llvm.org GIT mirror llvm / release_38 lib / Target / AMDGPU / AMDGPUInstructions.td
release_38

Tree @release_38 (Download .tar.gz)

AMDGPUInstructions.td @release_38raw · 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
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
//===-- AMDGPUInstructions.td - Common instruction defs ---*- tablegen -*-===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains instruction defs that are common to all hw codegen
// targets.
//
//===----------------------------------------------------------------------===//

class AMDGPUInst <dag outs, dag ins, string asm, list<dag> pattern> : Instruction {
  field bit isRegisterLoad = 0;
  field bit isRegisterStore = 0;

  let Namespace = "AMDGPU";
  let OutOperandList = outs;
  let InOperandList = ins;
  let AsmString = asm;
  let Pattern = pattern;
  let Itinerary = NullALU;

  let TSFlags{63} = isRegisterLoad;
  let TSFlags{62} = isRegisterStore;
}

class AMDGPUShaderInst <dag outs, dag ins, string asm, list<dag> pattern>
    : AMDGPUInst<outs, ins, asm, pattern> {

  field bits<32> Inst = 0xffffffff;

}

def FP32Denormals : Predicate<"Subtarget.hasFP32Denormals()">;
def FP64Denormals : Predicate<"Subtarget.hasFP64Denormals()">;
def UnsafeFPMath : Predicate<"TM.Options.UnsafeFPMath">;

def InstFlag : OperandWithDefaultOps <i32, (ops (i32 0))>;
def ADDRIndirect : ComplexPattern<iPTR, 2, "SelectADDRIndirect", [], []>;

let OperandType = "OPERAND_IMMEDIATE" in {

def u32imm : Operand<i32> {
  let PrintMethod = "printU32ImmOperand";
}

def u16imm : Operand<i16> {
  let PrintMethod = "printU16ImmOperand";
}

def u8imm : Operand<i8> {
  let PrintMethod = "printU8ImmOperand";
}

} // End OperandType = "OPERAND_IMMEDIATE"

//===--------------------------------------------------------------------===//
// Custom Operands
//===--------------------------------------------------------------------===//
def brtarget   : Operand<OtherVT>;

//===----------------------------------------------------------------------===//
// PatLeafs for floating-point comparisons
//===----------------------------------------------------------------------===//

def COND_OEQ : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETOEQ || N->get() == ISD::SETEQ;}]
>;

def COND_ONE : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETONE || N->get() == ISD::SETNE;}]
>;

def COND_OGT : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETOGT || N->get() == ISD::SETGT;}]
>;

def COND_OGE : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETOGE || N->get() == ISD::SETGE;}]
>;

def COND_OLT : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETOLT || N->get() == ISD::SETLT;}]
>;

def COND_OLE : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETOLE || N->get() == ISD::SETLE;}]
>;


def COND_O : PatLeaf <(cond), [{return N->get() == ISD::SETO;}]>;
def COND_UO : PatLeaf <(cond), [{return N->get() == ISD::SETUO;}]>;

//===----------------------------------------------------------------------===//
// PatLeafs for unsigned / unordered comparisons
//===----------------------------------------------------------------------===//

def COND_UEQ : PatLeaf <(cond), [{return N->get() == ISD::SETUEQ;}]>;
def COND_UNE : PatLeaf <(cond), [{return N->get() == ISD::SETUNE;}]>;
def COND_UGT : PatLeaf <(cond), [{return N->get() == ISD::SETUGT;}]>;
def COND_UGE : PatLeaf <(cond), [{return N->get() == ISD::SETUGE;}]>;
def COND_ULT : PatLeaf <(cond), [{return N->get() == ISD::SETULT;}]>;
def COND_ULE : PatLeaf <(cond), [{return N->get() == ISD::SETULE;}]>;

// XXX - For some reason R600 version is preferring to use unordered
// for setne?
def COND_UNE_NE : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETUNE || N->get() == ISD::SETNE;}]
>;

//===----------------------------------------------------------------------===//
// PatLeafs for signed comparisons
//===----------------------------------------------------------------------===//

def COND_SGT : PatLeaf <(cond), [{return N->get() == ISD::SETGT;}]>;
def COND_SGE : PatLeaf <(cond), [{return N->get() == ISD::SETGE;}]>;
def COND_SLT : PatLeaf <(cond), [{return N->get() == ISD::SETLT;}]>;
def COND_SLE : PatLeaf <(cond), [{return N->get() == ISD::SETLE;}]>;

//===----------------------------------------------------------------------===//
// PatLeafs for integer equality
//===----------------------------------------------------------------------===//

def COND_EQ : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETEQ || N->get() == ISD::SETUEQ;}]
>;

def COND_NE : PatLeaf <
  (cond),
  [{return N->get() == ISD::SETNE || N->get() == ISD::SETUNE;}]
>;

def COND_NULL : PatLeaf <
  (cond),
  [{(void)N; return false;}]
>;

//===----------------------------------------------------------------------===//
// Load/Store Pattern Fragments
//===----------------------------------------------------------------------===//

class PrivateMemOp <dag ops, dag frag> : PatFrag <ops, frag, [{
  return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::PRIVATE_ADDRESS;
}]>;

class PrivateLoad <SDPatternOperator op> : PrivateMemOp <
  (ops node:$ptr), (op node:$ptr)
>;

class PrivateStore <SDPatternOperator op> : PrivateMemOp <
  (ops node:$value, node:$ptr), (op node:$value, node:$ptr)
>;

def load_private : PrivateLoad <load>;

def truncstorei8_private : PrivateStore <truncstorei8>;
def truncstorei16_private : PrivateStore <truncstorei16>;
def store_private : PrivateStore <store>;

def global_store : PatFrag<(ops node:$val, node:$ptr),
    (store node:$val, node:$ptr), [{
        return isGlobalStore(dyn_cast<StoreSDNode>(N));
}]>;

// Global address space loads
def global_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{
    return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

// Constant address space loads
def constant_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{
    return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

class AZExtLoadBase <SDPatternOperator ld_node>: PatFrag<(ops node:$ptr),
                                              (ld_node node:$ptr), [{
  LoadSDNode *L = cast<LoadSDNode>(N);
  return L->getExtensionType() == ISD::ZEXTLOAD ||
         L->getExtensionType() == ISD::EXTLOAD;
}]>;

def az_extload : AZExtLoadBase <unindexedload>;

def az_extloadi8 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
  return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i8;
}]>;

def az_extloadi8_global : PatFrag<(ops node:$ptr), (az_extloadi8 node:$ptr), [{
    return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def sextloadi8_global : PatFrag<(ops node:$ptr), (sextloadi8 node:$ptr), [{
    return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def az_extloadi8_constant : PatFrag<(ops node:$ptr), (az_extloadi8 node:$ptr), [{
    return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

def sextloadi8_constant : PatFrag<(ops node:$ptr), (sextloadi8 node:$ptr), [{
    return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

def az_extloadi8_local : PatFrag<(ops node:$ptr), (az_extloadi8 node:$ptr), [{
    return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def sextloadi8_local : PatFrag<(ops node:$ptr), (sextloadi8 node:$ptr), [{
    return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def extloadi8_private : PrivateLoad <az_extloadi8>;
def sextloadi8_private : PrivateLoad <sextloadi8>;

def az_extloadi16 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
  return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
}]>;

def az_extloadi16_global : PatFrag<(ops node:$ptr), (az_extloadi16 node:$ptr), [{
    return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def sextloadi16_global : PatFrag<(ops node:$ptr), (sextloadi16 node:$ptr), [{
    return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def az_extloadi16_constant : PatFrag<(ops node:$ptr), (az_extloadi16 node:$ptr), [{
    return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

def sextloadi16_constant : PatFrag<(ops node:$ptr), (sextloadi16 node:$ptr), [{
    return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

def az_extloadi16_local : PatFrag<(ops node:$ptr), (az_extloadi16 node:$ptr), [{
    return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def sextloadi16_local : PatFrag<(ops node:$ptr), (sextloadi16 node:$ptr), [{
    return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def extloadi16_private : PrivateLoad <az_extloadi16>;
def sextloadi16_private : PrivateLoad <sextloadi16>;

def az_extloadi32 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
  return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
}]>;

def az_extloadi32_global : PatFrag<(ops node:$ptr),
                                   (az_extloadi32 node:$ptr), [{
  return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;

def az_extloadi32_flat : PatFrag<(ops node:$ptr),
                                   (az_extloadi32 node:$ptr), [{
  return isFlatLoad(dyn_cast<LoadSDNode>(N));
}]>;

def az_extloadi32_constant : PatFrag<(ops node:$ptr),
                                     (az_extloadi32 node:$ptr), [{
  return isConstantLoad(dyn_cast<LoadSDNode>(N), -1);
}]>;

def truncstorei8_global : PatFrag<(ops node:$val, node:$ptr),
                                  (truncstorei8 node:$val, node:$ptr), [{
  return isGlobalStore(dyn_cast<StoreSDNode>(N));
}]>;

def truncstorei16_global : PatFrag<(ops node:$val, node:$ptr),
                                  (truncstorei16 node:$val, node:$ptr), [{
  return isGlobalStore(dyn_cast<StoreSDNode>(N));
}]>;

def local_store : PatFrag<(ops node:$val, node:$ptr),
                             (store node:$val, node:$ptr), [{
  return isLocalStore(dyn_cast<StoreSDNode>(N));
}]>;

def truncstorei8_local : PatFrag<(ops node:$val, node:$ptr),
                                  (truncstorei8 node:$val, node:$ptr), [{
  return isLocalStore(dyn_cast<StoreSDNode>(N));
}]>;

def truncstorei16_local : PatFrag<(ops node:$val, node:$ptr),
                                  (truncstorei16 node:$val, node:$ptr), [{
  return isLocalStore(dyn_cast<StoreSDNode>(N));
}]>;

def local_load : PatFrag<(ops node:$ptr), (load node:$ptr), [{
    return isLocalLoad(dyn_cast<LoadSDNode>(N));
}]>;

class Aligned8Bytes <dag ops, dag frag> : PatFrag <ops, frag, [{
    return cast<MemSDNode>(N)->getAlignment() % 8 == 0;
}]>;

def local_load_aligned8bytes : Aligned8Bytes <
  (ops node:$ptr), (local_load node:$ptr)
>;

def local_store_aligned8bytes : Aligned8Bytes <
  (ops node:$val, node:$ptr), (local_store node:$val, node:$ptr)
>;

class local_binary_atomic_op<SDNode atomic_op> :
  PatFrag<(ops node:$ptr, node:$value),
    (atomic_op node:$ptr, node:$value), [{
  return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS;
}]>;


def atomic_swap_local : local_binary_atomic_op<atomic_swap>;
def atomic_load_add_local : local_binary_atomic_op<atomic_load_add>;
def atomic_load_sub_local : local_binary_atomic_op<atomic_load_sub>;
def atomic_load_and_local : local_binary_atomic_op<atomic_load_and>;
def atomic_load_or_local : local_binary_atomic_op<atomic_load_or>;
def atomic_load_xor_local : local_binary_atomic_op<atomic_load_xor>;
def atomic_load_nand_local : local_binary_atomic_op<atomic_load_nand>;
def atomic_load_min_local : local_binary_atomic_op<atomic_load_min>;
def atomic_load_max_local : local_binary_atomic_op<atomic_load_max>;
def atomic_load_umin_local : local_binary_atomic_op<atomic_load_umin>;
def atomic_load_umax_local : local_binary_atomic_op<atomic_load_umax>;

def mskor_global : PatFrag<(ops node:$val, node:$ptr),
                            (AMDGPUstore_mskor node:$val, node:$ptr), [{
  return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;
}]>;

multiclass AtomicCmpSwapLocal <SDNode cmp_swap_node> {

  def _32_local : PatFrag <
    (ops node:$ptr, node:$cmp, node:$swap),
    (cmp_swap_node node:$ptr, node:$cmp, node:$swap), [{
      AtomicSDNode *AN = cast<AtomicSDNode>(N);
      return AN->getMemoryVT() == MVT::i32 &&
             AN->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS;
  }]>;

  def _64_local : PatFrag<
    (ops node:$ptr, node:$cmp, node:$swap),
    (cmp_swap_node node:$ptr, node:$cmp, node:$swap), [{
      AtomicSDNode *AN = cast<AtomicSDNode>(N);
      return AN->getMemoryVT() == MVT::i64 &&
             AN->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS;
  }]>;
}

defm atomic_cmp_swap : AtomicCmpSwapLocal <atomic_cmp_swap>;

def mskor_flat : PatFrag<(ops node:$val, node:$ptr),
                            (AMDGPUstore_mskor node:$val, node:$ptr), [{
  return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::FLAT_ADDRESS;
}]>;

class global_binary_atomic_op<SDNode atomic_op> : PatFrag<
  (ops node:$ptr, node:$value),
  (atomic_op node:$ptr, node:$value),
  [{return cast<MemSDNode>(N)->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;}]
>;

def atomic_swap_global : global_binary_atomic_op<atomic_swap>;
def atomic_add_global : global_binary_atomic_op<atomic_load_add>;
def atomic_and_global : global_binary_atomic_op<atomic_load_and>;
def atomic_max_global : global_binary_atomic_op<atomic_load_max>;
def atomic_min_global : global_binary_atomic_op<atomic_load_min>;
def atomic_or_global : global_binary_atomic_op<atomic_load_or>;
def atomic_sub_global : global_binary_atomic_op<atomic_load_sub>;
def atomic_umax_global : global_binary_atomic_op<atomic_load_umax>;
def atomic_umin_global : global_binary_atomic_op<atomic_load_umin>;
def atomic_xor_global : global_binary_atomic_op<atomic_load_xor>;

//===----------------------------------------------------------------------===//
// Misc Pattern Fragments
//===----------------------------------------------------------------------===//

class Constants {
int TWO_PI = 0x40c90fdb;
int PI = 0x40490fdb;
int TWO_PI_INV = 0x3e22f983;
int FP_UINT_MAX_PLUS_1 = 0x4f800000;    // 1 << 32 in floating point encoding
int FP32_NEG_ONE = 0xbf800000;
int FP32_ONE = 0x3f800000;
}
def CONST : Constants;

def FP_ZERO : PatLeaf <
  (fpimm),
  [{return N->getValueAPF().isZero();}]
>;

def FP_ONE : PatLeaf <
  (fpimm),
  [{return N->isExactlyValue(1.0);}]
>;

def FP_HALF : PatLeaf <
  (fpimm),
  [{return N->isExactlyValue(0.5);}]
>;

let isCodeGenOnly = 1, isPseudo = 1 in {

let usesCustomInserter = 1  in {

class CLAMP <RegisterClass rc> : AMDGPUShaderInst <
  (outs rc:$dst),
  (ins rc:$src0),
  "CLAMP $dst, $src0",
  [(set f32:$dst, (AMDGPUclamp f32:$src0, (f32 FP_ZERO), (f32 FP_ONE)))]
>;

class FABS <RegisterClass rc> : AMDGPUShaderInst <
  (outs rc:$dst),
  (ins rc:$src0),
  "FABS $dst, $src0",
  [(set f32:$dst, (fabs f32:$src0))]
>;

class FNEG <RegisterClass rc> : AMDGPUShaderInst <
  (outs rc:$dst),
  (ins rc:$src0),
  "FNEG $dst, $src0",
  [(set f32:$dst, (fneg f32:$src0))]
>;

} // usesCustomInserter = 1

multiclass RegisterLoadStore <RegisterClass dstClass, Operand addrClass,
                    ComplexPattern addrPat> {
let UseNamedOperandTable = 1 in {

  def RegisterLoad : AMDGPUShaderInst <
    (outs dstClass:$dst),
    (ins addrClass:$addr, i32imm:$chan),
    "RegisterLoad $dst, $addr",
    [(set i32:$dst, (AMDGPUregister_load addrPat:$addr, (i32 timm:$chan)))]
  > {
    let isRegisterLoad = 1;
  }

  def RegisterStore : AMDGPUShaderInst <
    (outs),
    (ins dstClass:$val, addrClass:$addr, i32imm:$chan),
    "RegisterStore $val, $addr",
    [(AMDGPUregister_store i32:$val, addrPat:$addr, (i32 timm:$chan))]
  > {
    let isRegisterStore = 1;
  }
}
}

} // End isCodeGenOnly = 1, isPseudo = 1

/* Generic helper patterns for intrinsics */
/* -------------------------------------- */

class POW_Common <AMDGPUInst log_ieee, AMDGPUInst exp_ieee, AMDGPUInst mul>
  : Pat <
  (fpow f32:$src0, f32:$src1),
  (exp_ieee (mul f32:$src1, (log_ieee f32:$src0)))
>;

/* Other helper patterns */
/* --------------------- */

/* Extract element pattern */
class Extract_Element <ValueType sub_type, ValueType vec_type, int sub_idx,
                       SubRegIndex sub_reg>
  : Pat<
  (sub_type (extractelt vec_type:$src, sub_idx)),
  (EXTRACT_SUBREG $src, sub_reg)
>;

/* Insert element pattern */
class Insert_Element <ValueType elem_type, ValueType vec_type,
                      int sub_idx, SubRegIndex sub_reg>
  : Pat <
  (insertelt vec_type:$vec, elem_type:$elem, sub_idx),
  (INSERT_SUBREG $vec, $elem, sub_reg)
>;

// XXX: Convert to new syntax and use COPY_TO_REG, once the DFAPacketizer
// can handle COPY instructions.
// bitconvert pattern
class BitConvert <ValueType dt, ValueType st, RegisterClass rc> : Pat <
  (dt (bitconvert (st rc:$src0))),
  (dt rc:$src0)
>;

// XXX: Convert to new syntax and use COPY_TO_REG, once the DFAPacketizer
// can handle COPY instructions.
class DwordAddrPat<ValueType vt, RegisterClass rc> : Pat <
  (vt (AMDGPUdwordaddr (vt rc:$addr))),
  (vt rc:$addr)
>;

// BFI_INT patterns

multiclass BFIPatterns <Instruction BFI_INT,
                        Instruction LoadImm32,
                        RegisterClass RC64> {
  // Definition from ISA doc:
  // (y & x) | (z & ~x)
  def : Pat <
    (or (and i32:$y, i32:$x), (and i32:$z, (not i32:$x))),
    (BFI_INT $x, $y, $z)
  >;

  // SHA-256 Ch function
  // z ^ (x & (y ^ z))
  def : Pat <
    (xor i32:$z, (and i32:$x, (xor i32:$y, i32:$z))),
    (BFI_INT $x, $y, $z)
  >;

  def : Pat <
    (fcopysign f32:$src0, f32:$src1),
    (BFI_INT (LoadImm32 0x7fffffff), $src0, $src1)
  >;

  def : Pat <
    (f64 (fcopysign f64:$src0, f64:$src1)),
    (REG_SEQUENCE RC64,
      (i32 (EXTRACT_SUBREG $src0, sub0)), sub0,
      (BFI_INT (LoadImm32 0x7fffffff),
               (i32 (EXTRACT_SUBREG $src0, sub1)),
               (i32 (EXTRACT_SUBREG $src1, sub1))), sub1)
  >;
}

// SHA-256 Ma patterns

// ((x & z) | (y & (x | z))) -> BFI_INT (XOR x, y), z, y
class SHA256MaPattern <Instruction BFI_INT, Instruction XOR> : Pat <
  (or (and i32:$x, i32:$z), (and i32:$y, (or i32:$x, i32:$z))),
  (BFI_INT (XOR i32:$x, i32:$y), i32:$z, i32:$y)
>;

// Bitfield extract patterns

def IMMZeroBasedBitfieldMask : PatLeaf <(imm), [{
  return isMask_32(N->getZExtValue());
}]>;

def IMMPopCount : SDNodeXForm<imm, [{
  return CurDAG->getTargetConstant(countPopulation(N->getZExtValue()), SDLoc(N),
                                   MVT::i32);
}]>;

class BFEPattern <Instruction BFE, Instruction MOV> : Pat <
  (i32 (and (i32 (srl i32:$src, i32:$rshift)), IMMZeroBasedBitfieldMask:$mask)),
  (BFE $src, $rshift, (MOV (i32 (IMMPopCount $mask))))
>;

// rotr pattern
class ROTRPattern <Instruction BIT_ALIGN> : Pat <
  (rotr i32:$src0, i32:$src1),
  (BIT_ALIGN $src0, $src0, $src1)
>;

// 24-bit arithmetic patterns
def umul24 : PatFrag <(ops node:$x, node:$y), (mul node:$x, node:$y)>;

// Special conversion patterns

def cvt_rpi_i32_f32 : PatFrag <
  (ops node:$src),
  (fp_to_sint (ffloor (fadd $src, FP_HALF))),
  [{ (void) N; return TM.Options.NoNaNsFPMath; }]
>;

def cvt_flr_i32_f32 : PatFrag <
  (ops node:$src),
  (fp_to_sint (ffloor $src)),
  [{ (void)N; return TM.Options.NoNaNsFPMath; }]
>;

/*
class UMUL24Pattern <Instruction UMUL24> : Pat <
  (mul U24:$x, U24:$y),
  (UMUL24 $x, $y)
>;
*/

class IMad24Pat<Instruction Inst> : Pat <
  (add (AMDGPUmul_i24 i32:$src0, i32:$src1), i32:$src2),
  (Inst $src0, $src1, $src2)
>;

class UMad24Pat<Instruction Inst> : Pat <
  (add (AMDGPUmul_u24 i32:$src0, i32:$src1), i32:$src2),
  (Inst $src0, $src1, $src2)
>;

multiclass Expand24IBitOps<Instruction MulInst, Instruction AddInst> {
  def _expand_imad24 : Pat <
    (AMDGPUmad_i24 i32:$src0, i32:$src1, i32:$src2),
    (AddInst (MulInst $src0, $src1), $src2)
  >;

  def _expand_imul24 : Pat <
    (AMDGPUmul_i24 i32:$src0, i32:$src1),
    (MulInst $src0, $src1)
  >;
}

multiclass Expand24UBitOps<Instruction MulInst, Instruction AddInst> {
  def _expand_umad24 : Pat <
    (AMDGPUmad_u24 i32:$src0, i32:$src1, i32:$src2),
    (AddInst (MulInst $src0, $src1), $src2)
  >;

  def _expand_umul24 : Pat <
    (AMDGPUmul_u24 i32:$src0, i32:$src1),
    (MulInst $src0, $src1)
  >;
}

class RcpPat<Instruction RcpInst, ValueType vt> : Pat <
  (fdiv FP_ONE, vt:$src),
  (RcpInst $src)
>;

class RsqPat<Instruction RsqInst, ValueType vt> : Pat <
  (AMDGPUrcp (fsqrt vt:$src)),
  (RsqInst $src)
>;

include "R600Instructions.td"
include "R700Instructions.td"
include "EvergreenInstructions.td"
include "CaymanInstructions.td"

include "SIInstrInfo.td"