blob: 1433f7bbc943bf844345d298fb001bb2f4093f19 [file] [log] [blame]
Chris Lattnerd12c27c2010-01-05 06:09:35 +00001//===- InstCombineMulDivRem.cpp -------------------------------------------===//
2//
3// The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file implements the visit functions for mul, fmul, sdiv, udiv, fdiv,
11// srem, urem, frem.
12//
13//===----------------------------------------------------------------------===//
14
15#include "InstCombine.h"
16#include "llvm/IntrinsicInst.h"
17#include "llvm/Support/PatternMatch.h"
18using namespace llvm;
19using namespace PatternMatch;
20
21/// SubOne - Subtract one from a ConstantInt.
22static Constant *SubOne(ConstantInt *C) {
23 return ConstantInt::get(C->getContext(), C->getValue()-1);
24}
25
26/// MultiplyOverflows - True if the multiply can not be expressed in an int
27/// this size.
28static bool MultiplyOverflows(ConstantInt *C1, ConstantInt *C2, bool sign) {
29 uint32_t W = C1->getBitWidth();
30 APInt LHSExt = C1->getValue(), RHSExt = C2->getValue();
31 if (sign) {
32 LHSExt.sext(W * 2);
33 RHSExt.sext(W * 2);
34 } else {
35 LHSExt.zext(W * 2);
36 RHSExt.zext(W * 2);
37 }
38
39 APInt MulExt = LHSExt * RHSExt;
40
41 if (!sign)
42 return MulExt.ugt(APInt::getLowBitsSet(W * 2, W));
43
44 APInt Min = APInt::getSignedMinValue(W).sext(W * 2);
45 APInt Max = APInt::getSignedMaxValue(W).sext(W * 2);
46 return MulExt.slt(Min) || MulExt.sgt(Max);
47}
48
49Instruction *InstCombiner::visitMul(BinaryOperator &I) {
Duncan Sands096aa792010-11-13 15:10:37 +000050 bool Changed = SimplifyAssociativeOrCommutative(I);
Chris Lattnerd12c27c2010-01-05 06:09:35 +000051 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
52
53 if (isa<UndefValue>(Op1)) // undef * X -> 0
54 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
55
56 // Simplify mul instructions with a constant RHS.
57 if (Constant *Op1C = dyn_cast<Constant>(Op1)) {
58 if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1C)) {
59
60 // ((X << C1)*C2) == (X * (C2 << C1))
61 if (BinaryOperator *SI = dyn_cast<BinaryOperator>(Op0))
62 if (SI->getOpcode() == Instruction::Shl)
63 if (Constant *ShOp = dyn_cast<Constant>(SI->getOperand(1)))
64 return BinaryOperator::CreateMul(SI->getOperand(0),
65 ConstantExpr::getShl(CI, ShOp));
66
67 if (CI->isZero())
68 return ReplaceInstUsesWith(I, Op1C); // X * 0 == 0
69 if (CI->equalsInt(1)) // X * 1 == X
70 return ReplaceInstUsesWith(I, Op0);
71 if (CI->isAllOnesValue()) // X * -1 == 0 - X
72 return BinaryOperator::CreateNeg(Op0, I.getName());
73
74 const APInt& Val = cast<ConstantInt>(CI)->getValue();
75 if (Val.isPowerOf2()) { // Replace X*(2^C) with X << C
76 return BinaryOperator::CreateShl(Op0,
77 ConstantInt::get(Op0->getType(), Val.logBase2()));
78 }
Duncan Sands1df98592010-02-16 11:11:14 +000079 } else if (Op1C->getType()->isVectorTy()) {
Chris Lattnerd12c27c2010-01-05 06:09:35 +000080 if (Op1C->isNullValue())
81 return ReplaceInstUsesWith(I, Op1C);
82
83 if (ConstantVector *Op1V = dyn_cast<ConstantVector>(Op1C)) {
84 if (Op1V->isAllOnesValue()) // X * -1 == 0 - X
85 return BinaryOperator::CreateNeg(Op0, I.getName());
86
87 // As above, vector X*splat(1.0) -> X in all defined cases.
88 if (Constant *Splat = Op1V->getSplatValue()) {
89 if (ConstantInt *CI = dyn_cast<ConstantInt>(Splat))
90 if (CI->equalsInt(1))
91 return ReplaceInstUsesWith(I, Op0);
92 }
93 }
94 }
95
96 if (BinaryOperator *Op0I = dyn_cast<BinaryOperator>(Op0))
97 if (Op0I->getOpcode() == Instruction::Add && Op0I->hasOneUse() &&
98 isa<ConstantInt>(Op0I->getOperand(1)) && isa<ConstantInt>(Op1C)) {
99 // Canonicalize (X+C1)*C2 -> X*C2+C1*C2.
100 Value *Add = Builder->CreateMul(Op0I->getOperand(0), Op1C, "tmp");
101 Value *C1C2 = Builder->CreateMul(Op1C, Op0I->getOperand(1));
102 return BinaryOperator::CreateAdd(Add, C1C2);
103
104 }
105
106 // Try to fold constant mul into select arguments.
107 if (SelectInst *SI = dyn_cast<SelectInst>(Op0))
108 if (Instruction *R = FoldOpIntoSelect(I, SI))
109 return R;
110
111 if (isa<PHINode>(Op0))
112 if (Instruction *NV = FoldOpIntoPhi(I))
113 return NV;
114 }
115
116 if (Value *Op0v = dyn_castNegVal(Op0)) // -X * -Y = X*Y
117 if (Value *Op1v = dyn_castNegVal(Op1))
118 return BinaryOperator::CreateMul(Op0v, Op1v);
119
120 // (X / Y) * Y = X - (X % Y)
121 // (X / Y) * -Y = (X % Y) - X
122 {
123 Value *Op1C = Op1;
124 BinaryOperator *BO = dyn_cast<BinaryOperator>(Op0);
125 if (!BO ||
126 (BO->getOpcode() != Instruction::UDiv &&
127 BO->getOpcode() != Instruction::SDiv)) {
128 Op1C = Op0;
129 BO = dyn_cast<BinaryOperator>(Op1);
130 }
131 Value *Neg = dyn_castNegVal(Op1C);
132 if (BO && BO->hasOneUse() &&
133 (BO->getOperand(1) == Op1C || BO->getOperand(1) == Neg) &&
134 (BO->getOpcode() == Instruction::UDiv ||
135 BO->getOpcode() == Instruction::SDiv)) {
136 Value *Op0BO = BO->getOperand(0), *Op1BO = BO->getOperand(1);
137
138 // If the division is exact, X % Y is zero.
139 if (SDivOperator *SDiv = dyn_cast<SDivOperator>(BO))
140 if (SDiv->isExact()) {
141 if (Op1BO == Op1C)
142 return ReplaceInstUsesWith(I, Op0BO);
143 return BinaryOperator::CreateNeg(Op0BO);
144 }
145
146 Value *Rem;
147 if (BO->getOpcode() == Instruction::UDiv)
148 Rem = Builder->CreateURem(Op0BO, Op1BO);
149 else
150 Rem = Builder->CreateSRem(Op0BO, Op1BO);
151 Rem->takeName(BO);
152
153 if (Op1BO == Op1C)
154 return BinaryOperator::CreateSub(Op0BO, Rem);
155 return BinaryOperator::CreateSub(Rem, Op0BO);
156 }
157 }
158
159 /// i1 mul -> i1 and.
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000160 if (I.getType()->isIntegerTy(1))
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000161 return BinaryOperator::CreateAnd(Op0, Op1);
162
163 // X*(1 << Y) --> X << Y
164 // (1 << Y)*X --> X << Y
165 {
166 Value *Y;
167 if (match(Op0, m_Shl(m_One(), m_Value(Y))))
168 return BinaryOperator::CreateShl(Op1, Y);
169 if (match(Op1, m_Shl(m_One(), m_Value(Y))))
170 return BinaryOperator::CreateShl(Op0, Y);
171 }
172
173 // If one of the operands of the multiply is a cast from a boolean value, then
174 // we know the bool is either zero or one, so this is a 'masking' multiply.
175 // X * Y (where Y is 0 or 1) -> X & (0-Y)
Duncan Sands1df98592010-02-16 11:11:14 +0000176 if (!I.getType()->isVectorTy()) {
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000177 // -2 is "-1 << 1" so it is all bits set except the low one.
178 APInt Negative2(I.getType()->getPrimitiveSizeInBits(), (uint64_t)-2, true);
179
180 Value *BoolCast = 0, *OtherOp = 0;
181 if (MaskedValueIsZero(Op0, Negative2))
182 BoolCast = Op0, OtherOp = Op1;
183 else if (MaskedValueIsZero(Op1, Negative2))
184 BoolCast = Op1, OtherOp = Op0;
185
186 if (BoolCast) {
187 Value *V = Builder->CreateSub(Constant::getNullValue(I.getType()),
188 BoolCast, "tmp");
189 return BinaryOperator::CreateAnd(V, OtherOp);
190 }
191 }
192
193 return Changed ? &I : 0;
194}
195
196Instruction *InstCombiner::visitFMul(BinaryOperator &I) {
Duncan Sands096aa792010-11-13 15:10:37 +0000197 bool Changed = SimplifyAssociativeOrCommutative(I);
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000198 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
199
200 // Simplify mul instructions with a constant RHS...
201 if (Constant *Op1C = dyn_cast<Constant>(Op1)) {
202 if (ConstantFP *Op1F = dyn_cast<ConstantFP>(Op1C)) {
203 // "In IEEE floating point, x*1 is not equivalent to x for nans. However,
204 // ANSI says we can drop signals, so we can do this anyway." (from GCC)
205 if (Op1F->isExactlyValue(1.0))
Dan Gohmana9445e12010-03-02 01:11:08 +0000206 return ReplaceInstUsesWith(I, Op0); // Eliminate 'fmul double %X, 1.0'
Duncan Sands1df98592010-02-16 11:11:14 +0000207 } else if (Op1C->getType()->isVectorTy()) {
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000208 if (ConstantVector *Op1V = dyn_cast<ConstantVector>(Op1C)) {
209 // As above, vector X*splat(1.0) -> X in all defined cases.
210 if (Constant *Splat = Op1V->getSplatValue()) {
211 if (ConstantFP *F = dyn_cast<ConstantFP>(Splat))
212 if (F->isExactlyValue(1.0))
213 return ReplaceInstUsesWith(I, Op0);
214 }
215 }
216 }
217
218 // Try to fold constant mul into select arguments.
219 if (SelectInst *SI = dyn_cast<SelectInst>(Op0))
220 if (Instruction *R = FoldOpIntoSelect(I, SI))
221 return R;
222
223 if (isa<PHINode>(Op0))
224 if (Instruction *NV = FoldOpIntoPhi(I))
225 return NV;
226 }
227
228 if (Value *Op0v = dyn_castFNegVal(Op0)) // -X * -Y = X*Y
229 if (Value *Op1v = dyn_castFNegVal(Op1))
230 return BinaryOperator::CreateFMul(Op0v, Op1v);
231
232 return Changed ? &I : 0;
233}
234
235/// SimplifyDivRemOfSelect - Try to fold a divide or remainder of a select
236/// instruction.
237bool InstCombiner::SimplifyDivRemOfSelect(BinaryOperator &I) {
238 SelectInst *SI = cast<SelectInst>(I.getOperand(1));
239
240 // div/rem X, (Cond ? 0 : Y) -> div/rem X, Y
241 int NonNullOperand = -1;
242 if (Constant *ST = dyn_cast<Constant>(SI->getOperand(1)))
243 if (ST->isNullValue())
244 NonNullOperand = 2;
245 // div/rem X, (Cond ? Y : 0) -> div/rem X, Y
246 if (Constant *ST = dyn_cast<Constant>(SI->getOperand(2)))
247 if (ST->isNullValue())
248 NonNullOperand = 1;
249
250 if (NonNullOperand == -1)
251 return false;
252
253 Value *SelectCond = SI->getOperand(0);
254
255 // Change the div/rem to use 'Y' instead of the select.
256 I.setOperand(1, SI->getOperand(NonNullOperand));
257
258 // Okay, we know we replace the operand of the div/rem with 'Y' with no
259 // problem. However, the select, or the condition of the select may have
260 // multiple uses. Based on our knowledge that the operand must be non-zero,
261 // propagate the known value for the select into other uses of it, and
262 // propagate a known value of the condition into its other users.
263
264 // If the select and condition only have a single use, don't bother with this,
265 // early exit.
266 if (SI->use_empty() && SelectCond->hasOneUse())
267 return true;
268
269 // Scan the current block backward, looking for other uses of SI.
270 BasicBlock::iterator BBI = &I, BBFront = I.getParent()->begin();
271
272 while (BBI != BBFront) {
273 --BBI;
274 // If we found a call to a function, we can't assume it will return, so
275 // information from below it cannot be propagated above it.
276 if (isa<CallInst>(BBI) && !isa<IntrinsicInst>(BBI))
277 break;
278
279 // Replace uses of the select or its condition with the known values.
280 for (Instruction::op_iterator I = BBI->op_begin(), E = BBI->op_end();
281 I != E; ++I) {
282 if (*I == SI) {
283 *I = SI->getOperand(NonNullOperand);
284 Worklist.Add(BBI);
285 } else if (*I == SelectCond) {
286 *I = NonNullOperand == 1 ? ConstantInt::getTrue(BBI->getContext()) :
287 ConstantInt::getFalse(BBI->getContext());
288 Worklist.Add(BBI);
289 }
290 }
291
292 // If we past the instruction, quit looking for it.
293 if (&*BBI == SI)
294 SI = 0;
295 if (&*BBI == SelectCond)
296 SelectCond = 0;
297
298 // If we ran out of things to eliminate, break out of the loop.
299 if (SelectCond == 0 && SI == 0)
300 break;
301
302 }
303 return true;
304}
305
306
307/// This function implements the transforms on div instructions that work
308/// regardless of the kind of div instruction it is (udiv, sdiv, or fdiv). It is
309/// used by the visitors to those instructions.
310/// @brief Transforms common to all three div instructions
311Instruction *InstCombiner::commonDivTransforms(BinaryOperator &I) {
312 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
313
314 // undef / X -> 0 for integer.
315 // undef / X -> undef for FP (the undef could be a snan).
316 if (isa<UndefValue>(Op0)) {
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000317 if (Op0->getType()->isFPOrFPVectorTy())
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000318 return ReplaceInstUsesWith(I, Op0);
319 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
320 }
321
322 // X / undef -> undef
323 if (isa<UndefValue>(Op1))
324 return ReplaceInstUsesWith(I, Op1);
325
326 return 0;
327}
328
329/// This function implements the transforms common to both integer division
330/// instructions (udiv and sdiv). It is called by the visitors to those integer
331/// division instructions.
332/// @brief Common integer divide transforms
333Instruction *InstCombiner::commonIDivTransforms(BinaryOperator &I) {
334 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
335
336 // (sdiv X, X) --> 1 (udiv X, X) --> 1
337 if (Op0 == Op1) {
338 if (const VectorType *Ty = dyn_cast<VectorType>(I.getType())) {
339 Constant *CI = ConstantInt::get(Ty->getElementType(), 1);
340 std::vector<Constant*> Elts(Ty->getNumElements(), CI);
341 return ReplaceInstUsesWith(I, ConstantVector::get(Elts));
342 }
343
344 Constant *CI = ConstantInt::get(I.getType(), 1);
345 return ReplaceInstUsesWith(I, CI);
346 }
347
348 if (Instruction *Common = commonDivTransforms(I))
349 return Common;
350
351 // Handle cases involving: [su]div X, (select Cond, Y, Z)
352 // This does not apply for fdiv.
353 if (isa<SelectInst>(Op1) && SimplifyDivRemOfSelect(I))
354 return &I;
355
356 if (ConstantInt *RHS = dyn_cast<ConstantInt>(Op1)) {
357 // div X, 1 == X
358 if (RHS->equalsInt(1))
359 return ReplaceInstUsesWith(I, Op0);
360
361 // (X / C1) / C2 -> X / (C1*C2)
362 if (Instruction *LHS = dyn_cast<Instruction>(Op0))
363 if (Instruction::BinaryOps(LHS->getOpcode()) == I.getOpcode())
364 if (ConstantInt *LHSRHS = dyn_cast<ConstantInt>(LHS->getOperand(1))) {
365 if (MultiplyOverflows(RHS, LHSRHS,
366 I.getOpcode()==Instruction::SDiv))
367 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
368 else
369 return BinaryOperator::Create(I.getOpcode(), LHS->getOperand(0),
370 ConstantExpr::getMul(RHS, LHSRHS));
371 }
372
373 if (!RHS->isZero()) { // avoid X udiv 0
374 if (SelectInst *SI = dyn_cast<SelectInst>(Op0))
375 if (Instruction *R = FoldOpIntoSelect(I, SI))
376 return R;
377 if (isa<PHINode>(Op0))
378 if (Instruction *NV = FoldOpIntoPhi(I))
379 return NV;
380 }
381 }
382
383 // 0 / X == 0, we don't need to preserve faults!
384 if (ConstantInt *LHS = dyn_cast<ConstantInt>(Op0))
385 if (LHS->equalsInt(0))
386 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
387
388 // It can't be division by zero, hence it must be division by one.
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000389 if (I.getType()->isIntegerTy(1))
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000390 return ReplaceInstUsesWith(I, Op0);
391
392 if (ConstantVector *Op1V = dyn_cast<ConstantVector>(Op1)) {
393 if (ConstantInt *X = cast_or_null<ConstantInt>(Op1V->getSplatValue()))
394 // div X, 1 == X
395 if (X->isOne())
396 return ReplaceInstUsesWith(I, Op0);
397 }
398
399 return 0;
400}
401
402Instruction *InstCombiner::visitUDiv(BinaryOperator &I) {
403 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
404
405 // Handle the integer div common cases
406 if (Instruction *Common = commonIDivTransforms(I))
407 return Common;
408
409 if (ConstantInt *C = dyn_cast<ConstantInt>(Op1)) {
Owen Anderson5b396202010-01-17 06:49:03 +0000410 // X udiv 2^C -> X >> C
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000411 // Check to see if this is an unsigned division with an exact power of 2,
412 // if so, convert to a right shift.
413 if (C->getValue().isPowerOf2()) // 0 not included in isPowerOf2
414 return BinaryOperator::CreateLShr(Op0,
415 ConstantInt::get(Op0->getType(), C->getValue().logBase2()));
416
417 // X udiv C, where C >= signbit
418 if (C->getValue().isNegative()) {
419 Value *IC = Builder->CreateICmpULT( Op0, C);
420 return SelectInst::Create(IC, Constant::getNullValue(I.getType()),
421 ConstantInt::get(I.getType(), 1));
422 }
423 }
424
425 // X udiv (C1 << N), where C1 is "1<<C2" --> X >> (N+C2)
426 if (BinaryOperator *RHSI = dyn_cast<BinaryOperator>(I.getOperand(1))) {
427 if (RHSI->getOpcode() == Instruction::Shl &&
428 isa<ConstantInt>(RHSI->getOperand(0))) {
429 const APInt& C1 = cast<ConstantInt>(RHSI->getOperand(0))->getValue();
430 if (C1.isPowerOf2()) {
431 Value *N = RHSI->getOperand(1);
432 const Type *NTy = N->getType();
433 if (uint32_t C2 = C1.logBase2())
434 N = Builder->CreateAdd(N, ConstantInt::get(NTy, C2), "tmp");
435 return BinaryOperator::CreateLShr(Op0, N);
436 }
437 }
438 }
439
440 // udiv X, (Select Cond, C1, C2) --> Select Cond, (shr X, C1), (shr X, C2)
441 // where C1&C2 are powers of two.
442 if (SelectInst *SI = dyn_cast<SelectInst>(Op1))
443 if (ConstantInt *STO = dyn_cast<ConstantInt>(SI->getOperand(1)))
444 if (ConstantInt *SFO = dyn_cast<ConstantInt>(SI->getOperand(2))) {
445 const APInt &TVA = STO->getValue(), &FVA = SFO->getValue();
446 if (TVA.isPowerOf2() && FVA.isPowerOf2()) {
447 // Compute the shift amounts
448 uint32_t TSA = TVA.logBase2(), FSA = FVA.logBase2();
449 // Construct the "on true" case of the select
450 Constant *TC = ConstantInt::get(Op0->getType(), TSA);
451 Value *TSI = Builder->CreateLShr(Op0, TC, SI->getName()+".t");
452
453 // Construct the "on false" case of the select
454 Constant *FC = ConstantInt::get(Op0->getType(), FSA);
455 Value *FSI = Builder->CreateLShr(Op0, FC, SI->getName()+".f");
456
457 // construct the select instruction and return it.
458 return SelectInst::Create(SI->getOperand(0), TSI, FSI, SI->getName());
459 }
460 }
461 return 0;
462}
463
464Instruction *InstCombiner::visitSDiv(BinaryOperator &I) {
465 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
466
467 // Handle the integer div common cases
468 if (Instruction *Common = commonIDivTransforms(I))
469 return Common;
470
471 if (ConstantInt *RHS = dyn_cast<ConstantInt>(Op1)) {
472 // sdiv X, -1 == -X
473 if (RHS->isAllOnesValue())
474 return BinaryOperator::CreateNeg(Op0);
475
476 // sdiv X, C --> ashr X, log2(C)
477 if (cast<SDivOperator>(&I)->isExact() &&
478 RHS->getValue().isNonNegative() &&
479 RHS->getValue().isPowerOf2()) {
480 Value *ShAmt = llvm::ConstantInt::get(RHS->getType(),
481 RHS->getValue().exactLogBase2());
482 return BinaryOperator::CreateAShr(Op0, ShAmt, I.getName());
483 }
484
485 // -X/C --> X/-C provided the negation doesn't overflow.
486 if (SubOperator *Sub = dyn_cast<SubOperator>(Op0))
487 if (isa<Constant>(Sub->getOperand(0)) &&
488 cast<Constant>(Sub->getOperand(0))->isNullValue() &&
489 Sub->hasNoSignedWrap())
490 return BinaryOperator::CreateSDiv(Sub->getOperand(1),
491 ConstantExpr::getNeg(RHS));
492 }
493
494 // If the sign bits of both operands are zero (i.e. we can prove they are
495 // unsigned inputs), turn this into a udiv.
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000496 if (I.getType()->isIntegerTy()) {
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000497 APInt Mask(APInt::getSignBit(I.getType()->getPrimitiveSizeInBits()));
498 if (MaskedValueIsZero(Op0, Mask)) {
499 if (MaskedValueIsZero(Op1, Mask)) {
500 // X sdiv Y -> X udiv Y, iff X and Y don't have sign bit set
501 return BinaryOperator::CreateUDiv(Op0, Op1, I.getName());
502 }
503 ConstantInt *ShiftedInt;
504 if (match(Op1, m_Shl(m_ConstantInt(ShiftedInt), m_Value())) &&
505 ShiftedInt->getValue().isPowerOf2()) {
506 // X sdiv (1 << Y) -> X udiv (1 << Y) ( -> X u>> Y)
507 // Safe because the only negative value (1 << Y) can take on is
508 // INT_MIN, and X sdiv INT_MIN == X udiv INT_MIN == 0 if X doesn't have
509 // the sign bit set.
510 return BinaryOperator::CreateUDiv(Op0, Op1, I.getName());
511 }
512 }
513 }
514
515 return 0;
516}
517
518Instruction *InstCombiner::visitFDiv(BinaryOperator &I) {
519 return commonDivTransforms(I);
520}
521
522/// This function implements the transforms on rem instructions that work
523/// regardless of the kind of rem instruction it is (urem, srem, or frem). It
524/// is used by the visitors to those instructions.
525/// @brief Transforms common to all three rem instructions
526Instruction *InstCombiner::commonRemTransforms(BinaryOperator &I) {
527 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
528
529 if (isa<UndefValue>(Op0)) { // undef % X -> 0
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000530 if (I.getType()->isFPOrFPVectorTy())
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000531 return ReplaceInstUsesWith(I, Op0); // X % undef -> undef (could be SNaN)
532 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
533 }
534 if (isa<UndefValue>(Op1))
535 return ReplaceInstUsesWith(I, Op1); // X % undef -> undef
536
537 // Handle cases involving: rem X, (select Cond, Y, Z)
538 if (isa<SelectInst>(Op1) && SimplifyDivRemOfSelect(I))
539 return &I;
540
541 return 0;
542}
543
544/// This function implements the transforms common to both integer remainder
545/// instructions (urem and srem). It is called by the visitors to those integer
546/// remainder instructions.
547/// @brief Common integer remainder transforms
548Instruction *InstCombiner::commonIRemTransforms(BinaryOperator &I) {
549 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
550
551 if (Instruction *common = commonRemTransforms(I))
552 return common;
553
Benjamin Kramer1951a5b2010-11-17 19:11:46 +0000554 // X % X == 0
555 if (Op0 == Op1)
556 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
557
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000558 // 0 % X == 0 for integer, we don't need to preserve faults!
559 if (Constant *LHS = dyn_cast<Constant>(Op0))
560 if (LHS->isNullValue())
561 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
562
563 if (ConstantInt *RHS = dyn_cast<ConstantInt>(Op1)) {
564 // X % 0 == undef, we don't need to preserve faults!
565 if (RHS->equalsInt(0))
566 return ReplaceInstUsesWith(I, UndefValue::get(I.getType()));
567
568 if (RHS->equalsInt(1)) // X % 1 == 0
569 return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
570
571 if (Instruction *Op0I = dyn_cast<Instruction>(Op0)) {
572 if (SelectInst *SI = dyn_cast<SelectInst>(Op0I)) {
573 if (Instruction *R = FoldOpIntoSelect(I, SI))
574 return R;
575 } else if (isa<PHINode>(Op0I)) {
576 if (Instruction *NV = FoldOpIntoPhi(I))
577 return NV;
578 }
579
580 // See if we can fold away this rem instruction.
581 if (SimplifyDemandedInstructionBits(I))
582 return &I;
583 }
584 }
585
586 return 0;
587}
588
589Instruction *InstCombiner::visitURem(BinaryOperator &I) {
590 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
591
592 if (Instruction *common = commonIRemTransforms(I))
593 return common;
594
595 if (ConstantInt *RHS = dyn_cast<ConstantInt>(Op1)) {
596 // X urem C^2 -> X and C
597 // Check to see if this is an unsigned remainder with an exact power of 2,
598 // if so, convert to a bitwise and.
599 if (ConstantInt *C = dyn_cast<ConstantInt>(RHS))
600 if (C->getValue().isPowerOf2())
601 return BinaryOperator::CreateAnd(Op0, SubOne(C));
602 }
603
604 if (Instruction *RHSI = dyn_cast<Instruction>(I.getOperand(1))) {
605 // Turn A % (C << N), where C is 2^k, into A & ((C << N)-1)
606 if (RHSI->getOpcode() == Instruction::Shl &&
607 isa<ConstantInt>(RHSI->getOperand(0))) {
608 if (cast<ConstantInt>(RHSI->getOperand(0))->getValue().isPowerOf2()) {
609 Constant *N1 = Constant::getAllOnesValue(I.getType());
610 Value *Add = Builder->CreateAdd(RHSI, N1, "tmp");
611 return BinaryOperator::CreateAnd(Op0, Add);
612 }
613 }
614 }
615
616 // urem X, (select Cond, 2^C1, 2^C2) --> select Cond, (and X, C1), (and X, C2)
617 // where C1&C2 are powers of two.
618 if (SelectInst *SI = dyn_cast<SelectInst>(Op1)) {
619 if (ConstantInt *STO = dyn_cast<ConstantInt>(SI->getOperand(1)))
620 if (ConstantInt *SFO = dyn_cast<ConstantInt>(SI->getOperand(2))) {
621 // STO == 0 and SFO == 0 handled above.
622 if ((STO->getValue().isPowerOf2()) &&
623 (SFO->getValue().isPowerOf2())) {
624 Value *TrueAnd = Builder->CreateAnd(Op0, SubOne(STO),
625 SI->getName()+".t");
626 Value *FalseAnd = Builder->CreateAnd(Op0, SubOne(SFO),
627 SI->getName()+".f");
628 return SelectInst::Create(SI->getOperand(0), TrueAnd, FalseAnd);
629 }
630 }
631 }
632
633 return 0;
634}
635
636Instruction *InstCombiner::visitSRem(BinaryOperator &I) {
637 Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
638
639 // Handle the integer rem common cases
640 if (Instruction *Common = commonIRemTransforms(I))
641 return Common;
642
643 if (Value *RHSNeg = dyn_castNegVal(Op1))
644 if (!isa<Constant>(RHSNeg) ||
645 (isa<ConstantInt>(RHSNeg) &&
646 cast<ConstantInt>(RHSNeg)->getValue().isStrictlyPositive())) {
647 // X % -Y -> X % Y
648 Worklist.AddValue(I.getOperand(1));
649 I.setOperand(1, RHSNeg);
650 return &I;
651 }
652
653 // If the sign bits of both operands are zero (i.e. we can prove they are
654 // unsigned inputs), turn this into a urem.
Duncan Sandsb0bc6c32010-02-15 16:12:20 +0000655 if (I.getType()->isIntegerTy()) {
Chris Lattnerd12c27c2010-01-05 06:09:35 +0000656 APInt Mask(APInt::getSignBit(I.getType()->getPrimitiveSizeInBits()));
657 if (MaskedValueIsZero(Op1, Mask) && MaskedValueIsZero(Op0, Mask)) {
658 // X srem Y -> X urem Y, iff X and Y don't have sign bit set
659 return BinaryOperator::CreateURem(Op0, Op1, I.getName());
660 }
661 }
662
663 // If it's a constant vector, flip any negative values positive.
664 if (ConstantVector *RHSV = dyn_cast<ConstantVector>(Op1)) {
665 unsigned VWidth = RHSV->getNumOperands();
666
667 bool hasNegative = false;
668 for (unsigned i = 0; !hasNegative && i != VWidth; ++i)
669 if (ConstantInt *RHS = dyn_cast<ConstantInt>(RHSV->getOperand(i)))
670 if (RHS->getValue().isNegative())
671 hasNegative = true;
672
673 if (hasNegative) {
674 std::vector<Constant *> Elts(VWidth);
675 for (unsigned i = 0; i != VWidth; ++i) {
676 if (ConstantInt *RHS = dyn_cast<ConstantInt>(RHSV->getOperand(i))) {
677 if (RHS->getValue().isNegative())
678 Elts[i] = cast<ConstantInt>(ConstantExpr::getNeg(RHS));
679 else
680 Elts[i] = RHS;
681 }
682 }
683
684 Constant *NewRHSV = ConstantVector::get(Elts);
685 if (NewRHSV != RHSV) {
686 Worklist.AddValue(I.getOperand(1));
687 I.setOperand(1, NewRHSV);
688 return &I;
689 }
690 }
691 }
692
693 return 0;
694}
695
696Instruction *InstCombiner::visitFRem(BinaryOperator &I) {
697 return commonRemTransforms(I);
698}
699