aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/llvm14/lib/Target/X86/X86InstrVecCompiler.td
blob: 2429aa113fb193320797e76eba0093ff29c710ac (plain) (blame)
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
463
464
465
466
467
468
469
470
471
472
473
//===- X86InstrVecCompiler.td - Vector Compiler 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 file describes the various vector pseudo instructions used by the
// compiler, as well as Pat patterns used during instruction selection.
//
//===----------------------------------------------------------------------===//

//===----------------------------------------------------------------------===//
//  Non-instruction patterns
//===----------------------------------------------------------------------===//

let Predicates = [NoAVX512] in {
  // A vector extract of the first f32/f64 position is a subregister copy
  def : Pat<(f32 (extractelt (v4f32 VR128:$src), (iPTR 0))),
            (COPY_TO_REGCLASS (v4f32 VR128:$src), FR32)>;
  def : Pat<(f64 (extractelt (v2f64 VR128:$src), (iPTR 0))),
            (COPY_TO_REGCLASS (v2f64 VR128:$src), FR64)>;
}

let Predicates = [HasAVX512] in {
  // A vector extract of the first f32/f64 position is a subregister copy
  def : Pat<(f16 (extractelt (v8f16 VR128X:$src), (iPTR 0))),
            (COPY_TO_REGCLASS (v8f16 VR128X:$src), FR16X)>;
  def : Pat<(f32 (extractelt (v4f32 VR128X:$src), (iPTR 0))),
            (COPY_TO_REGCLASS (v4f32 VR128X:$src), FR32X)>;
  def : Pat<(f64 (extractelt (v2f64 VR128X:$src), (iPTR 0))),
            (COPY_TO_REGCLASS (v2f64 VR128X:$src), FR64X)>;
}

let Predicates = [NoVLX] in {
  def : Pat<(v8f16 (scalar_to_vector FR16X:$src)),
            (COPY_TO_REGCLASS FR16X:$src, VR128)>;
  // Implicitly promote a 32-bit scalar to a vector.
  def : Pat<(v4f32 (scalar_to_vector FR32:$src)),
            (COPY_TO_REGCLASS FR32:$src, VR128)>;
  // Implicitly promote a 64-bit scalar to a vector.
  def : Pat<(v2f64 (scalar_to_vector FR64:$src)),
            (COPY_TO_REGCLASS FR64:$src, VR128)>;
}

let Predicates = [HasVLX] in {
  def : Pat<(v8f16 (scalar_to_vector FR16X:$src)),
            (COPY_TO_REGCLASS FR16X:$src, VR128X)>;
  // Implicitly promote a 32-bit scalar to a vector.
  def : Pat<(v4f32 (scalar_to_vector FR32X:$src)),
            (COPY_TO_REGCLASS FR32X:$src, VR128X)>;
  // Implicitly promote a 64-bit scalar to a vector.
  def : Pat<(v2f64 (scalar_to_vector FR64X:$src)),
            (COPY_TO_REGCLASS FR64X:$src, VR128X)>;
}

//===----------------------------------------------------------------------===//
// Subvector tricks
//===----------------------------------------------------------------------===//

// Patterns for insert_subvector/extract_subvector to/from index=0
multiclass subvector_subreg_lowering<RegisterClass subRC, ValueType subVT,
                                     RegisterClass RC, ValueType VT,
                                     SubRegIndex subIdx> {
  def : Pat<(subVT (extract_subvector (VT RC:$src), (iPTR 0))),
            (subVT (EXTRACT_SUBREG RC:$src, subIdx))>;

  def : Pat<(VT (insert_subvector undef, subRC:$src, (iPTR 0))),
            (VT (INSERT_SUBREG (IMPLICIT_DEF), subRC:$src, subIdx))>;
}

// A 128-bit subvector extract from the first 256-bit vector position is a
// subregister copy that needs no instruction. Likewise, a 128-bit subvector
// insert to the first 256-bit vector position is a subregister copy that needs
// no instruction.
defm : subvector_subreg_lowering<VR128, v4i32, VR256, v8i32,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v4f32, VR256, v8f32,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v2i64, VR256, v4i64,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v2f64, VR256, v4f64,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v8i16, VR256, v16i16, sub_xmm>;
defm : subvector_subreg_lowering<VR128, v16i8, VR256, v32i8,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v8f16, VR256, v16f16, sub_xmm>;

// A 128-bit subvector extract from the first 512-bit vector position is a
// subregister copy that needs no instruction. Likewise, a 128-bit subvector
// insert to the first 512-bit vector position is a subregister copy that needs
// no instruction.
defm : subvector_subreg_lowering<VR128, v4i32, VR512, v16i32, sub_xmm>;
defm : subvector_subreg_lowering<VR128, v4f32, VR512, v16f32, sub_xmm>;
defm : subvector_subreg_lowering<VR128, v2i64, VR512, v8i64,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v2f64, VR512, v8f64,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v8i16, VR512, v32i16, sub_xmm>;
defm : subvector_subreg_lowering<VR128, v16i8, VR512, v64i8,  sub_xmm>;
defm : subvector_subreg_lowering<VR128, v8f16, VR512, v32f16, sub_xmm>;

// A 128-bit subvector extract from the first 512-bit vector position is a
// subregister copy that needs no instruction. Likewise, a 128-bit subvector
// insert to the first 512-bit vector position is a subregister copy that needs
// no instruction.
defm : subvector_subreg_lowering<VR256, v8i32,  VR512, v16i32, sub_ymm>;
defm : subvector_subreg_lowering<VR256, v8f32,  VR512, v16f32, sub_ymm>;
defm : subvector_subreg_lowering<VR256, v4i64,  VR512, v8i64,  sub_ymm>;
defm : subvector_subreg_lowering<VR256, v4f64,  VR512, v8f64,  sub_ymm>;
defm : subvector_subreg_lowering<VR256, v16i16, VR512, v32i16, sub_ymm>;
defm : subvector_subreg_lowering<VR256, v32i8,  VR512, v64i8,  sub_ymm>;
defm : subvector_subreg_lowering<VR256, v16f16, VR512, v32f16, sub_ymm>;


// If we're inserting into an all zeros vector, just use a plain move which
// will zero the upper bits. A post-isel hook will take care of removing
// any moves that we can prove are unnecessary.
multiclass subvec_zero_lowering<string MoveStr,
                                RegisterClass RC, ValueType DstTy,
                                ValueType SrcTy, SubRegIndex SubIdx> {
  def : Pat<(DstTy (insert_subvector immAllZerosV,
                                     (SrcTy RC:$src), (iPTR 0))),
            (SUBREG_TO_REG (i64 0),
             (SrcTy (!cast<Instruction>("VMOV"#MoveStr#"rr") RC:$src)), SubIdx)>;
}

let Predicates = [HasAVX, NoVLX] in {
  defm : subvec_zero_lowering<"APD", VR128, v4f64, v2f64, sub_xmm>;
  defm : subvec_zero_lowering<"APS", VR128, v8f32, v4f32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v4i64, v2i64, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v8i32, v4i32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v16i16, v8i16, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v32i8, v16i8, sub_xmm>;
}

let Predicates = [HasVLX] in {
  defm : subvec_zero_lowering<"APDZ128", VR128X, v4f64, v2f64, sub_xmm>;
  defm : subvec_zero_lowering<"APSZ128", VR128X, v8f32, v4f32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v4i64, v2i64, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v8i32, v4i32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v16i16, v8i16, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v32i8, v16i8, sub_xmm>;

  defm : subvec_zero_lowering<"APDZ128", VR128X, v8f64, v2f64, sub_xmm>;
  defm : subvec_zero_lowering<"APSZ128", VR128X, v16f32, v4f32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v8i64, v2i64, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v16i32, v4i32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v32i16, v8i16, sub_xmm>;
  defm : subvec_zero_lowering<"DQA64Z128", VR128X, v64i8, v16i8, sub_xmm>;

  defm : subvec_zero_lowering<"APDZ256", VR256X, v8f64, v4f64, sub_ymm>;
  defm : subvec_zero_lowering<"APSZ256", VR256X, v16f32, v8f32, sub_ymm>;
  defm : subvec_zero_lowering<"DQA64Z256", VR256X, v8i64, v4i64, sub_ymm>;
  defm : subvec_zero_lowering<"DQA64Z256", VR256X, v16i32, v8i32, sub_ymm>;
  defm : subvec_zero_lowering<"DQA64Z256", VR256X, v32i16, v16i16, sub_ymm>;
  defm : subvec_zero_lowering<"DQA64Z256", VR256X, v64i8, v32i8, sub_ymm>;
}

let Predicates = [HasAVX512, NoVLX] in {
  defm : subvec_zero_lowering<"APD", VR128, v8f64, v2f64, sub_xmm>;
  defm : subvec_zero_lowering<"APS", VR128, v16f32, v4f32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v8i64, v2i64, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v16i32, v4i32, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v32i16, v8i16, sub_xmm>;
  defm : subvec_zero_lowering<"DQA", VR128, v64i8, v16i8, sub_xmm>;

  defm : subvec_zero_lowering<"APDY", VR256, v8f64, v4f64, sub_ymm>;
  defm : subvec_zero_lowering<"APSY", VR256, v16f32, v8f32, sub_ymm>;
  defm : subvec_zero_lowering<"DQAY", VR256, v8i64, v4i64, sub_ymm>;
  defm : subvec_zero_lowering<"DQAY", VR256, v16i32, v8i32, sub_ymm>;
  defm : subvec_zero_lowering<"DQAY", VR256, v32i16, v16i16, sub_ymm>;
  defm : subvec_zero_lowering<"DQAY", VR256, v64i8, v32i8, sub_ymm>;
}

let Predicates = [HasFP16, HasVLX] in {
  defm : subvec_zero_lowering<"APSZ128", VR128X, v16f16, v8f16, sub_xmm>;
  defm : subvec_zero_lowering<"APSZ128", VR128X, v32f16, v8f16, sub_xmm>;
  defm : subvec_zero_lowering<"APSZ256", VR256X, v32f16, v16f16, sub_ymm>;
}

class maskzeroupper<ValueType vt, RegisterClass RC> :
  PatLeaf<(vt RC:$src), [{
    return isMaskZeroExtended(N);
  }]>;

def maskzeroupperv1i1  : maskzeroupper<v1i1,  VK1>;
def maskzeroupperv2i1  : maskzeroupper<v2i1,  VK2>;
def maskzeroupperv4i1  : maskzeroupper<v4i1,  VK4>;
def maskzeroupperv8i1  : maskzeroupper<v8i1,  VK8>;
def maskzeroupperv16i1 : maskzeroupper<v16i1, VK16>;
def maskzeroupperv32i1 : maskzeroupper<v32i1, VK32>;

// The patterns determine if we can depend on the upper bits of a mask register
// being zeroed by the previous operation so that we can skip explicit
// zeroing.
let Predicates = [HasBWI] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     maskzeroupperv1i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK1:$src, VK32)>;
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     maskzeroupperv8i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK8:$src, VK32)>;
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     maskzeroupperv16i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK16:$src, VK32)>;

  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv1i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK1:$src, VK64)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv8i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK8:$src, VK64)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv16i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK16:$src, VK64)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv32i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK32:$src, VK64)>;
}

let Predicates = [HasAVX512] in {
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     maskzeroupperv1i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK1:$src, VK16)>;
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     maskzeroupperv8i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK8:$src, VK16)>;
}

let Predicates = [HasDQI] in {
  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    maskzeroupperv1i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK1:$src, VK8)>;
}

let Predicates = [HasVLX, HasDQI] in {
  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    maskzeroupperv2i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK2:$src, VK8)>;
  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    maskzeroupperv4i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK4:$src, VK8)>;
}

let Predicates = [HasVLX] in {
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     maskzeroupperv2i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK2:$src, VK16)>;
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     maskzeroupperv4i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK4:$src, VK16)>;
}

let Predicates = [HasBWI, HasVLX] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     maskzeroupperv2i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK2:$src, VK32)>;
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     maskzeroupperv4i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK4:$src, VK32)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv2i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK2:$src, VK64)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     maskzeroupperv4i1:$src, (iPTR 0))),
            (COPY_TO_REGCLASS VK4:$src, VK64)>;
}

// If the bits are not zero we have to fall back to explicitly zeroing by
// using shifts.
let Predicates = [HasAVX512] in {
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     (v1i1 VK1:$mask), (iPTR 0))),
            (KSHIFTRWri (KSHIFTLWri (COPY_TO_REGCLASS VK1:$mask, VK16),
                                    (i8 15)), (i8 15))>;

  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     (v2i1 VK2:$mask), (iPTR 0))),
            (KSHIFTRWri (KSHIFTLWri (COPY_TO_REGCLASS VK2:$mask, VK16),
                                    (i8 14)), (i8 14))>;

  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     (v4i1 VK4:$mask), (iPTR 0))),
            (KSHIFTRWri (KSHIFTLWri (COPY_TO_REGCLASS VK4:$mask, VK16),
                                    (i8 12)), (i8 12))>;
}

let Predicates = [HasAVX512, NoDQI] in {
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (KSHIFTRWri (KSHIFTLWri (COPY_TO_REGCLASS VK8:$mask, VK16),
                                    (i8 8)), (i8 8))>;
}

let Predicates = [HasDQI] in {
  def : Pat<(v16i1 (insert_subvector (v16i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVBkk VK8:$mask), VK16)>;

  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    (v1i1 VK1:$mask), (iPTR 0))),
            (KSHIFTRBri (KSHIFTLBri (COPY_TO_REGCLASS VK1:$mask, VK8),
                                    (i8 7)), (i8 7))>;
  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    (v2i1 VK2:$mask), (iPTR 0))),
            (KSHIFTRBri (KSHIFTLBri (COPY_TO_REGCLASS VK2:$mask, VK8),
                                    (i8 6)), (i8 6))>;
  def : Pat<(v8i1 (insert_subvector (v8i1 immAllZerosV),
                                    (v4i1 VK4:$mask), (iPTR 0))),
            (KSHIFTRBri (KSHIFTLBri (COPY_TO_REGCLASS VK4:$mask, VK8),
                                    (i8 4)), (i8 4))>;
}

let Predicates = [HasBWI] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v16i1 VK16:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVWkk VK16:$mask), VK32)>;

  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v16i1 VK16:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVWkk VK16:$mask), VK64)>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v32i1 VK32:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVDkk VK32:$mask), VK64)>;
}

let Predicates = [HasBWI, NoDQI] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (KSHIFTRDri (KSHIFTLDri (COPY_TO_REGCLASS VK8:$mask, VK32),
                                    (i8 24)), (i8 24))>;

  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (KSHIFTRQri (KSHIFTLQri (COPY_TO_REGCLASS VK8:$mask, VK64),
                                    (i8 56)), (i8 56))>;
}

let Predicates = [HasBWI, HasDQI] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVBkk VK8:$mask), VK32)>;

  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v8i1 VK8:$mask), (iPTR 0))),
            (COPY_TO_REGCLASS (KMOVBkk VK8:$mask), VK64)>;
}

let Predicates = [HasBWI] in {
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v1i1 VK1:$mask), (iPTR 0))),
            (KSHIFTRDri (KSHIFTLDri (COPY_TO_REGCLASS VK1:$mask, VK32),
                                    (i8 31)), (i8 31))>;
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v2i1 VK2:$mask), (iPTR 0))),
            (KSHIFTRDri (KSHIFTLDri (COPY_TO_REGCLASS VK2:$mask, VK32),
                                    (i8 30)), (i8 30))>;
  def : Pat<(v32i1 (insert_subvector (v32i1 immAllZerosV),
                                     (v4i1 VK4:$mask), (iPTR 0))),
            (KSHIFTRDri (KSHIFTLDri (COPY_TO_REGCLASS VK4:$mask, VK32),
                                    (i8 28)), (i8 28))>;

  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v1i1 VK1:$mask), (iPTR 0))),
            (KSHIFTRQri (KSHIFTLQri (COPY_TO_REGCLASS VK1:$mask, VK64),
                                    (i8 63)), (i8 63))>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v2i1 VK2:$mask), (iPTR 0))),
            (KSHIFTRQri (KSHIFTLQri (COPY_TO_REGCLASS VK2:$mask, VK64),
                                    (i8 62)), (i8 62))>;
  def : Pat<(v64i1 (insert_subvector (v64i1 immAllZerosV),
                                     (v4i1 VK4:$mask), (iPTR 0))),
            (KSHIFTRQri (KSHIFTLQri (COPY_TO_REGCLASS VK4:$mask, VK64),
                                    (i8 60)), (i8 60))>;
}

//===----------------------------------------------------------------------===//
// Extra selection patterns for f128, f128mem

// movaps is shorter than movdqa. movaps is in SSE and movdqa is in SSE2.
let Predicates = [NoAVX] in {
def : Pat<(alignedstore (f128 VR128:$src), addr:$dst),
          (MOVAPSmr addr:$dst, VR128:$src)>;
def : Pat<(store (f128 VR128:$src), addr:$dst),
          (MOVUPSmr addr:$dst, VR128:$src)>;

def : Pat<(alignedloadf128 addr:$src),
          (MOVAPSrm addr:$src)>;
def : Pat<(loadf128 addr:$src),
          (MOVUPSrm addr:$src)>;
}

let Predicates = [HasAVX, NoVLX] in {
def : Pat<(alignedstore (f128 VR128:$src), addr:$dst),
          (VMOVAPSmr addr:$dst, VR128:$src)>;
def : Pat<(store (f128 VR128:$src), addr:$dst),
          (VMOVUPSmr addr:$dst, VR128:$src)>;

def : Pat<(alignedloadf128 addr:$src),
          (VMOVAPSrm addr:$src)>;
def : Pat<(loadf128 addr:$src),
          (VMOVUPSrm addr:$src)>;
}

let Predicates = [HasVLX] in {
def : Pat<(alignedstore (f128 VR128X:$src), addr:$dst),
          (VMOVAPSZ128mr addr:$dst, VR128X:$src)>;
def : Pat<(store (f128 VR128X:$src), addr:$dst),
          (VMOVUPSZ128mr addr:$dst, VR128X:$src)>;

def : Pat<(alignedloadf128 addr:$src),
          (VMOVAPSZ128rm addr:$src)>;
def : Pat<(loadf128 addr:$src),
          (VMOVUPSZ128rm addr:$src)>;
}

let Predicates = [UseSSE1] in {
// andps is shorter than andpd or pand. andps is SSE and andpd/pand are in SSE2
def : Pat<(f128 (X86fand VR128:$src1, (memopf128 addr:$src2))),
          (ANDPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fand VR128:$src1, VR128:$src2)),
          (ANDPSrr VR128:$src1, VR128:$src2)>;

def : Pat<(f128 (X86for VR128:$src1, (memopf128 addr:$src2))),
          (ORPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86for VR128:$src1, VR128:$src2)),
          (ORPSrr VR128:$src1, VR128:$src2)>;

def : Pat<(f128 (X86fxor VR128:$src1, (memopf128 addr:$src2))),
          (XORPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fxor VR128:$src1, VR128:$src2)),
          (XORPSrr VR128:$src1, VR128:$src2)>;
}

let Predicates = [HasAVX, NoVLX] in {
// andps is shorter than andpd or pand. andps is SSE and andpd/pand are in SSE2
def : Pat<(f128 (X86fand VR128:$src1, (loadf128 addr:$src2))),
          (VANDPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fand VR128:$src1, VR128:$src2)),
          (VANDPSrr VR128:$src1, VR128:$src2)>;

def : Pat<(f128 (X86for VR128:$src1, (loadf128 addr:$src2))),
          (VORPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86for VR128:$src1, VR128:$src2)),
          (VORPSrr VR128:$src1, VR128:$src2)>;

def : Pat<(f128 (X86fxor VR128:$src1, (loadf128 addr:$src2))),
          (VXORPSrm VR128:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fxor VR128:$src1, VR128:$src2)),
          (VXORPSrr VR128:$src1, VR128:$src2)>;
}

let Predicates = [HasVLX] in {
// andps is shorter than andpd or pand. andps is SSE and andpd/pand are in SSE2
def : Pat<(f128 (X86fand VR128X:$src1, (loadf128 addr:$src2))),
          (VANDPSZ128rm VR128X:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fand VR128X:$src1, VR128X:$src2)),
          (VANDPSZ128rr VR128X:$src1, VR128X:$src2)>;

def : Pat<(f128 (X86for VR128X:$src1, (loadf128 addr:$src2))),
          (VORPSZ128rm VR128X:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86for VR128X:$src1, VR128X:$src2)),
          (VORPSZ128rr VR128X:$src1, VR128X:$src2)>;

def : Pat<(f128 (X86fxor VR128X:$src1, (loadf128 addr:$src2))),
          (VXORPSZ128rm VR128X:$src1, f128mem:$src2)>;

def : Pat<(f128 (X86fxor VR128X:$src1, VR128X:$src2)),
          (VXORPSZ128rr VR128X:$src1, VR128X:$src2)>;
}