ ret <8 x i32> %out
+}
+
+define <8 x i32> @foldv8i32u() {
+; AVX-LABEL: foldv8i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [23,0,32,24,0,29,27,25]
+; AVX-NEXT: retq
+ %out = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>, i1 -1)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @foldv16i16() {
+; AVX-LABEL: foldv16i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [7,0,16,8,16,13,11,9,0,8,15,14,13,12,11,10]
+; AVX-NEXT: retq
+ %out = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>, i1 0)
+ ret <16 x i16> %out
+}
+
+define <16 x i16> @foldv16i16u() {
+; AVX-LABEL: foldv16i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [7,0,16,8,16,13,11,9,0,8,15,14,13,12,11,10]
+; AVX-NEXT: retq
+ %out = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>, i1 -1)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @foldv32i8() {
+; AVX-LABEL: foldv32i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2,1,0,8,8,0,0,0,0,0,0,0,0,6,5,5,1]
+; AVX-NEXT: retq
+ %out = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>, i1 0)
+ ret <32 x i8> %out
+}
+
+define <32 x i8> @foldv32i8u() {
+; AVX-LABEL: foldv32i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,8,0,8,5,3,1,0,0,7,6,5,4,3,2,1,0,8,8,0,0,0,0,0,0,0,0,6,5,5,1]
+; AVX-NEXT: retq
+ %out = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>, i1 -1)
+ ret <32 x i8> %out
+}
+
+declare <4 x i64> @llvm.ctlz.v4i64(<4 x i64>, i1)
+declare <8 x i32> @llvm.ctlz.v8i32(<8 x i32>, i1)
+declare <16 x i16> @llvm.ctlz.v16i16(<16 x i16>, i1)
+declare <32 x i8> @llvm.ctlz.v32i8(<32 x i8>, i1)
diff --git a/test/CodeGen/X86/vector-popcnt-128.ll b/test/CodeGen/X86/vector-popcnt-128.ll
new file mode 100644
index 000000000000..fef445de04ab
--- /dev/null
+++ b/test/CodeGen/X86/vector-popcnt-128.ll
@@ -0,0 +1,462 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <2 x i64> @testv2i64(<2 x i64> %in) {
+; SSE2-LABEL: testv2i64:
+; SSE2: # BB#0:
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlq $1, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: psubq %xmm1, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [3689348814741910323,3689348814741910323]
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: psrlq $2, %xmm0
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: paddq %xmm2, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlq $4, %xmm1
+; SSE2-NEXT: paddq %xmm0, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: pxor %xmm0, %xmm0
+; SSE2-NEXT: psadbw %xmm0, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv2i64:
+; SSE3: # BB#0:
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlq $1, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: psubq %xmm1, %xmm0
+; SSE3-NEXT: movdqa {{.*#+}} xmm1 = [3689348814741910323,3689348814741910323]
+; SSE3-NEXT: movdqa %xmm0, %xmm2
+; SSE3-NEXT: pand %xmm1, %xmm2
+; SSE3-NEXT: psrlq $2, %xmm0
+; SSE3-NEXT: pand %xmm1, %xmm0
+; SSE3-NEXT: paddq %xmm2, %xmm0
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlq $4, %xmm1
+; SSE3-NEXT: paddq %xmm0, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: pxor %xmm0, %xmm0
+; SSE3-NEXT: psadbw %xmm0, %xmm1
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv2i64:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT: movdqa %xmm0, %xmm2
+; SSSE3-NEXT: pand %xmm1, %xmm2
+; SSSE3-NEXT: movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT: movdqa %xmm3, %xmm4
+; SSSE3-NEXT: pshufb %xmm2, %xmm4
+; SSSE3-NEXT: psrlw $4, %xmm0
+; SSSE3-NEXT: pand %xmm1, %xmm0
+; SSSE3-NEXT: pshufb %xmm0, %xmm3
+; SSSE3-NEXT: paddb %xmm4, %xmm3
+; SSSE3-NEXT: pxor %xmm0, %xmm0
+; SSSE3-NEXT: psadbw %xmm3, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv2i64:
+; SSE41: # BB#0:
+; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm0, %xmm2
+; SSE41-NEXT: pand %xmm1, %xmm2
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT: movdqa %xmm3, %xmm4
+; SSE41-NEXT: pshufb %xmm2, %xmm4
+; SSE41-NEXT: psrlw $4, %xmm0
+; SSE41-NEXT: pand %xmm1, %xmm0
+; SSE41-NEXT: pshufb %xmm0, %xmm3
+; SSE41-NEXT: paddb %xmm4, %xmm3
+; SSE41-NEXT: pxor %xmm0, %xmm0
+; SSE41-NEXT: psadbw %xmm3, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm2
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX-NEXT: vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX-NEXT: vpsadbw %xmm0, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %in)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @testv4i32(<4 x i32> %in) {
+; SSE2-LABEL: testv4i32:
+; SSE2: # BB#0:
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrld $1, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: psubd %xmm1, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [858993459,858993459,858993459,858993459]
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: psrld $2, %xmm0
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrld $4, %xmm1
+; SSE2-NEXT: paddd %xmm0, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: pxor %xmm0, %xmm0
+; SSE2-NEXT: movdqa %xmm1, %xmm2
+; SSE2-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT: psadbw %xmm0, %xmm2
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT: psadbw %xmm0, %xmm1
+; SSE2-NEXT: packuswb %xmm2, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv4i32:
+; SSE3: # BB#0:
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrld $1, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: psubd %xmm1, %xmm0
+; SSE3-NEXT: movdqa {{.*#+}} xmm1 = [858993459,858993459,858993459,858993459]
+; SSE3-NEXT: movdqa %xmm0, %xmm2
+; SSE3-NEXT: pand %xmm1, %xmm2
+; SSE3-NEXT: psrld $2, %xmm0
+; SSE3-NEXT: pand %xmm1, %xmm0
+; SSE3-NEXT: paddd %xmm2, %xmm0
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrld $4, %xmm1
+; SSE3-NEXT: paddd %xmm0, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: pxor %xmm0, %xmm0
+; SSE3-NEXT: movdqa %xmm1, %xmm2
+; SSE3-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT: psadbw %xmm0, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: psadbw %xmm0, %xmm1
+; SSE3-NEXT: packuswb %xmm2, %xmm1
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv4i32:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT: movdqa %xmm0, %xmm3
+; SSSE3-NEXT: pand %xmm2, %xmm3
+; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT: movdqa %xmm1, %xmm4
+; SSSE3-NEXT: pshufb %xmm3, %xmm4
+; SSSE3-NEXT: psrlw $4, %xmm0
+; SSSE3-NEXT: pand %xmm2, %xmm0
+; SSSE3-NEXT: pshufb %xmm0, %xmm1
+; SSSE3-NEXT: paddb %xmm4, %xmm1
+; SSSE3-NEXT: pxor %xmm0, %xmm0
+; SSSE3-NEXT: movdqa %xmm1, %xmm2
+; SSSE3-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT: psadbw %xmm0, %xmm2
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT: psadbw %xmm0, %xmm1
+; SSSE3-NEXT: packuswb %xmm2, %xmm1
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv4i32:
+; SSE41: # BB#0:
+; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm0, %xmm3
+; SSE41-NEXT: pand %xmm2, %xmm3
+; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT: movdqa %xmm1, %xmm4
+; SSE41-NEXT: pshufb %xmm3, %xmm4
+; SSE41-NEXT: psrlw $4, %xmm0
+; SSE41-NEXT: pand %xmm2, %xmm0
+; SSE41-NEXT: pshufb %xmm0, %xmm1
+; SSE41-NEXT: paddb %xmm4, %xmm1
+; SSE41-NEXT: pxor %xmm0, %xmm0
+; SSE41-NEXT: movdqa %xmm1, %xmm2
+; SSE41-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE41-NEXT: psadbw %xmm0, %xmm2
+; SSE41-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE41-NEXT: psadbw %xmm0, %xmm1
+; SSE41-NEXT: packuswb %xmm2, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm2
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX-NEXT: vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX-NEXT: vpsadbw %xmm2, %xmm1, %xmm2
+; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX-NEXT: vpsadbw %xmm0, %xmm1, %xmm0
+; AVX-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %in)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @testv8i16(<8 x i16> %in) {
+; SSE2-LABEL: testv8i16:
+; SSE2: # BB#0:
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlw $1, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: psubw %xmm1, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [13107,13107,13107,13107,13107,13107,13107,13107]
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: psrlw $2, %xmm0
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: paddw %xmm2, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlw $4, %xmm1
+; SSE2-NEXT: paddw %xmm0, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: psllw $8, %xmm0
+; SSE2-NEXT: paddb %xmm1, %xmm0
+; SSE2-NEXT: psrlw $8, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv8i16:
+; SSE3: # BB#0:
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlw $1, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: psubw %xmm1, %xmm0
+; SSE3-NEXT: movdqa {{.*#+}} xmm1 = [13107,13107,13107,13107,13107,13107,13107,13107]
+; SSE3-NEXT: movdqa %xmm0, %xmm2
+; SSE3-NEXT: pand %xmm1, %xmm2
+; SSE3-NEXT: psrlw $2, %xmm0
+; SSE3-NEXT: pand %xmm1, %xmm0
+; SSE3-NEXT: paddw %xmm2, %xmm0
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlw $4, %xmm1
+; SSE3-NEXT: paddw %xmm0, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: psllw $8, %xmm0
+; SSE3-NEXT: paddb %xmm1, %xmm0
+; SSE3-NEXT: psrlw $8, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv8i16:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT: movdqa %xmm0, %xmm2
+; SSSE3-NEXT: pand %xmm1, %xmm2
+; SSSE3-NEXT: movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT: movdqa %xmm3, %xmm4
+; SSSE3-NEXT: pshufb %xmm2, %xmm4
+; SSSE3-NEXT: psrlw $4, %xmm0
+; SSSE3-NEXT: pand %xmm1, %xmm0
+; SSSE3-NEXT: pshufb %xmm0, %xmm3
+; SSSE3-NEXT: paddb %xmm4, %xmm3
+; SSSE3-NEXT: movdqa %xmm3, %xmm0
+; SSSE3-NEXT: psllw $8, %xmm0
+; SSSE3-NEXT: paddb %xmm3, %xmm0
+; SSSE3-NEXT: psrlw $8, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv8i16:
+; SSE41: # BB#0:
+; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm0, %xmm2
+; SSE41-NEXT: pand %xmm1, %xmm2
+; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT: movdqa %xmm3, %xmm4
+; SSE41-NEXT: pshufb %xmm2, %xmm4
+; SSE41-NEXT: psrlw $4, %xmm0
+; SSE41-NEXT: pand %xmm1, %xmm0
+; SSE41-NEXT: pshufb %xmm0, %xmm3
+; SSE41-NEXT: paddb %xmm4, %xmm3
+; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: psllw $8, %xmm0
+; SSE41-NEXT: paddb %xmm3, %xmm0
+; SSE41-NEXT: psrlw $8, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm2
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX-NEXT: vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT: vpsllw $8, %xmm0, %xmm1
+; AVX-NEXT: vpaddb %xmm0, %xmm1, %xmm0
+; AVX-NEXT: vpsrlw $8, %xmm0, %xmm0
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %in)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @testv16i8(<16 x i8> %in) {
+; SSE2-LABEL: testv16i8:
+; SSE2: # BB#0:
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlw $1, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: psubb %xmm1, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: psrlw $2, %xmm0
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: paddb %xmm2, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrlw $4, %xmm1
+; SSE2-NEXT: paddb %xmm0, %xmm1
+; SSE2-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv16i8:
+; SSE3: # BB#0:
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlw $1, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: psubb %xmm1, %xmm0
+; SSE3-NEXT: movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT: movdqa %xmm0, %xmm2
+; SSE3-NEXT: pand %xmm1, %xmm2
+; SSE3-NEXT: psrlw $2, %xmm0
+; SSE3-NEXT: pand %xmm1, %xmm0
+; SSE3-NEXT: paddb %xmm2, %xmm0
+; SSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE3-NEXT: psrlw $4, %xmm1
+; SSE3-NEXT: paddb %xmm0, %xmm1
+; SSE3-NEXT: pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv16i8:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT: movdqa %xmm0, %xmm3
+; SSSE3-NEXT: pand %xmm2, %xmm3
+; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT: movdqa %xmm1, %xmm4
+; SSSE3-NEXT: pshufb %xmm3, %xmm4
+; SSSE3-NEXT: psrlw $4, %xmm0
+; SSSE3-NEXT: pand %xmm2, %xmm0
+; SSSE3-NEXT: pshufb %xmm0, %xmm1
+; SSSE3-NEXT: paddb %xmm4, %xmm1
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv16i8:
+; SSE41: # BB#0:
+; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT: movdqa %xmm0, %xmm3
+; SSE41-NEXT: pand %xmm2, %xmm3
+; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT: movdqa %xmm1, %xmm4
+; SSE41-NEXT: pshufb %xmm3, %xmm4
+; SSE41-NEXT: psrlw $4, %xmm0
+; SSE41-NEXT: pand %xmm2, %xmm0
+; SSE41-NEXT: pshufb %xmm0, %xmm1
+; SSE41-NEXT: paddb %xmm4, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm2
+; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX-NEXT: vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX-NEXT: vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %in)
+ ret <16 x i8> %out
+}
+
+define <2 x i64> @foldv2i64() {
+; SSE-LABEL: foldv2i64:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,64]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [1,64]
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> <i64 256, i64 -1>)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @foldv4i32() {
+; SSE-LABEL: foldv4i32:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,32,0,8]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [1,32,0,8]
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> <i32 256, i32 -1, i32 0, i32 255>)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @foldv8i16() {
+; SSE-LABEL: foldv8i16:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,16,0,8,0,3,2,3]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [1,16,0,8,0,3,2,3]
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88>)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @foldv16i8() {
+; SSE-LABEL: foldv16i8:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [0,8,0,8,0,3,2,3,7,7,1,1,1,1,1,1]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [0,8,0,8,0,3,2,3,7,7,1,1,1,1,1,1]
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32>)
+ ret <16 x i8> %out
+}
+
+declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>)
+declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>)
+declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16>)
+declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>)
diff --git a/test/CodeGen/X86/vector-popcnt-256.ll b/test/CodeGen/X86/vector-popcnt-256.ll
new file mode 100644
index 000000000000..7ce4f712483a
--- /dev/null
+++ b/test/CodeGen/X86/vector-popcnt-256.ll
@@ -0,0 +1,220 @@
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <4 x i64> @testv4i64(<4 x i64> %in) {
+; AVX1-LABEL: testv4i64:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpsadbw %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm5
+; AVX1-NEXT: vpshufb %xmm5, %xmm4, %xmm5
+; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX1-NEXT: vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT: vpsadbw %xmm0, %xmm3, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv4i64:
+; AVX2: # BB#0:
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1
+; AVX2-NEXT: vpsadbw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: retq
+ %out = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %in)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @testv8i32(<8 x i32> %in) {
+; AVX1-LABEL: testv8i32:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm5 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX1-NEXT: vpsadbw %xmm5, %xmm3, %xmm5
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX1-NEXT: vpsadbw %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vpackuswb %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm5
+; AVX1-NEXT: vpshufb %xmm5, %xmm4, %xmm5
+; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX1-NEXT: vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; AVX1-NEXT: vpsadbw %xmm2, %xmm3, %xmm2
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX1-NEXT: vpsadbw %xmm0, %xmm3, %xmm0
+; AVX1-NEXT: vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv8i32:
+; AVX2: # BB#0:
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1
+; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT: vpsadbw %ymm2, %ymm1, %ymm2
+; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT: vpsadbw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %in)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @testv16i16(<16 x i16> %in) {
+; AVX1-LABEL: testv16i16:
+; AVX1: # BB#0:
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT: vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT: vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT: vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT: vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT: vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv16i16:
+; AVX2: # BB#0:
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %in)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @testv32i8(<32 x i8> %in) {
+; AVX1-LABEL: testv32i8:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT: vandps %xmm2, %xmm1, %xmm3
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT: vpsrlw $4, %xmm1, %xmm1
+; AVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vandps %xmm2, %xmm0, %xmm3
+; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT: vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX1-NEXT: vpaddb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv32i8:
+; AVX2: # BB#0:
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT: vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %in)
+ ret <32 x i8> %out
+}
+
+define <4 x i64> @foldv4i64() {
+; AVX-LABEL: foldv4i64:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [1,64,0,8]
+; AVX-NEXT: retq
+ %out = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @foldv8i32() {
+; AVX-LABEL: foldv8i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [1,32,0,8,16,3,2,3]
+; AVX-NEXT: retq
+ %out = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @foldv16i16() {
+; AVX-LABEL: foldv16i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [1,16,0,8,0,3,2,3,15,7,1,1,1,1,1,1]
+; AVX-NEXT: retq
+ %out = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @foldv32i8() {
+; AVX-LABEL: foldv32i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [0,8,0,8,0,3,2,3,7,7,1,1,1,1,1,1,1,1,0,0,1,2,3,4,5,6,7,8,2,2,3,7]
+; AVX-NEXT: retq
+ %out = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>)
+ ret <32 x i8> %out
+}
+
+declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>)
+declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>)
+declare <16 x i16> @llvm.ctpop.v16i16(<16 x i16>)
+declare <32 x i8> @llvm.ctpop.v32i8(<32 x i8>)
diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 1b42a637907c..944ec4b8d3ac 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -843,7 +843,6 @@ define <4 x i64> @insert_mem_and_zero_v4i64(i64* %ptr) {
define <4 x double> @insert_reg_and_zero_v4f64(double %a) {
; ALL-LABEL: insert_reg_and_zero_v4f64:
; ALL: # BB#0:
-; ALL-NEXT: # kill: XMM0<def> XMM0<kill> YMM0<def>
; ALL-NEXT: vxorpd %ymm1, %ymm1, %ymm1
; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
; ALL-NEXT: retq
diff --git a/test/CodeGen/X86/vector-shuffle-512-v8.ll b/test/CodeGen/X86/vector-shuffle-512-v8.ll
index 62d4af7809b6..8dc76231856a 100644
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -15,9 +15,8 @@ define <8 x double> @shuffle_v8f64_00000000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00000010(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00000010:
; ALL: # BB#0:
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
ret <8 x double> %shuffle
@@ -26,9 +25,8 @@ define <8 x double> @shuffle_v8f64_00000010(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00000200(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00000200:
; ALL: # BB#0:
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -37,9 +35,8 @@ define <8 x double> @shuffle_v8f64_00000200(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00003000(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00003000:
; ALL: # BB#0:
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,0,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -48,11 +45,8 @@ define <8 x double> @shuffle_v8f64_00003000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00040000(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00040000:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vbroadcastsd %xmm1, %ymm1
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -61,11 +55,8 @@ define <8 x double> @shuffle_v8f64_00040000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00500000(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00500000:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -74,11 +65,8 @@ define <8 x double> @shuffle_v8f64_00500000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_06000000(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_06000000:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2],ymm0[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,0]
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -87,11 +75,11 @@ define <8 x double> @shuffle_v8f64_06000000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_70000000(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_70000000:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,0,0,0]
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpxord %zmm1, %zmm1, %zmm1
+; ALL-NEXT: movl $7, %eax
+; ALL-NEXT: vpinsrq $0, %rax, %xmm1, %xmm2
+; ALL-NEXT: vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x double> %shuffle
@@ -100,10 +88,7 @@ define <8 x double> @shuffle_v8f64_70000000(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_01014545(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_01014545:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermpd $68, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
ret <8 x double> %shuffle
@@ -112,9 +97,8 @@ define <8 x double> @shuffle_v8f64_01014545(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00112233(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00112233:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,0,1,1]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,3,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3>
ret <8 x double> %shuffle
@@ -123,9 +107,8 @@ define <8 x double> @shuffle_v8f64_00112233(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00001111(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00001111:
; ALL: # BB#0:
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,1,1]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
ret <8 x double> %shuffle
@@ -134,11 +117,7 @@ define <8 x double> @shuffle_v8f64_00001111(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_81a3c5e7(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_81a3c5e7:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vshufpd $170, %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7>
ret <8 x double> %shuffle
@@ -147,10 +126,9 @@ define <8 x double> @shuffle_v8f64_81a3c5e7(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_08080808(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_08080808:
; ALL: # BB#0:
-; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vbroadcastsd %xmm1, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 0, i32 8, i32 0, i32 8>
ret <8 x double> %shuffle
@@ -159,15 +137,9 @@ define <8 x double> @shuffle_v8f64_08080808(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_08084c4c(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_08084c4c:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vinsertf128 $1, %xmm2, %ymm2, %ymm2
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vbroadcastsd %xmm3, %ymm3
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[3]
-; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vbroadcastsd %xmm1, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 4, i32 12, i32 4, i32 12>
ret <8 x double> %shuffle
@@ -176,13 +148,9 @@ define <8 x double> @shuffle_v8f64_08084c4c(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_8823cc67(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_8823cc67:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vbroadcastsd %xmm3, %ymm3
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; ALL-NEXT: vbroadcastsd %xmm1, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 8, i32 2, i32 3, i32 12, i32 12, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -191,13 +159,9 @@ define <8 x double> @shuffle_v8f64_8823cc67(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_9832dc76(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_9832dc76:
; ALL: # BB#0:
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,0,3,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm1
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
ret <8 x double> %shuffle
@@ -206,13 +170,9 @@ define <8 x double> @shuffle_v8f64_9832dc76(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_9810dc54(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_9810dc54:
; ALL: # BB#0:
-; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm2
-; ALL-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,0,3,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm1
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
ret <8 x double> %shuffle
@@ -221,15 +181,9 @@ define <8 x double> @shuffle_v8f64_9810dc54(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_08194c5d(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_08194c5d:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm3
-; ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
ret <8 x double> %shuffle
@@ -238,15 +192,9 @@ define <8 x double> @shuffle_v8f64_08194c5d(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_2a3b6e7f(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_2a3b6e7f:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm3
-; ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[2,1,3,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
ret <8 x double> %shuffle
@@ -255,13 +203,9 @@ define <8 x double> @shuffle_v8f64_2a3b6e7f(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_08192a3b(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_08192a3b:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm1[0,2,2,3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[2,1,3,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
ret <8 x double> %shuffle
@@ -270,11 +214,9 @@ define <8 x double> @shuffle_v8f64_08192a3b(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_08991abb(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_08991abb:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm1[0,0,1,1]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,2,3,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
ret <8 x double> %shuffle
@@ -283,12 +225,9 @@ define <8 x double> @shuffle_v8f64_08991abb(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_091b2d3f(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_091b2d3f:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[2,1,3,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 9, i32 1, i32 11, i32 2, i32 13, i32 3, i32 15>
ret <8 x double> %shuffle
@@ -297,11 +236,9 @@ define <8 x double> @shuffle_v8f64_091b2d3f(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_09ab1def(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_09ab1def:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermilpd {{.*#+}} ymm3 = ymm0[1,0,2,2]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
ret <8 x double> %shuffle
@@ -310,10 +247,7 @@ define <8 x double> @shuffle_v8f64_09ab1def(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00014445(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00014445:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $64, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
ret <8 x double> %shuffle
@@ -322,10 +256,7 @@ define <8 x double> @shuffle_v8f64_00014445(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00204464(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00204464:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $32, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
ret <8 x double> %shuffle
@@ -334,10 +265,7 @@ define <8 x double> @shuffle_v8f64_00204464(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_03004744(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_03004744:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $12, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -346,10 +274,7 @@ define <8 x double> @shuffle_v8f64_03004744(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10005444(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10005444:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $1, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -358,10 +283,7 @@ define <8 x double> @shuffle_v8f64_10005444(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_22006644(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_22006644:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $10, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -370,10 +292,7 @@ define <8 x double> @shuffle_v8f64_22006644(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_33307774(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_33307774:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $63, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
ret <8 x double> %shuffle
@@ -382,10 +301,7 @@ define <8 x double> @shuffle_v8f64_33307774(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_32107654(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_32107654:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermpd $27, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
ret <8 x double> %shuffle
@@ -394,10 +310,7 @@ define <8 x double> @shuffle_v8f64_32107654(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00234467:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $136, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -406,10 +319,7 @@ define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00224466(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00224466:
; ALL: # BB#0:
-; ALL-NEXT: vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $0, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
ret <8 x double> %shuffle
@@ -418,10 +328,7 @@ define <8 x double> @shuffle_v8f64_00224466(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10325476(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10325476:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $85, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
ret <8 x double> %shuffle
@@ -430,10 +337,7 @@ define <8 x double> @shuffle_v8f64_10325476(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_11335577(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_11335577:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $255, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
ret <8 x double> %shuffle
@@ -442,10 +346,7 @@ define <8 x double> @shuffle_v8f64_11335577(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10235467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10235467:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,0,2,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $153, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -454,10 +355,7 @@ define <8 x double> @shuffle_v8f64_10235467(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10225466(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10225466:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,0,2,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $17, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
ret <8 x double> %shuffle
@@ -466,10 +364,8 @@ define <8 x double> @shuffle_v8f64_10225466(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00015444(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00015444:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 5, i32 4, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -478,10 +374,8 @@ define <8 x double> @shuffle_v8f64_00015444(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00204644(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00204644:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 6, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -490,10 +384,8 @@ define <8 x double> @shuffle_v8f64_00204644(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_03004474(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_03004474:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,3,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 4, i32 7, i32 4>
ret <8 x double> %shuffle
@@ -502,10 +394,8 @@ define <8 x double> @shuffle_v8f64_03004474(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10004444(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10004444:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -514,10 +404,8 @@ define <8 x double> @shuffle_v8f64_10004444(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_22006446(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_22006446:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,0,0,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 4, i32 4, i32 6>
ret <8 x double> %shuffle
@@ -526,10 +414,8 @@ define <8 x double> @shuffle_v8f64_22006446(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_33307474(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_33307474:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,0,3,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 4, i32 7, i32 4>
ret <8 x double> %shuffle
@@ -538,9 +424,8 @@ define <8 x double> @shuffle_v8f64_33307474(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_32104567(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_32104567:
; ALL: # BB#0:
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -549,10 +434,8 @@ define <8 x double> @shuffle_v8f64_32104567(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00236744(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00236744:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 6, i32 7, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -561,10 +444,8 @@ define <8 x double> @shuffle_v8f64_00236744(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00226644(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00226644:
; ALL: # BB#0:
-; ALL-NEXT: vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 6, i32 6, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -573,9 +454,7 @@ define <8 x double> @shuffle_v8f64_00226644(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_10324567(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10324567:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $165, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -584,9 +463,7 @@ define <8 x double> @shuffle_v8f64_10324567(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_11334567(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_11334567:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $175, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -595,9 +472,7 @@ define <8 x double> @shuffle_v8f64_11334567(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_01235467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_01235467:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $154, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -606,9 +481,7 @@ define <8 x double> @shuffle_v8f64_01235467(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_01235466:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $26, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 6>
ret <8 x double> %shuffle
@@ -617,10 +490,8 @@ define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_002u6u44(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_002u6u44:
; ALL: # BB#0:
-; ALL-NEXT: vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,0,0]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 undef, i32 6, i32 undef, i32 4, i32 4>
ret <8 x double> %shuffle
@@ -629,10 +500,8 @@ define <8 x double> @shuffle_v8f64_002u6u44(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_00uu66uu(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00uu66uu:
; ALL: # BB#0:
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermpd %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 undef, i32 undef, i32 6, i32 6, i32 undef, i32 undef>
ret <8 x double> %shuffle
@@ -641,9 +510,7 @@ define <8 x double> @shuffle_v8f64_00uu66uu(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_103245uu(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_103245uu:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $37, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
ret <8 x double> %shuffle
@@ -652,9 +519,7 @@ define <8 x double> @shuffle_v8f64_103245uu(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_1133uu67(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_1133uu67:
; ALL: # BB#0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $143, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
ret <8 x double> %shuffle
@@ -663,9 +528,7 @@ define <8 x double> @shuffle_v8f64_1133uu67(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_0uu354uu(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_0uu354uu:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $24, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 undef, i32 undef, i32 3, i32 5, i32 4, i32 undef, i32 undef>
ret <8 x double> %shuffle
@@ -674,9 +537,7 @@ define <8 x double> @shuffle_v8f64_0uu354uu(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_uuu3uu66(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_uuu3uu66:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vmovddup {{.*#+}} ymm1 = ymm1[0,0,2,2]
-; ALL-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $8, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
ret <8 x double> %shuffle
@@ -685,16 +546,9 @@ define <8 x double> @shuffle_v8f64_uuu3uu66(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_c348cda0(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_c348cda0:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm2[0,1]
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vbroadcastsd %xmm1, %ymm4
-; ALL-NEXT: vblendpd {{.*#+}} ymm4 = ymm3[0,1,2],ymm4[3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm2 = ymm4[0],ymm2[1,2],ymm4[3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3]
-; ALL-NEXT: vbroadcastsd %xmm0, %ymm0
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 12, i32 3, i32 4, i32 8, i32 12, i32 13, i32 10, i32 0>
ret <8 x double> %shuffle
@@ -703,17 +557,9 @@ define <8 x double> @shuffle_v8f64_c348cda0(<8 x double> %a, <8 x double> %b) {
define <8 x double> @shuffle_v8f64_f511235a(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_f511235a:
; ALL: # BB#0:
-; ALL-NEXT: vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vblendpd {{.*#+}} ymm3 = ymm0[0],ymm2[1],ymm0[2,3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[2,3,1,3]
-; ALL-NEXT: vmovddup {{.*#+}} ymm4 = ymm1[0,0,2,2]
-; ALL-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3]
-; ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,1]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
-; ALL-NEXT: vextractf64x4 $1, %zmm1, %ymm1
-; ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,1,2,3]
-; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; ALL-NEXT: vinsertf64x4 $1, %ymm3, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2pd %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 15, i32 5, i32 1, i32 1, i32 2, i32 3, i32 5, i32 10>
ret <8 x double> %shuffle
@@ -731,9 +577,8 @@ define <8 x i64> @shuffle_v8i64_00000000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00000010(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00000010:
; ALL: # BB#0:
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
ret <8 x i64> %shuffle
@@ -742,9 +587,8 @@ define <8 x i64> @shuffle_v8i64_00000010(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00000200(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00000200:
; ALL: # BB#0:
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -753,9 +597,8 @@ define <8 x i64> @shuffle_v8i64_00000200(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00003000(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00003000:
; ALL: # BB#0:
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,0,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -764,11 +607,8 @@ define <8 x i64> @shuffle_v8i64_00003000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00040000(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00040000:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpbroadcastq %xmm1, %ymm1
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm0
-; ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -777,11 +617,8 @@ define <8 x i64> @shuffle_v8i64_00040000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00500000(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00500000:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -790,11 +627,8 @@ define <8 x i64> @shuffle_v8i64_00500000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_06000000(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_06000000:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,0,0]
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -803,11 +637,11 @@ define <8 x i64> @shuffle_v8i64_06000000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_70000000(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_70000000:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[3,0,0,0]
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpxord %zmm1, %zmm1, %zmm1
+; ALL-NEXT: movl $7, %eax
+; ALL-NEXT: vpinsrq $0, %rax, %xmm1, %xmm2
+; ALL-NEXT: vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
ret <8 x i64> %shuffle
@@ -816,10 +650,7 @@ define <8 x i64> @shuffle_v8i64_70000000(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_01014545(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_01014545:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vinserti128 $1, %xmm1, %ymm1, %ymm1
-; ALL-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermq $68, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
ret <8 x i64> %shuffle
@@ -828,9 +659,8 @@ define <8 x i64> @shuffle_v8i64_01014545(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00112233(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00112233:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,1,1]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,3,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3>
ret <8 x i64> %shuffle
@@ -839,9 +669,8 @@ define <8 x i64> @shuffle_v8i64_00112233(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00001111(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00001111:
; ALL: # BB#0:
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,1,1,1]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
ret <8 x i64> %shuffle
@@ -850,11 +679,7 @@ define <8 x i64> @shuffle_v8i64_00001111(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_81a3c5e7(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_81a3c5e7:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vshufpd $170, %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7>
ret <8 x i64> %shuffle
@@ -863,10 +688,9 @@ define <8 x i64> @shuffle_v8i64_81a3c5e7(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_08080808(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_08080808:
; ALL: # BB#0:
-; ALL-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vpbroadcastq %xmm1, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 0, i32 8, i32 0, i32 8>
ret <8 x i64> %shuffle
@@ -875,15 +699,9 @@ define <8 x i64> @shuffle_v8i64_08080808(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_08084c4c(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_08084c4c:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vinserti128 $1, %xmm2, %ymm2, %ymm2
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vpbroadcastq %xmm3, %ymm3
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
-; ALL-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; ALL-NEXT: vpbroadcastq %xmm1, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 4, i32 12, i32 4, i32 12>
ret <8 x i64> %shuffle
@@ -892,13 +710,9 @@ define <8 x i64> @shuffle_v8i64_08084c4c(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_8823cc67(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_8823cc67:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm2
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT: vpbroadcastq %xmm3, %ymm3
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; ALL-NEXT: vpbroadcastq %xmm1, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 8, i32 8, i32 2, i32 3, i32 12, i32 12, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -907,13 +721,9 @@ define <8 x i64> @shuffle_v8i64_8823cc67(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_9832dc76(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_9832dc76:
; ALL: # BB#0:
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm1
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
ret <8 x i64> %shuffle
@@ -922,13 +732,9 @@ define <8 x i64> @shuffle_v8i64_9832dc76(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_9810dc54(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_9810dc54:
; ALL: # BB#0:
-; ALL-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm2
-; ALL-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm1
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
-; ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
ret <8 x i64> %shuffle
@@ -937,15 +743,9 @@ define <8 x i64> @shuffle_v8i64_9810dc54(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_08194c5d(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_08194c5d:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
-; ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,1,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
ret <8 x i64> %shuffle
@@ -954,15 +754,9 @@ define <8 x i64> @shuffle_v8i64_08194c5d(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_2a3b6e7f(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_2a3b6e7f:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm3
-; ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,1,3,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
ret <8 x i64> %shuffle
@@ -971,13 +765,9 @@ define <8 x i64> @shuffle_v8i64_2a3b6e7f(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_08192a3b(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_08192a3b:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm2 = ymm1[0,2,2,3]
-; ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,1,3,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
ret <8 x i64> %shuffle
@@ -986,11 +776,9 @@ define <8 x i64> @shuffle_v8i64_08192a3b(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_08991abb(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_08991abb:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm2 = ymm1[0,0,1,1]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,2,3,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
ret <8 x i64> %shuffle
@@ -999,12 +787,9 @@ define <8 x i64> @shuffle_v8i64_08991abb(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_091b2d3f(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_091b2d3f:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,1,3,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm1, %zmm0, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 9, i32 1, i32 11, i32 2, i32 13, i32 3, i32 15>
ret <8 x i64> %shuffle
@@ -1013,11 +798,9 @@ define <8 x i64> @shuffle_v8i64_091b2d3f(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_09ab1def(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_09ab1def:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpshufd {{.*#+}} ymm3 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
ret <8 x i64> %shuffle
@@ -1026,10 +809,7 @@ define <8 x i64> @shuffle_v8i64_09ab1def(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00014445(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00014445:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $64, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
ret <8 x i64> %shuffle
@@ -1038,10 +818,7 @@ define <8 x i64> @shuffle_v8i64_00014445(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00204464(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00204464:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,2,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $32, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
ret <8 x i64> %shuffle
@@ -1050,10 +827,7 @@ define <8 x i64> @shuffle_v8i64_00204464(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_03004744(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_03004744:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,3,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $12, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1062,10 +836,7 @@ define <8 x i64> @shuffle_v8i64_03004744(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10005444(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10005444:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,0,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $1, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1074,10 +845,7 @@ define <8 x i64> @shuffle_v8i64_10005444(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_22006644(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_22006644:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $10, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1086,10 +854,7 @@ define <8 x i64> @shuffle_v8i64_22006644(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_33307774(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_33307774:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,3,3,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $63, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
ret <8 x i64> %shuffle
@@ -1098,10 +863,7 @@ define <8 x i64> @shuffle_v8i64_33307774(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_32107654(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_32107654:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,1,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermq $27, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
ret <8 x i64> %shuffle
@@ -1110,10 +872,7 @@ define <8 x i64> @shuffle_v8i64_32107654(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00234467(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00234467:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,2,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $136, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1122,10 +881,7 @@ define <8 x i64> @shuffle_v8i64_00234467(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00224466(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00224466:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $0, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
ret <8 x i64> %shuffle
@@ -1134,10 +890,7 @@ define <8 x i64> @shuffle_v8i64_00224466(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10325476(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10325476:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $85, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
ret <8 x i64> %shuffle
@@ -1146,10 +899,7 @@ define <8 x i64> @shuffle_v8i64_10325476(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_11335577(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_11335577:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $255, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
ret <8 x i64> %shuffle
@@ -1158,10 +908,7 @@ define <8 x i64> @shuffle_v8i64_11335577(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10235467(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10235467:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[1,0,2,3]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,0,2,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $153, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1170,10 +917,7 @@ define <8 x i64> @shuffle_v8i64_10235467(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10225466(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10225466:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[1,0,2,2]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,0,2,2]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $17, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
ret <8 x i64> %shuffle
@@ -1182,10 +926,8 @@ define <8 x i64> @shuffle_v8i64_10225466(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00015444(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00015444:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,0,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 5, i32 4, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1194,10 +936,8 @@ define <8 x i64> @shuffle_v8i64_00015444(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00204644(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00204644:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 6, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1206,10 +946,8 @@ define <8 x i64> @shuffle_v8i64_00204644(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_03004474(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_03004474:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,3,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 4, i32 7, i32 4>
ret <8 x i64> %shuffle
@@ -1218,10 +956,8 @@ define <8 x i64> @shuffle_v8i64_03004474(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10004444(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10004444:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1230,10 +966,8 @@ define <8 x i64> @shuffle_v8i64_10004444(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_22006446(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_22006446:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,0,0,2]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 4, i32 4, i32 6>
ret <8 x i64> %shuffle
@@ -1242,10 +976,8 @@ define <8 x i64> @shuffle_v8i64_22006446(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_33307474(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_33307474:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,0,3,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 4, i32 7, i32 4>
ret <8 x i64> %shuffle
@@ -1254,9 +986,8 @@ define <8 x i64> @shuffle_v8i64_33307474(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_32104567(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_32104567:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1265,10 +996,8 @@ define <8 x i64> @shuffle_v8i64_32104567(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00236744(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00236744:
; ALL: # BB#0:
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 6, i32 7, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1277,10 +1006,8 @@ define <8 x i64> @shuffle_v8i64_00236744(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00226644(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00226644:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 6, i32 6, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1289,9 +1016,7 @@ define <8 x i64> @shuffle_v8i64_00226644(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_10324567(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_10324567:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $165, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1300,9 +1025,7 @@ define <8 x i64> @shuffle_v8i64_10324567(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_11334567(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_11334567:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $175, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1311,9 +1034,7 @@ define <8 x i64> @shuffle_v8i64_11334567(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_01235467(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_01235467:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,0,2,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $154, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1322,9 +1043,7 @@ define <8 x i64> @shuffle_v8i64_01235467(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_01235466(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_01235466:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; ALL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $26, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 6>
ret <8 x i64> %shuffle
@@ -1333,10 +1052,8 @@ define <8 x i64> @shuffle_v8i64_01235466(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_002u6u44(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_002u6u44:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,0,0]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 undef, i32 6, i32 undef, i32 4, i32 4>
ret <8 x i64> %shuffle
@@ -1345,10 +1062,8 @@ define <8 x i64> @shuffle_v8i64_002u6u44(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_00uu66uu(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_00uu66uu:
; ALL: # BB#0:
-; ALL-NEXT: vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm1
+; ALL-NEXT: vpermq %zmm0, %zmm1, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 undef, i32 undef, i32 6, i32 6, i32 undef, i32 undef>
ret <8 x i64> %shuffle
@@ -1357,9 +1072,7 @@ define <8 x i64> @shuffle_v8i64_00uu66uu(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_103245uu(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_103245uu:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $37, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
ret <8 x i64> %shuffle
@@ -1368,9 +1081,7 @@ define <8 x i64> @shuffle_v8i64_103245uu(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_1133uu67(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_1133uu67:
; ALL: # BB#0:
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT: vpermilpd $143, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
ret <8 x i64> %shuffle
@@ -1379,9 +1090,7 @@ define <8 x i64> @shuffle_v8i64_1133uu67(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_0uu354uu(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_0uu354uu:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $24, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 undef, i32 undef, i32 3, i32 5, i32 4, i32 undef, i32 undef>
ret <8 x i64> %shuffle
@@ -1390,9 +1099,7 @@ define <8 x i64> @shuffle_v8i64_0uu354uu(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_uuu3uu66(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_uuu3uu66:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; ALL-NEXT: vpermilpd $8, %zmm0, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
ret <8 x i64> %shuffle
@@ -1401,15 +1108,9 @@ define <8 x i64> @shuffle_v8i64_uuu3uu66(<8 x i64> %a, <8 x i64> %b) {
define <8 x i64> @shuffle_v8i64_6caa87e5(<8 x i64> %a, <8 x i64> %b) {
; ALL-LABEL: shuffle_v8i64_6caa87e5:
; ALL: # BB#0:
-; ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; ALL-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
-; ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
-; ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; ALL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; ALL-NEXT: vmovdqa64 {{.*}}(%rip), %zmm2
+; ALL-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
+; ALL-NEXT: vmovaps %zmm2, %zmm0
; ALL-NEXT: retq
%shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 6, i32 12, i32 10, i32 10, i32 8, i32 7, i32 14, i32 5>
ret <8 x i64> %shuffle
diff --git a/test/CodeGen/X86/vector-tzcnt-128.ll b/test/CodeGen/X86/vector-tzcnt-128.ll
new file mode 100644
index 000000000000..422fe052d38b
--- /dev/null
+++ b/test/CodeGen/X86/vector-tzcnt-128.ll
@@ -0,0 +1,1788 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <2 x i64> @testv2i64(<2 x i64> %in) {
+; SSE2-LABEL: testv2i64:
+; SSE2: # BB#0:
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsfq %rax, %rax
+; SSE2-NEXT: movl $64, %ecx
+; SSE2-NEXT: cmoveq %rcx, %rax
+; SSE2-NEXT: movd %rax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsfq %rax, %rax
+; SSE2-NEXT: cmoveq %rcx, %rax
+; SSE2-NEXT: movd %rax, %xmm0
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv2i64:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsfq %rax, %rax
+; SSE3-NEXT: movl $64, %ecx
+; SSE3-NEXT: cmoveq %rcx, %rax
+; SSE3-NEXT: movd %rax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsfq %rax, %rax
+; SSE3-NEXT: cmoveq %rcx, %rax
+; SSE3-NEXT: movd %rax, %xmm0
+; SSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv2i64:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsfq %rax, %rax
+; SSSE3-NEXT: movl $64, %ecx
+; SSSE3-NEXT: cmoveq %rcx, %rax
+; SSSE3-NEXT: movd %rax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsfq %rax, %rax
+; SSSE3-NEXT: cmoveq %rcx, %rax
+; SSSE3-NEXT: movd %rax, %xmm0
+; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv2i64:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrq $1, %xmm0, %rax
+; SSE41-NEXT: bsfq %rax, %rax
+; SSE41-NEXT: movl $64, %ecx
+; SSE41-NEXT: cmoveq %rcx, %rax
+; SSE41-NEXT: movd %rax, %xmm1
+; SSE41-NEXT: movd %xmm0, %rax
+; SSE41-NEXT: bsfq %rax, %rax
+; SSE41-NEXT: cmoveq %rcx, %rax
+; SSE41-NEXT: movd %rax, %xmm0
+; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrq $1, %xmm0, %rax
+; AVX-NEXT: bsfq %rax, %rax
+; AVX-NEXT: movl $64, %ecx
+; AVX-NEXT: cmoveq %rcx, %rax
+; AVX-NEXT: vmovq %rax, %xmm1
+; AVX-NEXT: vmovq %xmm0, %rax
+; AVX-NEXT: bsfq %rax, %rax
+; AVX-NEXT: cmoveq %rcx, %rax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %in, i1 0)
+ ret <2 x i64> %out
+}
+
+define <2 x i64> @testv2i64u(<2 x i64> %in) {
+; SSE2-LABEL: testv2i64u:
+; SSE2: # BB#0:
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsfq %rax, %rax
+; SSE2-NEXT: movd %rax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %rax
+; SSE2-NEXT: bsfq %rax, %rax
+; SSE2-NEXT: movd %rax, %xmm0
+; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv2i64u:
+; SSE3: # BB#0:
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsfq %rax, %rax
+; SSE3-NEXT: movd %rax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %rax
+; SSE3-NEXT: bsfq %rax, %rax
+; SSE3-NEXT: movd %rax, %xmm0
+; SSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv2i64u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsfq %rax, %rax
+; SSSE3-NEXT: movd %rax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %rax
+; SSSE3-NEXT: bsfq %rax, %rax
+; SSSE3-NEXT: movd %rax, %xmm0
+; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv2i64u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrq $1, %xmm0, %rax
+; SSE41-NEXT: bsfq %rax, %rax
+; SSE41-NEXT: movd %rax, %xmm1
+; SSE41-NEXT: movd %xmm0, %rax
+; SSE41-NEXT: bsfq %rax, %rax
+; SSE41-NEXT: movd %rax, %xmm0
+; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv2i64u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrq $1, %xmm0, %rax
+; AVX-NEXT: bsfq %rax, %rax
+; AVX-NEXT: vmovq %rax, %xmm1
+; AVX-NEXT: vmovq %xmm0, %rax
+; AVX-NEXT: bsfq %rax, %rax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %in, i1 -1)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @testv4i32(<4 x i32> %in) {
+; SSE2-LABEL: testv4i32:
+; SSE2: # BB#0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movl $32, %ecx
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: cmovel %ecx, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv4i32:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movl $32, %ecx
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: cmovel %ecx, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv4i32:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSSE3-NEXT: movd %xmm1, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movl $32, %ecx
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSSE3-NEXT: movd %xmm2, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: cmovel %ecx, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv4i32:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrd $1, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: movl $32, %ecx
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: movd %edx, %xmm1
+; SSE41-NEXT: pinsrd $1, %eax, %xmm1
+; SSE41-NEXT: pextrd $2, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: pinsrd $2, %eax, %xmm1
+; SSE41-NEXT: pextrd $3, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: cmovel %ecx, %eax
+; SSE41-NEXT: pinsrd $3, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrd $1, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: movl $32, %ecx
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vmovd %edx, %xmm1
+; AVX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $2, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $3, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: cmovel %ecx, %eax
+; AVX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %in, i1 0)
+ ret <4 x i32> %out
+}
+
+define <4 x i32> @testv4i32u(<4 x i32> %in) {
+; SSE2-LABEL: testv4i32u:
+; SSE2: # BB#0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa %xmm1, %xmm0
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv4i32u:
+; SSE3: # BB#0:
+; SSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSE3-NEXT: movd %xmm1, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSE3-NEXT: movd %xmm2, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv4i32u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; SSSE3-NEXT: movd %xmm1, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
+; SSSE3-NEXT: movd %xmm2, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv4i32u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrd $1, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: movd %xmm0, %ecx
+; SSE41-NEXT: bsfl %ecx, %ecx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrd $1, %eax, %xmm1
+; SSE41-NEXT: pextrd $2, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrd $2, %eax, %xmm1
+; SSE41-NEXT: pextrd $3, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrd $3, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv4i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrd $1, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vmovd %xmm0, %ecx
+; AVX-NEXT: bsfl %ecx, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrd $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $2, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrd $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrd $3, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrd $3, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %in, i1 -1)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @testv8i16(<8 x i16> %in) {
+; SSE2-LABEL: testv8i16:
+; SSE2: # BB#0:
+; SSE2-NEXT: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %cx
+; SSE2-NEXT: movw $16, %ax
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: pextrw $3, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: pextrw $1, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: pextrw $2, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: bsfw %cx, %cx
+; SSE2-NEXT: cmovew %ax, %cx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv8i16:
+; SSE3: # BB#0:
+; SSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %cx
+; SSE3-NEXT: movw $16, %ax
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: pextrw $3, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE3-NEXT: pextrw $5, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE3-NEXT: pextrw $6, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: pextrw $2, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm3
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE3-NEXT: pextrw $4, %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: movd %xmm0, %ecx
+; SSE3-NEXT: bsfw %cx, %cx
+; SSE3-NEXT: cmovew %ax, %cx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv8i16:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %cx
+; SSSE3-NEXT: movw $16, %ax
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: pextrw $3, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSSE3-NEXT: pextrw $5, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: pextrw $1, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSSE3-NEXT: pextrw $6, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: pextrw $2, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm3
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSSE3-NEXT: pextrw $4, %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: movd %xmm0, %ecx
+; SSSE3-NEXT: bsfw %cx, %cx
+; SSSE3-NEXT: cmovew %ax, %cx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv8i16:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %cx
+; SSE41-NEXT: movw $16, %ax
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: movd %xmm0, %edx
+; SSE41-NEXT: bsfw %dx, %dx
+; SSE41-NEXT: cmovew %ax, %dx
+; SSE41-NEXT: movd %edx, %xmm1
+; SSE41-NEXT: pinsrw $1, %ecx, %xmm1
+; SSE41-NEXT: pextrw $2, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $2, %ecx, %xmm1
+; SSE41-NEXT: pextrw $3, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $3, %ecx, %xmm1
+; SSE41-NEXT: pextrw $4, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $4, %ecx, %xmm1
+; SSE41-NEXT: pextrw $5, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $5, %ecx, %xmm1
+; SSE41-NEXT: pextrw $6, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $6, %ecx, %xmm1
+; SSE41-NEXT: pextrw $7, %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: cmovew %ax, %cx
+; SSE41-NEXT: pinsrw $7, %ecx, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %cx
+; AVX-NEXT: movw $16, %ax
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vmovd %xmm0, %edx
+; AVX-NEXT: bsfw %dx, %dx
+; AVX-NEXT: cmovew %ax, %dx
+; AVX-NEXT: vmovd %edx, %xmm1
+; AVX-NEXT: vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: cmovew %ax, %cx
+; AVX-NEXT: vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %in, i1 0)
+ ret <8 x i16> %out
+}
+
+define <8 x i16> @testv8i16u(<8 x i16> %in) {
+; SSE2-LABEL: testv8i16u:
+; SSE2: # BB#0:
+; SSE2-NEXT: pextrw $7, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT: pextrw $5, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pextrw $1, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT: pextrw $6, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: pextrw $2, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: bsfw %ax, %ax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv8i16u:
+; SSE3: # BB#0:
+; SSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pextrw $3, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE3-NEXT: pextrw $5, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pextrw $1, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm1
+; SSE3-NEXT: movd %xmm0, %eax
+; SSE3-NEXT: bsfw %ax, %ax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv8i16u:
+; SSSE3: # BB#0:
+; SSSE3-NEXT: pextrw $7, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pextrw $3, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSSE3-NEXT: pextrw $5, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pextrw $1, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSSE3-NEXT: pextrw $6, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: pextrw $2, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSSE3-NEXT: pextrw $4, %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm1
+; SSSE3-NEXT: movd %xmm0, %eax
+; SSSE3-NEXT: bsfw %ax, %ax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv8i16u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrw $1, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: movd %xmm0, %ecx
+; SSE41-NEXT: bsfw %cx, %cx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrw $1, %eax, %xmm1
+; SSE41-NEXT: pextrw $2, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $2, %eax, %xmm1
+; SSE41-NEXT: pextrw $3, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $3, %eax, %xmm1
+; SSE41-NEXT: pextrw $4, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $4, %eax, %xmm1
+; SSE41-NEXT: pextrw $5, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $5, %eax, %xmm1
+; SSE41-NEXT: pextrw $6, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $6, %eax, %xmm1
+; SSE41-NEXT: pextrw $7, %xmm0, %eax
+; SSE41-NEXT: bsfw %ax, %ax
+; SSE41-NEXT: pinsrw $7, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv8i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrw $1, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vmovd %xmm0, %ecx
+; AVX-NEXT: bsfw %cx, %cx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrw $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $2, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $3, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $3, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $4, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $4, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $5, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $5, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $6, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrw $7, %xmm0, %eax
+; AVX-NEXT: bsfw %ax, %ax
+; AVX-NEXT: vpinsrw $7, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %in, i1 -1)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @testv16i8(<16 x i8> %in) {
+; SSE2-LABEL: testv16i8:
+; SSE2: # BB#0:
+; SSE2: pushq %rbp
+; SSE2: pushq %r14
+; SSE2: pushq %rbx
+; SSE2: movaps %xmm0, -16(%rsp)
+; SSE2-NEXT: movzbl -1(%rsp), %eax
+; SSE2-NEXT: bsfl %eax, %edx
+; SSE2-NEXT: movl $32, %eax
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: movl $8, %ecx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: movzbl -2(%rsp), %r14d
+; SSE2-NEXT: movzbl -3(%rsp), %ebx
+; SSE2-NEXT: movzbl -4(%rsp), %r9d
+; SSE2-NEXT: movzbl -5(%rsp), %edi
+; SSE2-NEXT: movzbl -6(%rsp), %r11d
+; SSE2-NEXT: movzbl -7(%rsp), %edx
+; SSE2-NEXT: movzbl -8(%rsp), %r8d
+; SSE2-NEXT: movzbl -9(%rsp), %esi
+; SSE2-NEXT: bsfl %esi, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsfl %edi, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm2
+; SSE2-NEXT: movzbl -10(%rsp), %edi
+; SSE2-NEXT: movzbl -11(%rsp), %esi
+; SSE2-NEXT: movzbl -12(%rsp), %r10d
+; SSE2-NEXT: movzbl -13(%rsp), %ebp
+; SSE2-NEXT: bsfl %ebp, %ebp
+; SSE2-NEXT: cmovel %eax, %ebp
+; SSE2-NEXT: cmpl $32, %ebp
+; SSE2-NEXT: cmovel %ecx, %ebp
+; SSE2-NEXT: movd %ebp, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: bsfl %ebx, %ebx
+; SSE2-NEXT: cmovel %eax, %ebx
+; SSE2-NEXT: cmpl $32, %ebx
+; SSE2-NEXT: cmovel %ecx, %ebx
+; SSE2-NEXT: movd %ebx, %xmm1
+; SSE2-NEXT: bsfl %esi, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: bsfl %edx, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm3
+; SSE2-NEXT: movzbl -14(%rsp), %edx
+; SSE2-NEXT: movzbl -15(%rsp), %esi
+; SSE2-NEXT: bsfl %esi, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsfl %r14d, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm0
+; SSE2-NEXT: bsfl %edi, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsfl %r11d, %esi
+; SSE2-NEXT: cmovel %eax, %esi
+; SSE2-NEXT: cmpl $32, %esi
+; SSE2-NEXT: cmovel %ecx, %esi
+; SSE2-NEXT: movd %esi, %xmm0
+; SSE2-NEXT: bsfl %edx, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE2-NEXT: bsfl %r9d, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: bsfl %r10d, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsfl %r8d, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm4
+; SSE2-NEXT: movzbl -16(%rsp), %edx
+; SSE2-NEXT: bsfl %edx, %edx
+; SSE2-NEXT: cmovel %eax, %edx
+; SSE2-NEXT: cmpl $32, %edx
+; SSE2-NEXT: cmovel %ecx, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: popq %r14
+; SSE2-NEXT: popq %rbp
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv16i8:
+; SSE3: # BB#0:
+; SSE3: pushq %rbp
+; SSE3: pushq %r14
+; SSE3: pushq %rbx
+; SSE3: movaps %xmm0, -16(%rsp)
+; SSE3-NEXT: movzbl -1(%rsp), %eax
+; SSE3-NEXT: bsfl %eax, %edx
+; SSE3-NEXT: movl $32, %eax
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: movl $8, %ecx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: movzbl -2(%rsp), %r14d
+; SSE3-NEXT: movzbl -3(%rsp), %ebx
+; SSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSE3-NEXT: movzbl -5(%rsp), %edi
+; SSE3-NEXT: movzbl -6(%rsp), %r11d
+; SSE3-NEXT: movzbl -7(%rsp), %edx
+; SSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSE3-NEXT: movzbl -9(%rsp), %esi
+; SSE3-NEXT: bsfl %esi, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsfl %edi, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm2
+; SSE3-NEXT: movzbl -10(%rsp), %edi
+; SSE3-NEXT: movzbl -11(%rsp), %esi
+; SSE3-NEXT: movzbl -12(%rsp), %r10d
+; SSE3-NEXT: movzbl -13(%rsp), %ebp
+; SSE3-NEXT: bsfl %ebp, %ebp
+; SSE3-NEXT: cmovel %eax, %ebp
+; SSE3-NEXT: cmpl $32, %ebp
+; SSE3-NEXT: cmovel %ecx, %ebp
+; SSE3-NEXT: movd %ebp, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: bsfl %ebx, %ebx
+; SSE3-NEXT: cmovel %eax, %ebx
+; SSE3-NEXT: cmpl $32, %ebx
+; SSE3-NEXT: cmovel %ecx, %ebx
+; SSE3-NEXT: movd %ebx, %xmm1
+; SSE3-NEXT: bsfl %esi, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE3-NEXT: bsfl %edx, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm3
+; SSE3-NEXT: movzbl -14(%rsp), %edx
+; SSE3-NEXT: movzbl -15(%rsp), %esi
+; SSE3-NEXT: bsfl %esi, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsfl %r14d, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm0
+; SSE3-NEXT: bsfl %edi, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsfl %r11d, %esi
+; SSE3-NEXT: cmovel %eax, %esi
+; SSE3-NEXT: cmpl $32, %esi
+; SSE3-NEXT: cmovel %ecx, %esi
+; SSE3-NEXT: movd %esi, %xmm0
+; SSE3-NEXT: bsfl %edx, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE3-NEXT: bsfl %r9d, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: bsfl %r10d, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsfl %r8d, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm4
+; SSE3-NEXT: movzbl -16(%rsp), %edx
+; SSE3-NEXT: bsfl %edx, %edx
+; SSE3-NEXT: cmovel %eax, %edx
+; SSE3-NEXT: cmpl $32, %edx
+; SSE3-NEXT: cmovel %ecx, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: popq %r14
+; SSE3-NEXT: popq %rbp
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv16i8:
+; SSSE3: # BB#0:
+; SSSE3: pushq %rbp
+; SSSE3: pushq %r14
+; SSSE3: pushq %rbx
+; SSSE3: movaps %xmm0, -16(%rsp)
+; SSSE3-NEXT: movzbl -1(%rsp), %eax
+; SSSE3-NEXT: bsfl %eax, %edx
+; SSSE3-NEXT: movl $32, %eax
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: movl $8, %ecx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: movzbl -2(%rsp), %r14d
+; SSSE3-NEXT: movzbl -3(%rsp), %ebx
+; SSSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSSE3-NEXT: movzbl -5(%rsp), %edi
+; SSSE3-NEXT: movzbl -6(%rsp), %r11d
+; SSSE3-NEXT: movzbl -7(%rsp), %edx
+; SSSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSSE3-NEXT: movzbl -9(%rsp), %esi
+; SSSE3-NEXT: bsfl %esi, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsfl %edi, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm2
+; SSSE3-NEXT: movzbl -10(%rsp), %edi
+; SSSE3-NEXT: movzbl -11(%rsp), %esi
+; SSSE3-NEXT: movzbl -12(%rsp), %r10d
+; SSSE3-NEXT: movzbl -13(%rsp), %ebp
+; SSSE3-NEXT: bsfl %ebp, %ebp
+; SSSE3-NEXT: cmovel %eax, %ebp
+; SSSE3-NEXT: cmpl $32, %ebp
+; SSSE3-NEXT: cmovel %ecx, %ebp
+; SSSE3-NEXT: movd %ebp, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: bsfl %ebx, %ebx
+; SSSE3-NEXT: cmovel %eax, %ebx
+; SSSE3-NEXT: cmpl $32, %ebx
+; SSSE3-NEXT: cmovel %ecx, %ebx
+; SSSE3-NEXT: movd %ebx, %xmm1
+; SSSE3-NEXT: bsfl %esi, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSSE3-NEXT: bsfl %edx, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm3
+; SSSE3-NEXT: movzbl -14(%rsp), %edx
+; SSSE3-NEXT: movzbl -15(%rsp), %esi
+; SSSE3-NEXT: bsfl %esi, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsfl %r14d, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm0
+; SSSE3-NEXT: bsfl %edi, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsfl %r11d, %esi
+; SSSE3-NEXT: cmovel %eax, %esi
+; SSSE3-NEXT: cmpl $32, %esi
+; SSSE3-NEXT: cmovel %ecx, %esi
+; SSSE3-NEXT: movd %esi, %xmm0
+; SSSE3-NEXT: bsfl %edx, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSSE3-NEXT: bsfl %r9d, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: bsfl %r10d, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsfl %r8d, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm4
+; SSSE3-NEXT: movzbl -16(%rsp), %edx
+; SSSE3-NEXT: bsfl %edx, %edx
+; SSSE3-NEXT: cmovel %eax, %edx
+; SSSE3-NEXT: cmpl $32, %edx
+; SSSE3-NEXT: cmovel %ecx, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: popq %rbx
+; SSSE3-NEXT: popq %r14
+; SSSE3-NEXT: popq %rbp
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv16i8:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %edx
+; SSE41-NEXT: movl $32, %eax
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: movl $8, %ecx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pextrb $0, %xmm0, %esi
+; SSE41-NEXT: bsfl %esi, %esi
+; SSE41-NEXT: cmovel %eax, %esi
+; SSE41-NEXT: cmpl $32, %esi
+; SSE41-NEXT: cmovel %ecx, %esi
+; SSE41-NEXT: movd %esi, %xmm1
+; SSE41-NEXT: pinsrb $1, %edx, %xmm1
+; SSE41-NEXT: pextrb $2, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $2, %edx, %xmm1
+; SSE41-NEXT: pextrb $3, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $3, %edx, %xmm1
+; SSE41-NEXT: pextrb $4, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $4, %edx, %xmm1
+; SSE41-NEXT: pextrb $5, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $5, %edx, %xmm1
+; SSE41-NEXT: pextrb $6, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $6, %edx, %xmm1
+; SSE41-NEXT: pextrb $7, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $7, %edx, %xmm1
+; SSE41-NEXT: pextrb $8, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $8, %edx, %xmm1
+; SSE41-NEXT: pextrb $9, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $9, %edx, %xmm1
+; SSE41-NEXT: pextrb $10, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $10, %edx, %xmm1
+; SSE41-NEXT: pextrb $11, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $11, %edx, %xmm1
+; SSE41-NEXT: pextrb $12, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $12, %edx, %xmm1
+; SSE41-NEXT: pextrb $13, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $13, %edx, %xmm1
+; SSE41-NEXT: pextrb $14, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $14, %edx, %xmm1
+; SSE41-NEXT: pextrb $15, %xmm0, %edx
+; SSE41-NEXT: bsfl %edx, %edx
+; SSE41-NEXT: cmovel %eax, %edx
+; SSE41-NEXT: cmpl $32, %edx
+; SSE41-NEXT: cmovel %ecx, %edx
+; SSE41-NEXT: pinsrb $15, %edx, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %edx
+; AVX-NEXT: movl $32, %eax
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: movl $8, %ecx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpextrb $0, %xmm0, %esi
+; AVX-NEXT: bsfl %esi, %esi
+; AVX-NEXT: cmovel %eax, %esi
+; AVX-NEXT: cmpl $32, %esi
+; AVX-NEXT: cmovel %ecx, %esi
+; AVX-NEXT: vmovd %esi, %xmm1
+; AVX-NEXT: vpinsrb $1, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $2, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $2, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $3, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $3, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $4, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $4, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $5, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $5, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $6, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $6, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $7, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $7, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $8, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $8, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $9, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $9, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $10, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $10, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $11, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $11, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $12, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $12, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $13, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $13, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $14, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $14, %edx, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $15, %xmm0, %edx
+; AVX-NEXT: bsfl %edx, %edx
+; AVX-NEXT: cmovel %eax, %edx
+; AVX-NEXT: cmpl $32, %edx
+; AVX-NEXT: cmovel %ecx, %edx
+; AVX-NEXT: vpinsrb $15, %edx, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %in, i1 0)
+ ret <16 x i8> %out
+}
+
+define <16 x i8> @testv16i8u(<16 x i8> %in) {
+; SSE2-LABEL: testv16i8u:
+; SSE2: # BB#0:
+; SSE2: pushq %rbx
+; SSE2: movaps %xmm0, -16(%rsp)
+; SSE2-NEXT: movzbl -1(%rsp), %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movzbl -2(%rsp), %r11d
+; SSE2-NEXT: movzbl -3(%rsp), %eax
+; SSE2-NEXT: movzbl -4(%rsp), %r9d
+; SSE2-NEXT: movzbl -5(%rsp), %edi
+; SSE2-NEXT: movzbl -6(%rsp), %r10d
+; SSE2-NEXT: movzbl -7(%rsp), %ecx
+; SSE2-NEXT: movzbl -8(%rsp), %r8d
+; SSE2-NEXT: movzbl -9(%rsp), %edx
+; SSE2-NEXT: bsfl %edx, %edx
+; SSE2-NEXT: movd %edx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: bsfl %edi, %edx
+; SSE2-NEXT: movd %edx, %xmm0
+; SSE2-NEXT: movzbl -10(%rsp), %edx
+; SSE2-NEXT: movzbl -11(%rsp), %esi
+; SSE2-NEXT: movzbl -12(%rsp), %edi
+; SSE2-NEXT: movzbl -13(%rsp), %ebx
+; SSE2-NEXT: bsfl %ebx, %ebx
+; SSE2-NEXT: movd %ebx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: bsfl %esi, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklbw %xmm0, %xmm3 # xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: bsfl %ecx, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: movzbl -14(%rsp), %eax
+; SSE2-NEXT: movzbl -15(%rsp), %ecx
+; SSE2-NEXT: bsfl %ecx, %ecx
+; SSE2-NEXT: movd %ecx, %xmm1
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT: bsfl %r11d, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: bsfl %edx, %ecx
+; SSE2-NEXT: movd %ecx, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: bsfl %r10d, %ecx
+; SSE2-NEXT: movd %ecx, %xmm0
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm3
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT: bsfl %r9d, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: bsfl %edi, %eax
+; SSE2-NEXT: movd %eax, %xmm2
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT: bsfl %r8d, %eax
+; SSE2-NEXT: movd %eax, %xmm4
+; SSE2-NEXT: movzbl -16(%rsp), %eax
+; SSE2-NEXT: bsfl %eax, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE2-NEXT: popq %rbx
+; SSE2-NEXT: retq
+;
+; SSE3-LABEL: testv16i8u:
+; SSE3: # BB#0:
+; SSE3: pushq %rbx
+; SSE3: movaps %xmm0, -16(%rsp)
+; SSE3-NEXT: movzbl -1(%rsp), %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: movzbl -2(%rsp), %r11d
+; SSE3-NEXT: movzbl -3(%rsp), %eax
+; SSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSE3-NEXT: movzbl -5(%rsp), %edi
+; SSE3-NEXT: movzbl -6(%rsp), %r10d
+; SSE3-NEXT: movzbl -7(%rsp), %ecx
+; SSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSE3-NEXT: movzbl -9(%rsp), %edx
+; SSE3-NEXT: bsfl %edx, %edx
+; SSE3-NEXT: movd %edx, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: bsfl %edi, %edx
+; SSE3-NEXT: movd %edx, %xmm0
+; SSE3-NEXT: movzbl -10(%rsp), %edx
+; SSE3-NEXT: movzbl -11(%rsp), %esi
+; SSE3-NEXT: movzbl -12(%rsp), %edi
+; SSE3-NEXT: movzbl -13(%rsp), %ebx
+; SSE3-NEXT: bsfl %ebx, %ebx
+; SSE3-NEXT: movd %ebx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: bsfl %esi, %eax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: punpcklbw %xmm0, %xmm3 # xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: bsfl %ecx, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: movzbl -14(%rsp), %eax
+; SSE3-NEXT: movzbl -15(%rsp), %ecx
+; SSE3-NEXT: bsfl %ecx, %ecx
+; SSE3-NEXT: movd %ecx, %xmm1
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE3-NEXT: bsfl %r11d, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: bsfl %edx, %ecx
+; SSE3-NEXT: movd %ecx, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: bsfl %r10d, %ecx
+; SSE3-NEXT: movd %ecx, %xmm0
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm3
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE3-NEXT: bsfl %r9d, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: bsfl %edi, %eax
+; SSE3-NEXT: movd %eax, %xmm2
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE3-NEXT: bsfl %r8d, %eax
+; SSE3-NEXT: movd %eax, %xmm4
+; SSE3-NEXT: movzbl -16(%rsp), %eax
+; SSE3-NEXT: bsfl %eax, %eax
+; SSE3-NEXT: movd %eax, %xmm0
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE3-NEXT: popq %rbx
+; SSE3-NEXT: retq
+;
+; SSSE3-LABEL: testv16i8u:
+; SSSE3: # BB#0:
+; SSSE3: pushq %rbx
+; SSSE3: movaps %xmm0, -16(%rsp)
+; SSSE3-NEXT: movzbl -1(%rsp), %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: movzbl -2(%rsp), %r11d
+; SSSE3-NEXT: movzbl -3(%rsp), %eax
+; SSSE3-NEXT: movzbl -4(%rsp), %r9d
+; SSSE3-NEXT: movzbl -5(%rsp), %edi
+; SSSE3-NEXT: movzbl -6(%rsp), %r10d
+; SSSE3-NEXT: movzbl -7(%rsp), %ecx
+; SSSE3-NEXT: movzbl -8(%rsp), %r8d
+; SSSE3-NEXT: movzbl -9(%rsp), %edx
+; SSSE3-NEXT: bsfl %edx, %edx
+; SSSE3-NEXT: movd %edx, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: bsfl %edi, %edx
+; SSSE3-NEXT: movd %edx, %xmm0
+; SSSE3-NEXT: movzbl -10(%rsp), %edx
+; SSSE3-NEXT: movzbl -11(%rsp), %esi
+; SSSE3-NEXT: movzbl -12(%rsp), %edi
+; SSSE3-NEXT: movzbl -13(%rsp), %ebx
+; SSSE3-NEXT: bsfl %ebx, %ebx
+; SSSE3-NEXT: movd %ebx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: bsfl %esi, %eax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: punpcklbw %xmm0, %xmm3 # xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: bsfl %ecx, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: movzbl -14(%rsp), %eax
+; SSSE3-NEXT: movzbl -15(%rsp), %ecx
+; SSSE3-NEXT: bsfl %ecx, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm1
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSSE3-NEXT: bsfl %r11d, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: bsfl %edx, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: bsfl %r10d, %ecx
+; SSSE3-NEXT: movd %ecx, %xmm0
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm3
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSSE3-NEXT: bsfl %r9d, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: bsfl %edi, %eax
+; SSSE3-NEXT: movd %eax, %xmm2
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSSE3-NEXT: bsfl %r8d, %eax
+; SSSE3-NEXT: movd %eax, %xmm4
+; SSSE3-NEXT: movzbl -16(%rsp), %eax
+; SSSE3-NEXT: bsfl %eax, %eax
+; SSSE3-NEXT: movd %eax, %xmm0
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSSE3-NEXT: popq %rbx
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: testv16i8u:
+; SSE41: # BB#0:
+; SSE41-NEXT: pextrb $1, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pextrb $0, %xmm0, %ecx
+; SSE41-NEXT: bsfl %ecx, %ecx
+; SSE41-NEXT: movd %ecx, %xmm1
+; SSE41-NEXT: pinsrb $1, %eax, %xmm1
+; SSE41-NEXT: pextrb $2, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $2, %eax, %xmm1
+; SSE41-NEXT: pextrb $3, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $3, %eax, %xmm1
+; SSE41-NEXT: pextrb $4, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $4, %eax, %xmm1
+; SSE41-NEXT: pextrb $5, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $5, %eax, %xmm1
+; SSE41-NEXT: pextrb $6, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $6, %eax, %xmm1
+; SSE41-NEXT: pextrb $7, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $7, %eax, %xmm1
+; SSE41-NEXT: pextrb $8, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $8, %eax, %xmm1
+; SSE41-NEXT: pextrb $9, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $9, %eax, %xmm1
+; SSE41-NEXT: pextrb $10, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $10, %eax, %xmm1
+; SSE41-NEXT: pextrb $11, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $11, %eax, %xmm1
+; SSE41-NEXT: pextrb $12, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $12, %eax, %xmm1
+; SSE41-NEXT: pextrb $13, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $13, %eax, %xmm1
+; SSE41-NEXT: pextrb $14, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $14, %eax, %xmm1
+; SSE41-NEXT: pextrb $15, %xmm0, %eax
+; SSE41-NEXT: bsfl %eax, %eax
+; SSE41-NEXT: pinsrb $15, %eax, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: testv16i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vpextrb $1, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX-NEXT: bsfl %ecx, %ecx
+; AVX-NEXT: vmovd %ecx, %xmm1
+; AVX-NEXT: vpinsrb $1, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $2, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $2, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $3, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $3, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $4, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $4, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $5, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $5, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $6, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $6, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $7, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $7, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $8, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $8, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $9, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $9, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $10, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $10, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $11, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $11, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $12, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $13, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $13, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $14, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX-NEXT: vpextrb $15, %xmm0, %eax
+; AVX-NEXT: bsfl %eax, %eax
+; AVX-NEXT: vpinsrb $15, %eax, %xmm1, %xmm0
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %in, i1 -1)
+ ret <16 x i8> %out
+}
+
+define <2 x i64> @foldv2i64() {
+; SSE-LABEL: foldv2i64:
+; SSE: # BB#0:
+; SSE-NEXT: movl $8, %eax
+; SSE-NEXT: movd %rax, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv2i64:
+; AVX: # BB#0:
+; AVX-NEXT: movl $8, %eax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> <i64 256, i64 -1>, i1 0)
+ ret <2 x i64> %out
+}
+
+define <2 x i64> @foldv2i64u() {
+; SSE-LABEL: foldv2i64u:
+; SSE: # BB#0:
+; SSE-NEXT: movl $8, %eax
+; SSE-NEXT: movd %rax, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv2i64u:
+; AVX: # BB#0:
+; AVX-NEXT: movl $8, %eax
+; AVX-NEXT: vmovq %rax, %xmm0
+; AVX-NEXT: retq
+ %out = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> <i64 256, i64 -1>, i1 -1)
+ ret <2 x i64> %out
+}
+
+define <4 x i32> @foldv4i32() {
+; SSE-LABEL: foldv4i32:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,32,0]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv4i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,32,0]
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> <i32 256, i32 -1, i32 0, i32 255>, i1 0)
+ ret <4 x i32> %out
+}
+
+define <4 x i32> @foldv4i32u() {
+; SSE-LABEL: foldv4i32u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,32,0]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv4i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,32,0]
+; AVX-NEXT: retq
+ %out = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> <i32 256, i32 -1, i32 0, i32 255>, i1 -1)
+ ret <4 x i32> %out
+}
+
+define <8 x i16> @foldv8i16() {
+; SSE-LABEL: foldv8i16:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,16,0,16,0,3,3]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv8i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,16,0,16,0,3,3]
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88>, i1 0)
+ ret <8 x i16> %out
+}
+
+define <8 x i16> @foldv8i16u() {
+; SSE-LABEL: foldv8i16u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,16,0,16,0,3,3]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv8i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,16,0,16,0,3,3]
+; AVX-NEXT: retq
+ %out = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88>, i1 -1)
+ ret <8 x i16> %out
+}
+
+define <16 x i8> @foldv16i8() {
+; SSE-LABEL: foldv16i8:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv16i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5]
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32>, i1 0)
+ ret <16 x i8> %out
+}
+
+define <16 x i8> @foldv16i8u() {
+; SSE-LABEL: foldv16i8u:
+; SSE: # BB#0:
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5]
+; SSE-NEXT: retq
+;
+; AVX-LABEL: foldv16i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5]
+; AVX-NEXT: retq
+ %out = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32>, i1 -1)
+ ret <16 x i8> %out
+}
+
+declare <2 x i64> @llvm.cttz.v2i64(<2 x i64>, i1)
+declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1)
+declare <8 x i16> @llvm.cttz.v8i16(<8 x i16>, i1)
+declare <16 x i8> @llvm.cttz.v16i8(<16 x i8>, i1)
diff --git a/test/CodeGen/X86/vector-tzcnt-256.ll b/test/CodeGen/X86/vector-tzcnt-256.ll
new file mode 100644
index 000000000000..8f744f79f85f
--- /dev/null
+++ b/test/CodeGen/X86/vector-tzcnt-256.ll
@@ -0,0 +1,1195 @@
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+
+target triple = "x86_64-unknown-unknown"
+
+define <4 x i64> @testv4i64(<4 x i64> %in) {
+; AVX1-LABEL: testv4i64:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrq $1, %xmm1, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: movl $64, %ecx
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm1, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: vmovq %rax, %xmm1
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT: vpextrq $1, %xmm0, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm0, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: cmoveq %rcx, %rax
+; AVX1-NEXT: vmovq %rax, %xmm0
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv4i64:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrq $1, %xmm1, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: movl $64, %ecx
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm1, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: vmovq %rax, %xmm1
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-NEXT: vpextrq $1, %xmm0, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm0, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: cmoveq %rcx, %rax
+; AVX2-NEXT: vmovq %rax, %xmm0
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %in, i1 0)
+ ret <4 x i64> %out
+}
+
+define <4 x i64> @testv4i64u(<4 x i64> %in) {
+; AVX1-LABEL: testv4i64u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrq $1, %xmm1, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm1, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: vmovq %rax, %xmm1
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT: vpextrq $1, %xmm0, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: vmovq %rax, %xmm2
+; AVX1-NEXT: vmovq %xmm0, %rax
+; AVX1-NEXT: bsfq %rax, %rax
+; AVX1-NEXT: vmovq %rax, %xmm0
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv4i64u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrq $1, %xmm1, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm1, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: vmovq %rax, %xmm1
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-NEXT: vpextrq $1, %xmm0, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: vmovq %rax, %xmm2
+; AVX2-NEXT: vmovq %xmm0, %rax
+; AVX2-NEXT: bsfq %rax, %rax
+; AVX2-NEXT: vmovq %rax, %xmm0
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %in, i1 -1)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @testv8i32(<8 x i32> %in) {
+; AVX1-LABEL: testv8i32:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %ecx
+; AVX1-NEXT: movl $32, %eax
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vmovd %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm1, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm1, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm0, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vmovd %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm0, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm0, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: cmovel %eax, %ecx
+; AVX1-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv8i32:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %ecx
+; AVX2-NEXT: movl $32, %eax
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vmovd %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm1, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm1, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm0, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vmovd %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm0, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm0, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: cmovel %eax, %ecx
+; AVX2-NEXT: vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %in, i1 0)
+ ret <8 x i32> %out
+}
+
+define <8 x i32> @testv8i32u(<8 x i32> %in) {
+; AVX1-LABEL: testv8i32u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vmovd %xmm1, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrd $1, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vmovd %xmm0, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $2, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrd $3, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv8i32u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vmovd %xmm1, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrd $1, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vmovd %xmm0, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrd $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $2, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrd $3, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %in, i1 -1)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @testv16i16(<16 x i16> %in) {
+; AVX1-LABEL: testv16i16:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %cx
+; AVX1-NEXT: movw $16, %ax
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vmovd %xmm1, %edx
+; AVX1-NEXT: bsfw %dx, %dx
+; AVX1-NEXT: cmovew %ax, %dx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vmovd %xmm0, %edx
+; AVX1-NEXT: bsfw %dx, %dx
+; AVX1-NEXT: cmovew %ax, %dx
+; AVX1-NEXT: vmovd %edx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: cmovew %ax, %cx
+; AVX1-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv16i16:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %cx
+; AVX2-NEXT: movw $16, %ax
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vmovd %xmm1, %edx
+; AVX2-NEXT: bsfw %dx, %dx
+; AVX2-NEXT: cmovew %ax, %dx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vmovd %xmm0, %edx
+; AVX2-NEXT: bsfw %dx, %dx
+; AVX2-NEXT: cmovew %ax, %dx
+; AVX2-NEXT: vmovd %edx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: cmovew %ax, %cx
+; AVX2-NEXT: vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %in, i1 0)
+ ret <16 x i16> %out
+}
+
+define <16 x i16> @testv16i16u(<16 x i16> %in) {
+; AVX1-LABEL: testv16i16u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vmovd %xmm1, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm1, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $7, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrw $1, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vmovd %xmm0, %ecx
+; AVX1-NEXT: bsfw %cx, %cx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $2, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $3, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $4, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $5, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $6, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrw $7, %xmm0, %eax
+; AVX1-NEXT: bsfw %ax, %ax
+; AVX1-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv16i16u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vmovd %xmm1, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm1, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $7, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrw $1, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vmovd %xmm0, %ecx
+; AVX2-NEXT: bsfw %cx, %cx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrw $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $2, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $3, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $4, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $5, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $6, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrw $7, %xmm0, %eax
+; AVX2-NEXT: bsfw %ax, %ax
+; AVX2-NEXT: vpinsrw $7, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %in, i1 -1)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @testv32i8(<32 x i8> %in) {
+; AVX1-LABEL: testv32i8:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %edx
+; AVX1-NEXT: movl $32, %eax
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: movl $8, %ecx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpextrb $0, %xmm1, %esi
+; AVX1-NEXT: bsfl %esi, %esi
+; AVX1-NEXT: cmovel %eax, %esi
+; AVX1-NEXT: cmpl $32, %esi
+; AVX1-NEXT: cmovel %ecx, %esi
+; AVX1-NEXT: vmovd %esi, %xmm2
+; AVX1-NEXT: vpinsrb $1, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $2, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $3, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $4, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $5, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $6, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $7, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $8, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $9, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $10, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $11, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $12, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $13, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $14, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm1, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $15, %edx, %xmm2, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpextrb $0, %xmm0, %esi
+; AVX1-NEXT: bsfl %esi, %esi
+; AVX1-NEXT: cmovel %eax, %esi
+; AVX1-NEXT: cmpl $32, %esi
+; AVX1-NEXT: cmovel %ecx, %esi
+; AVX1-NEXT: vmovd %esi, %xmm2
+; AVX1-NEXT: vpinsrb $1, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $2, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $3, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $4, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $5, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $6, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $7, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $8, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $9, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $10, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $11, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $12, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $13, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $14, %edx, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm0, %edx
+; AVX1-NEXT: bsfl %edx, %edx
+; AVX1-NEXT: cmovel %eax, %edx
+; AVX1-NEXT: cmpl $32, %edx
+; AVX1-NEXT: cmovel %ecx, %edx
+; AVX1-NEXT: vpinsrb $15, %edx, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv32i8:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %edx
+; AVX2-NEXT: movl $32, %eax
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: movl $8, %ecx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpextrb $0, %xmm1, %esi
+; AVX2-NEXT: bsfl %esi, %esi
+; AVX2-NEXT: cmovel %eax, %esi
+; AVX2-NEXT: cmpl $32, %esi
+; AVX2-NEXT: cmovel %ecx, %esi
+; AVX2-NEXT: vmovd %esi, %xmm2
+; AVX2-NEXT: vpinsrb $1, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $2, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $3, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $4, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $5, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $6, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $7, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $8, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $9, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $10, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $11, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $12, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $13, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $14, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm1, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $15, %edx, %xmm2, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpextrb $0, %xmm0, %esi
+; AVX2-NEXT: bsfl %esi, %esi
+; AVX2-NEXT: cmovel %eax, %esi
+; AVX2-NEXT: cmpl $32, %esi
+; AVX2-NEXT: cmovel %ecx, %esi
+; AVX2-NEXT: vmovd %esi, %xmm2
+; AVX2-NEXT: vpinsrb $1, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $2, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $3, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $4, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $5, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $6, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $7, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $8, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $9, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $10, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $11, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $12, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $13, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $14, %edx, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm0, %edx
+; AVX2-NEXT: bsfl %edx, %edx
+; AVX2-NEXT: cmovel %eax, %edx
+; AVX2-NEXT: cmpl $32, %edx
+; AVX2-NEXT: cmovel %ecx, %edx
+; AVX2-NEXT: vpinsrb $15, %edx, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> %in, i1 0)
+ ret <32 x i8> %out
+}
+
+define <32 x i8> @testv32i8u(<32 x i8> %in) {
+; AVX1-LABEL: testv32i8u:
+; AVX1: # BB#0:
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm1, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $15, %eax, %xmm2, %xmm1
+; AVX1-NEXT: vpextrb $1, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX1-NEXT: bsfl %ecx, %ecx
+; AVX1-NEXT: vmovd %ecx, %xmm2
+; AVX1-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $2, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $3, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $4, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $5, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $6, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $7, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $8, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $9, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $10, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $11, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $12, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $13, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $14, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX1-NEXT: vpextrb $15, %xmm0, %eax
+; AVX1-NEXT: bsfl %eax, %eax
+; AVX1-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: testv32i8u:
+; AVX2: # BB#0:
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpextrb $0, %xmm1, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm1, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $15, %eax, %xmm2, %xmm1
+; AVX2-NEXT: vpextrb $1, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpextrb $0, %xmm0, %ecx
+; AVX2-NEXT: bsfl %ecx, %ecx
+; AVX2-NEXT: vmovd %ecx, %xmm2
+; AVX2-NEXT: vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $2, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $3, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $4, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $5, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $6, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $7, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $8, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $9, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $10, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $11, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $12, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $13, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $14, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX2-NEXT: vpextrb $15, %xmm0, %eax
+; AVX2-NEXT: bsfl %eax, %eax
+; AVX2-NEXT: vpinsrb $15, %eax, %xmm2, %xmm0
+; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: retq
+ %out = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> %in, i1 -1)
+ ret <32 x i8> %out
+}
+
+define <4 x i64> @foldv4i64() {
+; AVX-LABEL: foldv4i64:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,64,0]
+; AVX-NEXT: retq
+ %out = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>, i1 0)
+ ret <4 x i64> %out
+}
+
+define <4 x i64> @foldv4i64u() {
+; AVX-LABEL: foldv4i64u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,64,0]
+; AVX-NEXT: retq
+ %out = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> <i64 256, i64 -1, i64 0, i64 255>, i1 -1)
+ ret <4 x i64> %out
+}
+
+define <8 x i32> @foldv8i32() {
+; AVX-LABEL: foldv8i32:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,32,0,16,0,3,3]
+; AVX-NEXT: retq
+ %out = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>, i1 0)
+ ret <8 x i32> %out
+}
+
+define <8 x i32> @foldv8i32u() {
+; AVX-LABEL: foldv8i32u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,32,0,16,0,3,3]
+; AVX-NEXT: retq
+ %out = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> <i32 256, i32 -1, i32 0, i32 255, i32 -65536, i32 7, i32 24, i32 88>, i1 -1)
+ ret <8 x i32> %out
+}
+
+define <16 x i16> @foldv16i16() {
+; AVX-LABEL: foldv16i16:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,16,0,16,0,3,3,1,1,0,1,2,3,4,5]
+; AVX-NEXT: retq
+ %out = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>, i1 0)
+ ret <16 x i16> %out
+}
+
+define <16 x i16> @foldv16i16u() {
+; AVX-LABEL: foldv16i16u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,16,0,16,0,3,3,1,1,0,1,2,3,4,5]
+; AVX-NEXT: retq
+ %out = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> <i16 256, i16 -1, i16 0, i16 255, i16 -65536, i16 7, i16 24, i16 88, i16 -2, i16 254, i16 1, i16 2, i16 4, i16 8, i16 16, i16 32>, i1 -1)
+ ret <16 x i16> %out
+}
+
+define <32 x i8> @foldv32i8() {
+; AVX-LABEL: foldv32i8:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5,6,7,8,8,7,6,5,4,3,2,1,0,0,0,0,0]
+; AVX-NEXT: retq
+ %out = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>, i1 0)
+ ret <32 x i8> %out
+}
+
+define <32 x i8> @foldv32i8u() {
+; AVX-LABEL: foldv32i8u:
+; AVX: # BB#0:
+; AVX-NEXT: vmovaps {{.*#+}} ymm0 = [8,0,8,0,8,0,3,3,1,1,0,1,2,3,4,5,6,7,8,8,7,6,5,4,3,2,1,0,0,0,0,0]
+; AVX-NEXT: retq
+ %out = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> <i8 256, i8 -1, i8 0, i8 255, i8 -65536, i8 7, i8 24, i8 88, i8 -2, i8 254, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 256, i8 -256, i8 -128, i8 -64, i8 -32, i8 -16, i8 -8, i8 -4, i8 -2, i8 -1, i8 3, i8 5, i8 7, i8 127>, i1 -1)
+ ret <32 x i8> %out
+}
+
+declare <4 x i64> @llvm.cttz.v4i64(<4 x i64>, i1)
+declare <8 x i32> @llvm.cttz.v8i32(<8 x i32>, i1)
+declare <16 x i16> @llvm.cttz.v16i16(<16 x i16>, i1)
+declare <32 x i8> @llvm.cttz.v32i8(<32 x i8>, i1)
diff --git a/test/CodeGen/X86/vector-zext.ll b/test/CodeGen/X86/vector-zext.ll
index 42781830ff2f..c64e17442675 100644
--- a/test/CodeGen/X86/vector-zext.ll
+++ b/test/CodeGen/X86/vector-zext.ll
@@ -9,7 +9,6 @@ define <8 x i32> @zext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone ssp
; SSE2: # BB#0: # %entry
; SSE2-NEXT: movdqa %xmm0, %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: # kill
; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE2-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
; SSE2-NEXT: pand .LCPI0_0(%rip), %xmm1
@@ -19,7 +18,6 @@ define <8 x i32> @zext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone ssp
; SSSE3: # BB#0: # %entry
; SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSSE3-NEXT: pxor %xmm2, %xmm2
-; SSSE3-NEXT: # kill
; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSSE3-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
; SSSE3-NEXT: pand .LCPI0_0(%rip), %xmm1
@@ -156,7 +154,6 @@ define <16 x i16> @zext_16i8_to_16i16(<16 x i8> %z) {
; SSE2: # BB#0: # %entry
; SSE2-NEXT: movdqa %xmm0, %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: # kill
; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE2-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE2-NEXT: pand .LCPI3_0(%rip), %xmm1
@@ -166,7 +163,6 @@ define <16 x i16> @zext_16i8_to_16i16(<16 x i8> %z) {
; SSSE3: # BB#0: # %entry
; SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSSE3-NEXT: pxor %xmm2, %xmm2
-; SSSE3-NEXT: # kill
; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSSE3-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSSE3-NEXT: pand .LCPI3_0(%rip), %xmm1
@@ -334,7 +330,6 @@ define <8 x i32> @shuf_zext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone
; SSE2: # BB#0: # %entry
; SSE2-NEXT: movdqa %xmm0, %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: # kill
; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE2-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE2-NEXT: retq
@@ -343,7 +338,6 @@ define <8 x i32> @shuf_zext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone
; SSSE3: # BB#0: # %entry
; SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSSE3-NEXT: pxor %xmm2, %xmm2
-; SSSE3-NEXT: # kill
; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSSE3-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSSE3-NEXT: retq
@@ -366,7 +360,6 @@ define <8 x i32> @shuf_zext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone
;
; AVX2-LABEL: shuf_zext_8i16_to_8i32:
; AVX2: # BB#0: # %entry
-; AVX2-NEXT: # kill
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: retq
entry:
@@ -380,7 +373,6 @@ define <4 x i64> @shuf_zext_4i32_to_4i64(<4 x i32> %A) nounwind uwtable readnone
; SSE2: # BB#0: # %entry
; SSE2-NEXT: movdqa %xmm0, %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: # kill
; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE2-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE2-NEXT: retq
@@ -389,7 +381,6 @@ define <4 x i64> @shuf_zext_4i32_to_4i64(<4 x i32> %A) nounwind uwtable readnone
; SSSE3: # BB#0: # %entry
; SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSSE3-NEXT: pxor %xmm2, %xmm2
-; SSSE3-NEXT: # kill
; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSSE3-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSSE3-NEXT: retq
@@ -413,7 +404,6 @@ define <4 x i64> @shuf_zext_4i32_to_4i64(<4 x i32> %A) nounwind uwtable readnone
;
; AVX2-LABEL: shuf_zext_4i32_to_4i64:
; AVX2: # BB#0: # %entry
-; AVX2-NEXT: # kill
; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; AVX2-NEXT: retq
entry:
diff --git a/test/CodeGen/X86/win32-eh-states.ll b/test/CodeGen/X86/win32-eh-states.ll
new file mode 100644
index 000000000000..8db127df6da7
--- /dev/null
+++ b/test/CodeGen/X86/win32-eh-states.ll
@@ -0,0 +1,112 @@
+; RUN: llc -mtriple=i686-pc-windows-msvc < %s | FileCheck %s
+
+; Based on this source:
+; extern "C" void may_throw(int);
+; void f() {
+; try {
+; may_throw(1);
+; try {
+; may_throw(2);
+; } catch (int) {
+; may_throw(3);
+; }
+; } catch (int) {
+; may_throw(4);
+; }
+; }
+
+%rtti.TypeDescriptor2 = type { i8**, i8*, [3 x i8] }
+%eh.CatchHandlerType = type { i32, i8* }
+
+declare void @may_throw(i32)
+declare i32 @__CxxFrameHandler3(...)
+declare void @llvm.eh.begincatch(i8*, i8*)
+declare void @llvm.eh.endcatch()
+declare i32 @llvm.eh.typeid.for(i8*)
+
+$"\01??_R0H@8" = comdat any
+
+@"\01??_7type_info@@6B@" = external constant i8*
+@"\01??_R0H@8" = linkonce_odr global %rtti.TypeDescriptor2 { i8** @"\01??_7type_info@@6B@", i8* null, [3 x i8] c".H\00" }, comdat
+@llvm.eh.handlertype.H.0 = private unnamed_addr constant %eh.CatchHandlerType { i32 0, i8* bitcast (%rtti.TypeDescriptor2* @"\01??_R0H@8" to i8*) }, section "llvm.metadata"
+
+define void @f() #0 {
+entry:
+ invoke void @may_throw(i32 1)
+ to label %invoke.cont unwind label %lpad
+
+invoke.cont: ; preds = %entry
+ invoke void @may_throw(i32 2)
+ to label %try.cont.9 unwind label %lpad.1
+
+try.cont.9: ; preds = %invoke.cont.3, %invoke.cont, %catch.7
+ ; FIXME: Something about our CFG breaks TailDuplication. This empy asm blocks
+ ; it so we can focus on testing the state numbering.
+ call void asm sideeffect "", "~{dirflag},~{fpsr},~{flags}"()
+ ret void
+
+lpad: ; preds = %catch, %entry
+ %0 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*)
+ catch %eh.CatchHandlerType* @llvm.eh.handlertype.H.0
+ %1 = extractvalue { i8*, i32 } %0, 0
+ %2 = extractvalue { i8*, i32 } %0, 1
+ br label %catch.dispatch.4
+
+lpad.1: ; preds = %invoke.cont
+ %3 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*)
+ catch i8* bitcast (%eh.CatchHandlerType* @llvm.eh.handlertype.H.0 to i8*)
+ %4 = extractvalue { i8*, i32 } %3, 0
+ %5 = extractvalue { i8*, i32 } %3, 1
+ %6 = tail call i32 @llvm.eh.typeid.for(i8* bitcast (%eh.CatchHandlerType* @llvm.eh.handlertype.H.0 to i8*)) #3
+ %matches = icmp eq i32 %5, %6
+ br i1 %matches, label %catch, label %catch.dispatch.4
+
+catch.dispatch.4: ; preds = %lpad.1, %lpad
+ %exn.slot.0 = phi i8* [ %4, %lpad.1 ], [ %1, %lpad ]
+ %ehselector.slot.0 = phi i32 [ %5, %lpad.1 ], [ %2, %lpad ]
+ %.pre = tail call i32 @llvm.eh.typeid.for(i8* bitcast (%eh.CatchHandlerType* @llvm.eh.handlertype.H.0 to i8*)) #3
+ %matches6 = icmp eq i32 %ehselector.slot.0, %.pre
+ br i1 %matches6, label %catch.7, label %eh.resume
+
+catch.7: ; preds = %catch.dispatch.4
+ tail call void @llvm.eh.begincatch(i8* %exn.slot.0, i8* null) #3
+ tail call void @may_throw(i32 4)
+ tail call void @llvm.eh.endcatch() #3
+ br label %try.cont.9
+
+catch: ; preds = %lpad.1
+ tail call void @llvm.eh.begincatch(i8* %4, i8* null) #3
+ invoke void @may_throw(i32 3)
+ to label %invoke.cont.3 unwind label %lpad
+
+invoke.cont.3: ; preds = %catch
+ tail call void @llvm.eh.endcatch() #3
+ br label %try.cont.9
+
+eh.resume: ; preds = %catch.dispatch.4
+ %lpad.val = insertvalue { i8*, i32 } undef, i8* %exn.slot.0, 0
+ %lpad.val.12 = insertvalue { i8*, i32 } %lpad.val, i32 %ehselector.slot.0, 1
+ resume { i8*, i32 } %lpad.val.12
+}
+
+; CHECK-LABEL: _f:
+; CHECK: movl $-1, [[state:[-0-9]+]](%ebp)
+; CHECK: movl $___ehhandler$f, {{.*}}
+;
+; CHECK: movl $0, [[state]](%ebp)
+; CHECK: movl $1, (%esp)
+; CHECK: calll _may_throw
+;
+; CHECK: movl $1, [[state]](%ebp)
+; CHECK: movl $2, (%esp)
+; CHECK: calll _may_throw
+
+; CHECK-LABEL: _f.catch:
+; CHECK: movl $4, Lf$frame_escape_{{[0-9]+.*}}
+; CHECK: movl $4, (%esp)
+; CHECK: calll _may_throw
+
+; CHECK-LABEL: _f.catch.1:
+; CHECK: movl $3, Lf$frame_escape_{{[0-9]+.*}}
+; CHECK: movl $3, (%esp)
+; CHECK: calll _may_throw
diff --git a/test/CodeGen/X86/win32-eh.ll b/test/CodeGen/X86/win32-eh.ll
index 4d3c34ed6c17..42c9d9e2240d 100644
--- a/test/CodeGen/X86/win32-eh.ll
+++ b/test/CodeGen/X86/win32-eh.ll
@@ -19,16 +19,18 @@ catchall:
}
; CHECK-LABEL: _use_except_handler3:
+; CHECK: pushl %ebp
+; CHECK: movl %esp, %ebp
; CHECK: subl ${{[0-9]+}}, %esp
-; CHECK: movl $-1, 12(%esp)
-; CHECK: movl $L__ehtable$use_except_handler3, 8(%esp)
-; CHECK: movl $__except_handler3, 4(%esp)
+; CHECK: movl $-1, -4(%ebp)
+; CHECK: movl $L__ehtable$use_except_handler3, -8(%ebp)
+; CHECK: leal -16(%ebp), %[[node:[^ ,]*]]
+; CHECK: movl $__except_handler3, -12(%ebp)
; CHECK: movl %fs:0, %[[next:[^ ,]*]]
-; CHECK: movl %[[next]], (%esp)
-; CHECK: leal (%esp), %[[node:[^ ,]*]]
+; CHECK: movl %[[next]], -16(%ebp)
; CHECK: movl %[[node]], %fs:0
; CHECK: calll _may_throw_or_crash
-; CHECK: movl (%esp), %[[next:[^ ,]*]]
+; CHECK: movl -16(%ebp), %[[next:[^ ,]*]]
; CHECK: movl %[[next]], %fs:0
; CHECK: retl
@@ -44,17 +46,21 @@ catchall:
}
; CHECK-LABEL: _use_except_handler4:
+; CHECK: pushl %ebp
+; CHECK: movl %esp, %ebp
; CHECK: subl ${{[0-9]+}}, %esp
-; CHECK: movl %esp, (%esp)
-; CHECK: movl $-2, 20(%esp)
-; CHECK: movl $L__ehtable$use_except_handler4, 4(%esp)
-; CHECK: leal 8(%esp), %[[node:[^ ,]*]]
-; CHECK: movl $__except_handler4, 12(%esp)
+; CHECK: movl %esp, -24(%ebp)
+; CHECK: movl $-2, -4(%ebp)
+; CHECK: movl $L__ehtable$use_except_handler4, %[[lsda:[^ ,]*]]
+; CHECK: xorl ___security_cookie, %[[lsda]]
+; CHECK: movl %[[lsda]], -8(%ebp)
+; CHECK: leal -16(%ebp), %[[node:[^ ,]*]]
+; CHECK: movl $__except_handler4, -12(%ebp)
; CHECK: movl %fs:0, %[[next:[^ ,]*]]
-; CHECK: movl %[[next]], 8(%esp)
+; CHECK: movl %[[next]], -16(%ebp)
; CHECK: movl %[[node]], %fs:0
; CHECK: calll _may_throw_or_crash
-; CHECK: movl 8(%esp), %[[next:[^ ,]*]]
+; CHECK: movl -16(%ebp), %[[next:[^ ,]*]]
; CHECK: movl %[[next]], %fs:0
; CHECK: retl
@@ -73,19 +79,34 @@ catchall:
}
; CHECK-LABEL: _use_CxxFrameHandler3:
+; CHECK: pushl %ebp
+; CHECK: movl %esp, %ebp
; CHECK: subl ${{[0-9]+}}, %esp
-; CHECK: movl %esp, (%esp)
-; CHECK: movl $-1, 12(%esp)
-; CHECK: leal 4(%esp), %[[node:[^ ,]*]]
-; CHECK: movl $___ehhandler$use_CxxFrameHandler3, 8(%esp)
+; CHECK: movl %esp, -16(%ebp)
+; CHECK: movl $-1, -4(%ebp)
+; CHECK: leal -12(%ebp), %[[node:[^ ,]*]]
+; CHECK: movl $___ehhandler$use_CxxFrameHandler3, -8(%ebp)
; CHECK: movl %fs:0, %[[next:[^ ,]*]]
-; CHECK: movl %[[next]], 4(%esp)
+; CHECK: movl %[[next]], -12(%ebp)
; CHECK: movl %[[node]], %fs:0
+; CHECK: movl $0, -4(%ebp)
; CHECK: calll _may_throw_or_crash
-; CHECK: movl 4(%esp), %[[next:[^ ,]*]]
+; CHECK: movl -12(%ebp), %[[next:[^ ,]*]]
; CHECK: movl %[[next]], %fs:0
; CHECK: retl
+; CHECK: .section .xdata,"dr"
+; CHECK-LABEL: L__ehtable$use_CxxFrameHandler3:
+; CHECK-NEXT: .long 429065506
+; CHECK-NEXT: .long 2
+; CHECK-NEXT: .long ($stateUnwindMap$use_CxxFrameHandler3)
+; CHECK-NEXT: .long 1
+; CHECK-NEXT: .long ($tryMap$use_CxxFrameHandler3)
+; CHECK-NEXT: .long 0
+; CHECK-NEXT: .long 0
+; CHECK-NEXT: .long 0
+; CHECK-NEXT: .long 1
+
; CHECK-LABEL: ___ehhandler$use_CxxFrameHandler3:
; CHECK: movl $L__ehtable$use_CxxFrameHandler3, %eax
; CHECK: jmp ___CxxFrameHandler3 # TAILCALL