llvm.org GIT mirror llvm / 8c9e52a lib / Target / ARM / ARMTargetTransformInfo.cpp
8c9e52a

Tree @8c9e52a (Download .tar.gz)

ARMTargetTransformInfo.cpp @8c9e52araw · 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
//===-- ARMTargetTransformInfo.cpp - ARM specific TTI pass ----------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
/// \file
/// This file implements a TargetTransformInfo analysis pass specific to the
/// ARM target machine. It uses the target's detailed information to provide
/// more precise answers to certain TTI queries, while letting the target
/// independent and default TTI implementations handle the rest.
///
//===----------------------------------------------------------------------===//

#define DEBUG_TYPE "armtti"
#include "ARM.h"
#include "ARMTargetMachine.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Support/Debug.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/CostTable.h"
using namespace llvm;

// Declare the pass initialization routine locally as target-specific passes
// don't havve a target-wide initialization entry point, and so we rely on the
// pass constructor initialization.
namespace llvm {
void initializeARMTTIPass(PassRegistry &);
}

namespace {

class ARMTTI : public ImmutablePass, public TargetTransformInfo {
  const ARMBaseTargetMachine *TM;
  const ARMSubtarget *ST;
  const ARMTargetLowering *TLI;

  /// Estimate the overhead of scalarizing an instruction. Insert and Extract
  /// are set if the result needs to be inserted and/or extracted from vectors.
  unsigned getScalarizationOverhead(Type *Ty, bool Insert, bool Extract) const;

public:
  ARMTTI() : ImmutablePass(ID), TM(0), ST(0), TLI(0) {
    llvm_unreachable("This pass cannot be directly constructed");
  }

  ARMTTI(const ARMBaseTargetMachine *TM)
      : ImmutablePass(ID), TM(TM), ST(TM->getSubtargetImpl()),
        TLI(TM->getTargetLowering()) {
    initializeARMTTIPass(*PassRegistry::getPassRegistry());
  }

  virtual void initializePass() {
    pushTTIStack(this);
  }

  virtual void finalizePass() {
    popTTIStack();
  }

  virtual void getAnalysisUsage(AnalysisUsage &AU) const {
    TargetTransformInfo::getAnalysisUsage(AU);
  }

  /// Pass identification.
  static char ID;

  /// Provide necessary pointer adjustments for the two base classes.
  virtual void *getAdjustedAnalysisPointer(const void *ID) {
    if (ID == &TargetTransformInfo::ID)
      return (TargetTransformInfo*)this;
    return this;
  }

  /// \name Scalar TTI Implementations
  /// @{

  virtual unsigned getIntImmCost(const APInt &Imm, Type *Ty) const;

  /// @}


  /// \name Vector TTI Implementations
  /// @{

  unsigned getNumberOfRegisters(bool Vector) const {
    if (Vector) {
      if (ST->hasNEON())
        return 16;
      return 0;
    }

    if (ST->isThumb1Only())
      return 8;
    return 16;
  }

  unsigned getRegisterBitWidth(bool Vector) const {
    if (Vector) {
      if (ST->hasNEON())
        return 128;
      return 0;
    }

    return 32;
  }

  unsigned getMaximumUnrollFactor() const {
    // These are out of order CPUs:
    if (ST->isCortexA15() || ST->isSwift())
      return 2;
    return 1;
  }

  unsigned getShuffleCost(ShuffleKind Kind, Type *Tp,
                          int Index, Type *SubTp) const;

  unsigned getCastInstrCost(unsigned Opcode, Type *Dst,
                                      Type *Src) const;

  unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy, Type *CondTy) const;

  unsigned getVectorInstrCost(unsigned Opcode, Type *Val, unsigned Index) const;

  unsigned getAddressComputationCost(Type *Val) const;
  /// @}
};

} // end anonymous namespace

INITIALIZE_AG_PASS(ARMTTI, TargetTransformInfo, "armtti",
                   "ARM Target Transform Info", true, true, false)
char ARMTTI::ID = 0;

ImmutablePass *
llvm::createARMTargetTransformInfoPass(const ARMBaseTargetMachine *TM) {
  return new ARMTTI(TM);
}


unsigned ARMTTI::getIntImmCost(const APInt &Imm, Type *Ty) const {
  assert(Ty->isIntegerTy());

  unsigned Bits = Ty->getPrimitiveSizeInBits();
  if (Bits == 0 || Bits > 32)
    return 4;

  int32_t SImmVal = Imm.getSExtValue();
  uint32_t ZImmVal = Imm.getZExtValue();
  if (!ST->isThumb()) {
    if ((SImmVal >= 0 && SImmVal < 65536) ||
        (ARM_AM::getSOImmVal(ZImmVal) != -1) ||
        (ARM_AM::getSOImmVal(~ZImmVal) != -1))
      return 1;
    return ST->hasV6T2Ops() ? 2 : 3;
  } else if (ST->isThumb2()) {
    if ((SImmVal >= 0 && SImmVal < 65536) ||
        (ARM_AM::getT2SOImmVal(ZImmVal) != -1) ||
        (ARM_AM::getT2SOImmVal(~ZImmVal) != -1))
      return 1;
    return ST->hasV6T2Ops() ? 2 : 3;
  } else /*Thumb1*/ {
    if (SImmVal >= 0 && SImmVal < 256)
      return 1;
    if ((~ZImmVal < 256) || ARM_AM::isThumbImmShiftedVal(ZImmVal))
      return 2;
    // Load from constantpool.
    return 3;
  }
  return 2;
}

unsigned ARMTTI::getCastInstrCost(unsigned Opcode, Type *Dst,
                                    Type *Src) const {
  int ISD = TLI->InstructionOpcodeToISD(Opcode);
  assert(ISD && "Invalid opcode");

  // Single to/from double precision conversions.
  static const CostTblEntry<MVT> NEONFltDblTbl[] = {
    // Vector fptrunc/fpext conversions.
    { ISD::FP_ROUND,   MVT::v2f64, 2 },
    { ISD::FP_EXTEND,  MVT::v2f32, 2 },
    { ISD::FP_EXTEND,  MVT::v4f32, 4 }
  };

  if (Src->isVectorTy() && ST->hasNEON() && (ISD == ISD::FP_ROUND ||
                                          ISD == ISD::FP_EXTEND)) {
    std::pair<unsigned, MVT> LT = TLI->getTypeLegalizationCost(Src);
    int Idx = CostTableLookup<MVT>(NEONFltDblTbl, array_lengthof(NEONFltDblTbl),
                                ISD, LT.second);
    if (Idx != -1)
      return LT.first * NEONFltDblTbl[Idx].Cost;
  }

  EVT SrcTy = TLI->getValueType(Src);
  EVT DstTy = TLI->getValueType(Dst);

  if (!SrcTy.isSimple() || !DstTy.isSimple())
    return TargetTransformInfo::getCastInstrCost(Opcode, Dst, Src);

  // Some arithmetic, load and store operations have specific instructions
  // to cast up/down their types automatically at no extra cost.
  // TODO: Get these tables to know at least what the related operations are.
  static const TypeConversionCostTblEntry<MVT> NEONVectorConversionTbl[] = {
    { ISD::SIGN_EXTEND, MVT::v4i32, MVT::v4i16, 0 },
    { ISD::ZERO_EXTEND, MVT::v4i32, MVT::v4i16, 0 },
    { ISD::SIGN_EXTEND, MVT::v2i64, MVT::v2i32, 1 },
    { ISD::ZERO_EXTEND, MVT::v2i64, MVT::v2i32, 1 },
    { ISD::TRUNCATE,    MVT::v4i32, MVT::v4i64, 0 },
    { ISD::TRUNCATE,    MVT::v4i16, MVT::v4i32, 1 },

    // The number of vmovl instructions for the extension.
    { ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i16, 3 },
    { ISD::ZERO_EXTEND, MVT::v4i64, MVT::v4i16, 3 },
    { ISD::SIGN_EXTEND, MVT::v8i32, MVT::v8i8, 3 },
    { ISD::ZERO_EXTEND, MVT::v8i32, MVT::v8i8, 3 },
    { ISD::SIGN_EXTEND, MVT::v8i64, MVT::v8i8, 7 },
    { ISD::ZERO_EXTEND, MVT::v8i64, MVT::v8i8, 7 },
    { ISD::SIGN_EXTEND, MVT::v8i64, MVT::v8i16, 6 },
    { ISD::ZERO_EXTEND, MVT::v8i64, MVT::v8i16, 6 },
    { ISD::SIGN_EXTEND, MVT::v16i32, MVT::v16i8, 6 },
    { ISD::ZERO_EXTEND, MVT::v16i32, MVT::v16i8, 6 },

    // Operations that we legalize using load/stores to the stack.
    { ISD::TRUNCATE,    MVT::v16i8, MVT::v16i32, 4*1 + 16*2 + 2*1 },
    { ISD::TRUNCATE,    MVT::v8i8, MVT::v8i32, 2*1 + 8*2 + 1 },

    // Vector float <-> i32 conversions.
    { ISD::SINT_TO_FP,  MVT::v4f32, MVT::v4i32, 1 },
    { ISD::UINT_TO_FP,  MVT::v4f32, MVT::v4i32, 1 },

    { ISD::SINT_TO_FP,  MVT::v2f32, MVT::v2i8, 3 },
    { ISD::UINT_TO_FP,  MVT::v2f32, MVT::v2i8, 3 },
    { ISD::SINT_TO_FP,  MVT::v2f32, MVT::v2i16, 2 },
    { ISD::UINT_TO_FP,  MVT::v2f32, MVT::v2i16, 2 },
    { ISD::SINT_TO_FP,  MVT::v2f32, MVT::v2i32, 1 },
    { ISD::UINT_TO_FP,  MVT::v2f32, MVT::v2i32, 1 },
    { ISD::SINT_TO_FP,  MVT::v4f32, MVT::v4i1, 3 },
    { ISD::UINT_TO_FP,  MVT::v4f32, MVT::v4i1, 3 },
    { ISD::SINT_TO_FP,  MVT::v4f32, MVT::v4i8, 3 },
    { ISD::UINT_TO_FP,  MVT::v4f32, MVT::v4i8, 3 },
    { ISD::SINT_TO_FP,  MVT::v4f32, MVT::v4i16, 2 },
    { ISD::UINT_TO_FP,  MVT::v4f32, MVT::v4i16, 2 },
    { ISD::SINT_TO_FP,  MVT::v8f32, MVT::v8i16, 4 },
    { ISD::UINT_TO_FP,  MVT::v8f32, MVT::v8i16, 4 },
    { ISD::SINT_TO_FP,  MVT::v8f32, MVT::v8i32, 2 },
    { ISD::UINT_TO_FP,  MVT::v8f32, MVT::v8i32, 2 },
    { ISD::SINT_TO_FP,  MVT::v16f32, MVT::v16i16, 8 },
    { ISD::UINT_TO_FP,  MVT::v16f32, MVT::v16i16, 8 },
    { ISD::SINT_TO_FP,  MVT::v16f32, MVT::v16i32, 4 },
    { ISD::UINT_TO_FP,  MVT::v16f32, MVT::v16i32, 4 },

    { ISD::FP_TO_SINT,  MVT::v4i32, MVT::v4f32, 1 },
    { ISD::FP_TO_UINT,  MVT::v4i32, MVT::v4f32, 1 },
    { ISD::FP_TO_SINT,  MVT::v4i8, MVT::v4f32, 3 },
    { ISD::FP_TO_UINT,  MVT::v4i8, MVT::v4f32, 3 },
    { ISD::FP_TO_SINT,  MVT::v4i16, MVT::v4f32, 2 },
    { ISD::FP_TO_UINT,  MVT::v4i16, MVT::v4f32, 2 },

    // Vector double <-> i32 conversions.
    { ISD::SINT_TO_FP,  MVT::v2f64, MVT::v2i32, 2 },
    { ISD::UINT_TO_FP,  MVT::v2f64, MVT::v2i32, 2 },

    { ISD::SINT_TO_FP,  MVT::v2f64, MVT::v2i8, 4 },
    { ISD::UINT_TO_FP,  MVT::v2f64, MVT::v2i8, 4 },
    { ISD::SINT_TO_FP,  MVT::v2f64, MVT::v2i16, 3 },
    { ISD::UINT_TO_FP,  MVT::v2f64, MVT::v2i16, 3 },
    { ISD::SINT_TO_FP,  MVT::v2f64, MVT::v2i32, 2 },
    { ISD::UINT_TO_FP,  MVT::v2f64, MVT::v2i32, 2 },

    { ISD::FP_TO_SINT,  MVT::v2i32, MVT::v2f64, 2 },
    { ISD::FP_TO_UINT,  MVT::v2i32, MVT::v2f64, 2 },
    { ISD::FP_TO_SINT,  MVT::v8i16, MVT::v8f32, 4 },
    { ISD::FP_TO_UINT,  MVT::v8i16, MVT::v8f32, 4 },
    { ISD::FP_TO_SINT,  MVT::v16i16, MVT::v16f32, 8 },
    { ISD::FP_TO_UINT,  MVT::v16i16, MVT::v16f32, 8 }
  };

  if (SrcTy.isVector() && ST->hasNEON()) {
    int Idx = ConvertCostTableLookup<MVT>(NEONVectorConversionTbl,
                                array_lengthof(NEONVectorConversionTbl),
                                ISD, DstTy.getSimpleVT(), SrcTy.getSimpleVT());
    if (Idx != -1)
      return NEONVectorConversionTbl[Idx].Cost;
  }

  // Scalar float to integer conversions.
  static const TypeConversionCostTblEntry<MVT> NEONFloatConversionTbl[] = {
    { ISD::FP_TO_SINT,  MVT::i1, MVT::f32, 2 },
    { ISD::FP_TO_UINT,  MVT::i1, MVT::f32, 2 },
    { ISD::FP_TO_SINT,  MVT::i1, MVT::f64, 2 },
    { ISD::FP_TO_UINT,  MVT::i1, MVT::f64, 2 },
    { ISD::FP_TO_SINT,  MVT::i8, MVT::f32, 2 },
    { ISD::FP_TO_UINT,  MVT::i8, MVT::f32, 2 },
    { ISD::FP_TO_SINT,  MVT::i8, MVT::f64, 2 },
    { ISD::FP_TO_UINT,  MVT::i8, MVT::f64, 2 },
    { ISD::FP_TO_SINT,  MVT::i16, MVT::f32, 2 },
    { ISD::FP_TO_UINT,  MVT::i16, MVT::f32, 2 },
    { ISD::FP_TO_SINT,  MVT::i16, MVT::f64, 2 },
    { ISD::FP_TO_UINT,  MVT::i16, MVT::f64, 2 },
    { ISD::FP_TO_SINT,  MVT::i32, MVT::f32, 2 },
    { ISD::FP_TO_UINT,  MVT::i32, MVT::f32, 2 },
    { ISD::FP_TO_SINT,  MVT::i32, MVT::f64, 2 },
    { ISD::FP_TO_UINT,  MVT::i32, MVT::f64, 2 },
    { ISD::FP_TO_SINT,  MVT::i64, MVT::f32, 10 },
    { ISD::FP_TO_UINT,  MVT::i64, MVT::f32, 10 },
    { ISD::FP_TO_SINT,  MVT::i64, MVT::f64, 10 },
    { ISD::FP_TO_UINT,  MVT::i64, MVT::f64, 10 }
  };
  if (SrcTy.isFloatingPoint() && ST->hasNEON()) {
    int Idx = ConvertCostTableLookup<MVT>(NEONFloatConversionTbl,
                                        array_lengthof(NEONFloatConversionTbl),
                                        ISD, DstTy.getSimpleVT(),
                                        SrcTy.getSimpleVT());
    if (Idx != -1)
        return NEONFloatConversionTbl[Idx].Cost;
  }

  // Scalar integer to float conversions.
  static const TypeConversionCostTblEntry<MVT> NEONIntegerConversionTbl[] = {
    { ISD::SINT_TO_FP,  MVT::f32, MVT::i1, 2 },
    { ISD::UINT_TO_FP,  MVT::f32, MVT::i1, 2 },
    { ISD::SINT_TO_FP,  MVT::f64, MVT::i1, 2 },
    { ISD::UINT_TO_FP,  MVT::f64, MVT::i1, 2 },
    { ISD::SINT_TO_FP,  MVT::f32, MVT::i8, 2 },
    { ISD::UINT_TO_FP,  MVT::f32, MVT::i8, 2 },
    { ISD::SINT_TO_FP,  MVT::f64, MVT::i8, 2 },
    { ISD::UINT_TO_FP,  MVT::f64, MVT::i8, 2 },
    { ISD::SINT_TO_FP,  MVT::f32, MVT::i16, 2 },
    { ISD::UINT_TO_FP,  MVT::f32, MVT::i16, 2 },
    { ISD::SINT_TO_FP,  MVT::f64, MVT::i16, 2 },
    { ISD::UINT_TO_FP,  MVT::f64, MVT::i16, 2 },
    { ISD::SINT_TO_FP,  MVT::f32, MVT::i32, 2 },
    { ISD::UINT_TO_FP,  MVT::f32, MVT::i32, 2 },
    { ISD::SINT_TO_FP,  MVT::f64, MVT::i32, 2 },
    { ISD::UINT_TO_FP,  MVT::f64, MVT::i32, 2 },
    { ISD::SINT_TO_FP,  MVT::f32, MVT::i64, 10 },
    { ISD::UINT_TO_FP,  MVT::f32, MVT::i64, 10 },
    { ISD::SINT_TO_FP,  MVT::f64, MVT::i64, 10 },
    { ISD::UINT_TO_FP,  MVT::f64, MVT::i64, 10 }
  };

  if (SrcTy.isInteger() && ST->hasNEON()) {
    int Idx = ConvertCostTableLookup<MVT>(NEONIntegerConversionTbl,
                                       array_lengthof(NEONIntegerConversionTbl),
                                       ISD, DstTy.getSimpleVT(),
                                       SrcTy.getSimpleVT());
    if (Idx != -1)
      return NEONIntegerConversionTbl[Idx].Cost;
  }

  // Scalar integer conversion costs.
  static const TypeConversionCostTblEntry<MVT> ARMIntegerConversionTbl[] = {
    // i16 -> i64 requires two dependent operations.
    { ISD::SIGN_EXTEND, MVT::i64, MVT::i16, 2 },

    // Truncates on i64 are assumed to be free.
    { ISD::TRUNCATE,    MVT::i32, MVT::i64, 0 },
    { ISD::TRUNCATE,    MVT::i16, MVT::i64, 0 },
    { ISD::TRUNCATE,    MVT::i8,  MVT::i64, 0 },
    { ISD::TRUNCATE,    MVT::i1,  MVT::i64, 0 }
  };

  if (SrcTy.isInteger()) {
    int Idx =
      ConvertCostTableLookup<MVT>(ARMIntegerConversionTbl,
                                  array_lengthof(ARMIntegerConversionTbl),
                                  ISD, DstTy.getSimpleVT(),
                                  SrcTy.getSimpleVT());
    if (Idx != -1)
      return ARMIntegerConversionTbl[Idx].Cost;
  }

  return TargetTransformInfo::getCastInstrCost(Opcode, Dst, Src);
}

unsigned ARMTTI::getVectorInstrCost(unsigned Opcode, Type *ValTy,
                                    unsigned Index) const {
  // Penalize inserting into an D-subregister. We end up with a three times
  // lower estimated throughput on swift.
  if (ST->isSwift() &&
      Opcode == Instruction::InsertElement &&
      ValTy->isVectorTy() &&
      ValTy->getScalarSizeInBits() <= 32)
    return 3;

  return TargetTransformInfo::getVectorInstrCost(Opcode, ValTy, Index);
}

unsigned ARMTTI::getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
                                    Type *CondTy) const {

  int ISD = TLI->InstructionOpcodeToISD(Opcode);
  // On NEON a a vector select gets lowered to vbsl.
  if (ST->hasNEON() && ValTy->isVectorTy() && ISD == ISD::SELECT) {
    // Lowering of some vector selects is currently far from perfect.
    static const TypeConversionCostTblEntry<MVT> NEONVectorSelectTbl[] = {
      { ISD::SELECT, MVT::v16i1, MVT::v16i16, 2*16 + 1 + 3*1 + 4*1 },
      { ISD::SELECT, MVT::v8i1, MVT::v8i32, 4*8 + 1*3 + 1*4 + 1*2 },
      { ISD::SELECT, MVT::v16i1, MVT::v16i32, 4*16 + 1*6 + 1*8 + 1*4 },
      { ISD::SELECT, MVT::v4i1, MVT::v4i64, 4*4 + 1*2 + 1 },
      { ISD::SELECT, MVT::v8i1, MVT::v8i64, 50 },
      { ISD::SELECT, MVT::v16i1, MVT::v16i64, 100 }
    };

    EVT SelCondTy = TLI->getValueType(CondTy);
    EVT SelValTy = TLI->getValueType(ValTy);
    int Idx = ConvertCostTableLookup<MVT>(NEONVectorSelectTbl,
                                          array_lengthof(NEONVectorSelectTbl),
                                          ISD, SelCondTy.getSimpleVT(),
                                          SelValTy.getSimpleVT());
    if (Idx != -1)
      return NEONVectorSelectTbl[Idx].Cost;

    std::pair<unsigned, MVT> LT = TLI->getTypeLegalizationCost(ValTy);
    return LT.first;
  }

  return TargetTransformInfo::getCmpSelInstrCost(Opcode, ValTy, CondTy);
}

unsigned ARMTTI::getAddressComputationCost(Type *Ty) const {
  // In many cases the address computation is not merged into the instruction
  // addressing mode.
  return 1;
}

unsigned ARMTTI::getShuffleCost(ShuffleKind Kind, Type *Tp, int Index,
                                Type *SubTp) const {
  // We only handle costs of reverse shuffles for now.
  if (Kind != SK_Reverse)
    return TargetTransformInfo::getShuffleCost(Kind, Tp, Index, SubTp);

  static const CostTblEntry<MVT> NEONShuffleTbl[] = {
    // Reverse shuffle cost one instruction if we are shuffling within a double
    // word (vrev) or two if we shuffle a quad word (vrev, vext).
    { ISD::VECTOR_SHUFFLE, MVT::v2i32, 1 },
    { ISD::VECTOR_SHUFFLE, MVT::v2f32, 1 },
    { ISD::VECTOR_SHUFFLE, MVT::v2i64, 1 },
    { ISD::VECTOR_SHUFFLE, MVT::v2f64, 1 },

    { ISD::VECTOR_SHUFFLE, MVT::v4i32, 2 },
    { ISD::VECTOR_SHUFFLE, MVT::v4f32, 2 },
    { ISD::VECTOR_SHUFFLE, MVT::v8i16, 2 },
    { ISD::VECTOR_SHUFFLE, MVT::v16i8, 2 }
  };

  std::pair<unsigned, MVT> LT = TLI->getTypeLegalizationCost(Tp);

  int Idx = CostTableLookup<MVT>(NEONShuffleTbl, array_lengthof(NEONShuffleTbl),
                                 ISD::VECTOR_SHUFFLE, LT.second);
  if (Idx == -1)
    return TargetTransformInfo::getShuffleCost(Kind, Tp, Index, SubTp);

  return LT.first * NEONShuffleTbl[Idx].Cost;
}