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
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
|
//===-- X86InstrFMA.td - FMA Instruction Set ---------------*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes FMA (Fused Multiply-Add) instructions.
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// FMA3 - Intel 3 operand Fused Multiply-Add instructions
//===----------------------------------------------------------------------===//
let Constraints = "$src1 = $dst" in {
multiclass fma3p_rm<bits<8> opc, string OpcodeStr> {
let neverHasSideEffects = 1 in {
def r : FMA3<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
let mayLoad = 1 in
def m : FMA3<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, f128mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
def rY : FMA3<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
let mayLoad = 1 in
def mY : FMA3<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, f256mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
} // neverHasSideEffects = 1
}
// Intrinsic for 213 pattern
multiclass fma3p_rm_int<bits<8> opc, string OpcodeStr,
PatFrag MemFrag128, PatFrag MemFrag256,
SDNode Op213, ValueType OpVT128, ValueType OpVT256> {
def r : FMA3<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR128:$dst, (OpVT128 (Op213 VR128:$src2,
VR128:$src1, VR128:$src3)))]>;
def m : FMA3<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, f128mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR128:$dst, (OpVT128 (Op213 VR128:$src2, VR128:$src1,
(MemFrag128 addr:$src3))))]>;
def rY : FMA3<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR256:$dst, (OpVT256 (Op213 VR256:$src2, VR256:$src1,
VR256:$src3)))]>;
def mY : FMA3<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, f256mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR256:$dst,
(OpVT256 (Op213 VR256:$src2, VR256:$src1,
(MemFrag256 addr:$src3))))]>;
}
} // Constraints = "$src1 = $dst"
multiclass fma3p_forms<bits<8> opc132, bits<8> opc213, bits<8> opc231,
string OpcodeStr, string PackTy,
PatFrag MemFrag128, PatFrag MemFrag256,
SDNode Op, ValueType OpTy128, ValueType OpTy256> {
defm r213 : fma3p_rm_int <opc213, !strconcat(OpcodeStr,
!strconcat("213", PackTy)), MemFrag128, MemFrag256,
Op, OpTy128, OpTy256>;
defm r132 : fma3p_rm <opc132,
!strconcat(OpcodeStr, !strconcat("132", PackTy))>;
defm r231 : fma3p_rm <opc231,
!strconcat(OpcodeStr, !strconcat("231", PackTy))>;
}
// Fused Multiply-Add
let ExeDomain = SSEPackedSingle in {
defm VFMADDPS : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "ps", memopv4f32,
memopv8f32, X86Fmadd, v4f32, v8f32>;
defm VFMSUBPS : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "ps", memopv4f32,
memopv8f32, X86Fmsub, v4f32, v8f32>;
defm VFMADDSUBPS : fma3p_forms<0x96, 0xA6, 0xB6, "vfmaddsub", "ps",
memopv4f32, memopv8f32, X86Fmaddsub,
v4f32, v8f32>;
defm VFMSUBADDPS : fma3p_forms<0x97, 0xA7, 0xB7, "vfmsubadd", "ps",
memopv4f32, memopv8f32, X86Fmsubadd,
v4f32, v8f32>;
}
let ExeDomain = SSEPackedDouble in {
defm VFMADDPD : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "pd", memopv2f64,
memopv4f64, X86Fmadd, v2f64, v4f64>, VEX_W;
defm VFMSUBPD : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "pd", memopv2f64,
memopv4f64, X86Fmsub, v2f64, v4f64>, VEX_W;
defm VFMADDSUBPD : fma3p_forms<0x96, 0xA6, 0xB6, "vfmaddsub", "pd",
memopv2f64, memopv4f64, X86Fmaddsub,
v2f64, v4f64>, VEX_W;
defm VFMSUBADDPD : fma3p_forms<0x97, 0xA7, 0xB7, "vfmsubadd", "pd",
memopv2f64, memopv4f64, X86Fmsubadd,
v2f64, v4f64>, VEX_W;
}
// Fused Negative Multiply-Add
let ExeDomain = SSEPackedSingle in {
defm VFNMADDPS : fma3p_forms<0x9C, 0xAC, 0xBC, "vfnmadd", "ps", memopv4f32,
memopv8f32, X86Fnmadd, v4f32, v8f32>;
defm VFNMSUBPS : fma3p_forms<0x9E, 0xAE, 0xBE, "vfnmsub", "ps", memopv4f32,
memopv8f32, X86Fnmsub, v4f32, v8f32>;
}
let ExeDomain = SSEPackedDouble in {
defm VFNMADDPD : fma3p_forms<0x9C, 0xAC, 0xBC, "vfnmadd", "pd", memopv2f64,
memopv4f64, X86Fnmadd, v2f64, v4f64>, VEX_W;
defm VFNMSUBPD : fma3p_forms<0x9E, 0xAE, 0xBE, "vfnmsub", "pd",
memopv2f64, memopv4f64, X86Fnmsub, v2f64,
v4f64>, VEX_W;
}
let Predicates = [HasFMA] in {
def : Pat<(int_x86_fma_vfmadd_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFMADDPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmadd_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFMADDPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsub_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFMSUBPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmsub_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFMSUBPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFMADDSUBPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFMADDSUBPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFMSUBADDPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFMSUBADDPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmadd_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMADDPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmadd_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFMADDPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsub_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMSUBPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmsub_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFMSUBPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMADDSUBPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFMADDSUBPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMSUBADDPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFMSUBADDPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmadd_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFMADDPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmadd_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFMADDPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsub_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFMSUBPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmsub_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFMSUBPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFMADDSUBPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFMADDSUBPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFMSUBADDPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFMSUBADDPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmadd_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMADDPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmadd_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFMADDPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsub_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMSUBPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmsub_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFMSUBPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMADDSUBPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmaddsub_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFMADDSUBPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFMSUBADDPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfmsubadd_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFMSUBADDPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFNMADDPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFNMADDPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_ps VR128:$src2, VR128:$src1, VR128:$src3),
(VFNMSUBPSr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_ps VR128:$src2, VR128:$src1,
(memopv4f32 addr:$src3)),
(VFNMSUBPSr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFNMADDPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFNMADDPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_ps_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFNMSUBPSr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_ps_256 VR256:$src2, VR256:$src1,
(memopv8f32 addr:$src3)),
(VFNMSUBPSr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFNMADDPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFNMADDPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_pd VR128:$src2, VR128:$src1, VR128:$src3),
(VFNMSUBPDr213r VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_pd VR128:$src2, VR128:$src1,
(memopv2f64 addr:$src3)),
(VFNMSUBPDr213m VR128:$src1, VR128:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFNMADDPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfnmadd_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFNMADDPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_pd_256 VR256:$src2, VR256:$src1, VR256:$src3),
(VFNMSUBPDr213rY VR256:$src1, VR256:$src2, VR256:$src3)>;
def : Pat<(int_x86_fma_vfnmsub_pd_256 VR256:$src2, VR256:$src1,
(memopv4f64 addr:$src3)),
(VFNMSUBPDr213mY VR256:$src1, VR256:$src2, addr:$src3)>;
} // Predicates = [HasFMA]
let Constraints = "$src1 = $dst" in {
multiclass fma3s_rm<bits<8> opc, string OpcodeStr, X86MemOperand x86memop,
RegisterClass RC> {
let neverHasSideEffects = 1 in {
def r : FMA3<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, RC:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
let mayLoad = 1 in
def m : FMA3<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, RC:$src2, x86memop:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
} // neverHasSideEffects = 1
}
multiclass fma3s_rm_int<bits<8> opc, string OpcodeStr, Operand memop,
ComplexPattern mem_cpat, Intrinsic IntId,
RegisterClass RC, SDNode OpNode, ValueType OpVT> {
def r_Int : FMA3<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR128:$dst, (IntId VR128:$src2, VR128:$src1,
VR128:$src3))]>;
def m_Int : FMA3<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, memop:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR128:$dst,
(IntId VR128:$src2, VR128:$src1, mem_cpat:$src3))]>;
def r : FMA3<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, RC:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set RC:$dst,
(OpVT (OpNode RC:$src2, RC:$src1, RC:$src3)))]>;
let mayLoad = 1 in
def m : FMA3<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, RC:$src2, memop:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"), []>;
}
} // Constraints = "$src1 = $dst"
multiclass fma3s_forms<bits<8> opc132, bits<8> opc213, bits<8> opc231,
string OpStr, Intrinsic IntF32, Intrinsic IntF64,
SDNode OpNode> {
defm SSr132 : fma3s_rm<opc132, !strconcat(OpStr, "132ss"), f32mem, FR32>;
defm SSr231 : fma3s_rm<opc231, !strconcat(OpStr, "231ss"), f32mem, FR32>;
defm SDr132 : fma3s_rm<opc132, !strconcat(OpStr, "132sd"), f64mem, FR64>,
VEX_W;
defm SDr231 : fma3s_rm<opc231, !strconcat(OpStr, "231sd"), f64mem, FR64>,
VEX_W;
defm SSr213 : fma3s_rm_int <opc213, !strconcat(OpStr, "213ss"), ssmem,
sse_load_f32, IntF32, FR32, OpNode, f32>;
defm SDr213 : fma3s_rm_int <opc213, !strconcat(OpStr, "213sd"), sdmem,
sse_load_f64, IntF64, FR64, OpNode, f64>, VEX_W;
}
defm VFMADD : fma3s_forms<0x99, 0xA9, 0xB9, "vfmadd", int_x86_fma_vfmadd_ss,
int_x86_fma_vfmadd_sd, X86Fmadd>, VEX_LIG;
defm VFMSUB : fma3s_forms<0x9B, 0xAB, 0xBB, "vfmsub", int_x86_fma_vfmsub_ss,
int_x86_fma_vfmsub_sd, X86Fmsub>, VEX_LIG;
defm VFNMADD : fma3s_forms<0x9D, 0xAD, 0xBD, "vfnmadd", int_x86_fma_vfnmadd_ss,
int_x86_fma_vfnmadd_sd, X86Fnmadd>, VEX_LIG;
defm VFNMSUB : fma3s_forms<0x9F, 0xAF, 0xBF, "vfnmsub", int_x86_fma_vfnmsub_ss,
int_x86_fma_vfnmsub_sd, X86Fnmsub>, VEX_LIG;
//===----------------------------------------------------------------------===//
// FMA4 - AMD 4 operand Fused Multiply-Add instructions
//===----------------------------------------------------------------------===//
multiclass fma4s<bits<8> opc, string OpcodeStr, Operand memop,
ComplexPattern mem_cpat, Intrinsic Int> {
def rr : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, VR128:$src2, VR128:$src3))]>, VEX_W, MemOp4;
def rm : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, memop:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, VR128:$src2, mem_cpat:$src3))]>, VEX_W, MemOp4;
def mr : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, memop:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, mem_cpat:$src2, VR128:$src3))]>;
// For disassembler
let isCodeGenOnly = 1 in
def rr_REV : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), []>;
}
multiclass fma4p<bits<8> opc, string OpcodeStr,
Intrinsic Int128, Intrinsic Int256,
PatFrag ld_frag128, PatFrag ld_frag256> {
def rr : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int128 VR128:$src1, VR128:$src2, VR128:$src3))]>, VEX_W, MemOp4;
def rm : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, f128mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst, (Int128 VR128:$src1, VR128:$src2,
(ld_frag128 addr:$src3)))]>, VEX_W, MemOp4;
def mr : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f128mem:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int128 VR128:$src1, (ld_frag128 addr:$src2), VR128:$src3))]>;
def rrY : FMA4<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst,
(Int256 VR256:$src1, VR256:$src2, VR256:$src3))]>, VEX_W, MemOp4;
def rmY : FMA4<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, f256mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst, (Int256 VR256:$src1, VR256:$src2,
(ld_frag256 addr:$src3)))]>, VEX_W, MemOp4;
def mrY : FMA4<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst,
(Int256 VR256:$src1, (ld_frag256 addr:$src2), VR256:$src3))]>;
// For disassembler
let isCodeGenOnly = 1 in {
def rr_REV : FMA4<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), []>;
def rrY_REV : FMA4<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), []>;
} // isCodeGenOnly = 1
}
let Predicates = [HasFMA4] in {
defm VFMADDSS4 : fma4s<0x6A, "vfmaddss", ssmem, sse_load_f32,
int_x86_fma_vfmadd_ss>;
defm VFMADDSD4 : fma4s<0x6B, "vfmaddsd", sdmem, sse_load_f64,
int_x86_fma_vfmadd_sd>;
defm VFMADDPS4 : fma4p<0x68, "vfmaddps", int_x86_fma_vfmadd_ps,
int_x86_fma_vfmadd_ps_256, memopv4f32, memopv8f32>;
defm VFMADDPD4 : fma4p<0x69, "vfmaddpd", int_x86_fma_vfmadd_pd,
int_x86_fma_vfmadd_pd_256, memopv2f64, memopv4f64>;
defm VFMSUBSS4 : fma4s<0x6E, "vfmsubss", ssmem, sse_load_f32,
int_x86_fma_vfmsub_ss>;
defm VFMSUBSD4 : fma4s<0x6F, "vfmsubsd", sdmem, sse_load_f64,
int_x86_fma_vfmsub_sd>;
defm VFMSUBPS4 : fma4p<0x6C, "vfmsubps", int_x86_fma_vfmsub_ps,
int_x86_fma_vfmsub_ps_256, memopv4f32, memopv8f32>;
defm VFMSUBPD4 : fma4p<0x6D, "vfmsubpd", int_x86_fma_vfmsub_pd,
int_x86_fma_vfmsub_pd_256, memopv2f64, memopv4f64>;
defm VFNMADDSS4 : fma4s<0x7A, "vfnmaddss", ssmem, sse_load_f32,
int_x86_fma_vfnmadd_ss>;
defm VFNMADDSD4 : fma4s<0x7B, "vfnmaddsd", sdmem, sse_load_f64,
int_x86_fma_vfnmadd_sd>;
defm VFNMADDPS4 : fma4p<0x78, "vfnmaddps", int_x86_fma_vfnmadd_ps,
int_x86_fma_vfnmadd_ps_256, memopv4f32, memopv8f32>;
defm VFNMADDPD4 : fma4p<0x79, "vfnmaddpd", int_x86_fma_vfnmadd_pd,
int_x86_fma_vfnmadd_pd_256, memopv2f64, memopv4f64>;
defm VFNMSUBSS4 : fma4s<0x7E, "vfnmsubss", ssmem, sse_load_f32,
int_x86_fma_vfnmsub_ss>;
defm VFNMSUBSD4 : fma4s<0x7F, "vfnmsubsd", sdmem, sse_load_f64,
int_x86_fma_vfnmsub_sd>;
defm VFNMSUBPS4 : fma4p<0x7C, "vfnmsubps", int_x86_fma_vfnmsub_ps,
int_x86_fma_vfnmsub_ps_256, memopv4f32, memopv8f32>;
defm VFNMSUBPD4 : fma4p<0x7D, "vfnmsubpd", int_x86_fma_vfnmsub_pd,
int_x86_fma_vfnmsub_pd_256, memopv2f64, memopv4f64>;
defm VFMADDSUBPS4 : fma4p<0x5C, "vfmaddsubps", int_x86_fma_vfmaddsub_ps,
int_x86_fma_vfmaddsub_ps_256, memopv4f32, memopv8f32>;
defm VFMADDSUBPD4 : fma4p<0x5D, "vfmaddsubpd", int_x86_fma_vfmaddsub_pd,
int_x86_fma_vfmaddsub_pd_256, memopv2f64, memopv4f64>;
defm VFMSUBADDPS4 : fma4p<0x5E, "vfmsubaddps", int_x86_fma_vfmsubadd_ps,
int_x86_fma_vfmsubadd_ps_256, memopv4f32, memopv8f32>;
defm VFMSUBADDPD4 : fma4p<0x5F, "vfmsubaddpd", int_x86_fma_vfmsubadd_pd,
int_x86_fma_vfmsubadd_pd_256, memopv2f64, memopv4f64>;
} // HasFMA4
|
