llvm.org GIT mirror llvm / release_50 lib / Target / AMDGPU / FLATInstructions.td
release_50

Tree @release_50 (Download .tar.gz)

FLATInstructions.td @release_50raw · history · 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
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
//===-- FLATInstructions.td - FLAT Instruction Defintions -----------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//

def FLATAtomic : ComplexPattern<i64, 3, "SelectFlatAtomic", [], [], -10>;
def FLATOffset : ComplexPattern<i64, 3, "SelectFlat", [], [], -10>;

//===----------------------------------------------------------------------===//
// FLAT classes
//===----------------------------------------------------------------------===//

class FLAT_Pseudo<string opName, dag outs, dag ins,
                  string asmOps, list<dag> pattern=[]> :
  InstSI<outs, ins, "", pattern>,
  SIMCInstr<opName, SIEncodingFamily.NONE> {

  let isPseudo = 1;
  let isCodeGenOnly = 1;

  let SubtargetPredicate = isCIVI;

  let FLAT = 1;
  // Internally, FLAT instruction are executed as both an LDS and a
  // Buffer instruction; so, they increment both VM_CNT and LGKM_CNT
  // and are not considered done until both have been decremented.
  let VM_CNT = 1;
  let LGKM_CNT = 1;

  let UseNamedOperandTable = 1;
  let hasSideEffects = 0;
  let SchedRW = [WriteVMEM];

  string Mnemonic = opName;
  string AsmOperands = asmOps;

  bits<1> is_flat_global = 0;
  bits<1> is_flat_scratch = 0;

  bits<1> has_vdst = 1;
  bits<1> has_data = 1;
  bits<1> has_glc  = 1;
  bits<1> glcValue = 0;

  // TODO: M0 if it could possibly access LDS (before gfx9? only)?
  let Uses = !if(is_flat_global, [EXEC], [EXEC, FLAT_SCR]);
}

class FLAT_Real <bits<7> op, FLAT_Pseudo ps> :
  InstSI <ps.OutOperandList, ps.InOperandList, ps.Mnemonic # ps.AsmOperands, []>,
  Enc64 {

  let isPseudo = 0;
  let isCodeGenOnly = 0;

  // copy relevant pseudo op flags
  let SubtargetPredicate = ps.SubtargetPredicate;
  let AsmMatchConverter  = ps.AsmMatchConverter;
  let TSFlags = ps.TSFlags;
  let UseNamedOperandTable = ps.UseNamedOperandTable;

  // encoding fields
  bits<8> vaddr;
  bits<8> vdata;
  bits<8> vdst;
  bits<1> slc;
  bits<1> glc;

  // Only valid on gfx9
  bits<1> lds = 0; // XXX - What does this actually do?

  // Segment, 00=flat, 01=scratch, 10=global, 11=reserved
  bits<2> seg = !if(ps.is_flat_global, 0b10,
                  !if(ps.is_flat_scratch, 0b01, 0));

  // Signed offset. Highest bit ignored for flat and treated as 12-bit
  // unsigned for flat acceses.
  bits<13> offset;
  bits<1> nv = 0; // XXX - What does this actually do?

  // We don't use tfe right now, and it was removed in gfx9.
  bits<1> tfe = 0;

  // Only valid on GFX9+
  let Inst{12-0} = offset;
  let Inst{13} = lds;
  let Inst{15-14} = seg;

  let Inst{16}    = !if(ps.has_glc, glc, ps.glcValue);
  let Inst{17}    = slc;
  let Inst{24-18} = op;
  let Inst{31-26} = 0x37; // Encoding.
  let Inst{39-32} = vaddr;
  let Inst{47-40} = !if(ps.has_data, vdata, ?);
  // 54-48 is reserved.
  let Inst{55}    = nv; // nv on GFX9+, TFE before.
  let Inst{63-56} = !if(ps.has_vdst, vdst, ?);
}

class FLAT_Load_Pseudo <string opName, RegisterClass regClass,
  bit HasSignedOffset = 0> : FLAT_Pseudo<
  opName,
  (outs regClass:$vdst),
  !if(HasSignedOffset,
    (ins VReg_64:$vaddr, offset_s13:$offset, GLC:$glc, slc:$slc),
    (ins VReg_64:$vaddr, offset_u12:$offset, GLC:$glc, slc:$slc)),
  " $vdst, $vaddr$offset$glc$slc"> {
  let has_data = 0;
  let mayLoad = 1;
}

class FLAT_Global_Load_Pseudo<string opName, RegisterClass regClass> :
  FLAT_Load_Pseudo<opName, regClass, 1> {
  let is_flat_global = 1;
}

class FLAT_Scratch_Load_Pseudo<string opName, RegisterClass regClass> :
  FLAT_Load_Pseudo<opName, regClass, 1> {
  let is_flat_scratch = 1;
}

class FLAT_Store_Pseudo <string opName, RegisterClass vdataClass,
  bit HasSignedOffset = 0> : FLAT_Pseudo<
  opName,
  (outs),
  !if(HasSignedOffset,
    (ins VReg_64:$vaddr, vdataClass:$vdata, offset_s13:$offset, GLC:$glc, slc:$slc),
    (ins VReg_64:$vaddr, vdataClass:$vdata, offset_u12:$offset, GLC:$glc, slc:$slc)),
  " $vaddr, $vdata$offset$glc$slc"> {
  let mayLoad  = 0;
  let mayStore = 1;
  let has_vdst = 0;
}

class FLAT_Global_Store_Pseudo<string opName, RegisterClass regClass> :
  FLAT_Store_Pseudo<opName, regClass, 1> {
  let is_flat_global = 1;
}

class FLAT_Scratch_Store_Pseudo<string opName, RegisterClass regClass> :
  FLAT_Store_Pseudo<opName, regClass, 1> {
  let is_flat_scratch = 1;
}

multiclass FLAT_Atomic_Pseudo<
  string opName,
  RegisterClass vdst_rc,
  ValueType vt,
  SDPatternOperator atomic = null_frag,
  ValueType data_vt = vt,
  RegisterClass data_rc = vdst_rc,
  bit HasSignedOffset = 0> {

  def "" : FLAT_Pseudo <opName,
    (outs),
    !if(HasSignedOffset,
      (ins VReg_64:$vaddr, data_rc:$vdata, offset_s13:$offset, slc:$slc),
      (ins VReg_64:$vaddr, data_rc:$vdata, offset_u12:$offset, slc:$slc)),
    " $vaddr, $vdata$offset$slc",
    []>,
    AtomicNoRet <NAME, 0> {
    let mayLoad = 1;
    let mayStore = 1;
    let has_glc  = 0;
    let glcValue = 0;
    let has_vdst = 0;
    let PseudoInstr = NAME;
  }

  def _RTN : FLAT_Pseudo <opName,
    (outs vdst_rc:$vdst),
    !if(HasSignedOffset,
      (ins VReg_64:$vaddr, data_rc:$vdata, offset_s13:$offset, slc:$slc),
      (ins VReg_64:$vaddr, data_rc:$vdata, offset_u12:$offset, slc:$slc)),
    " $vdst, $vaddr, $vdata$offset glc$slc",
    [(set vt:$vdst,
      (atomic (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc), data_vt:$vdata))]>,
    AtomicNoRet <NAME, 1> {
    let mayLoad  = 1;
    let mayStore = 1;
    let hasPostISelHook = 1;
    let has_glc  = 0;
    let glcValue = 1;
    let PseudoInstr = NAME # "_RTN";
  }
}

class flat_binary_atomic_op<SDNode atomic_op> : PatFrag<
  (ops node:$ptr, node:$value),
  (atomic_op node:$ptr, node:$value),
  [{return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUASI.FLAT_ADDRESS;}]
>;

def atomic_cmp_swap_flat : flat_binary_atomic_op<AMDGPUatomic_cmp_swap>;
def atomic_swap_flat     : flat_binary_atomic_op<atomic_swap>;
def atomic_add_flat      : flat_binary_atomic_op<atomic_load_add>;
def atomic_and_flat      : flat_binary_atomic_op<atomic_load_and>;
def atomic_max_flat      : flat_binary_atomic_op<atomic_load_max>;
def atomic_min_flat      : flat_binary_atomic_op<atomic_load_min>;
def atomic_or_flat       : flat_binary_atomic_op<atomic_load_or>;
def atomic_sub_flat      : flat_binary_atomic_op<atomic_load_sub>;
def atomic_umax_flat     : flat_binary_atomic_op<atomic_load_umax>;
def atomic_umin_flat     : flat_binary_atomic_op<atomic_load_umin>;
def atomic_xor_flat      : flat_binary_atomic_op<atomic_load_xor>;
def atomic_inc_flat      : flat_binary_atomic_op<SIatomic_inc>;
def atomic_dec_flat      : flat_binary_atomic_op<SIatomic_dec>;



//===----------------------------------------------------------------------===//
// Flat Instructions
//===----------------------------------------------------------------------===//

def FLAT_LOAD_UBYTE    : FLAT_Load_Pseudo <"flat_load_ubyte", VGPR_32>;
def FLAT_LOAD_SBYTE    : FLAT_Load_Pseudo <"flat_load_sbyte", VGPR_32>;
def FLAT_LOAD_USHORT   : FLAT_Load_Pseudo <"flat_load_ushort", VGPR_32>;
def FLAT_LOAD_SSHORT   : FLAT_Load_Pseudo <"flat_load_sshort", VGPR_32>;
def FLAT_LOAD_DWORD    : FLAT_Load_Pseudo <"flat_load_dword", VGPR_32>;
def FLAT_LOAD_DWORDX2  : FLAT_Load_Pseudo <"flat_load_dwordx2", VReg_64>;
def FLAT_LOAD_DWORDX4  : FLAT_Load_Pseudo <"flat_load_dwordx4", VReg_128>;
def FLAT_LOAD_DWORDX3  : FLAT_Load_Pseudo <"flat_load_dwordx3", VReg_96>;

def FLAT_STORE_BYTE    : FLAT_Store_Pseudo <"flat_store_byte", VGPR_32>;
def FLAT_STORE_SHORT   : FLAT_Store_Pseudo <"flat_store_short", VGPR_32>;
def FLAT_STORE_DWORD   : FLAT_Store_Pseudo <"flat_store_dword", VGPR_32>;
def FLAT_STORE_DWORDX2 : FLAT_Store_Pseudo <"flat_store_dwordx2", VReg_64>;
def FLAT_STORE_DWORDX4 : FLAT_Store_Pseudo <"flat_store_dwordx4", VReg_128>;
def FLAT_STORE_DWORDX3 : FLAT_Store_Pseudo <"flat_store_dwordx3", VReg_96>;

defm FLAT_ATOMIC_CMPSWAP    : FLAT_Atomic_Pseudo <"flat_atomic_cmpswap",
                                VGPR_32, i32, atomic_cmp_swap_flat,
                                v2i32, VReg_64>;

defm FLAT_ATOMIC_CMPSWAP_X2 : FLAT_Atomic_Pseudo <"flat_atomic_cmpswap_x2",
                                VReg_64, i64, atomic_cmp_swap_flat,
                                v2i64, VReg_128>;

defm FLAT_ATOMIC_SWAP       : FLAT_Atomic_Pseudo <"flat_atomic_swap",
                                VGPR_32, i32, atomic_swap_flat>;

defm FLAT_ATOMIC_SWAP_X2    : FLAT_Atomic_Pseudo <"flat_atomic_swap_x2",
                                VReg_64, i64, atomic_swap_flat>;

defm FLAT_ATOMIC_ADD        : FLAT_Atomic_Pseudo <"flat_atomic_add",
                                VGPR_32, i32, atomic_add_flat>;

defm FLAT_ATOMIC_SUB        : FLAT_Atomic_Pseudo <"flat_atomic_sub",
                                VGPR_32, i32, atomic_sub_flat>;

defm FLAT_ATOMIC_SMIN       : FLAT_Atomic_Pseudo <"flat_atomic_smin",
                                VGPR_32, i32, atomic_min_flat>;

defm FLAT_ATOMIC_UMIN       : FLAT_Atomic_Pseudo <"flat_atomic_umin",
                                VGPR_32, i32, atomic_umin_flat>;

defm FLAT_ATOMIC_SMAX       : FLAT_Atomic_Pseudo <"flat_atomic_smax",
                                VGPR_32, i32, atomic_max_flat>;

defm FLAT_ATOMIC_UMAX       : FLAT_Atomic_Pseudo <"flat_atomic_umax",
                                VGPR_32, i32, atomic_umax_flat>;

defm FLAT_ATOMIC_AND        : FLAT_Atomic_Pseudo <"flat_atomic_and",
                                VGPR_32, i32, atomic_and_flat>;

defm FLAT_ATOMIC_OR         : FLAT_Atomic_Pseudo <"flat_atomic_or",
                                VGPR_32, i32, atomic_or_flat>;

defm FLAT_ATOMIC_XOR        : FLAT_Atomic_Pseudo <"flat_atomic_xor",
                                VGPR_32, i32, atomic_xor_flat>;

defm FLAT_ATOMIC_INC        : FLAT_Atomic_Pseudo <"flat_atomic_inc",
                                VGPR_32, i32, atomic_inc_flat>;

defm FLAT_ATOMIC_DEC        : FLAT_Atomic_Pseudo <"flat_atomic_dec",
                                VGPR_32, i32, atomic_dec_flat>;

defm FLAT_ATOMIC_ADD_X2     : FLAT_Atomic_Pseudo <"flat_atomic_add_x2",
                                VReg_64, i64, atomic_add_flat>;

defm FLAT_ATOMIC_SUB_X2     : FLAT_Atomic_Pseudo <"flat_atomic_sub_x2",
                                VReg_64, i64, atomic_sub_flat>;

defm FLAT_ATOMIC_SMIN_X2    : FLAT_Atomic_Pseudo <"flat_atomic_smin_x2",
                                VReg_64, i64, atomic_min_flat>;

defm FLAT_ATOMIC_UMIN_X2    : FLAT_Atomic_Pseudo <"flat_atomic_umin_x2",
                                VReg_64, i64, atomic_umin_flat>;

defm FLAT_ATOMIC_SMAX_X2    : FLAT_Atomic_Pseudo <"flat_atomic_smax_x2",
                                VReg_64, i64, atomic_max_flat>;

defm FLAT_ATOMIC_UMAX_X2    : FLAT_Atomic_Pseudo <"flat_atomic_umax_x2",
                                VReg_64, i64, atomic_umax_flat>;

defm FLAT_ATOMIC_AND_X2     : FLAT_Atomic_Pseudo <"flat_atomic_and_x2",
                                VReg_64, i64, atomic_and_flat>;

defm FLAT_ATOMIC_OR_X2      : FLAT_Atomic_Pseudo <"flat_atomic_or_x2",
                                VReg_64, i64, atomic_or_flat>;

defm FLAT_ATOMIC_XOR_X2     : FLAT_Atomic_Pseudo <"flat_atomic_xor_x2",
                                VReg_64, i64, atomic_xor_flat>;

defm FLAT_ATOMIC_INC_X2     : FLAT_Atomic_Pseudo <"flat_atomic_inc_x2",
                                VReg_64, i64, atomic_inc_flat>;

defm FLAT_ATOMIC_DEC_X2     : FLAT_Atomic_Pseudo <"flat_atomic_dec_x2",
                                VReg_64, i64, atomic_dec_flat>;

let SubtargetPredicate = isCI in { // CI Only flat instructions : FIXME Only?

defm FLAT_ATOMIC_FCMPSWAP    : FLAT_Atomic_Pseudo <"flat_atomic_fcmpswap",
                                VGPR_32, f32, null_frag, v2f32, VReg_64>;

defm FLAT_ATOMIC_FCMPSWAP_X2 : FLAT_Atomic_Pseudo <"flat_atomic_fcmpswap_x2",
                                VReg_64, f64, null_frag, v2f64, VReg_128>;

defm FLAT_ATOMIC_FMIN        : FLAT_Atomic_Pseudo <"flat_atomic_fmin",
                                VGPR_32, f32>;

defm FLAT_ATOMIC_FMAX        : FLAT_Atomic_Pseudo <"flat_atomic_fmax",
                                VGPR_32, f32>;

defm FLAT_ATOMIC_FMIN_X2     : FLAT_Atomic_Pseudo <"flat_atomic_fmin_x2",
                                VReg_64, f64>;

defm FLAT_ATOMIC_FMAX_X2     : FLAT_Atomic_Pseudo <"flat_atomic_fmax_x2",
                                VReg_64, f64>;

} // End SubtargetPredicate = isCI

let SubtargetPredicate = HasFlatGlobalInsts in {
def GLOBAL_LOAD_UBYTE    : FLAT_Global_Load_Pseudo <"global_load_ubyte", VGPR_32>;
def GLOBAL_LOAD_SBYTE    : FLAT_Global_Load_Pseudo <"global_load_sbyte", VGPR_32>;
def GLOBAL_LOAD_USHORT   : FLAT_Global_Load_Pseudo <"global_load_ushort", VGPR_32>;
def GLOBAL_LOAD_SSHORT   : FLAT_Global_Load_Pseudo <"global_load_sshort", VGPR_32>;
def GLOBAL_LOAD_DWORD    : FLAT_Global_Load_Pseudo <"global_load_dword", VGPR_32>;
def GLOBAL_LOAD_DWORDX2  : FLAT_Global_Load_Pseudo <"global_load_dwordx2", VReg_64>;
def GLOBAL_LOAD_DWORDX3  : FLAT_Global_Load_Pseudo <"global_load_dwordx3", VReg_96>;
def GLOBAL_LOAD_DWORDX4  : FLAT_Global_Load_Pseudo <"global_load_dwordx4", VReg_128>;

def GLOBAL_STORE_BYTE    : FLAT_Global_Store_Pseudo <"global_store_byte", VGPR_32>;
def GLOBAL_STORE_SHORT   : FLAT_Global_Store_Pseudo <"global_store_short", VGPR_32>;
def GLOBAL_STORE_DWORD   : FLAT_Global_Store_Pseudo <"global_store_dword", VGPR_32>;
def GLOBAL_STORE_DWORDX2 : FLAT_Global_Store_Pseudo <"global_store_dwordx2", VReg_64>;
def GLOBAL_STORE_DWORDX3 : FLAT_Global_Store_Pseudo <"global_store_dwordx3", VReg_96>;
def GLOBAL_STORE_DWORDX4 : FLAT_Global_Store_Pseudo <"global_store_dwordx4", VReg_128>;

} // End SubtargetPredicate = HasFlatGlobalInsts


//===----------------------------------------------------------------------===//
// Flat Patterns
//===----------------------------------------------------------------------===//

class flat_ld <SDPatternOperator ld> : PatFrag<(ops node:$ptr),
                                               (ld node:$ptr), [{
  auto const AS = cast<MemSDNode>(N)->getAddressSpace();
  return AS == AMDGPUASI.FLAT_ADDRESS ||
         AS == AMDGPUASI.GLOBAL_ADDRESS ||
         AS == AMDGPUASI.CONSTANT_ADDRESS;
}]>;

class flat_st <SDPatternOperator st> : PatFrag<(ops node:$val, node:$ptr),
                                               (st node:$val, node:$ptr), [{
  auto const AS = cast<MemSDNode>(N)->getAddressSpace();
  return AS == AMDGPUASI.FLAT_ADDRESS ||
         AS == AMDGPUASI.GLOBAL_ADDRESS;
}]>;

def atomic_flat_load   : flat_ld <atomic_load>;
def flat_load          : flat_ld <load>;
def flat_az_extloadi8  : flat_ld <az_extloadi8>;
def flat_sextloadi8    : flat_ld <sextloadi8>;
def flat_az_extloadi16 : flat_ld <az_extloadi16>;
def flat_sextloadi16   : flat_ld <sextloadi16>;

def atomic_flat_store  : flat_st <atomic_store>;
def flat_store         : flat_st <store>;
def flat_truncstorei8  : flat_st <truncstorei8>;
def flat_truncstorei16 : flat_st <truncstorei16>;

// Patterns for global loads with no offset.
class FlatLoadPat <FLAT_Pseudo inst, SDPatternOperator node, ValueType vt> : Pat <
  (vt (node (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc))),
  (inst $vaddr, $offset, 0, $slc)
>;

class FlatLoadAtomicPat <FLAT_Pseudo inst, SDPatternOperator node, ValueType vt> : Pat <
  (vt (node (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc))),
  (inst $vaddr, $offset, 1, $slc)
>;

class FlatStorePat <FLAT_Pseudo inst, SDPatternOperator node, ValueType vt> : Pat <
  (node vt:$data, (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc)),
  (inst $vaddr, $data, $offset, 0, $slc)
>;

class FlatStoreAtomicPat <FLAT_Pseudo inst, SDPatternOperator node, ValueType vt> : Pat <
  // atomic store follows atomic binop convention so the address comes
  // first.
  (node (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc), vt:$data),
  (inst $vaddr, $data, $offset, 1, $slc)
>;

class FlatAtomicPat <FLAT_Pseudo inst, SDPatternOperator node, ValueType vt,
                     ValueType data_vt = vt> : Pat <
  (vt (node (FLATAtomic i64:$vaddr, i16:$offset, i1:$slc), data_vt:$data)),
  (inst $vaddr, $data, $offset, $slc)
>;

let Predicates = [isCIVI] in {

def : FlatLoadPat <FLAT_LOAD_UBYTE, flat_az_extloadi8, i32>;
def : FlatLoadPat <FLAT_LOAD_SBYTE, flat_sextloadi8, i32>;
def : FlatLoadPat <FLAT_LOAD_UBYTE, flat_az_extloadi8, i16>;
def : FlatLoadPat <FLAT_LOAD_SBYTE, flat_sextloadi8, i16>;
def : FlatLoadPat <FLAT_LOAD_USHORT, flat_az_extloadi16, i32>;
def : FlatLoadPat <FLAT_LOAD_SSHORT, flat_sextloadi16, i32>;
def : FlatLoadPat <FLAT_LOAD_DWORD, flat_load, i32>;
def : FlatLoadPat <FLAT_LOAD_DWORDX2, flat_load, v2i32>;
def : FlatLoadPat <FLAT_LOAD_DWORDX4, flat_load, v4i32>;

def : FlatLoadAtomicPat <FLAT_LOAD_DWORD, atomic_flat_load, i32>;
def : FlatLoadAtomicPat <FLAT_LOAD_DWORDX2, atomic_flat_load, i64>;

def : FlatStorePat <FLAT_STORE_BYTE, flat_truncstorei8, i32>;
def : FlatStorePat <FLAT_STORE_SHORT, flat_truncstorei16, i32>;
def : FlatStorePat <FLAT_STORE_DWORD, flat_store, i32>;
def : FlatStorePat <FLAT_STORE_DWORDX2, flat_store, v2i32>;
def : FlatStorePat <FLAT_STORE_DWORDX4, flat_store, v4i32>;

def : FlatStoreAtomicPat <FLAT_STORE_DWORD, atomic_flat_store, i32>;
def : FlatStoreAtomicPat <FLAT_STORE_DWORDX2, atomic_flat_store, i64>;

def : FlatAtomicPat <FLAT_ATOMIC_ADD_RTN, atomic_add_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SUB_RTN, atomic_sub_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_INC_RTN, atomic_inc_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_DEC_RTN, atomic_dec_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_AND_RTN, atomic_and_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SMAX_RTN, atomic_max_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_UMAX_RTN, atomic_umax_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SMIN_RTN, atomic_min_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_UMIN_RTN, atomic_umin_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_OR_RTN, atomic_or_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SWAP_RTN, atomic_swap_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_CMPSWAP_RTN, AMDGPUatomic_cmp_swap_global, i32, v2i32>;
def : FlatAtomicPat <FLAT_ATOMIC_XOR_RTN, atomic_xor_global, i32>;

def : FlatAtomicPat <FLAT_ATOMIC_ADD_X2_RTN, atomic_add_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_SUB_X2_RTN, atomic_sub_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_INC_X2_RTN, atomic_inc_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_DEC_X2_RTN, atomic_dec_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_AND_X2_RTN, atomic_and_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_SMAX_X2_RTN, atomic_max_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_UMAX_X2_RTN, atomic_umax_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_SMIN_X2_RTN, atomic_min_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_UMIN_X2_RTN, atomic_umin_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_OR_X2_RTN, atomic_or_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_SWAP_X2_RTN, atomic_swap_global, i64>;
def : FlatAtomicPat <FLAT_ATOMIC_CMPSWAP_X2_RTN, AMDGPUatomic_cmp_swap_global, i64, v2i64>;
def : FlatAtomicPat <FLAT_ATOMIC_XOR_X2_RTN, atomic_xor_global, i64>;

} // End Predicates = [isCIVI]

let Predicates = [isVI] in {
  def : FlatStorePat <FLAT_STORE_BYTE, flat_truncstorei8, i16>;
  def : FlatStorePat <FLAT_STORE_SHORT, flat_store, i16>;
}


//===----------------------------------------------------------------------===//
// Target
//===----------------------------------------------------------------------===//

//===----------------------------------------------------------------------===//
// CI
//===----------------------------------------------------------------------===//

class FLAT_Real_ci <bits<7> op, FLAT_Pseudo ps> :
  FLAT_Real <op, ps>,
  SIMCInstr <ps.PseudoInstr, SIEncodingFamily.SI> {
  let AssemblerPredicate = isCIOnly;
  let DecoderNamespace="CI";
}

def FLAT_LOAD_UBYTE_ci         : FLAT_Real_ci <0x8,  FLAT_LOAD_UBYTE>;
def FLAT_LOAD_SBYTE_ci         : FLAT_Real_ci <0x9,  FLAT_LOAD_SBYTE>;
def FLAT_LOAD_USHORT_ci        : FLAT_Real_ci <0xa,  FLAT_LOAD_USHORT>;
def FLAT_LOAD_SSHORT_ci        : FLAT_Real_ci <0xb,  FLAT_LOAD_SSHORT>;
def FLAT_LOAD_DWORD_ci         : FLAT_Real_ci <0xc,  FLAT_LOAD_DWORD>;
def FLAT_LOAD_DWORDX2_ci       : FLAT_Real_ci <0xd,  FLAT_LOAD_DWORDX2>;
def FLAT_LOAD_DWORDX4_ci       : FLAT_Real_ci <0xe,  FLAT_LOAD_DWORDX4>;
def FLAT_LOAD_DWORDX3_ci       : FLAT_Real_ci <0xf,  FLAT_LOAD_DWORDX3>;

def FLAT_STORE_BYTE_ci         : FLAT_Real_ci <0x18, FLAT_STORE_BYTE>;
def FLAT_STORE_SHORT_ci        : FLAT_Real_ci <0x1a, FLAT_STORE_SHORT>;
def FLAT_STORE_DWORD_ci        : FLAT_Real_ci <0x1c, FLAT_STORE_DWORD>;
def FLAT_STORE_DWORDX2_ci      : FLAT_Real_ci <0x1d, FLAT_STORE_DWORDX2>;
def FLAT_STORE_DWORDX4_ci      : FLAT_Real_ci <0x1e, FLAT_STORE_DWORDX4>;
def FLAT_STORE_DWORDX3_ci      : FLAT_Real_ci <0x1f, FLAT_STORE_DWORDX3>;

multiclass FLAT_Real_Atomics_ci <bits<7> op, FLAT_Pseudo ps> {
  def _ci     : FLAT_Real_ci<op, !cast<FLAT_Pseudo>(ps.PseudoInstr)>;
  def _RTN_ci : FLAT_Real_ci<op, !cast<FLAT_Pseudo>(ps.PseudoInstr # "_RTN")>;
}

defm FLAT_ATOMIC_SWAP          : FLAT_Real_Atomics_ci <0x30, FLAT_ATOMIC_SWAP>;
defm FLAT_ATOMIC_CMPSWAP       : FLAT_Real_Atomics_ci <0x31, FLAT_ATOMIC_CMPSWAP>;
defm FLAT_ATOMIC_ADD           : FLAT_Real_Atomics_ci <0x32, FLAT_ATOMIC_ADD>;
defm FLAT_ATOMIC_SUB           : FLAT_Real_Atomics_ci <0x33, FLAT_ATOMIC_SUB>;
defm FLAT_ATOMIC_SMIN          : FLAT_Real_Atomics_ci <0x35, FLAT_ATOMIC_SMIN>;
defm FLAT_ATOMIC_UMIN          : FLAT_Real_Atomics_ci <0x36, FLAT_ATOMIC_UMIN>;
defm FLAT_ATOMIC_SMAX          : FLAT_Real_Atomics_ci <0x37, FLAT_ATOMIC_SMAX>;
defm FLAT_ATOMIC_UMAX          : FLAT_Real_Atomics_ci <0x38, FLAT_ATOMIC_UMAX>;
defm FLAT_ATOMIC_AND           : FLAT_Real_Atomics_ci <0x39, FLAT_ATOMIC_AND>;
defm FLAT_ATOMIC_OR            : FLAT_Real_Atomics_ci <0x3a, FLAT_ATOMIC_OR>;
defm FLAT_ATOMIC_XOR           : FLAT_Real_Atomics_ci <0x3b, FLAT_ATOMIC_XOR>;
defm FLAT_ATOMIC_INC           : FLAT_Real_Atomics_ci <0x3c, FLAT_ATOMIC_INC>;
defm FLAT_ATOMIC_DEC           : FLAT_Real_Atomics_ci <0x3d, FLAT_ATOMIC_DEC>;
defm FLAT_ATOMIC_SWAP_X2       : FLAT_Real_Atomics_ci <0x50, FLAT_ATOMIC_SWAP_X2>;
defm FLAT_ATOMIC_CMPSWAP_X2    : FLAT_Real_Atomics_ci <0x51, FLAT_ATOMIC_CMPSWAP_X2>;
defm FLAT_ATOMIC_ADD_X2        : FLAT_Real_Atomics_ci <0x52, FLAT_ATOMIC_ADD_X2>;
defm FLAT_ATOMIC_SUB_X2        : FLAT_Real_Atomics_ci <0x53, FLAT_ATOMIC_SUB_X2>;
defm FLAT_ATOMIC_SMIN_X2       : FLAT_Real_Atomics_ci <0x55, FLAT_ATOMIC_SMIN_X2>;
defm FLAT_ATOMIC_UMIN_X2       : FLAT_Real_Atomics_ci <0x56, FLAT_ATOMIC_UMIN_X2>;
defm FLAT_ATOMIC_SMAX_X2       : FLAT_Real_Atomics_ci <0x57, FLAT_ATOMIC_SMAX_X2>;
defm FLAT_ATOMIC_UMAX_X2       : FLAT_Real_Atomics_ci <0x58, FLAT_ATOMIC_UMAX_X2>;
defm FLAT_ATOMIC_AND_X2        : FLAT_Real_Atomics_ci <0x59, FLAT_ATOMIC_AND_X2>;
defm FLAT_ATOMIC_OR_X2         : FLAT_Real_Atomics_ci <0x5a, FLAT_ATOMIC_OR_X2>;
defm FLAT_ATOMIC_XOR_X2        : FLAT_Real_Atomics_ci <0x5b, FLAT_ATOMIC_XOR_X2>;
defm FLAT_ATOMIC_INC_X2        : FLAT_Real_Atomics_ci <0x5c, FLAT_ATOMIC_INC_X2>;
defm FLAT_ATOMIC_DEC_X2        : FLAT_Real_Atomics_ci <0x5d, FLAT_ATOMIC_DEC_X2>;

// CI Only flat instructions
defm FLAT_ATOMIC_FCMPSWAP      : FLAT_Real_Atomics_ci <0x3e, FLAT_ATOMIC_FCMPSWAP>;
defm FLAT_ATOMIC_FMIN          : FLAT_Real_Atomics_ci <0x3f, FLAT_ATOMIC_FMIN>;
defm FLAT_ATOMIC_FMAX          : FLAT_Real_Atomics_ci <0x40, FLAT_ATOMIC_FMAX>;
defm FLAT_ATOMIC_FCMPSWAP_X2   : FLAT_Real_Atomics_ci <0x5e, FLAT_ATOMIC_FCMPSWAP_X2>;
defm FLAT_ATOMIC_FMIN_X2       : FLAT_Real_Atomics_ci <0x5f, FLAT_ATOMIC_FMIN_X2>;
defm FLAT_ATOMIC_FMAX_X2       : FLAT_Real_Atomics_ci <0x60, FLAT_ATOMIC_FMAX_X2>;


//===----------------------------------------------------------------------===//
// VI
//===----------------------------------------------------------------------===//

class FLAT_Real_vi <bits<7> op, FLAT_Pseudo ps> :
  FLAT_Real <op, ps>,
  SIMCInstr <ps.PseudoInstr, SIEncodingFamily.VI> {
  let AssemblerPredicate = isVI;
  let DecoderNamespace="VI";
}

def FLAT_LOAD_UBYTE_vi         : FLAT_Real_vi <0x10, FLAT_LOAD_UBYTE>;
def FLAT_LOAD_SBYTE_vi         : FLAT_Real_vi <0x11, FLAT_LOAD_SBYTE>;
def FLAT_LOAD_USHORT_vi        : FLAT_Real_vi <0x12, FLAT_LOAD_USHORT>;
def FLAT_LOAD_SSHORT_vi        : FLAT_Real_vi <0x13, FLAT_LOAD_SSHORT>;
def FLAT_LOAD_DWORD_vi         : FLAT_Real_vi <0x14, FLAT_LOAD_DWORD>;
def FLAT_LOAD_DWORDX2_vi       : FLAT_Real_vi <0x15, FLAT_LOAD_DWORDX2>;
def FLAT_LOAD_DWORDX4_vi       : FLAT_Real_vi <0x17, FLAT_LOAD_DWORDX4>;
def FLAT_LOAD_DWORDX3_vi       : FLAT_Real_vi <0x16, FLAT_LOAD_DWORDX3>;

def FLAT_STORE_BYTE_vi         : FLAT_Real_vi <0x18, FLAT_STORE_BYTE>;
def FLAT_STORE_SHORT_vi        : FLAT_Real_vi <0x1a, FLAT_STORE_SHORT>;
def FLAT_STORE_DWORD_vi        : FLAT_Real_vi <0x1c, FLAT_STORE_DWORD>;
def FLAT_STORE_DWORDX2_vi      : FLAT_Real_vi <0x1d, FLAT_STORE_DWORDX2>;
def FLAT_STORE_DWORDX4_vi      : FLAT_Real_vi <0x1f, FLAT_STORE_DWORDX4>;
def FLAT_STORE_DWORDX3_vi      : FLAT_Real_vi <0x1e, FLAT_STORE_DWORDX3>;

multiclass FLAT_Real_Atomics_vi <bits<7> op, FLAT_Pseudo ps> {
  def _vi     : FLAT_Real_vi<op, !cast<FLAT_Pseudo>(ps.PseudoInstr)>;
  def _RTN_vi : FLAT_Real_vi<op, !cast<FLAT_Pseudo>(ps.PseudoInstr # "_RTN")>;
}

defm FLAT_ATOMIC_SWAP       : FLAT_Real_Atomics_vi <0x40, FLAT_ATOMIC_SWAP>;
defm FLAT_ATOMIC_CMPSWAP    : FLAT_Real_Atomics_vi <0x41, FLAT_ATOMIC_CMPSWAP>;
defm FLAT_ATOMIC_ADD        : FLAT_Real_Atomics_vi <0x42, FLAT_ATOMIC_ADD>;
defm FLAT_ATOMIC_SUB        : FLAT_Real_Atomics_vi <0x43, FLAT_ATOMIC_SUB>;
defm FLAT_ATOMIC_SMIN       : FLAT_Real_Atomics_vi <0x44, FLAT_ATOMIC_SMIN>;
defm FLAT_ATOMIC_UMIN       : FLAT_Real_Atomics_vi <0x45, FLAT_ATOMIC_UMIN>;
defm FLAT_ATOMIC_SMAX       : FLAT_Real_Atomics_vi <0x46, FLAT_ATOMIC_SMAX>;
defm FLAT_ATOMIC_UMAX       : FLAT_Real_Atomics_vi <0x47, FLAT_ATOMIC_UMAX>;
defm FLAT_ATOMIC_AND        : FLAT_Real_Atomics_vi <0x48, FLAT_ATOMIC_AND>;
defm FLAT_ATOMIC_OR         : FLAT_Real_Atomics_vi <0x49, FLAT_ATOMIC_OR>;
defm FLAT_ATOMIC_XOR        : FLAT_Real_Atomics_vi <0x4a, FLAT_ATOMIC_XOR>;
defm FLAT_ATOMIC_INC        : FLAT_Real_Atomics_vi <0x4b, FLAT_ATOMIC_INC>;
defm FLAT_ATOMIC_DEC        : FLAT_Real_Atomics_vi <0x4c, FLAT_ATOMIC_DEC>;
defm FLAT_ATOMIC_SWAP_X2    : FLAT_Real_Atomics_vi <0x60, FLAT_ATOMIC_SWAP_X2>;
defm FLAT_ATOMIC_CMPSWAP_X2 : FLAT_Real_Atomics_vi <0x61, FLAT_ATOMIC_CMPSWAP_X2>;
defm FLAT_ATOMIC_ADD_X2     : FLAT_Real_Atomics_vi <0x62, FLAT_ATOMIC_ADD_X2>;
defm FLAT_ATOMIC_SUB_X2     : FLAT_Real_Atomics_vi <0x63, FLAT_ATOMIC_SUB_X2>;
defm FLAT_ATOMIC_SMIN_X2    : FLAT_Real_Atomics_vi <0x64, FLAT_ATOMIC_SMIN_X2>;
defm FLAT_ATOMIC_UMIN_X2    : FLAT_Real_Atomics_vi <0x65, FLAT_ATOMIC_UMIN_X2>;
defm FLAT_ATOMIC_SMAX_X2    : FLAT_Real_Atomics_vi <0x66, FLAT_ATOMIC_SMAX_X2>;
defm FLAT_ATOMIC_UMAX_X2    : FLAT_Real_Atomics_vi <0x67, FLAT_ATOMIC_UMAX_X2>;
defm FLAT_ATOMIC_AND_X2     : FLAT_Real_Atomics_vi <0x68, FLAT_ATOMIC_AND_X2>;
defm FLAT_ATOMIC_OR_X2      : FLAT_Real_Atomics_vi <0x69, FLAT_ATOMIC_OR_X2>;
defm FLAT_ATOMIC_XOR_X2     : FLAT_Real_Atomics_vi <0x6a, FLAT_ATOMIC_XOR_X2>;
defm FLAT_ATOMIC_INC_X2     : FLAT_Real_Atomics_vi <0x6b, FLAT_ATOMIC_INC_X2>;
defm FLAT_ATOMIC_DEC_X2     : FLAT_Real_Atomics_vi <0x6c, FLAT_ATOMIC_DEC_X2>;

def GLOBAL_LOAD_UBYTE_vi : FLAT_Real_vi <0x10, GLOBAL_LOAD_UBYTE>;
def GLOBAL_LOAD_SBYTE_vi : FLAT_Real_vi <0x11, GLOBAL_LOAD_SBYTE>;
def GLOBAL_LOAD_USHORT_vi : FLAT_Real_vi <0x12, GLOBAL_LOAD_USHORT>;
def GLOBAL_LOAD_SSHORT_vi : FLAT_Real_vi <0x13, GLOBAL_LOAD_SSHORT>;
def GLOBAL_LOAD_DWORD_vi : FLAT_Real_vi <0x14, GLOBAL_LOAD_DWORD>;
def GLOBAL_LOAD_DWORDX2_vi : FLAT_Real_vi <0x15, GLOBAL_LOAD_DWORDX2>;
def GLOBAL_LOAD_DWORDX4_vi : FLAT_Real_vi <0x17, GLOBAL_LOAD_DWORDX4>;
def GLOBAL_LOAD_DWORDX3_vi : FLAT_Real_vi <0x16, GLOBAL_LOAD_DWORDX3>;

def GLOBAL_STORE_BYTE_vi : FLAT_Real_vi <0x18, GLOBAL_STORE_BYTE>;
def GLOBAL_STORE_SHORT_vi : FLAT_Real_vi <0x1a, GLOBAL_STORE_SHORT>;
def GLOBAL_STORE_DWORD_vi : FLAT_Real_vi <0x1c, GLOBAL_STORE_DWORD>;
def GLOBAL_STORE_DWORDX2_vi : FLAT_Real_vi <0x1d, GLOBAL_STORE_DWORDX2>;
def GLOBAL_STORE_DWORDX4_vi : FLAT_Real_vi <0x1f, GLOBAL_STORE_DWORDX4>;
def GLOBAL_STORE_DWORDX3_vi : FLAT_Real_vi <0x1e, GLOBAL_STORE_DWORDX3>;