1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
|
//===-- AMDGPUGIsel.td - AMDGPU GlobalISel Patterns---------*- tablegen -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
// This files contains patterns that should only be used by GlobalISel. For
// example patterns for V_* instructions that have S_* equivalents.
// SelectionDAG does not support selecting V_* instructions.
//===----------------------------------------------------------------------===//
include "AMDGPU.td"
include "AMDGPUCombine.td"
def sd_vsrc0 : ComplexPattern<i32, 1, "">;
def gi_vsrc0 :
GIComplexOperandMatcher<s32, "selectVSRC0">,
GIComplexPatternEquiv<sd_vsrc0>;
def sd_vcsrc : ComplexPattern<i32, 1, "">;
def gi_vcsrc :
GIComplexOperandMatcher<s32, "selectVCSRC">,
GIComplexPatternEquiv<sd_vcsrc>;
def gi_vop3mods0 :
GIComplexOperandMatcher<s32, "selectVOP3Mods0">,
GIComplexPatternEquiv<VOP3Mods0>;
def gi_vop3mods :
GIComplexOperandMatcher<s32, "selectVOP3Mods">,
GIComplexPatternEquiv<VOP3Mods>;
def gi_vop3modsnoncanonicalizing :
GIComplexOperandMatcher<s32, "selectVOP3ModsNonCanonicalizing">,
GIComplexPatternEquiv<VOP3ModsNonCanonicalizing>;
def gi_vop3_no_mods :
GIComplexOperandMatcher<s32, "selectVOP3NoMods">,
GIComplexPatternEquiv<VOP3NoMods>;
def gi_vop3omods :
GIComplexOperandMatcher<s32, "selectVOP3OMods">,
GIComplexPatternEquiv<VOP3OMods>;
def gi_vop3pmods :
GIComplexOperandMatcher<s32, "selectVOP3PMods">,
GIComplexPatternEquiv<VOP3PMods>;
def gi_vop3pmodsdot :
GIComplexOperandMatcher<s32, "selectVOP3PModsDOT">,
GIComplexPatternEquiv<VOP3PModsDOT>;
def gi_dotiuvop3pmods :
GIComplexOperandMatcher<s32, "selectDotIUVOP3PMods">,
GIComplexPatternEquiv<DotIUVOP3PMods>;
def gi_wmmaopselvop3pmods :
GIComplexOperandMatcher<s32, "selectWMMAOpSelVOP3PMods">,
GIComplexPatternEquiv<WMMAOpSelVOP3PMods>;
def gi_vop3opselmods :
GIComplexOperandMatcher<s32, "selectVOP3OpSelMods">,
GIComplexPatternEquiv<VOP3OpSelMods>;
def gi_vinterpmods :
GIComplexOperandMatcher<s32, "selectVINTERPMods">,
GIComplexPatternEquiv<VINTERPMods>;
def gi_vinterpmods_hi :
GIComplexOperandMatcher<s32, "selectVINTERPModsHi">,
GIComplexPatternEquiv<VINTERPModsHi>;
// FIXME: Why do we have both VOP3OpSel and VOP3OpSelMods?
def gi_vop3opsel :
GIComplexOperandMatcher<s32, "selectVOP3OpSelMods">,
GIComplexPatternEquiv<VOP3OpSel>;
def gi_smrd_imm :
GIComplexOperandMatcher<s64, "selectSmrdImm">,
GIComplexPatternEquiv<SMRDImm>;
def gi_smrd_imm32 :
GIComplexOperandMatcher<s64, "selectSmrdImm32">,
GIComplexPatternEquiv<SMRDImm32>;
def gi_smrd_sgpr :
GIComplexOperandMatcher<s64, "selectSmrdSgpr">,
GIComplexPatternEquiv<SMRDSgpr>;
def gi_smrd_sgpr_imm :
GIComplexOperandMatcher<s64, "selectSmrdSgprImm">,
GIComplexPatternEquiv<SMRDSgprImm>;
def gi_flat_offset :
GIComplexOperandMatcher<s64, "selectFlatOffset">,
GIComplexPatternEquiv<FlatOffset>;
def gi_global_offset :
GIComplexOperandMatcher<s64, "selectGlobalOffset">,
GIComplexPatternEquiv<GlobalOffset>;
def gi_global_saddr :
GIComplexOperandMatcher<s64, "selectGlobalSAddr">,
GIComplexPatternEquiv<GlobalSAddr>;
def gi_mubuf_scratch_offset :
GIComplexOperandMatcher<s32, "selectMUBUFScratchOffset">,
GIComplexPatternEquiv<MUBUFScratchOffset>;
def gi_buf_soffset :
GIComplexOperandMatcher<s32, "selectBUFSOffset">,
GIComplexPatternEquiv<BUFSOffset>;
def gi_mubuf_scratch_offen :
GIComplexOperandMatcher<s32, "selectMUBUFScratchOffen">,
GIComplexPatternEquiv<MUBUFScratchOffen>;
def gi_flat_scratch_offset :
GIComplexOperandMatcher<s32, "selectScratchOffset">,
GIComplexPatternEquiv<ScratchOffset>;
def gi_flat_scratch_saddr :
GIComplexOperandMatcher<s32, "selectScratchSAddr">,
GIComplexPatternEquiv<ScratchSAddr>;
def gi_flat_scratch_svaddr :
GIComplexOperandMatcher<s32, "selectScratchSVAddr">,
GIComplexPatternEquiv<ScratchSVAddr>;
def gi_ds_1addr_1offset :
GIComplexOperandMatcher<s32, "selectDS1Addr1Offset">,
GIComplexPatternEquiv<DS1Addr1Offset>;
def gi_ds_64bit_4byte_aligned :
GIComplexOperandMatcher<s64, "selectDS64Bit4ByteAligned">,
GIComplexPatternEquiv<DS64Bit4ByteAligned>;
def gi_ds_128bit_8byte_aligned :
GIComplexOperandMatcher<s64, "selectDS128Bit8ByteAligned">,
GIComplexPatternEquiv<DS128Bit8ByteAligned>;
def gi_mubuf_addr64 :
GIComplexOperandMatcher<s64, "selectMUBUFAddr64">,
GIComplexPatternEquiv<MUBUFAddr64>;
def gi_mubuf_offset :
GIComplexOperandMatcher<s64, "selectMUBUFOffset">,
GIComplexPatternEquiv<MUBUFOffset>;
def gi_smrd_buffer_imm :
GIComplexOperandMatcher<s64, "selectSMRDBufferImm">,
GIComplexPatternEquiv<SMRDBufferImm>;
def gi_smrd_buffer_imm32 :
GIComplexOperandMatcher<s64, "selectSMRDBufferImm32">,
GIComplexPatternEquiv<SMRDBufferImm32>;
def gi_smrd_buffer_sgpr_imm :
GIComplexOperandMatcher<s64, "selectSMRDBufferSgprImm">,
GIComplexPatternEquiv<SMRDBufferSgprImm>;
def gi_vop3_mad_mix_mods :
GIComplexOperandMatcher<s64, "selectVOP3PMadMixMods">,
GIComplexPatternEquiv<VOP3PMadMixMods>;
def gi_vop3_mad_mix_mods_ext :
GIComplexOperandMatcher<s64, "selectVOP3PMadMixModsExt">,
GIComplexPatternEquiv<VOP3PMadMixModsExt>;
// Separate load nodes are defined to glue m0 initialization in
// SelectionDAG. The GISel selector can just insert m0 initialization
// directly before selecting a glue-less load, so hide this
// distinction.
def : GINodeEquiv<G_LOAD, AMDGPUld_glue> {
let CheckMMOIsNonAtomic = 1;
let IfSignExtend = G_SEXTLOAD;
let IfZeroExtend = G_ZEXTLOAD;
}
def : GINodeEquiv<G_STORE, AMDGPUst_glue> {
let CheckMMOIsNonAtomic = 1;
}
def : GINodeEquiv<G_LOAD, AMDGPUatomic_ld_glue> {
bit CheckMMOIsAtomic = 1;
}
def : GINodeEquiv<G_STORE, AMDGPUatomic_st_glue> {
bit CheckMMOIsAtomic = 1;
}
def : GINodeEquiv<G_ATOMIC_CMPXCHG, atomic_cmp_swap_glue>;
def : GINodeEquiv<G_ATOMICRMW_XCHG, atomic_swap_glue>;
def : GINodeEquiv<G_ATOMICRMW_ADD, atomic_load_add_glue>;
def : GINodeEquiv<G_ATOMICRMW_SUB, atomic_load_sub_glue>;
def : GINodeEquiv<G_ATOMICRMW_AND, atomic_load_and_glue>;
def : GINodeEquiv<G_ATOMICRMW_OR, atomic_load_or_glue>;
def : GINodeEquiv<G_ATOMICRMW_XOR, atomic_load_xor_glue>;
def : GINodeEquiv<G_ATOMICRMW_MIN, atomic_load_min_glue>;
def : GINodeEquiv<G_ATOMICRMW_MAX, atomic_load_max_glue>;
def : GINodeEquiv<G_ATOMICRMW_UMIN, atomic_load_umin_glue>;
def : GINodeEquiv<G_ATOMICRMW_UMAX, atomic_load_umax_glue>;
def : GINodeEquiv<G_ATOMICRMW_FADD, atomic_load_fadd_glue>;
def : GINodeEquiv<G_AMDGPU_FFBH_U32, AMDGPUffbh_u32_impl>;
def : GINodeEquiv<G_AMDGPU_FFBL_B32, AMDGPUffbl_b32_impl>;
def : GINodeEquiv<G_AMDGPU_FMIN_LEGACY, AMDGPUfmin_legacy>;
def : GINodeEquiv<G_AMDGPU_FMAX_LEGACY, AMDGPUfmax_legacy>;
def : GINodeEquiv<G_AMDGPU_RCP_IFLAG, AMDGPUrcp_iflag>;
def : GINodeEquiv<G_AMDGPU_CVT_F32_UBYTE0, AMDGPUcvt_f32_ubyte0>;
def : GINodeEquiv<G_AMDGPU_CVT_F32_UBYTE1, AMDGPUcvt_f32_ubyte1>;
def : GINodeEquiv<G_AMDGPU_CVT_F32_UBYTE2, AMDGPUcvt_f32_ubyte2>;
def : GINodeEquiv<G_AMDGPU_CVT_F32_UBYTE3, AMDGPUcvt_f32_ubyte3>;
def : GINodeEquiv<G_AMDGPU_CVT_PK_I16_I32, AMDGPUpk_i16_i32_impl>;
def : GINodeEquiv<G_AMDGPU_SMED3, AMDGPUsmed3>;
def : GINodeEquiv<G_AMDGPU_UMED3, AMDGPUumed3>;
def : GINodeEquiv<G_AMDGPU_FMED3, AMDGPUfmed3_impl>;
def : GINodeEquiv<G_AMDGPU_CLAMP, AMDGPUclamp>;
def : GINodeEquiv<G_AMDGPU_ATOMIC_CMPXCHG, AMDGPUatomic_cmp_swap>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD, SIbuffer_load>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_USHORT, SIbuffer_load_ushort>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_UBYTE, SIbuffer_load_ubyte>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_SSHORT, SIbuffer_load_short>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_SBYTE, SIbuffer_load_byte>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_FORMAT, SIbuffer_load_format>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_FORMAT_TFE, SIbuffer_load_format_tfe>;
def : GINodeEquiv<G_AMDGPU_BUFFER_LOAD_FORMAT_D16, SIbuffer_load_format_d16>;
def : GINodeEquiv<G_AMDGPU_TBUFFER_LOAD_FORMAT, SItbuffer_load>;
def : GINodeEquiv<G_AMDGPU_TBUFFER_LOAD_FORMAT_D16, SItbuffer_load_d16>;
def : GINodeEquiv<G_AMDGPU_BUFFER_STORE, SIbuffer_store>;
def : GINodeEquiv<G_AMDGPU_BUFFER_STORE_SHORT, SIbuffer_store_short>;
def : GINodeEquiv<G_AMDGPU_BUFFER_STORE_BYTE, SIbuffer_store_byte>;
def : GINodeEquiv<G_AMDGPU_BUFFER_STORE_FORMAT, SIbuffer_store_format>;
def : GINodeEquiv<G_AMDGPU_BUFFER_STORE_FORMAT_D16, SIbuffer_store_format_d16>;
def : GINodeEquiv<G_AMDGPU_TBUFFER_STORE_FORMAT, SItbuffer_store>;
def : GINodeEquiv<G_AMDGPU_TBUFFER_STORE_FORMAT_D16, SItbuffer_store_d16>;
// FIXME: Check MMO is atomic
def : GINodeEquiv<G_ATOMICRMW_UINC_WRAP, atomic_load_uinc_wrap_glue>;
def : GINodeEquiv<G_ATOMICRMW_UDEC_WRAP, atomic_load_udec_wrap_glue>;
def : GINodeEquiv<G_AMDGPU_ATOMIC_FMIN, SIatomic_fmin>;
def : GINodeEquiv<G_AMDGPU_ATOMIC_FMAX, SIatomic_fmax>;
def : GINodeEquiv<G_AMDGPU_ATOMIC_FMIN, atomic_load_fmin_glue>;
def : GINodeEquiv<G_AMDGPU_ATOMIC_FMAX, atomic_load_fmax_glue>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_SWAP, SIbuffer_atomic_swap>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_ADD, SIbuffer_atomic_add>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_SUB, SIbuffer_atomic_sub>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_SMIN, SIbuffer_atomic_smin>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_UMIN, SIbuffer_atomic_umin>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_SMAX, SIbuffer_atomic_smax>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_UMAX, SIbuffer_atomic_umax>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_AND, SIbuffer_atomic_and>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_OR, SIbuffer_atomic_or>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_XOR, SIbuffer_atomic_xor>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_INC, SIbuffer_atomic_inc>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_DEC, SIbuffer_atomic_dec>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_FADD, SIbuffer_atomic_fadd>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_FMIN, SIbuffer_atomic_fmin>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_FMAX, SIbuffer_atomic_fmax>;
def : GINodeEquiv<G_AMDGPU_BUFFER_ATOMIC_CMPSWAP, SIbuffer_atomic_cmpswap>;
def : GINodeEquiv<G_AMDGPU_S_BUFFER_LOAD, SIsbuffer_load>;
def : GINodeEquiv<G_FPTRUNC_ROUND_UPWARD, SIfptrunc_round_upward>;
def : GINodeEquiv<G_FPTRUNC_ROUND_DOWNWARD, SIfptrunc_round_downward>;
class GISelSop2Pat <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src0_vt SReg_32:$src0), (src1_vt SReg_32:$src1))),
(inst src0_vt:$src0, src1_vt:$src1)
>;
class GISelVop2Pat <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src0_vt (sd_vsrc0 src0_vt:$src0)), (src1_vt VGPR_32:$src1))),
(inst src0_vt:$src0, src1_vt:$src1)
>;
class GISelVop2CommutePat <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src1_vt VGPR_32:$src1), (src0_vt (sd_vsrc0 src0_vt:$src0)))),
(inst src0_vt:$src0, src1_vt:$src1)
>;
class GISelVop3Pat2 <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src0_vt (sd_vcsrc src0_vt:$src0)), (src1_vt (sd_vcsrc src1_vt:$src1)))),
(inst src0_vt:$src0, src1_vt:$src1)
>;
class GISelVop3Pat2CommutePat <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src0_vt (sd_vcsrc src0_vt:$src0)), (src1_vt (sd_vcsrc src1_vt:$src1)))),
(inst src0_vt:$src1, src1_vt:$src0)
>;
class GISelVop3Pat2ModsPat <
SDPatternOperator node,
Instruction inst,
ValueType dst_vt,
ValueType src0_vt = dst_vt, ValueType src1_vt = src0_vt> : GCNPat <
(dst_vt (node (src0_vt (VOP3Mods0 src0_vt:$src0, i32:$src0_modifiers, i1:$clamp, i32:$omods)),
(src1_vt (VOP3Mods src1_vt:$src1, i32:$src1_modifiers)))),
(inst i32:$src0_modifiers, src0_vt:$src0,
i32:$src1_modifiers, src1_vt:$src1, $clamp, $omods)
>;
multiclass GISelVop2IntrPat <
SDPatternOperator node, Instruction inst,
ValueType dst_vt, ValueType src_vt = dst_vt> {
def : GISelVop2Pat <node, inst, dst_vt, src_vt>;
// FIXME: Intrinsics aren't marked as commutable, so we need to add an explicit
// pattern to handle commuting. This is another reason why legalizing to a
// generic machine instruction may be better that matching the intrinsic
// directly.
def : GISelVop2CommutePat <node, inst, dst_vt, src_vt>;
}
// Since GlobalISel is more flexible then SelectionDAG, I think we can get
// away with adding patterns for integer types and not legalizing all
// loads and stores to vector types. This should help simplify the load/store
// legalization.
foreach Ty = [i64, p0, p1, p4] in {
defm : SMRD_Pattern <"S_LOAD_DWORDX2", Ty>;
}
def gi_as_i32timm : GICustomOperandRenderer<"renderTruncTImm">,
GISDNodeXFormEquiv<as_i32timm>;
def gi_as_i16timm : GICustomOperandRenderer<"renderTruncTImm">,
GISDNodeXFormEquiv<as_i16timm>;
def gi_as_i8timm : GICustomOperandRenderer<"renderTruncTImm">,
GISDNodeXFormEquiv<as_i8timm>;
def gi_as_i1timm : GICustomOperandRenderer<"renderTruncTImm">,
GISDNodeXFormEquiv<as_i1timm>;
def gi_NegateImm : GICustomOperandRenderer<"renderNegateImm">,
GISDNodeXFormEquiv<NegateImm>;
def gi_bitcast_fpimm_to_i32 : GICustomOperandRenderer<"renderBitcastImm">,
GISDNodeXFormEquiv<bitcast_fpimm_to_i32>;
def gi_IMMPopCount : GICustomOperandRenderer<"renderPopcntImm">,
GISDNodeXFormEquiv<IMMPopCount>;
def gi_extract_cpol : GICustomOperandRenderer<"renderExtractCPol">,
GISDNodeXFormEquiv<extract_cpol>;
def gi_extract_swz : GICustomOperandRenderer<"renderExtractSWZ">,
GISDNodeXFormEquiv<extract_swz>;
def gi_set_glc : GICustomOperandRenderer<"renderSetGLC">,
GISDNodeXFormEquiv<set_glc>;
def gi_frameindex_to_targetframeindex : GICustomOperandRenderer<"renderFrameIndex">,
GISDNodeXFormEquiv<frameindex_to_targetframeindex>;
def gi_fp_pow2_to_exponent : GICustomOperandRenderer<"renderFPPow2ToExponent">,
GISDNodeXFormEquiv<FPPow2ToExponentXForm>;
|
