| 1 | //===-- AMDGPUCodeGenPrepare.cpp ------------------------------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | // |
| 9 | /// \file |
| 10 | /// This pass does misc. AMDGPU optimizations on IR before instruction |
| 11 | /// selection. |
| 12 | // |
| 13 | //===----------------------------------------------------------------------===// |
| 14 | |
| 15 | #include "AMDGPU.h" |
| 16 | #include "AMDGPUTargetMachine.h" |
| 17 | #include "SIModeRegisterDefaults.h" |
| 18 | #include "llvm/ADT/SetVector.h" |
| 19 | #include "llvm/Analysis/AssumptionCache.h" |
| 20 | #include "llvm/Analysis/ConstantFolding.h" |
| 21 | #include "llvm/Analysis/TargetLibraryInfo.h" |
| 22 | #include "llvm/Analysis/TargetTransformInfo.h" |
| 23 | #include "llvm/Analysis/UniformityAnalysis.h" |
| 24 | #include "llvm/Analysis/ValueTracking.h" |
| 25 | #include "llvm/CodeGen/TargetPassConfig.h" |
| 26 | #include "llvm/IR/Dominators.h" |
| 27 | #include "llvm/IR/IRBuilder.h" |
| 28 | #include "llvm/IR/InstVisitor.h" |
| 29 | #include "llvm/IR/IntrinsicsAMDGPU.h" |
| 30 | #include "llvm/IR/PatternMatch.h" |
| 31 | #include "llvm/IR/ValueHandle.h" |
| 32 | #include "llvm/InitializePasses.h" |
| 33 | #include "llvm/Pass.h" |
| 34 | #include "llvm/Support/KnownBits.h" |
| 35 | #include "llvm/Support/KnownFPClass.h" |
| 36 | #include "llvm/Transforms/Utils/BasicBlockUtils.h" |
| 37 | #include "llvm/Transforms/Utils/IntegerDivision.h" |
| 38 | #include "llvm/Transforms/Utils/Local.h" |
| 39 | |
| 40 | #define DEBUG_TYPE "amdgpu-codegenprepare" |
| 41 | |
| 42 | using namespace llvm; |
| 43 | using namespace llvm::PatternMatch; |
| 44 | |
| 45 | namespace { |
| 46 | |
| 47 | static cl::opt<bool> WidenLoads( |
| 48 | "amdgpu-codegenprepare-widen-constant-loads" , |
| 49 | cl::desc("Widen sub-dword constant address space loads in AMDGPUCodeGenPrepare" ), |
| 50 | cl::ReallyHidden, |
| 51 | cl::init(Val: false)); |
| 52 | |
| 53 | static cl::opt<bool> |
| 54 | BreakLargePHIs("amdgpu-codegenprepare-break-large-phis" , |
| 55 | cl::desc("Break large PHI nodes for DAGISel" ), |
| 56 | cl::ReallyHidden, cl::init(Val: true)); |
| 57 | |
| 58 | static cl::opt<bool> |
| 59 | ForceBreakLargePHIs("amdgpu-codegenprepare-force-break-large-phis" , |
| 60 | cl::desc("For testing purposes, always break large " |
| 61 | "PHIs even if it isn't profitable." ), |
| 62 | cl::ReallyHidden, cl::init(Val: false)); |
| 63 | |
| 64 | static cl::opt<unsigned> BreakLargePHIsThreshold( |
| 65 | "amdgpu-codegenprepare-break-large-phis-threshold" , |
| 66 | cl::desc("Minimum type size in bits for breaking large PHI nodes" ), |
| 67 | cl::ReallyHidden, cl::init(Val: 32)); |
| 68 | |
| 69 | static cl::opt<bool> UseMul24Intrin( |
| 70 | "amdgpu-codegenprepare-mul24" , |
| 71 | cl::desc("Introduce mul24 intrinsics in AMDGPUCodeGenPrepare" ), |
| 72 | cl::ReallyHidden, |
| 73 | cl::init(Val: true)); |
| 74 | |
| 75 | // Legalize 64-bit division by using the generic IR expansion. |
| 76 | static cl::opt<bool> ExpandDiv64InIR( |
| 77 | "amdgpu-codegenprepare-expand-div64" , |
| 78 | cl::desc("Expand 64-bit division in AMDGPUCodeGenPrepare" ), |
| 79 | cl::ReallyHidden, |
| 80 | cl::init(Val: false)); |
| 81 | |
| 82 | // Leave all division operations as they are. This supersedes ExpandDiv64InIR |
| 83 | // and is used for testing the legalizer. |
| 84 | static cl::opt<bool> DisableIDivExpand( |
| 85 | "amdgpu-codegenprepare-disable-idiv-expansion" , |
| 86 | cl::desc("Prevent expanding integer division in AMDGPUCodeGenPrepare" ), |
| 87 | cl::ReallyHidden, |
| 88 | cl::init(Val: false)); |
| 89 | |
| 90 | // Disable processing of fdiv so we can better test the backend implementations. |
| 91 | static cl::opt<bool> DisableFDivExpand( |
| 92 | "amdgpu-codegenprepare-disable-fdiv-expansion" , |
| 93 | cl::desc("Prevent expanding floating point division in AMDGPUCodeGenPrepare" ), |
| 94 | cl::ReallyHidden, |
| 95 | cl::init(Val: false)); |
| 96 | |
| 97 | class AMDGPUCodeGenPrepareImpl |
| 98 | : public InstVisitor<AMDGPUCodeGenPrepareImpl, bool> { |
| 99 | public: |
| 100 | Function &F; |
| 101 | const GCNSubtarget &ST; |
| 102 | const AMDGPUTargetMachine &TM; |
| 103 | const TargetLibraryInfo *TLI; |
| 104 | const UniformityInfo &UA; |
| 105 | const DataLayout &DL; |
| 106 | SimplifyQuery SQ; |
| 107 | const bool HasFP32DenormalFlush; |
| 108 | bool FlowChanged = false; |
| 109 | mutable Function *SqrtF32 = nullptr; |
| 110 | mutable Function *LdexpF32 = nullptr; |
| 111 | mutable SmallVector<WeakVH> DeadVals; |
| 112 | |
| 113 | DenseMap<const PHINode *, bool> BreakPhiNodesCache; |
| 114 | |
| 115 | AMDGPUCodeGenPrepareImpl(Function &F, const AMDGPUTargetMachine &TM, |
| 116 | const TargetLibraryInfo *TLI, AssumptionCache *AC, |
| 117 | const DominatorTree *DT, const UniformityInfo &UA) |
| 118 | : F(F), ST(TM.getSubtarget<GCNSubtarget>(F)), TM(TM), TLI(TLI), UA(UA), |
| 119 | DL(F.getDataLayout()), SQ(DL, TLI, DT, AC), |
| 120 | HasFP32DenormalFlush(SIModeRegisterDefaults(F, ST).FP32Denormals == |
| 121 | DenormalMode::getPreserveSign()) {} |
| 122 | |
| 123 | Function *getSqrtF32() const { |
| 124 | if (SqrtF32) |
| 125 | return SqrtF32; |
| 126 | |
| 127 | LLVMContext &Ctx = F.getContext(); |
| 128 | SqrtF32 = Intrinsic::getOrInsertDeclaration( |
| 129 | M: F.getParent(), id: Intrinsic::amdgcn_sqrt, Tys: {Type::getFloatTy(C&: Ctx)}); |
| 130 | return SqrtF32; |
| 131 | } |
| 132 | |
| 133 | Function *getLdexpF32() const { |
| 134 | if (LdexpF32) |
| 135 | return LdexpF32; |
| 136 | |
| 137 | LLVMContext &Ctx = F.getContext(); |
| 138 | LdexpF32 = Intrinsic::getOrInsertDeclaration( |
| 139 | M: F.getParent(), id: Intrinsic::ldexp, |
| 140 | Tys: {Type::getFloatTy(C&: Ctx), Type::getInt32Ty(C&: Ctx)}); |
| 141 | return LdexpF32; |
| 142 | } |
| 143 | |
| 144 | bool canBreakPHINode(const PHINode &I); |
| 145 | |
| 146 | /// Return true if \p T is a legal scalar floating point type. |
| 147 | bool isLegalFloatingTy(const Type *T) const; |
| 148 | |
| 149 | /// Wrapper to pass all the arguments to computeKnownFPClass |
| 150 | KnownFPClass computeKnownFPClass(const Value *V, FPClassTest Interested, |
| 151 | const Instruction *CtxI) const { |
| 152 | return llvm::computeKnownFPClass(V, InterestedClasses: Interested, |
| 153 | SQ: SQ.getWithInstruction(I: CtxI)); |
| 154 | } |
| 155 | |
| 156 | bool canIgnoreDenormalInput(const Value *V, const Instruction *CtxI) const { |
| 157 | return HasFP32DenormalFlush || |
| 158 | computeKnownFPClass(V, Interested: fcSubnormal, CtxI).isKnownNeverSubnormal(); |
| 159 | } |
| 160 | |
| 161 | /// \returns The minimum number of bits needed to store the value of \Op as an |
| 162 | /// unsigned integer. Truncating to this size and then zero-extending to |
| 163 | /// the original will not change the value. |
| 164 | unsigned numBitsUnsigned(Value *Op, const Instruction *CtxI) const; |
| 165 | |
| 166 | /// \returns The minimum number of bits needed to store the value of \Op as a |
| 167 | /// signed integer. Truncating to this size and then sign-extending to |
| 168 | /// the original size will not change the value. |
| 169 | unsigned numBitsSigned(Value *Op, const Instruction *CtxI) const; |
| 170 | |
| 171 | /// Replace mul instructions with llvm.amdgcn.mul.u24 or llvm.amdgcn.mul.s24. |
| 172 | /// SelectionDAG has an issue where an and asserting the bits are known |
| 173 | bool replaceMulWithMul24(BinaryOperator &I) const; |
| 174 | |
| 175 | /// Perform same function as equivalently named function in DAGCombiner. Since |
| 176 | /// we expand some divisions here, we need to perform this before obscuring. |
| 177 | bool foldBinOpIntoSelect(BinaryOperator &I) const; |
| 178 | |
| 179 | bool divHasSpecialOptimization(BinaryOperator &I, |
| 180 | Value *Num, Value *Den) const; |
| 181 | unsigned getDivNumBits(BinaryOperator &I, Value *Num, Value *Den, |
| 182 | unsigned MaxDivBits, bool Signed) const; |
| 183 | |
| 184 | /// Expands 24 bit div or rem. |
| 185 | Value* expandDivRem24(IRBuilder<> &Builder, BinaryOperator &I, |
| 186 | Value *Num, Value *Den, |
| 187 | bool IsDiv, bool IsSigned) const; |
| 188 | |
| 189 | Value *expandDivRem24Impl(IRBuilder<> &Builder, BinaryOperator &I, |
| 190 | Value *Num, Value *Den, unsigned NumBits, |
| 191 | bool IsDiv, bool IsSigned) const; |
| 192 | |
| 193 | /// Expands 32 bit div or rem. |
| 194 | Value* expandDivRem32(IRBuilder<> &Builder, BinaryOperator &I, |
| 195 | Value *Num, Value *Den) const; |
| 196 | |
| 197 | Value *shrinkDivRem64(IRBuilder<> &Builder, BinaryOperator &I, |
| 198 | Value *Num, Value *Den) const; |
| 199 | void expandDivRem64(BinaryOperator &I) const; |
| 200 | |
| 201 | /// Widen a scalar load. |
| 202 | /// |
| 203 | /// \details \p Widen scalar load for uniform, small type loads from constant |
| 204 | // memory / to a full 32-bits and then truncate the input to allow a scalar |
| 205 | // load instead of a vector load. |
| 206 | // |
| 207 | /// \returns True. |
| 208 | |
| 209 | bool canWidenScalarExtLoad(LoadInst &I) const; |
| 210 | |
| 211 | Value *matchFractPat(IntrinsicInst &I); |
| 212 | Value *applyFractPat(IRBuilder<> &Builder, Value *FractArg); |
| 213 | |
| 214 | bool canOptimizeWithRsq(FastMathFlags DivFMF, FastMathFlags SqrtFMF) const; |
| 215 | |
| 216 | Value *optimizeWithRsq(IRBuilder<> &Builder, Value *Num, Value *Den, |
| 217 | FastMathFlags DivFMF, FastMathFlags SqrtFMF, |
| 218 | const Instruction *CtxI) const; |
| 219 | |
| 220 | Value *optimizeWithRcp(IRBuilder<> &Builder, Value *Num, Value *Den, |
| 221 | FastMathFlags FMF, const Instruction *CtxI) const; |
| 222 | Value *optimizeWithFDivFast(IRBuilder<> &Builder, Value *Num, Value *Den, |
| 223 | float ReqdAccuracy) const; |
| 224 | |
| 225 | Value *visitFDivElement(IRBuilder<> &Builder, Value *Num, Value *Den, |
| 226 | FastMathFlags DivFMF, FastMathFlags SqrtFMF, |
| 227 | Value *RsqOp, const Instruction *FDiv, |
| 228 | float ReqdAccuracy) const; |
| 229 | |
| 230 | std::pair<Value *, Value *> getFrexpResults(IRBuilder<> &Builder, |
| 231 | Value *Src) const; |
| 232 | |
| 233 | Value *emitRcpIEEE1ULP(IRBuilder<> &Builder, Value *Src, |
| 234 | bool IsNegative) const; |
| 235 | Value *emitFrexpDiv(IRBuilder<> &Builder, Value *LHS, Value *RHS, |
| 236 | FastMathFlags FMF) const; |
| 237 | Value *emitSqrtIEEE2ULP(IRBuilder<> &Builder, Value *Src, |
| 238 | FastMathFlags FMF) const; |
| 239 | Value *emitRsqF64(IRBuilder<> &Builder, Value *X, FastMathFlags SqrtFMF, |
| 240 | FastMathFlags DivFMF, const Instruction *CtxI, |
| 241 | bool IsNegative) const; |
| 242 | |
| 243 | CallInst *createWorkitemIdX(IRBuilder<> &B) const; |
| 244 | void replaceWithWorkitemIdX(Instruction &I) const; |
| 245 | void replaceWithMaskedWorkitemIdX(Instruction &I, unsigned WaveSize) const; |
| 246 | bool tryReplaceWithWorkitemId(Instruction &I, unsigned Wave) const; |
| 247 | |
| 248 | bool tryNarrowMathIfNoOverflow(Instruction *I); |
| 249 | |
| 250 | public: |
| 251 | bool visitFDiv(BinaryOperator &I); |
| 252 | |
| 253 | bool visitInstruction(Instruction &I) { return false; } |
| 254 | bool visitBinaryOperator(BinaryOperator &I); |
| 255 | bool visitLoadInst(LoadInst &I); |
| 256 | bool visitSelectInst(SelectInst &I); |
| 257 | bool visitPHINode(PHINode &I); |
| 258 | bool visitAddrSpaceCastInst(AddrSpaceCastInst &I); |
| 259 | |
| 260 | bool visitIntrinsicInst(IntrinsicInst &I); |
| 261 | bool visitFMinLike(IntrinsicInst &I); |
| 262 | bool visitSqrt(IntrinsicInst &I); |
| 263 | bool visitMbcntLo(IntrinsicInst &I) const; |
| 264 | bool visitMbcntHi(IntrinsicInst &I) const; |
| 265 | bool run(); |
| 266 | }; |
| 267 | |
| 268 | class AMDGPUCodeGenPrepare : public FunctionPass { |
| 269 | public: |
| 270 | static char ID; |
| 271 | AMDGPUCodeGenPrepare() : FunctionPass(ID) {} |
| 272 | void getAnalysisUsage(AnalysisUsage &AU) const override { |
| 273 | AU.addRequired<AssumptionCacheTracker>(); |
| 274 | AU.addRequired<UniformityInfoWrapperPass>(); |
| 275 | AU.addRequired<TargetLibraryInfoWrapperPass>(); |
| 276 | |
| 277 | // FIXME: Division expansion needs to preserve the dominator tree. |
| 278 | if (!ExpandDiv64InIR) |
| 279 | AU.setPreservesAll(); |
| 280 | } |
| 281 | bool runOnFunction(Function &F) override; |
| 282 | StringRef getPassName() const override { return "AMDGPU IR optimizations" ; } |
| 283 | }; |
| 284 | |
| 285 | } // end anonymous namespace |
| 286 | |
| 287 | bool AMDGPUCodeGenPrepareImpl::run() { |
| 288 | BreakPhiNodesCache.clear(); |
| 289 | bool MadeChange = false; |
| 290 | |
| 291 | // Need to use make_early_inc_range because integer division expansion is |
| 292 | // handled by Transform/Utils, and it can delete instructions such as the |
| 293 | // terminator of the BB. |
| 294 | for (BasicBlock &BB : reverse(C&: F)) { |
| 295 | for (Instruction &I : make_early_inc_range(Range: reverse(C&: BB))) { |
| 296 | if (!isInstructionTriviallyDead(I: &I, TLI)) |
| 297 | MadeChange |= visit(I); |
| 298 | } |
| 299 | } |
| 300 | |
| 301 | while (!DeadVals.empty()) { |
| 302 | if (auto *I = dyn_cast_or_null<Instruction>(Val: DeadVals.pop_back_val())) |
| 303 | RecursivelyDeleteTriviallyDeadInstructions(V: I, TLI); |
| 304 | } |
| 305 | |
| 306 | return MadeChange; |
| 307 | } |
| 308 | |
| 309 | bool AMDGPUCodeGenPrepareImpl::isLegalFloatingTy(const Type *Ty) const { |
| 310 | return Ty->isFloatTy() || Ty->isDoubleTy() || |
| 311 | (Ty->isHalfTy() && ST.has16BitInsts()); |
| 312 | } |
| 313 | |
| 314 | bool AMDGPUCodeGenPrepareImpl::canWidenScalarExtLoad(LoadInst &I) const { |
| 315 | Type *Ty = I.getType(); |
| 316 | int TySize = DL.getTypeSizeInBits(Ty); |
| 317 | Align Alignment = DL.getValueOrABITypeAlignment(Alignment: I.getAlign(), Ty); |
| 318 | |
| 319 | return I.isSimple() && TySize < 32 && Alignment >= 4 && UA.isUniform(I: &I); |
| 320 | } |
| 321 | |
| 322 | unsigned |
| 323 | AMDGPUCodeGenPrepareImpl::numBitsUnsigned(Value *Op, |
| 324 | const Instruction *CtxI) const { |
| 325 | return computeKnownBits(V: Op, Q: SQ.getWithInstruction(I: CtxI)).countMaxActiveBits(); |
| 326 | } |
| 327 | |
| 328 | unsigned |
| 329 | AMDGPUCodeGenPrepareImpl::numBitsSigned(Value *Op, |
| 330 | const Instruction *CtxI) const { |
| 331 | return ComputeMaxSignificantBits(Op, DL: SQ.DL, AC: SQ.AC, CxtI: CtxI, DT: SQ.DT); |
| 332 | } |
| 333 | |
| 334 | static void (IRBuilder<> &Builder, |
| 335 | SmallVectorImpl<Value *> &Values, Value *V) { |
| 336 | auto *VT = dyn_cast<FixedVectorType>(Val: V->getType()); |
| 337 | if (!VT) { |
| 338 | Values.push_back(Elt: V); |
| 339 | return; |
| 340 | } |
| 341 | |
| 342 | for (int I = 0, E = VT->getNumElements(); I != E; ++I) |
| 343 | Values.push_back(Elt: Builder.CreateExtractElement(Vec: V, Idx: I)); |
| 344 | } |
| 345 | |
| 346 | static Value *insertValues(IRBuilder<> &Builder, |
| 347 | Type *Ty, |
| 348 | SmallVectorImpl<Value *> &Values) { |
| 349 | if (!Ty->isVectorTy()) { |
| 350 | assert(Values.size() == 1); |
| 351 | return Values[0]; |
| 352 | } |
| 353 | |
| 354 | Value *NewVal = PoisonValue::get(T: Ty); |
| 355 | for (int I = 0, E = Values.size(); I != E; ++I) |
| 356 | NewVal = Builder.CreateInsertElement(Vec: NewVal, NewElt: Values[I], Idx: I); |
| 357 | |
| 358 | return NewVal; |
| 359 | } |
| 360 | |
| 361 | bool AMDGPUCodeGenPrepareImpl::replaceMulWithMul24(BinaryOperator &I) const { |
| 362 | if (I.getOpcode() != Instruction::Mul) |
| 363 | return false; |
| 364 | |
| 365 | Type *Ty = I.getType(); |
| 366 | unsigned Size = Ty->getScalarSizeInBits(); |
| 367 | if (Size <= 16 && ST.has16BitInsts()) |
| 368 | return false; |
| 369 | |
| 370 | // Prefer scalar if this could be s_mul_i32 |
| 371 | if (UA.isUniform(I: &I)) |
| 372 | return false; |
| 373 | |
| 374 | Value *LHS = I.getOperand(i_nocapture: 0); |
| 375 | Value *RHS = I.getOperand(i_nocapture: 1); |
| 376 | IRBuilder<> Builder(&I); |
| 377 | Builder.SetCurrentDebugLocation(I.getDebugLoc()); |
| 378 | |
| 379 | unsigned LHSBits = 0, RHSBits = 0; |
| 380 | bool IsSigned = false; |
| 381 | |
| 382 | if (ST.hasMulU24() && (LHSBits = numBitsUnsigned(Op: LHS, CtxI: &I)) <= 24 && |
| 383 | (RHSBits = numBitsUnsigned(Op: RHS, CtxI: &I)) <= 24) { |
| 384 | IsSigned = false; |
| 385 | |
| 386 | } else if (ST.hasMulI24() && (LHSBits = numBitsSigned(Op: LHS, CtxI: &I)) <= 24 && |
| 387 | (RHSBits = numBitsSigned(Op: RHS, CtxI: &I)) <= 24) { |
| 388 | IsSigned = true; |
| 389 | |
| 390 | } else |
| 391 | return false; |
| 392 | |
| 393 | SmallVector<Value *, 4> LHSVals; |
| 394 | SmallVector<Value *, 4> RHSVals; |
| 395 | SmallVector<Value *, 4> ResultVals; |
| 396 | extractValues(Builder, Values&: LHSVals, V: LHS); |
| 397 | extractValues(Builder, Values&: RHSVals, V: RHS); |
| 398 | |
| 399 | IntegerType *I32Ty = Builder.getInt32Ty(); |
| 400 | IntegerType *IntrinTy = Size > 32 ? Builder.getInt64Ty() : I32Ty; |
| 401 | Type *DstTy = LHSVals[0]->getType(); |
| 402 | |
| 403 | for (int I = 0, E = LHSVals.size(); I != E; ++I) { |
| 404 | Value *LHS = IsSigned ? Builder.CreateSExtOrTrunc(V: LHSVals[I], DestTy: I32Ty) |
| 405 | : Builder.CreateZExtOrTrunc(V: LHSVals[I], DestTy: I32Ty); |
| 406 | Value *RHS = IsSigned ? Builder.CreateSExtOrTrunc(V: RHSVals[I], DestTy: I32Ty) |
| 407 | : Builder.CreateZExtOrTrunc(V: RHSVals[I], DestTy: I32Ty); |
| 408 | Intrinsic::ID ID = |
| 409 | IsSigned ? Intrinsic::amdgcn_mul_i24 : Intrinsic::amdgcn_mul_u24; |
| 410 | Value *Result = Builder.CreateIntrinsic(ID, Types: {IntrinTy}, Args: {LHS, RHS}); |
| 411 | Result = IsSigned ? Builder.CreateSExtOrTrunc(V: Result, DestTy: DstTy) |
| 412 | : Builder.CreateZExtOrTrunc(V: Result, DestTy: DstTy); |
| 413 | ResultVals.push_back(Elt: Result); |
| 414 | } |
| 415 | |
| 416 | Value *NewVal = insertValues(Builder, Ty, Values&: ResultVals); |
| 417 | NewVal->takeName(V: &I); |
| 418 | I.replaceAllUsesWith(V: NewVal); |
| 419 | DeadVals.push_back(Elt: &I); |
| 420 | |
| 421 | return true; |
| 422 | } |
| 423 | |
| 424 | // Find a select instruction, which may have been casted. This is mostly to deal |
| 425 | // with cases where i16 selects were promoted here to i32. |
| 426 | static SelectInst *findSelectThroughCast(Value *V, CastInst *&Cast) { |
| 427 | Cast = nullptr; |
| 428 | if (SelectInst *Sel = dyn_cast<SelectInst>(Val: V)) |
| 429 | return Sel; |
| 430 | |
| 431 | if ((Cast = dyn_cast<CastInst>(Val: V))) { |
| 432 | if (SelectInst *Sel = dyn_cast<SelectInst>(Val: Cast->getOperand(i_nocapture: 0))) |
| 433 | return Sel; |
| 434 | } |
| 435 | |
| 436 | return nullptr; |
| 437 | } |
| 438 | |
| 439 | bool AMDGPUCodeGenPrepareImpl::foldBinOpIntoSelect(BinaryOperator &BO) const { |
| 440 | // Don't do this unless the old select is going away. We want to eliminate the |
| 441 | // binary operator, not replace a binop with a select. |
| 442 | int SelOpNo = 0; |
| 443 | |
| 444 | CastInst *CastOp; |
| 445 | |
| 446 | // TODO: Should probably try to handle some cases with multiple |
| 447 | // users. Duplicating the select may be profitable for division. |
| 448 | SelectInst *Sel = findSelectThroughCast(V: BO.getOperand(i_nocapture: 0), Cast&: CastOp); |
| 449 | if (!Sel || !Sel->hasOneUse()) { |
| 450 | SelOpNo = 1; |
| 451 | Sel = findSelectThroughCast(V: BO.getOperand(i_nocapture: 1), Cast&: CastOp); |
| 452 | } |
| 453 | |
| 454 | if (!Sel || !Sel->hasOneUse()) |
| 455 | return false; |
| 456 | |
| 457 | Constant *CT = dyn_cast<Constant>(Val: Sel->getTrueValue()); |
| 458 | Constant *CF = dyn_cast<Constant>(Val: Sel->getFalseValue()); |
| 459 | Constant *CBO = dyn_cast<Constant>(Val: BO.getOperand(i_nocapture: SelOpNo ^ 1)); |
| 460 | if (!CBO || !CT || !CF) |
| 461 | return false; |
| 462 | |
| 463 | if (CastOp) { |
| 464 | if (!CastOp->hasOneUse()) |
| 465 | return false; |
| 466 | CT = ConstantFoldCastOperand(Opcode: CastOp->getOpcode(), C: CT, DestTy: BO.getType(), DL); |
| 467 | CF = ConstantFoldCastOperand(Opcode: CastOp->getOpcode(), C: CF, DestTy: BO.getType(), DL); |
| 468 | } |
| 469 | |
| 470 | // TODO: Handle special 0/-1 cases DAG combine does, although we only really |
| 471 | // need to handle divisions here. |
| 472 | Constant *FoldedT = |
| 473 | SelOpNo ? ConstantFoldBinaryOpOperands(Opcode: BO.getOpcode(), LHS: CBO, RHS: CT, DL) |
| 474 | : ConstantFoldBinaryOpOperands(Opcode: BO.getOpcode(), LHS: CT, RHS: CBO, DL); |
| 475 | if (!FoldedT || isa<ConstantExpr>(Val: FoldedT)) |
| 476 | return false; |
| 477 | |
| 478 | Constant *FoldedF = |
| 479 | SelOpNo ? ConstantFoldBinaryOpOperands(Opcode: BO.getOpcode(), LHS: CBO, RHS: CF, DL) |
| 480 | : ConstantFoldBinaryOpOperands(Opcode: BO.getOpcode(), LHS: CF, RHS: CBO, DL); |
| 481 | if (!FoldedF || isa<ConstantExpr>(Val: FoldedF)) |
| 482 | return false; |
| 483 | |
| 484 | IRBuilder<> Builder(&BO); |
| 485 | Builder.SetCurrentDebugLocation(BO.getDebugLoc()); |
| 486 | if (const FPMathOperator *FPOp = dyn_cast<const FPMathOperator>(Val: &BO)) |
| 487 | Builder.setFastMathFlags(FPOp->getFastMathFlags()); |
| 488 | |
| 489 | Value *NewSelect = Builder.CreateSelect(C: Sel->getCondition(), |
| 490 | True: FoldedT, False: FoldedF); |
| 491 | NewSelect->takeName(V: &BO); |
| 492 | BO.replaceAllUsesWith(V: NewSelect); |
| 493 | DeadVals.push_back(Elt: &BO); |
| 494 | if (CastOp) |
| 495 | DeadVals.push_back(Elt: CastOp); |
| 496 | DeadVals.push_back(Elt: Sel); |
| 497 | return true; |
| 498 | } |
| 499 | |
| 500 | std::pair<Value *, Value *> |
| 501 | AMDGPUCodeGenPrepareImpl::getFrexpResults(IRBuilder<> &Builder, |
| 502 | Value *Src) const { |
| 503 | Type *Ty = Src->getType(); |
| 504 | Value *Frexp = Builder.CreateIntrinsic(ID: Intrinsic::frexp, |
| 505 | Types: {Ty, Builder.getInt32Ty()}, Args: Src); |
| 506 | Value *FrexpMant = Builder.CreateExtractValue(Agg: Frexp, Idxs: {0}); |
| 507 | |
| 508 | // Bypass the bug workaround for the exponent result since it doesn't matter. |
| 509 | // TODO: Does the bug workaround even really need to consider the exponent |
| 510 | // result? It's unspecified by the spec. |
| 511 | |
| 512 | Value *FrexpExp = |
| 513 | ST.hasFractBug() |
| 514 | ? Builder.CreateIntrinsic(ID: Intrinsic::amdgcn_frexp_exp, |
| 515 | Types: {Builder.getInt32Ty(), Ty}, Args: Src) |
| 516 | : Builder.CreateExtractValue(Agg: Frexp, Idxs: {1}); |
| 517 | return {FrexpMant, FrexpExp}; |
| 518 | } |
| 519 | |
| 520 | /// Emit an expansion of 1.0 / Src good for 1ulp that supports denormals. |
| 521 | Value *AMDGPUCodeGenPrepareImpl::emitRcpIEEE1ULP(IRBuilder<> &Builder, |
| 522 | Value *Src, |
| 523 | bool IsNegative) const { |
| 524 | // Same as for 1.0, but expand the sign out of the constant. |
| 525 | // -1.0 / x -> rcp (fneg x) |
| 526 | if (IsNegative) |
| 527 | Src = Builder.CreateFNeg(V: Src); |
| 528 | |
| 529 | // The rcp instruction doesn't support denormals, so scale the input |
| 530 | // out of the denormal range and convert at the end. |
| 531 | // |
| 532 | // Expand as 2^-n * (1.0 / (x * 2^n)) |
| 533 | |
| 534 | // TODO: Skip scaling if input is known never denormal and the input |
| 535 | // range won't underflow to denormal. The hard part is knowing the |
| 536 | // result. We need a range check, the result could be denormal for |
| 537 | // 0x1p+126 < den <= 0x1p+127. |
| 538 | auto [FrexpMant, FrexpExp] = getFrexpResults(Builder, Src); |
| 539 | Value *ScaleFactor = Builder.CreateNeg(V: FrexpExp); |
| 540 | Value *Rcp = Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rcp, V: FrexpMant); |
| 541 | return Builder.CreateCall(Callee: getLdexpF32(), Args: {Rcp, ScaleFactor}); |
| 542 | } |
| 543 | |
| 544 | /// Emit a 2ulp expansion for fdiv by using frexp for input scaling. |
| 545 | Value *AMDGPUCodeGenPrepareImpl::emitFrexpDiv(IRBuilder<> &Builder, Value *LHS, |
| 546 | Value *RHS, |
| 547 | FastMathFlags FMF) const { |
| 548 | // If we have have to work around the fract/frexp bug, we're worse off than |
| 549 | // using the fdiv.fast expansion. The full safe expansion is faster if we have |
| 550 | // fast FMA. |
| 551 | if (HasFP32DenormalFlush && ST.hasFractBug() && !ST.hasFastFMAF32() && |
| 552 | (!FMF.noNaNs() || !FMF.noInfs())) |
| 553 | return nullptr; |
| 554 | |
| 555 | // We're scaling the LHS to avoid a denormal input, and scale the denominator |
| 556 | // to avoid large values underflowing the result. |
| 557 | auto [FrexpMantRHS, FrexpExpRHS] = getFrexpResults(Builder, Src: RHS); |
| 558 | |
| 559 | Value *Rcp = |
| 560 | Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rcp, V: FrexpMantRHS); |
| 561 | |
| 562 | auto [FrexpMantLHS, FrexpExpLHS] = getFrexpResults(Builder, Src: LHS); |
| 563 | Value *Mul = Builder.CreateFMul(L: FrexpMantLHS, R: Rcp); |
| 564 | |
| 565 | // We multiplied by 2^N/2^M, so we need to multiply by 2^(N-M) to scale the |
| 566 | // result. |
| 567 | Value *ExpDiff = Builder.CreateSub(LHS: FrexpExpLHS, RHS: FrexpExpRHS); |
| 568 | return Builder.CreateCall(Callee: getLdexpF32(), Args: {Mul, ExpDiff}); |
| 569 | } |
| 570 | |
| 571 | /// Emit a sqrt that handles denormals and is accurate to 2ulp. |
| 572 | Value *AMDGPUCodeGenPrepareImpl::emitSqrtIEEE2ULP(IRBuilder<> &Builder, |
| 573 | Value *Src, |
| 574 | FastMathFlags FMF) const { |
| 575 | Type *Ty = Src->getType(); |
| 576 | APFloat SmallestNormal = |
| 577 | APFloat::getSmallestNormalized(Sem: Ty->getFltSemantics()); |
| 578 | Value *NeedScale = |
| 579 | Builder.CreateFCmpOLT(LHS: Src, RHS: ConstantFP::get(Ty, V: SmallestNormal)); |
| 580 | |
| 581 | ConstantInt *Zero = Builder.getInt32(C: 0); |
| 582 | Value *InputScaleFactor = |
| 583 | Builder.CreateSelect(C: NeedScale, True: Builder.getInt32(C: 32), False: Zero); |
| 584 | |
| 585 | Value *Scaled = Builder.CreateCall(Callee: getLdexpF32(), Args: {Src, InputScaleFactor}); |
| 586 | |
| 587 | Value *Sqrt = Builder.CreateCall(Callee: getSqrtF32(), Args: Scaled); |
| 588 | |
| 589 | Value *OutputScaleFactor = |
| 590 | Builder.CreateSelect(C: NeedScale, True: Builder.getInt32(C: -16), False: Zero); |
| 591 | return Builder.CreateCall(Callee: getLdexpF32(), Args: {Sqrt, OutputScaleFactor}); |
| 592 | } |
| 593 | |
| 594 | /// Emit an expansion of 1.0 / sqrt(Src) good for 1ulp that supports denormals. |
| 595 | static Value *emitRsqIEEE1ULP(IRBuilder<> &Builder, Value *Src, |
| 596 | bool IsNegative) { |
| 597 | // bool need_scale = x < 0x1p-126f; |
| 598 | // float input_scale = need_scale ? 0x1.0p+24f : 1.0f; |
| 599 | // float output_scale = need_scale ? 0x1.0p+12f : 1.0f; |
| 600 | // rsq(x * input_scale) * output_scale; |
| 601 | |
| 602 | Type *Ty = Src->getType(); |
| 603 | APFloat SmallestNormal = |
| 604 | APFloat::getSmallestNormalized(Sem: Ty->getFltSemantics()); |
| 605 | Value *NeedScale = |
| 606 | Builder.CreateFCmpOLT(LHS: Src, RHS: ConstantFP::get(Ty, V: SmallestNormal)); |
| 607 | Constant *One = ConstantFP::get(Ty, V: 1.0); |
| 608 | Constant *InputScale = ConstantFP::get(Ty, V: 0x1.0p+24); |
| 609 | Constant *OutputScale = |
| 610 | ConstantFP::get(Ty, V: IsNegative ? -0x1.0p+12 : 0x1.0p+12); |
| 611 | |
| 612 | Value *InputScaleFactor = Builder.CreateSelect(C: NeedScale, True: InputScale, False: One); |
| 613 | |
| 614 | Value *ScaledInput = Builder.CreateFMul(L: Src, R: InputScaleFactor); |
| 615 | Value *Rsq = Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rsq, V: ScaledInput); |
| 616 | Value *OutputScaleFactor = Builder.CreateSelect( |
| 617 | C: NeedScale, True: OutputScale, False: IsNegative ? ConstantFP::get(Ty, V: -1.0) : One); |
| 618 | |
| 619 | return Builder.CreateFMul(L: Rsq, R: OutputScaleFactor); |
| 620 | } |
| 621 | |
| 622 | /// Emit inverse sqrt expansion for f64 with a correction sequence on top of |
| 623 | /// v_rsq_f64. This should give a 1ulp result. |
| 624 | Value *AMDGPUCodeGenPrepareImpl::emitRsqF64(IRBuilder<> &Builder, Value *X, |
| 625 | FastMathFlags SqrtFMF, |
| 626 | FastMathFlags DivFMF, |
| 627 | const Instruction *CtxI, |
| 628 | bool IsNegative) const { |
| 629 | // rsq(x): |
| 630 | // double y0 = BUILTIN_AMDGPU_RSQRT_F64(x); |
| 631 | // double e = MATH_MAD(-y0 * (x == PINF_F64 || x == 0.0 ? y0 : x), y0, 1.0); |
| 632 | // return MATH_MAD(y0*e, MATH_MAD(e, 0.375, 0.5), y0); |
| 633 | // |
| 634 | // -rsq(x): |
| 635 | // double y0 = BUILTIN_AMDGPU_RSQRT_F64(x); |
| 636 | // double e = MATH_MAD(-y0 * (x == PINF_F64 || x == 0.0 ? y0 : x), y0, 1.0); |
| 637 | // return MATH_MAD(-y0*e, MATH_MAD(e, 0.375, 0.5), -y0); |
| 638 | // |
| 639 | // The rsq instruction handles the special cases correctly. We need to check |
| 640 | // for the edge case conditions to ensure the special case propagates through |
| 641 | // the later instructions. |
| 642 | |
| 643 | Value *Y0 = Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rsq, V: X); |
| 644 | |
| 645 | // Try to elide the edge case check. |
| 646 | // |
| 647 | // Fast math flags imply: |
| 648 | // sqrt ninf => !isinf(x) |
| 649 | // fdiv ninf => x != 0, !isinf(x) |
| 650 | bool MaybePosInf = !SqrtFMF.noInfs() && !DivFMF.noInfs(); |
| 651 | bool MaybeZero = !DivFMF.noInfs(); |
| 652 | |
| 653 | DenormalMode DenormMode; |
| 654 | FPClassTest Interested = fcNone; |
| 655 | if (MaybePosInf) |
| 656 | Interested = fcPosInf; |
| 657 | if (MaybeZero) |
| 658 | Interested |= fcZero; |
| 659 | |
| 660 | if (Interested != fcNone) { |
| 661 | KnownFPClass KnownSrc = computeKnownFPClass(V: X, Interested, CtxI); |
| 662 | if (KnownSrc.isKnownNeverPosInfinity()) |
| 663 | MaybePosInf = false; |
| 664 | |
| 665 | DenormMode = F.getDenormalMode(FPType: X->getType()->getFltSemantics()); |
| 666 | if (KnownSrc.isKnownNeverLogicalZero(Mode: DenormMode)) |
| 667 | MaybeZero = false; |
| 668 | } |
| 669 | |
| 670 | Value *SpecialOrRsq = X; |
| 671 | if (MaybeZero || MaybePosInf) { |
| 672 | Value *Cond; |
| 673 | if (MaybePosInf && MaybeZero) { |
| 674 | if (DenormMode.Input != DenormalMode::DenormalModeKind::Dynamic) { |
| 675 | FPClassTest TestMask = fcPosInf | fcZero; |
| 676 | if (DenormMode.inputsAreZero()) |
| 677 | TestMask |= fcSubnormal; |
| 678 | |
| 679 | Cond = Builder.createIsFPClass(FPNum: X, Test: TestMask); |
| 680 | } else { |
| 681 | // Avoid using llvm.is.fpclass for dynamic denormal mode, since it |
| 682 | // doesn't respect the floating-point environment. |
| 683 | Value *IsZero = |
| 684 | Builder.CreateFCmpOEQ(LHS: X, RHS: ConstantFP::getZero(Ty: X->getType())); |
| 685 | Value *IsInf = |
| 686 | Builder.CreateFCmpOEQ(LHS: X, RHS: ConstantFP::getInfinity(Ty: X->getType())); |
| 687 | Cond = Builder.CreateOr(LHS: IsZero, RHS: IsInf); |
| 688 | } |
| 689 | } else if (MaybeZero) { |
| 690 | Cond = Builder.CreateFCmpOEQ(LHS: X, RHS: ConstantFP::getZero(Ty: X->getType())); |
| 691 | } else { |
| 692 | Cond = Builder.CreateFCmpOEQ(LHS: X, RHS: ConstantFP::getInfinity(Ty: X->getType())); |
| 693 | } |
| 694 | |
| 695 | SpecialOrRsq = Builder.CreateSelect(C: Cond, True: Y0, False: X); |
| 696 | } |
| 697 | |
| 698 | Value *NegY0 = Builder.CreateFNeg(V: Y0); |
| 699 | Value *NegXY0 = Builder.CreateFMul(L: SpecialOrRsq, R: NegY0); |
| 700 | |
| 701 | // Could be fmuladd, but isFMAFasterThanFMulAndFAdd is always true for f64. |
| 702 | Value *E = Builder.CreateFMA(Factor1: NegXY0, Factor2: Y0, Summand: ConstantFP::get(Ty: X->getType(), V: 1.0)); |
| 703 | |
| 704 | Value *Y0E = Builder.CreateFMul(L: E, R: IsNegative ? NegY0 : Y0); |
| 705 | |
| 706 | Value *EFMA = Builder.CreateFMA(Factor1: E, Factor2: ConstantFP::get(Ty: X->getType(), V: 0.375), |
| 707 | Summand: ConstantFP::get(Ty: X->getType(), V: 0.5)); |
| 708 | |
| 709 | return Builder.CreateFMA(Factor1: Y0E, Factor2: EFMA, Summand: IsNegative ? NegY0 : Y0); |
| 710 | } |
| 711 | |
| 712 | bool AMDGPUCodeGenPrepareImpl::canOptimizeWithRsq(FastMathFlags DivFMF, |
| 713 | FastMathFlags SqrtFMF) const { |
| 714 | // The rsqrt contraction increases accuracy from ~2ulp to ~1ulp for f32 and |
| 715 | // f64. |
| 716 | return DivFMF.allowContract() && SqrtFMF.allowContract(); |
| 717 | } |
| 718 | |
| 719 | Value *AMDGPUCodeGenPrepareImpl::optimizeWithRsq( |
| 720 | IRBuilder<> &Builder, Value *Num, Value *Den, const FastMathFlags DivFMF, |
| 721 | const FastMathFlags SqrtFMF, const Instruction *CtxI) const { |
| 722 | // The rsqrt contraction increases accuracy from ~2ulp to ~1ulp. |
| 723 | assert(DivFMF.allowContract() && SqrtFMF.allowContract()); |
| 724 | |
| 725 | // rsq_f16 is accurate to 0.51 ulp. |
| 726 | // rsq_f32 is accurate for !fpmath >= 1.0ulp and denormals are flushed. |
| 727 | // rsq_f64 is never accurate. |
| 728 | const ConstantFP *CLHS = dyn_cast<ConstantFP>(Val: Num); |
| 729 | if (!CLHS) |
| 730 | return nullptr; |
| 731 | |
| 732 | bool IsNegative = false; |
| 733 | |
| 734 | // TODO: Handle other numerator values with arcp. |
| 735 | if (CLHS->isExactlyValue(V: 1.0) || (IsNegative = CLHS->isExactlyValue(V: -1.0))) { |
| 736 | // Add in the sqrt flags. |
| 737 | IRBuilder<>::FastMathFlagGuard Guard(Builder); |
| 738 | Builder.setFastMathFlags(DivFMF | SqrtFMF); |
| 739 | |
| 740 | if (Den->getType()->isFloatTy()) { |
| 741 | if ((DivFMF.approxFunc() && SqrtFMF.approxFunc()) || |
| 742 | canIgnoreDenormalInput(V: Den, CtxI)) { |
| 743 | Value *Result = |
| 744 | Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rsq, V: Den); |
| 745 | // -1.0 / sqrt(x) -> fneg(rsq(x)) |
| 746 | return IsNegative ? Builder.CreateFNeg(V: Result) : Result; |
| 747 | } |
| 748 | |
| 749 | return emitRsqIEEE1ULP(Builder, Src: Den, IsNegative); |
| 750 | } |
| 751 | |
| 752 | if (Den->getType()->isDoubleTy()) |
| 753 | return emitRsqF64(Builder, X: Den, SqrtFMF, DivFMF, CtxI, IsNegative); |
| 754 | } |
| 755 | |
| 756 | return nullptr; |
| 757 | } |
| 758 | |
| 759 | // Optimize fdiv with rcp: |
| 760 | // |
| 761 | // 1/x -> rcp(x) when rcp is sufficiently accurate or inaccurate rcp is |
| 762 | // allowed with afn. |
| 763 | // |
| 764 | // a/b -> a*rcp(b) when arcp is allowed, and we only need provide ULP 1.0 |
| 765 | Value * |
| 766 | AMDGPUCodeGenPrepareImpl::optimizeWithRcp(IRBuilder<> &Builder, Value *Num, |
| 767 | Value *Den, FastMathFlags FMF, |
| 768 | const Instruction *CtxI) const { |
| 769 | // rcp_f16 is accurate to 0.51 ulp. |
| 770 | // rcp_f32 is accurate for !fpmath >= 1.0ulp and denormals are flushed. |
| 771 | // rcp_f64 is never accurate. |
| 772 | assert(Den->getType()->isFloatTy()); |
| 773 | |
| 774 | if (const ConstantFP *CLHS = dyn_cast<ConstantFP>(Val: Num)) { |
| 775 | bool IsNegative = false; |
| 776 | if (CLHS->isExactlyValue(V: 1.0) || |
| 777 | (IsNegative = CLHS->isExactlyValue(V: -1.0))) { |
| 778 | Value *Src = Den; |
| 779 | |
| 780 | if (HasFP32DenormalFlush || FMF.approxFunc()) { |
| 781 | // -1.0 / x -> 1.0 / fneg(x) |
| 782 | if (IsNegative) |
| 783 | Src = Builder.CreateFNeg(V: Src); |
| 784 | |
| 785 | // v_rcp_f32 and v_rsq_f32 do not support denormals, and according to |
| 786 | // the CI documentation has a worst case error of 1 ulp. |
| 787 | // OpenCL requires <= 2.5 ulp for 1.0 / x, so it should always be OK |
| 788 | // to use it as long as we aren't trying to use denormals. |
| 789 | // |
| 790 | // v_rcp_f16 and v_rsq_f16 DO support denormals. |
| 791 | |
| 792 | // NOTE: v_sqrt and v_rcp will be combined to v_rsq later. So we don't |
| 793 | // insert rsq intrinsic here. |
| 794 | |
| 795 | // 1.0 / x -> rcp(x) |
| 796 | return Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rcp, V: Src); |
| 797 | } |
| 798 | |
| 799 | // TODO: If the input isn't denormal, and we know the input exponent isn't |
| 800 | // big enough to introduce a denormal we can avoid the scaling. |
| 801 | return emitRcpIEEE1ULP(Builder, Src, IsNegative); |
| 802 | } |
| 803 | } |
| 804 | |
| 805 | if (FMF.allowReciprocal()) { |
| 806 | // x / y -> x * (1.0 / y) |
| 807 | |
| 808 | // TODO: Could avoid denormal scaling and use raw rcp if we knew the output |
| 809 | // will never underflow. |
| 810 | if (HasFP32DenormalFlush || FMF.approxFunc()) { |
| 811 | Value *Recip = Builder.CreateUnaryIntrinsic(ID: Intrinsic::amdgcn_rcp, V: Den); |
| 812 | return Builder.CreateFMul(L: Num, R: Recip); |
| 813 | } |
| 814 | |
| 815 | Value *Recip = emitRcpIEEE1ULP(Builder, Src: Den, IsNegative: false); |
| 816 | return Builder.CreateFMul(L: Num, R: Recip); |
| 817 | } |
| 818 | |
| 819 | return nullptr; |
| 820 | } |
| 821 | |
| 822 | // optimize with fdiv.fast: |
| 823 | // |
| 824 | // a/b -> fdiv.fast(a, b) when !fpmath >= 2.5ulp with denormals flushed. |
| 825 | // |
| 826 | // 1/x -> fdiv.fast(1,x) when !fpmath >= 2.5ulp. |
| 827 | // |
| 828 | // NOTE: optimizeWithRcp should be tried first because rcp is the preference. |
| 829 | Value *AMDGPUCodeGenPrepareImpl::optimizeWithFDivFast( |
| 830 | IRBuilder<> &Builder, Value *Num, Value *Den, float ReqdAccuracy) const { |
| 831 | // fdiv.fast can achieve 2.5 ULP accuracy. |
| 832 | if (ReqdAccuracy < 2.5f) |
| 833 | return nullptr; |
| 834 | |
| 835 | // Only have fdiv.fast for f32. |
| 836 | assert(Den->getType()->isFloatTy()); |
| 837 | |
| 838 | bool NumIsOne = false; |
| 839 | if (const ConstantFP *CNum = dyn_cast<ConstantFP>(Val: Num)) { |
| 840 | if (CNum->isExactlyValue(V: +1.0) || CNum->isExactlyValue(V: -1.0)) |
| 841 | NumIsOne = true; |
| 842 | } |
| 843 | |
| 844 | // fdiv does not support denormals. But 1.0/x is always fine to use it. |
| 845 | // |
| 846 | // TODO: This works for any value with a specific known exponent range, don't |
| 847 | // just limit to constant 1. |
| 848 | if (!HasFP32DenormalFlush && !NumIsOne) |
| 849 | return nullptr; |
| 850 | |
| 851 | return Builder.CreateIntrinsic(ID: Intrinsic::amdgcn_fdiv_fast, Args: {Num, Den}); |
| 852 | } |
| 853 | |
| 854 | Value *AMDGPUCodeGenPrepareImpl::visitFDivElement( |
| 855 | IRBuilder<> &Builder, Value *Num, Value *Den, FastMathFlags DivFMF, |
| 856 | FastMathFlags SqrtFMF, Value *RsqOp, const Instruction *FDivInst, |
| 857 | float ReqdDivAccuracy) const { |
| 858 | if (RsqOp) { |
| 859 | Value *Rsq = |
| 860 | optimizeWithRsq(Builder, Num, Den: RsqOp, DivFMF, SqrtFMF, CtxI: FDivInst); |
| 861 | if (Rsq) |
| 862 | return Rsq; |
| 863 | } |
| 864 | |
| 865 | if (!Num->getType()->isFloatTy()) |
| 866 | return nullptr; |
| 867 | |
| 868 | Value *Rcp = optimizeWithRcp(Builder, Num, Den, FMF: DivFMF, CtxI: FDivInst); |
| 869 | if (Rcp) |
| 870 | return Rcp; |
| 871 | |
| 872 | // In the basic case fdiv_fast has the same instruction count as the frexp div |
| 873 | // expansion. Slightly prefer fdiv_fast since it ends in an fmul that can |
| 874 | // potentially be fused into a user. Also, materialization of the constants |
| 875 | // can be reused for multiple instances. |
| 876 | Value *FDivFast = optimizeWithFDivFast(Builder, Num, Den, ReqdAccuracy: ReqdDivAccuracy); |
| 877 | if (FDivFast) |
| 878 | return FDivFast; |
| 879 | |
| 880 | return emitFrexpDiv(Builder, LHS: Num, RHS: Den, FMF: DivFMF); |
| 881 | } |
| 882 | |
| 883 | // Optimizations is performed based on fpmath, fast math flags as well as |
| 884 | // denormals to optimize fdiv with either rcp or fdiv.fast. |
| 885 | // |
| 886 | // With rcp: |
| 887 | // 1/x -> rcp(x) when rcp is sufficiently accurate or inaccurate rcp is |
| 888 | // allowed with afn. |
| 889 | // |
| 890 | // a/b -> a*rcp(b) when inaccurate rcp is allowed with afn. |
| 891 | // |
| 892 | // With fdiv.fast: |
| 893 | // a/b -> fdiv.fast(a, b) when !fpmath >= 2.5ulp with denormals flushed. |
| 894 | // |
| 895 | // 1/x -> fdiv.fast(1,x) when !fpmath >= 2.5ulp. |
| 896 | // |
| 897 | // NOTE: rcp is the preference in cases that both are legal. |
| 898 | bool AMDGPUCodeGenPrepareImpl::visitFDiv(BinaryOperator &FDiv) { |
| 899 | if (DisableFDivExpand) |
| 900 | return false; |
| 901 | |
| 902 | Type *Ty = FDiv.getType()->getScalarType(); |
| 903 | const bool IsFloat = Ty->isFloatTy(); |
| 904 | if (!IsFloat && !Ty->isDoubleTy()) |
| 905 | return false; |
| 906 | |
| 907 | // The f64 rcp/rsq approximations are pretty inaccurate. We can do an |
| 908 | // expansion around them in codegen. f16 is good enough to always use. |
| 909 | |
| 910 | const FPMathOperator *FPOp = cast<const FPMathOperator>(Val: &FDiv); |
| 911 | const FastMathFlags DivFMF = FPOp->getFastMathFlags(); |
| 912 | const float ReqdAccuracy = FPOp->getFPAccuracy(); |
| 913 | |
| 914 | FastMathFlags SqrtFMF; |
| 915 | |
| 916 | Value *Num = FDiv.getOperand(i_nocapture: 0); |
| 917 | Value *Den = FDiv.getOperand(i_nocapture: 1); |
| 918 | |
| 919 | Value *RsqOp = nullptr; |
| 920 | auto *DenII = dyn_cast<IntrinsicInst>(Val: Den); |
| 921 | if (DenII && DenII->getIntrinsicID() == Intrinsic::sqrt && |
| 922 | DenII->hasOneUse()) { |
| 923 | const auto *SqrtOp = cast<FPMathOperator>(Val: DenII); |
| 924 | SqrtFMF = SqrtOp->getFastMathFlags(); |
| 925 | if (canOptimizeWithRsq(DivFMF, SqrtFMF)) |
| 926 | RsqOp = SqrtOp->getOperand(i: 0); |
| 927 | } |
| 928 | |
| 929 | // rcp path not yet implemented for f64. |
| 930 | if (!IsFloat && !RsqOp) |
| 931 | return false; |
| 932 | |
| 933 | // Inaccurate rcp is allowed with afn. |
| 934 | // |
| 935 | // Defer to codegen to handle this. |
| 936 | // |
| 937 | // TODO: Decide on an interpretation for interactions between afn + arcp + |
| 938 | // !fpmath, and make it consistent between here and codegen. For now, defer |
| 939 | // expansion of afn to codegen. The current interpretation is so aggressive we |
| 940 | // don't need any pre-consideration here when we have better information. A |
| 941 | // more conservative interpretation could use handling here. |
| 942 | const bool AllowInaccurateRcp = DivFMF.approxFunc(); |
| 943 | if (!RsqOp && AllowInaccurateRcp) |
| 944 | return false; |
| 945 | |
| 946 | // Defer the correct implementations to codegen. |
| 947 | if (IsFloat && ReqdAccuracy < 1.0f) |
| 948 | return false; |
| 949 | |
| 950 | IRBuilder<> Builder(FDiv.getParent(), std::next(x: FDiv.getIterator())); |
| 951 | Builder.setFastMathFlags(DivFMF); |
| 952 | Builder.SetCurrentDebugLocation(FDiv.getDebugLoc()); |
| 953 | |
| 954 | SmallVector<Value *, 4> NumVals; |
| 955 | SmallVector<Value *, 4> DenVals; |
| 956 | SmallVector<Value *, 4> RsqDenVals; |
| 957 | extractValues(Builder, Values&: NumVals, V: Num); |
| 958 | extractValues(Builder, Values&: DenVals, V: Den); |
| 959 | |
| 960 | if (RsqOp) |
| 961 | extractValues(Builder, Values&: RsqDenVals, V: RsqOp); |
| 962 | |
| 963 | SmallVector<Value *, 4> ResultVals(NumVals.size()); |
| 964 | for (int I = 0, E = NumVals.size(); I != E; ++I) { |
| 965 | Value *NumElt = NumVals[I]; |
| 966 | Value *DenElt = DenVals[I]; |
| 967 | Value *RsqDenElt = RsqOp ? RsqDenVals[I] : nullptr; |
| 968 | |
| 969 | Value *NewElt = |
| 970 | visitFDivElement(Builder, Num: NumElt, Den: DenElt, DivFMF, SqrtFMF, RsqOp: RsqDenElt, |
| 971 | FDivInst: cast<Instruction>(Val: FPOp), ReqdDivAccuracy: ReqdAccuracy); |
| 972 | if (!NewElt) { |
| 973 | // Keep the original, but scalarized. |
| 974 | |
| 975 | // This has the unfortunate side effect of sometimes scalarizing when |
| 976 | // we're not going to do anything. |
| 977 | NewElt = Builder.CreateFDiv(L: NumElt, R: DenElt); |
| 978 | if (auto *NewEltInst = dyn_cast<Instruction>(Val: NewElt)) |
| 979 | NewEltInst->copyMetadata(SrcInst: FDiv); |
| 980 | } |
| 981 | |
| 982 | ResultVals[I] = NewElt; |
| 983 | } |
| 984 | |
| 985 | Value *NewVal = insertValues(Builder, Ty: FDiv.getType(), Values&: ResultVals); |
| 986 | |
| 987 | if (NewVal) { |
| 988 | FDiv.replaceAllUsesWith(V: NewVal); |
| 989 | NewVal->takeName(V: &FDiv); |
| 990 | DeadVals.push_back(Elt: &FDiv); |
| 991 | } |
| 992 | |
| 993 | return true; |
| 994 | } |
| 995 | |
| 996 | static std::pair<Value*, Value*> getMul64(IRBuilder<> &Builder, |
| 997 | Value *LHS, Value *RHS) { |
| 998 | Type *I32Ty = Builder.getInt32Ty(); |
| 999 | Type *I64Ty = Builder.getInt64Ty(); |
| 1000 | |
| 1001 | Value *LHS_EXT64 = Builder.CreateZExt(V: LHS, DestTy: I64Ty); |
| 1002 | Value *RHS_EXT64 = Builder.CreateZExt(V: RHS, DestTy: I64Ty); |
| 1003 | Value *MUL64 = Builder.CreateMul(LHS: LHS_EXT64, RHS: RHS_EXT64); |
| 1004 | Value *Lo = Builder.CreateTrunc(V: MUL64, DestTy: I32Ty); |
| 1005 | Value *Hi = Builder.CreateLShr(LHS: MUL64, RHS: Builder.getInt64(C: 32)); |
| 1006 | Hi = Builder.CreateTrunc(V: Hi, DestTy: I32Ty); |
| 1007 | return std::pair(Lo, Hi); |
| 1008 | } |
| 1009 | |
| 1010 | static Value* getMulHu(IRBuilder<> &Builder, Value *LHS, Value *RHS) { |
| 1011 | return getMul64(Builder, LHS, RHS).second; |
| 1012 | } |
| 1013 | |
| 1014 | /// Figure out how many bits are really needed for this division. |
| 1015 | /// \p MaxDivBits is an optimization hint to bypass the second |
| 1016 | /// ComputeNumSignBits/computeKnownBits call if the first one is |
| 1017 | /// insufficient. |
| 1018 | unsigned AMDGPUCodeGenPrepareImpl::getDivNumBits(BinaryOperator &I, Value *Num, |
| 1019 | Value *Den, |
| 1020 | unsigned MaxDivBits, |
| 1021 | bool IsSigned) const { |
| 1022 | assert(Num->getType()->getScalarSizeInBits() == |
| 1023 | Den->getType()->getScalarSizeInBits()); |
| 1024 | unsigned SSBits = Num->getType()->getScalarSizeInBits(); |
| 1025 | if (IsSigned) { |
| 1026 | unsigned RHSSignBits = ComputeNumSignBits(Op: Den, DL: SQ.DL, AC: SQ.AC, CxtI: &I, DT: SQ.DT); |
| 1027 | // A sign bit needs to be reserved for shrinking. |
| 1028 | unsigned DivBits = SSBits - RHSSignBits + 1; |
| 1029 | if (DivBits > MaxDivBits) |
| 1030 | return SSBits; |
| 1031 | |
| 1032 | unsigned LHSSignBits = ComputeNumSignBits(Op: Num, DL: SQ.DL, AC: SQ.AC, CxtI: &I); |
| 1033 | |
| 1034 | unsigned SignBits = std::min(a: LHSSignBits, b: RHSSignBits); |
| 1035 | DivBits = SSBits - SignBits + 1; |
| 1036 | return DivBits; |
| 1037 | } |
| 1038 | |
| 1039 | // All bits are used for unsigned division for Num or Den in range |
| 1040 | // (SignedMax, UnsignedMax]. |
| 1041 | KnownBits Known = computeKnownBits(V: Den, Q: SQ.getWithInstruction(I: &I)); |
| 1042 | if (Known.isNegative() || !Known.isNonNegative()) |
| 1043 | return SSBits; |
| 1044 | unsigned RHSSignBits = Known.countMinLeadingZeros(); |
| 1045 | unsigned DivBits = SSBits - RHSSignBits; |
| 1046 | if (DivBits > MaxDivBits) |
| 1047 | return SSBits; |
| 1048 | |
| 1049 | Known = computeKnownBits(V: Num, Q: SQ.getWithInstruction(I: &I)); |
| 1050 | if (Known.isNegative() || !Known.isNonNegative()) |
| 1051 | return SSBits; |
| 1052 | unsigned LHSSignBits = Known.countMinLeadingZeros(); |
| 1053 | |
| 1054 | unsigned SignBits = std::min(a: LHSSignBits, b: RHSSignBits); |
| 1055 | DivBits = SSBits - SignBits; |
| 1056 | return DivBits; |
| 1057 | } |
| 1058 | |
| 1059 | // The fractional part of a float is enough to accurately represent up to |
| 1060 | // a 24-bit signed integer. |
| 1061 | Value *AMDGPUCodeGenPrepareImpl::expandDivRem24(IRBuilder<> &Builder, |
| 1062 | BinaryOperator &I, Value *Num, |
| 1063 | Value *Den, bool IsDiv, |
| 1064 | bool IsSigned) const { |
| 1065 | unsigned DivBits = getDivNumBits(I, Num, Den, MaxDivBits: 24, IsSigned); |
| 1066 | if (DivBits > 24) |
| 1067 | return nullptr; |
| 1068 | return expandDivRem24Impl(Builder, I, Num, Den, NumBits: DivBits, IsDiv, IsSigned); |
| 1069 | } |
| 1070 | |
| 1071 | Value *AMDGPUCodeGenPrepareImpl::expandDivRem24Impl( |
| 1072 | IRBuilder<> &Builder, BinaryOperator &I, Value *Num, Value *Den, |
| 1073 | unsigned DivBits, bool IsDiv, bool IsSigned) const { |
| 1074 | Type *I32Ty = Builder.getInt32Ty(); |
| 1075 | Num = Builder.CreateTrunc(V: Num, DestTy: I32Ty); |
| 1076 | Den = Builder.CreateTrunc(V: Den, DestTy: I32Ty); |
| 1077 | |
| 1078 | Type *F32Ty = Builder.getFloatTy(); |
| 1079 | ConstantInt *One = Builder.getInt32(C: 1); |
| 1080 | Value *JQ = One; |
| 1081 | |
| 1082 | if (IsSigned) { |
| 1083 | // char|short jq = ia ^ ib; |
| 1084 | JQ = Builder.CreateXor(LHS: Num, RHS: Den); |
| 1085 | |
| 1086 | // jq = jq >> (bitsize - 2) |
| 1087 | JQ = Builder.CreateAShr(LHS: JQ, RHS: Builder.getInt32(C: 30)); |
| 1088 | |
| 1089 | // jq = jq | 0x1 |
| 1090 | JQ = Builder.CreateOr(LHS: JQ, RHS: One); |
| 1091 | } |
| 1092 | |
| 1093 | // int ia = (int)LHS; |
| 1094 | Value *IA = Num; |
| 1095 | |
| 1096 | // int ib, (int)RHS; |
| 1097 | Value *IB = Den; |
| 1098 | |
| 1099 | // float fa = (float)ia; |
| 1100 | Value *FA = IsSigned ? Builder.CreateSIToFP(V: IA, DestTy: F32Ty) |
| 1101 | : Builder.CreateUIToFP(V: IA, DestTy: F32Ty); |
| 1102 | |
| 1103 | // float fb = (float)ib; |
| 1104 | Value *FB = IsSigned ? Builder.CreateSIToFP(V: IB,DestTy: F32Ty) |
| 1105 | : Builder.CreateUIToFP(V: IB,DestTy: F32Ty); |
| 1106 | |
| 1107 | Value *RCP = Builder.CreateIntrinsic(ID: Intrinsic::amdgcn_rcp, |
| 1108 | Types: Builder.getFloatTy(), Args: {FB}); |
| 1109 | Value *FQM = Builder.CreateFMul(L: FA, R: RCP); |
| 1110 | |
| 1111 | // fq = trunc(fqm); |
| 1112 | CallInst *FQ = Builder.CreateUnaryIntrinsic(ID: Intrinsic::trunc, V: FQM); |
| 1113 | FQ->copyFastMathFlags(FMF: Builder.getFastMathFlags()); |
| 1114 | |
| 1115 | // float fqneg = -fq; |
| 1116 | Value *FQNeg = Builder.CreateFNeg(V: FQ); |
| 1117 | |
| 1118 | // float fr = mad(fqneg, fb, fa); |
| 1119 | auto FMAD = !ST.hasMadMacF32Insts() |
| 1120 | ? Intrinsic::fma |
| 1121 | : (Intrinsic::ID)Intrinsic::amdgcn_fmad_ftz; |
| 1122 | Value *FR = Builder.CreateIntrinsic(ID: FMAD, |
| 1123 | Types: {FQNeg->getType()}, Args: {FQNeg, FB, FA}, FMFSource: FQ); |
| 1124 | |
| 1125 | // int iq = (int)fq; |
| 1126 | Value *IQ = IsSigned ? Builder.CreateFPToSI(V: FQ, DestTy: I32Ty) |
| 1127 | : Builder.CreateFPToUI(V: FQ, DestTy: I32Ty); |
| 1128 | |
| 1129 | // fr = fabs(fr); |
| 1130 | FR = Builder.CreateUnaryIntrinsic(ID: Intrinsic::fabs, V: FR, FMFSource: FQ); |
| 1131 | |
| 1132 | // fb = fabs(fb); |
| 1133 | FB = Builder.CreateUnaryIntrinsic(ID: Intrinsic::fabs, V: FB, FMFSource: FQ); |
| 1134 | |
| 1135 | // int cv = fr >= fb; |
| 1136 | Value *CV = Builder.CreateFCmpOGE(LHS: FR, RHS: FB); |
| 1137 | |
| 1138 | // jq = (cv ? jq : 0); |
| 1139 | JQ = Builder.CreateSelect(C: CV, True: JQ, False: Builder.getInt32(C: 0)); |
| 1140 | |
| 1141 | // dst = iq + jq; |
| 1142 | Value *Div = Builder.CreateAdd(LHS: IQ, RHS: JQ); |
| 1143 | |
| 1144 | Value *Res = Div; |
| 1145 | if (!IsDiv) { |
| 1146 | // Rem needs compensation, it's easier to recompute it |
| 1147 | Value *Rem = Builder.CreateMul(LHS: Div, RHS: Den); |
| 1148 | Res = Builder.CreateSub(LHS: Num, RHS: Rem); |
| 1149 | } |
| 1150 | |
| 1151 | if (DivBits != 0 && DivBits < 32) { |
| 1152 | // Extend in register from the number of bits this divide really is. |
| 1153 | if (IsSigned) { |
| 1154 | int InRegBits = 32 - DivBits; |
| 1155 | |
| 1156 | Res = Builder.CreateShl(LHS: Res, RHS: InRegBits); |
| 1157 | Res = Builder.CreateAShr(LHS: Res, RHS: InRegBits); |
| 1158 | } else { |
| 1159 | ConstantInt *TruncMask |
| 1160 | = Builder.getInt32(C: (UINT64_C(1) << DivBits) - 1); |
| 1161 | Res = Builder.CreateAnd(LHS: Res, RHS: TruncMask); |
| 1162 | } |
| 1163 | } |
| 1164 | |
| 1165 | return Res; |
| 1166 | } |
| 1167 | |
| 1168 | // Try to recognize special cases the DAG will emit special, better expansions |
| 1169 | // than the general expansion we do here. |
| 1170 | |
| 1171 | // TODO: It would be better to just directly handle those optimizations here. |
| 1172 | bool AMDGPUCodeGenPrepareImpl::divHasSpecialOptimization(BinaryOperator &I, |
| 1173 | Value *Num, |
| 1174 | Value *Den) const { |
| 1175 | if (Constant *C = dyn_cast<Constant>(Val: Den)) { |
| 1176 | // Arbitrary constants get a better expansion as long as a wider mulhi is |
| 1177 | // legal. |
| 1178 | if (C->getType()->getScalarSizeInBits() <= 32) |
| 1179 | return true; |
| 1180 | |
| 1181 | // TODO: Sdiv check for not exact for some reason. |
| 1182 | |
| 1183 | // If there's no wider mulhi, there's only a better expansion for powers of |
| 1184 | // two. |
| 1185 | // TODO: Should really know for each vector element. |
| 1186 | if (isKnownToBeAPowerOfTwo(V: C, OrZero: true, Q: SQ.getWithInstruction(I: &I))) |
| 1187 | return true; |
| 1188 | |
| 1189 | return false; |
| 1190 | } |
| 1191 | |
| 1192 | if (BinaryOperator *BinOpDen = dyn_cast<BinaryOperator>(Val: Den)) { |
| 1193 | // fold (udiv x, (shl c, y)) -> x >>u (log2(c)+y) iff c is power of 2 |
| 1194 | if (BinOpDen->getOpcode() == Instruction::Shl && |
| 1195 | isa<Constant>(Val: BinOpDen->getOperand(i_nocapture: 0)) && |
| 1196 | isKnownToBeAPowerOfTwo(V: BinOpDen->getOperand(i_nocapture: 0), OrZero: true, |
| 1197 | Q: SQ.getWithInstruction(I: &I))) { |
| 1198 | return true; |
| 1199 | } |
| 1200 | } |
| 1201 | |
| 1202 | return false; |
| 1203 | } |
| 1204 | |
| 1205 | static Value *getSign32(Value *V, IRBuilder<> &Builder, const DataLayout DL) { |
| 1206 | // Check whether the sign can be determined statically. |
| 1207 | KnownBits Known = computeKnownBits(V, DL); |
| 1208 | if (Known.isNegative()) |
| 1209 | return Constant::getAllOnesValue(Ty: V->getType()); |
| 1210 | if (Known.isNonNegative()) |
| 1211 | return Constant::getNullValue(Ty: V->getType()); |
| 1212 | return Builder.CreateAShr(LHS: V, RHS: Builder.getInt32(C: 31)); |
| 1213 | } |
| 1214 | |
| 1215 | Value *AMDGPUCodeGenPrepareImpl::expandDivRem32(IRBuilder<> &Builder, |
| 1216 | BinaryOperator &I, Value *X, |
| 1217 | Value *Y) const { |
| 1218 | Instruction::BinaryOps Opc = I.getOpcode(); |
| 1219 | assert(Opc == Instruction::URem || Opc == Instruction::UDiv || |
| 1220 | Opc == Instruction::SRem || Opc == Instruction::SDiv); |
| 1221 | |
| 1222 | FastMathFlags FMF; |
| 1223 | FMF.setFast(); |
| 1224 | Builder.setFastMathFlags(FMF); |
| 1225 | |
| 1226 | if (divHasSpecialOptimization(I, Num: X, Den: Y)) |
| 1227 | return nullptr; // Keep it for later optimization. |
| 1228 | |
| 1229 | bool IsDiv = Opc == Instruction::UDiv || Opc == Instruction::SDiv; |
| 1230 | bool IsSigned = Opc == Instruction::SRem || Opc == Instruction::SDiv; |
| 1231 | |
| 1232 | Type *Ty = X->getType(); |
| 1233 | Type *I32Ty = Builder.getInt32Ty(); |
| 1234 | Type *F32Ty = Builder.getFloatTy(); |
| 1235 | |
| 1236 | if (Ty->getScalarSizeInBits() != 32) { |
| 1237 | if (IsSigned) { |
| 1238 | X = Builder.CreateSExtOrTrunc(V: X, DestTy: I32Ty); |
| 1239 | Y = Builder.CreateSExtOrTrunc(V: Y, DestTy: I32Ty); |
| 1240 | } else { |
| 1241 | X = Builder.CreateZExtOrTrunc(V: X, DestTy: I32Ty); |
| 1242 | Y = Builder.CreateZExtOrTrunc(V: Y, DestTy: I32Ty); |
| 1243 | } |
| 1244 | } |
| 1245 | |
| 1246 | if (Value *Res = expandDivRem24(Builder, I, Num: X, Den: Y, IsDiv, IsSigned)) { |
| 1247 | return IsSigned ? Builder.CreateSExtOrTrunc(V: Res, DestTy: Ty) : |
| 1248 | Builder.CreateZExtOrTrunc(V: Res, DestTy: Ty); |
| 1249 | } |
| 1250 | |
| 1251 | ConstantInt *Zero = Builder.getInt32(C: 0); |
| 1252 | ConstantInt *One = Builder.getInt32(C: 1); |
| 1253 | |
| 1254 | Value *Sign = nullptr; |
| 1255 | if (IsSigned) { |
| 1256 | Value *SignX = getSign32(V: X, Builder, DL); |
| 1257 | Value *SignY = getSign32(V: Y, Builder, DL); |
| 1258 | // Remainder sign is the same as LHS |
| 1259 | Sign = IsDiv ? Builder.CreateXor(LHS: SignX, RHS: SignY) : SignX; |
| 1260 | |
| 1261 | X = Builder.CreateAdd(LHS: X, RHS: SignX); |
| 1262 | Y = Builder.CreateAdd(LHS: Y, RHS: SignY); |
| 1263 | |
| 1264 | X = Builder.CreateXor(LHS: X, RHS: SignX); |
| 1265 | Y = Builder.CreateXor(LHS: Y, RHS: SignY); |
| 1266 | } |
| 1267 | |
| 1268 | // The algorithm here is based on ideas from "Software Integer Division", Tom |
| 1269 | // Rodeheffer, August 2008. |
| 1270 | // |
| 1271 | // unsigned udiv(unsigned x, unsigned y) { |
| 1272 | // // Initial estimate of inv(y). The constant is less than 2^32 to ensure |
| 1273 | // // that this is a lower bound on inv(y), even if some of the calculations |
| 1274 | // // round up. |
| 1275 | // unsigned z = (unsigned)((4294967296.0 - 512.0) * v_rcp_f32((float)y)); |
| 1276 | // |
| 1277 | // // One round of UNR (Unsigned integer Newton-Raphson) to improve z. |
| 1278 | // // Empirically this is guaranteed to give a "two-y" lower bound on |
| 1279 | // // inv(y). |
| 1280 | // z += umulh(z, -y * z); |
| 1281 | // |
| 1282 | // // Quotient/remainder estimate. |
| 1283 | // unsigned q = umulh(x, z); |
| 1284 | // unsigned r = x - q * y; |
| 1285 | // |
| 1286 | // // Two rounds of quotient/remainder refinement. |
| 1287 | // if (r >= y) { |
| 1288 | // ++q; |
| 1289 | // r -= y; |
| 1290 | // } |
| 1291 | // if (r >= y) { |
| 1292 | // ++q; |
| 1293 | // r -= y; |
| 1294 | // } |
| 1295 | // |
| 1296 | // return q; |
| 1297 | // } |
| 1298 | |
| 1299 | // Initial estimate of inv(y). |
| 1300 | Value *FloatY = Builder.CreateUIToFP(V: Y, DestTy: F32Ty); |
| 1301 | Value *RcpY = Builder.CreateIntrinsic(ID: Intrinsic::amdgcn_rcp, Types: F32Ty, Args: {FloatY}); |
| 1302 | Constant *Scale = ConstantFP::get(Ty: F32Ty, V: llvm::bit_cast<float>(from: 0x4F7FFFFE)); |
| 1303 | Value *ScaledY = Builder.CreateFMul(L: RcpY, R: Scale); |
| 1304 | Value *Z = Builder.CreateFPToUI(V: ScaledY, DestTy: I32Ty); |
| 1305 | |
| 1306 | // One round of UNR. |
| 1307 | Value *NegY = Builder.CreateSub(LHS: Zero, RHS: Y); |
| 1308 | Value *NegYZ = Builder.CreateMul(LHS: NegY, RHS: Z); |
| 1309 | Z = Builder.CreateAdd(LHS: Z, RHS: getMulHu(Builder, LHS: Z, RHS: NegYZ)); |
| 1310 | |
| 1311 | // Quotient/remainder estimate. |
| 1312 | Value *Q = getMulHu(Builder, LHS: X, RHS: Z); |
| 1313 | Value *R = Builder.CreateSub(LHS: X, RHS: Builder.CreateMul(LHS: Q, RHS: Y)); |
| 1314 | |
| 1315 | // First quotient/remainder refinement. |
| 1316 | Value *Cond = Builder.CreateICmpUGE(LHS: R, RHS: Y); |
| 1317 | if (IsDiv) |
| 1318 | Q = Builder.CreateSelect(C: Cond, True: Builder.CreateAdd(LHS: Q, RHS: One), False: Q); |
| 1319 | R = Builder.CreateSelect(C: Cond, True: Builder.CreateSub(LHS: R, RHS: Y), False: R); |
| 1320 | |
| 1321 | // Second quotient/remainder refinement. |
| 1322 | Cond = Builder.CreateICmpUGE(LHS: R, RHS: Y); |
| 1323 | Value *Res; |
| 1324 | if (IsDiv) |
| 1325 | Res = Builder.CreateSelect(C: Cond, True: Builder.CreateAdd(LHS: Q, RHS: One), False: Q); |
| 1326 | else |
| 1327 | Res = Builder.CreateSelect(C: Cond, True: Builder.CreateSub(LHS: R, RHS: Y), False: R); |
| 1328 | |
| 1329 | if (IsSigned) { |
| 1330 | Res = Builder.CreateXor(LHS: Res, RHS: Sign); |
| 1331 | Res = Builder.CreateSub(LHS: Res, RHS: Sign); |
| 1332 | Res = Builder.CreateSExtOrTrunc(V: Res, DestTy: Ty); |
| 1333 | } else { |
| 1334 | Res = Builder.CreateZExtOrTrunc(V: Res, DestTy: Ty); |
| 1335 | } |
| 1336 | return Res; |
| 1337 | } |
| 1338 | |
| 1339 | Value *AMDGPUCodeGenPrepareImpl::shrinkDivRem64(IRBuilder<> &Builder, |
| 1340 | BinaryOperator &I, Value *Num, |
| 1341 | Value *Den) const { |
| 1342 | if (!ExpandDiv64InIR && divHasSpecialOptimization(I, Num, Den)) |
| 1343 | return nullptr; // Keep it for later optimization. |
| 1344 | |
| 1345 | Instruction::BinaryOps Opc = I.getOpcode(); |
| 1346 | |
| 1347 | bool IsDiv = Opc == Instruction::SDiv || Opc == Instruction::UDiv; |
| 1348 | bool IsSigned = Opc == Instruction::SDiv || Opc == Instruction::SRem; |
| 1349 | |
| 1350 | unsigned NumDivBits = getDivNumBits(I, Num, Den, MaxDivBits: 32, IsSigned); |
| 1351 | if (NumDivBits > 32) |
| 1352 | return nullptr; |
| 1353 | |
| 1354 | Value *Narrowed = nullptr; |
| 1355 | if (NumDivBits <= 24) { |
| 1356 | Narrowed = expandDivRem24Impl(Builder, I, Num, Den, DivBits: NumDivBits, |
| 1357 | IsDiv, IsSigned); |
| 1358 | } else if (NumDivBits <= 32) { |
| 1359 | Narrowed = expandDivRem32(Builder, I, X: Num, Y: Den); |
| 1360 | } |
| 1361 | |
| 1362 | if (Narrowed) { |
| 1363 | return IsSigned ? Builder.CreateSExt(V: Narrowed, DestTy: Num->getType()) : |
| 1364 | Builder.CreateZExt(V: Narrowed, DestTy: Num->getType()); |
| 1365 | } |
| 1366 | |
| 1367 | return nullptr; |
| 1368 | } |
| 1369 | |
| 1370 | void AMDGPUCodeGenPrepareImpl::expandDivRem64(BinaryOperator &I) const { |
| 1371 | Instruction::BinaryOps Opc = I.getOpcode(); |
| 1372 | // Do the general expansion. |
| 1373 | if (Opc == Instruction::UDiv || Opc == Instruction::SDiv) { |
| 1374 | expandDivisionUpTo64Bits(Div: &I); |
| 1375 | return; |
| 1376 | } |
| 1377 | |
| 1378 | if (Opc == Instruction::URem || Opc == Instruction::SRem) { |
| 1379 | expandRemainderUpTo64Bits(Rem: &I); |
| 1380 | return; |
| 1381 | } |
| 1382 | |
| 1383 | llvm_unreachable("not a division" ); |
| 1384 | } |
| 1385 | |
| 1386 | /* |
| 1387 | This will cause non-byte load in consistency, for example: |
| 1388 | ``` |
| 1389 | %load = load i1, ptr addrspace(4) %arg, align 4 |
| 1390 | %zext = zext i1 %load to |
| 1391 | i64 %add = add i64 %zext |
| 1392 | ``` |
| 1393 | Instead of creating `s_and_b32 s0, s0, 1`, |
| 1394 | it will create `s_and_b32 s0, s0, 0xff`. |
| 1395 | We accept this change since the non-byte load assumes the upper bits |
| 1396 | within the byte are all 0. |
| 1397 | */ |
| 1398 | bool AMDGPUCodeGenPrepareImpl::tryNarrowMathIfNoOverflow(Instruction *I) { |
| 1399 | unsigned Opc = I->getOpcode(); |
| 1400 | Type *OldType = I->getType(); |
| 1401 | |
| 1402 | if (Opc != Instruction::Add && Opc != Instruction::Mul) |
| 1403 | return false; |
| 1404 | |
| 1405 | unsigned OrigBit = OldType->getScalarSizeInBits(); |
| 1406 | |
| 1407 | if (Opc != Instruction::Add && Opc != Instruction::Mul) |
| 1408 | llvm_unreachable("Unexpected opcode, only valid for Instruction::Add and " |
| 1409 | "Instruction::Mul." ); |
| 1410 | |
| 1411 | unsigned MaxBitsNeeded = computeKnownBits(V: I, DL).countMaxActiveBits(); |
| 1412 | |
| 1413 | MaxBitsNeeded = std::max<unsigned>(a: bit_ceil(Value: MaxBitsNeeded), b: 8); |
| 1414 | Type *NewType = DL.getSmallestLegalIntType(C&: I->getContext(), Width: MaxBitsNeeded); |
| 1415 | if (!NewType) |
| 1416 | return false; |
| 1417 | unsigned NewBit = NewType->getIntegerBitWidth(); |
| 1418 | if (NewBit >= OrigBit) |
| 1419 | return false; |
| 1420 | NewType = I->getType()->getWithNewBitWidth(NewBitWidth: NewBit); |
| 1421 | |
| 1422 | // Old cost |
| 1423 | const TargetTransformInfo &TTI = TM.getTargetTransformInfo(F); |
| 1424 | InstructionCost OldCost = |
| 1425 | TTI.getArithmeticInstrCost(Opcode: Opc, Ty: OldType, CostKind: TTI::TCK_RecipThroughput); |
| 1426 | // New cost of new op |
| 1427 | InstructionCost NewCost = |
| 1428 | TTI.getArithmeticInstrCost(Opcode: Opc, Ty: NewType, CostKind: TTI::TCK_RecipThroughput); |
| 1429 | // New cost of narrowing 2 operands (use trunc) |
| 1430 | int NumOfNonConstOps = 2; |
| 1431 | if (isa<Constant>(Val: I->getOperand(i: 0)) || isa<Constant>(Val: I->getOperand(i: 1))) { |
| 1432 | // Cannot be both constant, should be propagated |
| 1433 | NumOfNonConstOps = 1; |
| 1434 | } |
| 1435 | NewCost += NumOfNonConstOps * TTI.getCastInstrCost(Opcode: Instruction::Trunc, |
| 1436 | Dst: NewType, Src: OldType, |
| 1437 | CCH: TTI.getCastContextHint(I), |
| 1438 | CostKind: TTI::TCK_RecipThroughput); |
| 1439 | // New cost of zext narrowed result to original type |
| 1440 | NewCost += |
| 1441 | TTI.getCastInstrCost(Opcode: Instruction::ZExt, Dst: OldType, Src: NewType, |
| 1442 | CCH: TTI.getCastContextHint(I), CostKind: TTI::TCK_RecipThroughput); |
| 1443 | if (NewCost >= OldCost) |
| 1444 | return false; |
| 1445 | |
| 1446 | IRBuilder<> Builder(I); |
| 1447 | Value *Trunc0 = Builder.CreateTrunc(V: I->getOperand(i: 0), DestTy: NewType); |
| 1448 | Value *Trunc1 = Builder.CreateTrunc(V: I->getOperand(i: 1), DestTy: NewType); |
| 1449 | Value *Arith = |
| 1450 | Builder.CreateBinOp(Opc: (Instruction::BinaryOps)Opc, LHS: Trunc0, RHS: Trunc1); |
| 1451 | |
| 1452 | Value *Zext = Builder.CreateZExt(V: Arith, DestTy: OldType); |
| 1453 | I->replaceAllUsesWith(V: Zext); |
| 1454 | DeadVals.push_back(Elt: I); |
| 1455 | return true; |
| 1456 | } |
| 1457 | |
| 1458 | bool AMDGPUCodeGenPrepareImpl::visitBinaryOperator(BinaryOperator &I) { |
| 1459 | if (foldBinOpIntoSelect(BO&: I)) |
| 1460 | return true; |
| 1461 | |
| 1462 | if (UseMul24Intrin && replaceMulWithMul24(I)) |
| 1463 | return true; |
| 1464 | if (tryNarrowMathIfNoOverflow(I: &I)) |
| 1465 | return true; |
| 1466 | |
| 1467 | bool Changed = false; |
| 1468 | Instruction::BinaryOps Opc = I.getOpcode(); |
| 1469 | Type *Ty = I.getType(); |
| 1470 | Value *NewDiv = nullptr; |
| 1471 | unsigned ScalarSize = Ty->getScalarSizeInBits(); |
| 1472 | |
| 1473 | SmallVector<BinaryOperator *, 8> Div64ToExpand; |
| 1474 | |
| 1475 | if ((Opc == Instruction::URem || Opc == Instruction::UDiv || |
| 1476 | Opc == Instruction::SRem || Opc == Instruction::SDiv) && |
| 1477 | ScalarSize <= 64 && |
| 1478 | !DisableIDivExpand) { |
| 1479 | Value *Num = I.getOperand(i_nocapture: 0); |
| 1480 | Value *Den = I.getOperand(i_nocapture: 1); |
| 1481 | IRBuilder<> Builder(&I); |
| 1482 | Builder.SetCurrentDebugLocation(I.getDebugLoc()); |
| 1483 | |
| 1484 | if (auto *VT = dyn_cast<FixedVectorType>(Val: Ty)) { |
| 1485 | NewDiv = PoisonValue::get(T: VT); |
| 1486 | |
| 1487 | for (unsigned N = 0, E = VT->getNumElements(); N != E; ++N) { |
| 1488 | Value *NumEltN = Builder.CreateExtractElement(Vec: Num, Idx: N); |
| 1489 | Value *DenEltN = Builder.CreateExtractElement(Vec: Den, Idx: N); |
| 1490 | |
| 1491 | Value *NewElt; |
| 1492 | if (ScalarSize <= 32) { |
| 1493 | NewElt = expandDivRem32(Builder, I, X: NumEltN, Y: DenEltN); |
| 1494 | if (!NewElt) |
| 1495 | NewElt = Builder.CreateBinOp(Opc, LHS: NumEltN, RHS: DenEltN); |
| 1496 | } else { |
| 1497 | // See if this 64-bit division can be shrunk to 32/24-bits before |
| 1498 | // producing the general expansion. |
| 1499 | NewElt = shrinkDivRem64(Builder, I, Num: NumEltN, Den: DenEltN); |
| 1500 | if (!NewElt) { |
| 1501 | // The general 64-bit expansion introduces control flow and doesn't |
| 1502 | // return the new value. Just insert a scalar copy and defer |
| 1503 | // expanding it. |
| 1504 | NewElt = Builder.CreateBinOp(Opc, LHS: NumEltN, RHS: DenEltN); |
| 1505 | // CreateBinOp does constant folding. If the operands are constant, |
| 1506 | // it will return a Constant instead of a BinaryOperator. |
| 1507 | if (auto *NewEltBO = dyn_cast<BinaryOperator>(Val: NewElt)) |
| 1508 | Div64ToExpand.push_back(Elt: NewEltBO); |
| 1509 | } |
| 1510 | } |
| 1511 | |
| 1512 | if (auto *NewEltI = dyn_cast<Instruction>(Val: NewElt)) |
| 1513 | NewEltI->copyIRFlags(V: &I); |
| 1514 | |
| 1515 | NewDiv = Builder.CreateInsertElement(Vec: NewDiv, NewElt, Idx: N); |
| 1516 | } |
| 1517 | } else { |
| 1518 | if (ScalarSize <= 32) |
| 1519 | NewDiv = expandDivRem32(Builder, I, X: Num, Y: Den); |
| 1520 | else { |
| 1521 | NewDiv = shrinkDivRem64(Builder, I, Num, Den); |
| 1522 | if (!NewDiv) |
| 1523 | Div64ToExpand.push_back(Elt: &I); |
| 1524 | } |
| 1525 | } |
| 1526 | |
| 1527 | if (NewDiv) { |
| 1528 | I.replaceAllUsesWith(V: NewDiv); |
| 1529 | DeadVals.push_back(Elt: &I); |
| 1530 | Changed = true; |
| 1531 | } |
| 1532 | } |
| 1533 | |
| 1534 | if (ExpandDiv64InIR) { |
| 1535 | // TODO: We get much worse code in specially handled constant cases. |
| 1536 | for (BinaryOperator *Div : Div64ToExpand) { |
| 1537 | expandDivRem64(I&: *Div); |
| 1538 | FlowChanged = true; |
| 1539 | Changed = true; |
| 1540 | } |
| 1541 | } |
| 1542 | |
| 1543 | return Changed; |
| 1544 | } |
| 1545 | |
| 1546 | bool AMDGPUCodeGenPrepareImpl::visitLoadInst(LoadInst &I) { |
| 1547 | if (!WidenLoads) |
| 1548 | return false; |
| 1549 | |
| 1550 | if ((I.getPointerAddressSpace() == AMDGPUAS::CONSTANT_ADDRESS || |
| 1551 | I.getPointerAddressSpace() == AMDGPUAS::CONSTANT_ADDRESS_32BIT) && |
| 1552 | canWidenScalarExtLoad(I)) { |
| 1553 | IRBuilder<> Builder(&I); |
| 1554 | Builder.SetCurrentDebugLocation(I.getDebugLoc()); |
| 1555 | |
| 1556 | Type *I32Ty = Builder.getInt32Ty(); |
| 1557 | LoadInst *WidenLoad = Builder.CreateLoad(Ty: I32Ty, Ptr: I.getPointerOperand()); |
| 1558 | WidenLoad->copyMetadata(SrcInst: I); |
| 1559 | |
| 1560 | // If we have range metadata, we need to convert the type, and not make |
| 1561 | // assumptions about the high bits. |
| 1562 | if (auto *Range = WidenLoad->getMetadata(KindID: LLVMContext::MD_range)) { |
| 1563 | ConstantInt *Lower = |
| 1564 | mdconst::extract<ConstantInt>(MD: Range->getOperand(I: 0)); |
| 1565 | |
| 1566 | if (Lower->isNullValue()) { |
| 1567 | WidenLoad->setMetadata(KindID: LLVMContext::MD_range, Node: nullptr); |
| 1568 | } else { |
| 1569 | Metadata *LowAndHigh[] = { |
| 1570 | ConstantAsMetadata::get(C: ConstantInt::get(Ty: I32Ty, V: Lower->getValue().zext(width: 32))), |
| 1571 | // Don't make assumptions about the high bits. |
| 1572 | ConstantAsMetadata::get(C: ConstantInt::get(Ty: I32Ty, V: 0)) |
| 1573 | }; |
| 1574 | |
| 1575 | WidenLoad->setMetadata(KindID: LLVMContext::MD_range, |
| 1576 | Node: MDNode::get(Context&: F.getContext(), MDs: LowAndHigh)); |
| 1577 | } |
| 1578 | } |
| 1579 | |
| 1580 | int TySize = DL.getTypeSizeInBits(Ty: I.getType()); |
| 1581 | Type *IntNTy = Builder.getIntNTy(N: TySize); |
| 1582 | Value *ValTrunc = Builder.CreateTrunc(V: WidenLoad, DestTy: IntNTy); |
| 1583 | Value *ValOrig = Builder.CreateBitCast(V: ValTrunc, DestTy: I.getType()); |
| 1584 | I.replaceAllUsesWith(V: ValOrig); |
| 1585 | DeadVals.push_back(Elt: &I); |
| 1586 | return true; |
| 1587 | } |
| 1588 | |
| 1589 | return false; |
| 1590 | } |
| 1591 | |
| 1592 | bool AMDGPUCodeGenPrepareImpl::visitSelectInst(SelectInst &I) { |
| 1593 | Value *Cond = I.getCondition(); |
| 1594 | Value *TrueVal = I.getTrueValue(); |
| 1595 | Value *FalseVal = I.getFalseValue(); |
| 1596 | Value *CmpVal; |
| 1597 | CmpPredicate Pred; |
| 1598 | |
| 1599 | // Match fract pattern with nan check. |
| 1600 | if (!match(V: Cond, P: m_FCmp(Pred, L: m_Value(V&: CmpVal), R: m_NonNaN()))) |
| 1601 | return false; |
| 1602 | |
| 1603 | FPMathOperator *FPOp = dyn_cast<FPMathOperator>(Val: &I); |
| 1604 | if (!FPOp) |
| 1605 | return false; |
| 1606 | |
| 1607 | IRBuilder<> Builder(&I); |
| 1608 | Builder.setFastMathFlags(FPOp->getFastMathFlags()); |
| 1609 | |
| 1610 | auto *IITrue = dyn_cast<IntrinsicInst>(Val: TrueVal); |
| 1611 | auto *IIFalse = dyn_cast<IntrinsicInst>(Val: FalseVal); |
| 1612 | |
| 1613 | Value *Fract = nullptr; |
| 1614 | if (Pred == FCmpInst::FCMP_UNO && TrueVal == CmpVal && IIFalse && |
| 1615 | CmpVal == matchFractPat(I&: *IIFalse)) { |
| 1616 | // isnan(x) ? x : fract(x) |
| 1617 | Fract = applyFractPat(Builder, FractArg: CmpVal); |
| 1618 | } else if (Pred == FCmpInst::FCMP_ORD && FalseVal == CmpVal && IITrue && |
| 1619 | CmpVal == matchFractPat(I&: *IITrue)) { |
| 1620 | // !isnan(x) ? fract(x) : x |
| 1621 | Fract = applyFractPat(Builder, FractArg: CmpVal); |
| 1622 | } else |
| 1623 | return false; |
| 1624 | |
| 1625 | Fract->takeName(V: &I); |
| 1626 | I.replaceAllUsesWith(V: Fract); |
| 1627 | DeadVals.push_back(Elt: &I); |
| 1628 | return true; |
| 1629 | } |
| 1630 | |
| 1631 | static bool areInSameBB(const Value *A, const Value *B) { |
| 1632 | const auto *IA = dyn_cast<Instruction>(Val: A); |
| 1633 | const auto *IB = dyn_cast<Instruction>(Val: B); |
| 1634 | return IA && IB && IA->getParent() == IB->getParent(); |
| 1635 | } |
| 1636 | |
| 1637 | // Helper for breaking large PHIs that returns true when an extractelement on V |
| 1638 | // is likely to be folded away by the DAG combiner. |
| 1639 | static bool isInterestingPHIIncomingValue(const Value *V) { |
| 1640 | const auto *FVT = dyn_cast<FixedVectorType>(Val: V->getType()); |
| 1641 | if (!FVT) |
| 1642 | return false; |
| 1643 | |
| 1644 | const Value *CurVal = V; |
| 1645 | |
| 1646 | // Check for insertelements, keeping track of the elements covered. |
| 1647 | BitVector EltsCovered(FVT->getNumElements()); |
| 1648 | while (const auto *IE = dyn_cast<InsertElementInst>(Val: CurVal)) { |
| 1649 | const auto *Idx = dyn_cast<ConstantInt>(Val: IE->getOperand(i_nocapture: 2)); |
| 1650 | |
| 1651 | // Non constant index/out of bounds index -> folding is unlikely. |
| 1652 | // The latter is more of a sanity check because canonical IR should just |
| 1653 | // have replaced those with poison. |
| 1654 | if (!Idx || Idx->getZExtValue() >= FVT->getNumElements()) |
| 1655 | return false; |
| 1656 | |
| 1657 | const auto *VecSrc = IE->getOperand(i_nocapture: 0); |
| 1658 | |
| 1659 | // If the vector source is another instruction, it must be in the same basic |
| 1660 | // block. Otherwise, the DAGCombiner won't see the whole thing and is |
| 1661 | // unlikely to be able to do anything interesting here. |
| 1662 | if (isa<Instruction>(Val: VecSrc) && !areInSameBB(A: VecSrc, B: IE)) |
| 1663 | return false; |
| 1664 | |
| 1665 | CurVal = VecSrc; |
| 1666 | EltsCovered.set(Idx->getZExtValue()); |
| 1667 | |
| 1668 | // All elements covered. |
| 1669 | if (EltsCovered.all()) |
| 1670 | return true; |
| 1671 | } |
| 1672 | |
| 1673 | // We either didn't find a single insertelement, or the insertelement chain |
| 1674 | // ended before all elements were covered. Check for other interesting values. |
| 1675 | |
| 1676 | // Constants are always interesting because we can just constant fold the |
| 1677 | // extractelements. |
| 1678 | if (isa<Constant>(Val: CurVal)) |
| 1679 | return true; |
| 1680 | |
| 1681 | // shufflevector is likely to be profitable if either operand is a constant, |
| 1682 | // or if either source is in the same block. |
| 1683 | // This is because shufflevector is most often lowered as a series of |
| 1684 | // insert/extract elements anyway. |
| 1685 | if (const auto *SV = dyn_cast<ShuffleVectorInst>(Val: CurVal)) { |
| 1686 | return isa<Constant>(Val: SV->getOperand(i_nocapture: 1)) || |
| 1687 | areInSameBB(A: SV, B: SV->getOperand(i_nocapture: 0)) || |
| 1688 | areInSameBB(A: SV, B: SV->getOperand(i_nocapture: 1)); |
| 1689 | } |
| 1690 | |
| 1691 | return false; |
| 1692 | } |
| 1693 | |
| 1694 | static void collectPHINodes(const PHINode &I, |
| 1695 | SmallPtrSet<const PHINode *, 8> &SeenPHIs) { |
| 1696 | const auto [It, Inserted] = SeenPHIs.insert(Ptr: &I); |
| 1697 | if (!Inserted) |
| 1698 | return; |
| 1699 | |
| 1700 | for (const Value *Inc : I.incoming_values()) { |
| 1701 | if (const auto *PhiInc = dyn_cast<PHINode>(Val: Inc)) |
| 1702 | collectPHINodes(I: *PhiInc, SeenPHIs); |
| 1703 | } |
| 1704 | |
| 1705 | for (const User *U : I.users()) { |
| 1706 | if (const auto *PhiU = dyn_cast<PHINode>(Val: U)) |
| 1707 | collectPHINodes(I: *PhiU, SeenPHIs); |
| 1708 | } |
| 1709 | } |
| 1710 | |
| 1711 | bool AMDGPUCodeGenPrepareImpl::canBreakPHINode(const PHINode &I) { |
| 1712 | // Check in the cache first. |
| 1713 | if (const auto It = BreakPhiNodesCache.find(Val: &I); |
| 1714 | It != BreakPhiNodesCache.end()) |
| 1715 | return It->second; |
| 1716 | |
| 1717 | // We consider PHI nodes as part of "chains", so given a PHI node I, we |
| 1718 | // recursively consider all its users and incoming values that are also PHI |
| 1719 | // nodes. We then make a decision about all of those PHIs at once. Either they |
| 1720 | // all get broken up, or none of them do. That way, we avoid cases where a |
| 1721 | // single PHI is/is not broken and we end up reforming/exploding a vector |
| 1722 | // multiple times, or even worse, doing it in a loop. |
| 1723 | SmallPtrSet<const PHINode *, 8> WorkList; |
| 1724 | collectPHINodes(I, SeenPHIs&: WorkList); |
| 1725 | |
| 1726 | #ifndef NDEBUG |
| 1727 | // Check that none of the PHI nodes in the worklist are in the map. If some of |
| 1728 | // them are, it means we're not good enough at collecting related PHIs. |
| 1729 | for (const PHINode *WLP : WorkList) { |
| 1730 | assert(BreakPhiNodesCache.count(WLP) == 0); |
| 1731 | } |
| 1732 | #endif |
| 1733 | |
| 1734 | // To consider a PHI profitable to break, we need to see some interesting |
| 1735 | // incoming values. At least 2/3rd (rounded up) of all PHIs in the worklist |
| 1736 | // must have one to consider all PHIs breakable. |
| 1737 | // |
| 1738 | // This threshold has been determined through performance testing. |
| 1739 | // |
| 1740 | // Note that the computation below is equivalent to |
| 1741 | // |
| 1742 | // (unsigned)ceil((K / 3.0) * 2) |
| 1743 | // |
| 1744 | // It's simply written this way to avoid mixing integral/FP arithmetic. |
| 1745 | const auto Threshold = (alignTo(Value: WorkList.size() * 2, Align: 3) / 3); |
| 1746 | unsigned NumBreakablePHIs = 0; |
| 1747 | bool CanBreak = false; |
| 1748 | for (const PHINode *Cur : WorkList) { |
| 1749 | // Don't break PHIs that have no interesting incoming values. That is, where |
| 1750 | // there is no clear opportunity to fold the "extractelement" instructions |
| 1751 | // we would add. |
| 1752 | // |
| 1753 | // Note: IC does not run after this pass, so we're only interested in the |
| 1754 | // foldings that the DAG combiner can do. |
| 1755 | if (any_of(Range: Cur->incoming_values(), P: isInterestingPHIIncomingValue)) { |
| 1756 | if (++NumBreakablePHIs >= Threshold) { |
| 1757 | CanBreak = true; |
| 1758 | break; |
| 1759 | } |
| 1760 | } |
| 1761 | } |
| 1762 | |
| 1763 | for (const PHINode *Cur : WorkList) |
| 1764 | BreakPhiNodesCache[Cur] = CanBreak; |
| 1765 | |
| 1766 | return CanBreak; |
| 1767 | } |
| 1768 | |
| 1769 | /// Helper class for "break large PHIs" (visitPHINode). |
| 1770 | /// |
| 1771 | /// This represents a slice of a PHI's incoming value, which is made up of: |
| 1772 | /// - The type of the slice (Ty) |
| 1773 | /// - The index in the incoming value's vector where the slice starts (Idx) |
| 1774 | /// - The number of elements in the slice (NumElts). |
| 1775 | /// It also keeps track of the NewPHI node inserted for this particular slice. |
| 1776 | /// |
| 1777 | /// Slice examples: |
| 1778 | /// <4 x i64> -> Split into four i64 slices. |
| 1779 | /// -> [i64, 0, 1], [i64, 1, 1], [i64, 2, 1], [i64, 3, 1] |
| 1780 | /// <5 x i16> -> Split into 2 <2 x i16> slices + a i16 tail. |
| 1781 | /// -> [<2 x i16>, 0, 2], [<2 x i16>, 2, 2], [i16, 4, 1] |
| 1782 | class VectorSlice { |
| 1783 | public: |
| 1784 | VectorSlice(Type *Ty, unsigned Idx, unsigned NumElts) |
| 1785 | : Ty(Ty), Idx(Idx), NumElts(NumElts) {} |
| 1786 | |
| 1787 | Type *Ty = nullptr; |
| 1788 | unsigned Idx = 0; |
| 1789 | unsigned NumElts = 0; |
| 1790 | PHINode *NewPHI = nullptr; |
| 1791 | |
| 1792 | /// Slice \p Inc according to the information contained within this slice. |
| 1793 | /// This is cached, so if called multiple times for the same \p BB & \p Inc |
| 1794 | /// pair, it returns the same Sliced value as well. |
| 1795 | /// |
| 1796 | /// Note this *intentionally* does not return the same value for, say, |
| 1797 | /// [%bb.0, %0] & [%bb.1, %0] as: |
| 1798 | /// - It could cause issues with dominance (e.g. if bb.1 is seen first, then |
| 1799 | /// the value in bb.1 may not be reachable from bb.0 if it's its |
| 1800 | /// predecessor.) |
| 1801 | /// - We also want to make our extract instructions as local as possible so |
| 1802 | /// the DAG has better chances of folding them out. Duplicating them like |
| 1803 | /// that is beneficial in that regard. |
| 1804 | /// |
| 1805 | /// This is both a minor optimization to avoid creating duplicate |
| 1806 | /// instructions, but also a requirement for correctness. It is not forbidden |
| 1807 | /// for a PHI node to have the same [BB, Val] pair multiple times. If we |
| 1808 | /// returned a new value each time, those previously identical pairs would all |
| 1809 | /// have different incoming values (from the same block) and it'd cause a "PHI |
| 1810 | /// node has multiple entries for the same basic block with different incoming |
| 1811 | /// values!" verifier error. |
| 1812 | Value *getSlicedVal(BasicBlock *BB, Value *Inc, StringRef NewValName) { |
| 1813 | Value *&Res = SlicedVals[{BB, Inc}]; |
| 1814 | if (Res) |
| 1815 | return Res; |
| 1816 | |
| 1817 | IRBuilder<> B(BB->getTerminator()); |
| 1818 | if (Instruction *IncInst = dyn_cast<Instruction>(Val: Inc)) |
| 1819 | B.SetCurrentDebugLocation(IncInst->getDebugLoc()); |
| 1820 | |
| 1821 | if (NumElts > 1) { |
| 1822 | SmallVector<int, 4> Mask; |
| 1823 | for (unsigned K = Idx; K < (Idx + NumElts); ++K) |
| 1824 | Mask.push_back(Elt: K); |
| 1825 | Res = B.CreateShuffleVector(V: Inc, Mask, Name: NewValName); |
| 1826 | } else |
| 1827 | Res = B.CreateExtractElement(Vec: Inc, Idx, Name: NewValName); |
| 1828 | |
| 1829 | return Res; |
| 1830 | } |
| 1831 | |
| 1832 | private: |
| 1833 | SmallDenseMap<std::pair<BasicBlock *, Value *>, Value *> SlicedVals; |
| 1834 | }; |
| 1835 | |
| 1836 | bool AMDGPUCodeGenPrepareImpl::visitPHINode(PHINode &I) { |
| 1837 | // Break-up fixed-vector PHIs into smaller pieces. |
| 1838 | // Default threshold is 32, so it breaks up any vector that's >32 bits into |
| 1839 | // its elements, or into 32-bit pieces (for 8/16 bit elts). |
| 1840 | // |
| 1841 | // This is only helpful for DAGISel because it doesn't handle large PHIs as |
| 1842 | // well as GlobalISel. DAGISel lowers PHIs by using CopyToReg/CopyFromReg. |
| 1843 | // With large, odd-sized PHIs we may end up needing many `build_vector` |
| 1844 | // operations with most elements being "undef". This inhibits a lot of |
| 1845 | // optimization opportunities and can result in unreasonably high register |
| 1846 | // pressure and the inevitable stack spilling. |
| 1847 | if (!BreakLargePHIs || getCGPassBuilderOption().EnableGlobalISelOption) |
| 1848 | return false; |
| 1849 | |
| 1850 | FixedVectorType *FVT = dyn_cast<FixedVectorType>(Val: I.getType()); |
| 1851 | if (!FVT || FVT->getNumElements() == 1 || |
| 1852 | DL.getTypeSizeInBits(Ty: FVT) <= BreakLargePHIsThreshold) |
| 1853 | return false; |
| 1854 | |
| 1855 | if (!ForceBreakLargePHIs && !canBreakPHINode(I)) |
| 1856 | return false; |
| 1857 | |
| 1858 | std::vector<VectorSlice> Slices; |
| 1859 | |
| 1860 | Type *EltTy = FVT->getElementType(); |
| 1861 | { |
| 1862 | unsigned Idx = 0; |
| 1863 | // For 8/16 bits type, don't scalarize fully but break it up into as many |
| 1864 | // 32-bit slices as we can, and scalarize the tail. |
| 1865 | const unsigned EltSize = DL.getTypeSizeInBits(Ty: EltTy); |
| 1866 | const unsigned NumElts = FVT->getNumElements(); |
| 1867 | if (EltSize == 8 || EltSize == 16) { |
| 1868 | const unsigned SubVecSize = (32 / EltSize); |
| 1869 | Type *SubVecTy = FixedVectorType::get(ElementType: EltTy, NumElts: SubVecSize); |
| 1870 | for (unsigned End = alignDown(Value: NumElts, Align: SubVecSize); Idx < End; |
| 1871 | Idx += SubVecSize) |
| 1872 | Slices.emplace_back(args&: SubVecTy, args&: Idx, args: SubVecSize); |
| 1873 | } |
| 1874 | |
| 1875 | // Scalarize all remaining elements. |
| 1876 | for (; Idx < NumElts; ++Idx) |
| 1877 | Slices.emplace_back(args&: EltTy, args&: Idx, args: 1); |
| 1878 | } |
| 1879 | |
| 1880 | assert(Slices.size() > 1); |
| 1881 | |
| 1882 | // Create one PHI per vector piece. The "VectorSlice" class takes care of |
| 1883 | // creating the necessary instruction to extract the relevant slices of each |
| 1884 | // incoming value. |
| 1885 | IRBuilder<> B(I.getParent()); |
| 1886 | B.SetCurrentDebugLocation(I.getDebugLoc()); |
| 1887 | |
| 1888 | unsigned IncNameSuffix = 0; |
| 1889 | for (VectorSlice &S : Slices) { |
| 1890 | // We need to reset the build on each iteration, because getSlicedVal may |
| 1891 | // have inserted something into I's BB. |
| 1892 | B.SetInsertPoint(I.getParent()->getFirstNonPHIIt()); |
| 1893 | S.NewPHI = B.CreatePHI(Ty: S.Ty, NumReservedValues: I.getNumIncomingValues()); |
| 1894 | |
| 1895 | for (const auto &[Idx, BB] : enumerate(First: I.blocks())) { |
| 1896 | S.NewPHI->addIncoming(V: S.getSlicedVal(BB, Inc: I.getIncomingValue(i: Idx), |
| 1897 | NewValName: "largephi.extractslice" + |
| 1898 | std::to_string(val: IncNameSuffix++)), |
| 1899 | BB); |
| 1900 | } |
| 1901 | } |
| 1902 | |
| 1903 | // And replace this PHI with a vector of all the previous PHI values. |
| 1904 | Value *Vec = PoisonValue::get(T: FVT); |
| 1905 | unsigned NameSuffix = 0; |
| 1906 | for (VectorSlice &S : Slices) { |
| 1907 | const auto ValName = "largephi.insertslice" + std::to_string(val: NameSuffix++); |
| 1908 | if (S.NumElts > 1) |
| 1909 | Vec = B.CreateInsertVector(DstType: FVT, SrcVec: Vec, SubVec: S.NewPHI, Idx: S.Idx, Name: ValName); |
| 1910 | else |
| 1911 | Vec = B.CreateInsertElement(Vec, NewElt: S.NewPHI, Idx: S.Idx, Name: ValName); |
| 1912 | } |
| 1913 | |
| 1914 | I.replaceAllUsesWith(V: Vec); |
| 1915 | DeadVals.push_back(Elt: &I); |
| 1916 | return true; |
| 1917 | } |
| 1918 | |
| 1919 | /// \param V Value to check |
| 1920 | /// \param DL DataLayout |
| 1921 | /// \param TM TargetMachine (TODO: remove once DL contains nullptr values) |
| 1922 | /// \param AS Target Address Space |
| 1923 | /// \return true if \p V cannot be the null value of \p AS, false otherwise. |
| 1924 | static bool isPtrKnownNeverNull(const Value *V, const DataLayout &DL, |
| 1925 | const AMDGPUTargetMachine &TM, unsigned AS) { |
| 1926 | // Pointer cannot be null if it's a block address, GV or alloca. |
| 1927 | // NOTE: We don't support extern_weak, but if we did, we'd need to check for |
| 1928 | // it as the symbol could be null in such cases. |
| 1929 | if (isa<BlockAddress, GlobalValue, AllocaInst>(Val: V)) |
| 1930 | return true; |
| 1931 | |
| 1932 | // Check nonnull arguments. |
| 1933 | if (const auto *Arg = dyn_cast<Argument>(Val: V); Arg && Arg->hasNonNullAttr()) |
| 1934 | return true; |
| 1935 | |
| 1936 | // Check nonnull loads. |
| 1937 | if (const auto *Load = dyn_cast<LoadInst>(Val: V); |
| 1938 | Load && Load->hasMetadata(KindID: LLVMContext::MD_nonnull)) |
| 1939 | return true; |
| 1940 | |
| 1941 | // getUnderlyingObject may have looked through another addrspacecast, although |
| 1942 | // the optimizable situations most likely folded out by now. |
| 1943 | if (AS != cast<PointerType>(Val: V->getType())->getAddressSpace()) |
| 1944 | return false; |
| 1945 | |
| 1946 | // TODO: Calls that return nonnull? |
| 1947 | |
| 1948 | // For all other things, use KnownBits. |
| 1949 | // We either use 0 or all bits set to indicate null, so check whether the |
| 1950 | // value can be zero or all ones. |
| 1951 | // |
| 1952 | // TODO: Use ValueTracking's isKnownNeverNull if it becomes aware that some |
| 1953 | // address spaces have non-zero null values. |
| 1954 | auto SrcPtrKB = computeKnownBits(V, DL); |
| 1955 | const auto NullVal = TM.getNullPointerValue(AddrSpace: AS); |
| 1956 | |
| 1957 | assert(SrcPtrKB.getBitWidth() == DL.getPointerSizeInBits(AS)); |
| 1958 | assert((NullVal == 0 || NullVal == -1) && |
| 1959 | "don't know how to check for this null value!" ); |
| 1960 | return NullVal ? !SrcPtrKB.getMaxValue().isAllOnes() : SrcPtrKB.isNonZero(); |
| 1961 | } |
| 1962 | |
| 1963 | bool AMDGPUCodeGenPrepareImpl::visitAddrSpaceCastInst(AddrSpaceCastInst &I) { |
| 1964 | // Intrinsic doesn't support vectors, also it seems that it's often difficult |
| 1965 | // to prove that a vector cannot have any nulls in it so it's unclear if it's |
| 1966 | // worth supporting. |
| 1967 | if (I.getType()->isVectorTy()) |
| 1968 | return false; |
| 1969 | |
| 1970 | // Check if this can be lowered to a amdgcn.addrspacecast.nonnull. |
| 1971 | // This is only worthwhile for casts from/to priv/local to flat. |
| 1972 | const unsigned SrcAS = I.getSrcAddressSpace(); |
| 1973 | const unsigned DstAS = I.getDestAddressSpace(); |
| 1974 | |
| 1975 | bool CanLower = false; |
| 1976 | if (SrcAS == AMDGPUAS::FLAT_ADDRESS) |
| 1977 | CanLower = (DstAS == AMDGPUAS::LOCAL_ADDRESS || |
| 1978 | DstAS == AMDGPUAS::PRIVATE_ADDRESS); |
| 1979 | else if (DstAS == AMDGPUAS::FLAT_ADDRESS) |
| 1980 | CanLower = (SrcAS == AMDGPUAS::LOCAL_ADDRESS || |
| 1981 | SrcAS == AMDGPUAS::PRIVATE_ADDRESS); |
| 1982 | if (!CanLower) |
| 1983 | return false; |
| 1984 | |
| 1985 | SmallVector<const Value *, 4> WorkList; |
| 1986 | getUnderlyingObjects(V: I.getOperand(i_nocapture: 0), Objects&: WorkList); |
| 1987 | if (!all_of(Range&: WorkList, P: [&](const Value *V) { |
| 1988 | return isPtrKnownNeverNull(V, DL, TM, AS: SrcAS); |
| 1989 | })) |
| 1990 | return false; |
| 1991 | |
| 1992 | IRBuilder<> B(&I); |
| 1993 | auto *Intrin = B.CreateIntrinsic( |
| 1994 | RetTy: I.getType(), ID: Intrinsic::amdgcn_addrspacecast_nonnull, Args: {I.getOperand(i_nocapture: 0)}); |
| 1995 | I.replaceAllUsesWith(V: Intrin); |
| 1996 | DeadVals.push_back(Elt: &I); |
| 1997 | return true; |
| 1998 | } |
| 1999 | |
| 2000 | bool AMDGPUCodeGenPrepareImpl::visitIntrinsicInst(IntrinsicInst &I) { |
| 2001 | switch (I.getIntrinsicID()) { |
| 2002 | case Intrinsic::minnum: |
| 2003 | case Intrinsic::minimumnum: |
| 2004 | case Intrinsic::minimum: |
| 2005 | return visitFMinLike(I); |
| 2006 | case Intrinsic::sqrt: |
| 2007 | return visitSqrt(I); |
| 2008 | case Intrinsic::amdgcn_mbcnt_lo: |
| 2009 | return visitMbcntLo(I); |
| 2010 | case Intrinsic::amdgcn_mbcnt_hi: |
| 2011 | return visitMbcntHi(I); |
| 2012 | default: |
| 2013 | return false; |
| 2014 | } |
| 2015 | } |
| 2016 | |
| 2017 | /// Match non-nan fract pattern. |
| 2018 | /// minnum(fsub(x, floor(x)), nextafter(1.0, -1.0)) |
| 2019 | /// minimumnum(fsub(x, floor(x)), nextafter(1.0, -1.0)) |
| 2020 | /// minimum(fsub(x, floor(x)), nextafter(1.0, -1.0)) |
| 2021 | /// |
| 2022 | /// If fract is a useful instruction for the subtarget. Does not account for the |
| 2023 | /// nan handling; the instruction has a nan check on the input value. |
| 2024 | Value *AMDGPUCodeGenPrepareImpl::matchFractPat(IntrinsicInst &I) { |
| 2025 | if (ST.hasFractBug()) |
| 2026 | return nullptr; |
| 2027 | |
| 2028 | Intrinsic::ID IID = I.getIntrinsicID(); |
| 2029 | |
| 2030 | // The value is only used in contexts where we know the input isn't a nan, so |
| 2031 | // any of the fmin variants are fine. |
| 2032 | if (IID != Intrinsic::minnum && IID != Intrinsic::minimum && |
| 2033 | IID != Intrinsic::minimumnum) |
| 2034 | return nullptr; |
| 2035 | |
| 2036 | Type *Ty = I.getType(); |
| 2037 | if (!isLegalFloatingTy(Ty: Ty->getScalarType())) |
| 2038 | return nullptr; |
| 2039 | |
| 2040 | Value *Arg0 = I.getArgOperand(i: 0); |
| 2041 | Value *Arg1 = I.getArgOperand(i: 1); |
| 2042 | |
| 2043 | const APFloat *C; |
| 2044 | if (!match(V: Arg1, P: m_APFloat(Res&: C))) |
| 2045 | return nullptr; |
| 2046 | |
| 2047 | APFloat One(1.0); |
| 2048 | bool LosesInfo; |
| 2049 | One.convert(ToSemantics: C->getSemantics(), RM: APFloat::rmNearestTiesToEven, losesInfo: &LosesInfo); |
| 2050 | |
| 2051 | // Match nextafter(1.0, -1) |
| 2052 | One.next(nextDown: true); |
| 2053 | if (One != *C) |
| 2054 | return nullptr; |
| 2055 | |
| 2056 | Value *FloorSrc; |
| 2057 | if (match(V: Arg0, P: m_FSub(L: m_Value(V&: FloorSrc), |
| 2058 | R: m_Intrinsic<Intrinsic::floor>(Op0: m_Deferred(V: FloorSrc))))) |
| 2059 | return FloorSrc; |
| 2060 | return nullptr; |
| 2061 | } |
| 2062 | |
| 2063 | Value *AMDGPUCodeGenPrepareImpl::applyFractPat(IRBuilder<> &Builder, |
| 2064 | Value *FractArg) { |
| 2065 | SmallVector<Value *, 4> FractVals; |
| 2066 | extractValues(Builder, Values&: FractVals, V: FractArg); |
| 2067 | |
| 2068 | SmallVector<Value *, 4> ResultVals(FractVals.size()); |
| 2069 | |
| 2070 | Type *Ty = FractArg->getType()->getScalarType(); |
| 2071 | for (unsigned I = 0, E = FractVals.size(); I != E; ++I) { |
| 2072 | ResultVals[I] = |
| 2073 | Builder.CreateIntrinsic(ID: Intrinsic::amdgcn_fract, Types: {Ty}, Args: {FractVals[I]}); |
| 2074 | } |
| 2075 | |
| 2076 | return insertValues(Builder, Ty: FractArg->getType(), Values&: ResultVals); |
| 2077 | } |
| 2078 | |
| 2079 | bool AMDGPUCodeGenPrepareImpl::visitFMinLike(IntrinsicInst &I) { |
| 2080 | Value *FractArg = matchFractPat(I); |
| 2081 | if (!FractArg) |
| 2082 | return false; |
| 2083 | |
| 2084 | // Match pattern for fract intrinsic in contexts where the nan check has been |
| 2085 | // optimized out (and hope the knowledge the source can't be nan wasn't lost). |
| 2086 | if (!I.hasNoNaNs() && !isKnownNeverNaN(V: FractArg, SQ: SQ.getWithInstruction(I: &I))) |
| 2087 | return false; |
| 2088 | |
| 2089 | IRBuilder<> Builder(&I); |
| 2090 | FastMathFlags FMF = I.getFastMathFlags(); |
| 2091 | FMF.setNoNaNs(); |
| 2092 | Builder.setFastMathFlags(FMF); |
| 2093 | |
| 2094 | Value *Fract = applyFractPat(Builder, FractArg); |
| 2095 | Fract->takeName(V: &I); |
| 2096 | I.replaceAllUsesWith(V: Fract); |
| 2097 | DeadVals.push_back(Elt: &I); |
| 2098 | return true; |
| 2099 | } |
| 2100 | |
| 2101 | // Expand llvm.sqrt.f32 calls with !fpmath metadata in a semi-fast way. |
| 2102 | bool AMDGPUCodeGenPrepareImpl::visitSqrt(IntrinsicInst &Sqrt) { |
| 2103 | Type *Ty = Sqrt.getType()->getScalarType(); |
| 2104 | if (!Ty->isFloatTy() && (!Ty->isHalfTy() || ST.has16BitInsts())) |
| 2105 | return false; |
| 2106 | |
| 2107 | const FPMathOperator *FPOp = cast<const FPMathOperator>(Val: &Sqrt); |
| 2108 | FastMathFlags SqrtFMF = FPOp->getFastMathFlags(); |
| 2109 | |
| 2110 | // We're trying to handle the fast-but-not-that-fast case only. The lowering |
| 2111 | // of fast llvm.sqrt will give the raw instruction anyway. |
| 2112 | if (SqrtFMF.approxFunc()) |
| 2113 | return false; |
| 2114 | |
| 2115 | const float ReqdAccuracy = FPOp->getFPAccuracy(); |
| 2116 | |
| 2117 | // Defer correctly rounded expansion to codegen. |
| 2118 | if (ReqdAccuracy < 1.0f) |
| 2119 | return false; |
| 2120 | |
| 2121 | Value *SrcVal = Sqrt.getOperand(i_nocapture: 0); |
| 2122 | bool CanTreatAsDAZ = canIgnoreDenormalInput(V: SrcVal, CtxI: &Sqrt); |
| 2123 | |
| 2124 | // The raw instruction is 1 ulp, but the correction for denormal handling |
| 2125 | // brings it to 2. |
| 2126 | if (!CanTreatAsDAZ && ReqdAccuracy < 2.0f) |
| 2127 | return false; |
| 2128 | |
| 2129 | IRBuilder<> Builder(&Sqrt); |
| 2130 | SmallVector<Value *, 4> SrcVals; |
| 2131 | extractValues(Builder, Values&: SrcVals, V: SrcVal); |
| 2132 | |
| 2133 | SmallVector<Value *, 4> ResultVals(SrcVals.size()); |
| 2134 | for (int I = 0, E = SrcVals.size(); I != E; ++I) { |
| 2135 | if (CanTreatAsDAZ) |
| 2136 | ResultVals[I] = Builder.CreateCall(Callee: getSqrtF32(), Args: SrcVals[I]); |
| 2137 | else |
| 2138 | ResultVals[I] = emitSqrtIEEE2ULP(Builder, Src: SrcVals[I], FMF: SqrtFMF); |
| 2139 | } |
| 2140 | |
| 2141 | Value *NewSqrt = insertValues(Builder, Ty: Sqrt.getType(), Values&: ResultVals); |
| 2142 | NewSqrt->takeName(V: &Sqrt); |
| 2143 | Sqrt.replaceAllUsesWith(V: NewSqrt); |
| 2144 | DeadVals.push_back(Elt: &Sqrt); |
| 2145 | return true; |
| 2146 | } |
| 2147 | |
| 2148 | bool AMDGPUCodeGenPrepare::runOnFunction(Function &F) { |
| 2149 | if (skipFunction(F)) |
| 2150 | return false; |
| 2151 | |
| 2152 | auto *TPC = getAnalysisIfAvailable<TargetPassConfig>(); |
| 2153 | if (!TPC) |
| 2154 | return false; |
| 2155 | |
| 2156 | const AMDGPUTargetMachine &TM = TPC->getTM<AMDGPUTargetMachine>(); |
| 2157 | const TargetLibraryInfo *TLI = |
| 2158 | &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F); |
| 2159 | AssumptionCache *AC = |
| 2160 | &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F); |
| 2161 | auto *DTWP = getAnalysisIfAvailable<DominatorTreeWrapperPass>(); |
| 2162 | const DominatorTree *DT = DTWP ? &DTWP->getDomTree() : nullptr; |
| 2163 | const UniformityInfo &UA = |
| 2164 | getAnalysis<UniformityInfoWrapperPass>().getUniformityInfo(); |
| 2165 | return AMDGPUCodeGenPrepareImpl(F, TM, TLI, AC, DT, UA).run(); |
| 2166 | } |
| 2167 | |
| 2168 | PreservedAnalyses AMDGPUCodeGenPreparePass::run(Function &F, |
| 2169 | FunctionAnalysisManager &FAM) { |
| 2170 | const AMDGPUTargetMachine &ATM = static_cast<const AMDGPUTargetMachine &>(TM); |
| 2171 | const TargetLibraryInfo *TLI = &FAM.getResult<TargetLibraryAnalysis>(IR&: F); |
| 2172 | AssumptionCache *AC = &FAM.getResult<AssumptionAnalysis>(IR&: F); |
| 2173 | const DominatorTree *DT = FAM.getCachedResult<DominatorTreeAnalysis>(IR&: F); |
| 2174 | const UniformityInfo &UA = FAM.getResult<UniformityInfoAnalysis>(IR&: F); |
| 2175 | AMDGPUCodeGenPrepareImpl Impl(F, ATM, TLI, AC, DT, UA); |
| 2176 | if (!Impl.run()) |
| 2177 | return PreservedAnalyses::all(); |
| 2178 | PreservedAnalyses PA = PreservedAnalyses::none(); |
| 2179 | if (!Impl.FlowChanged) |
| 2180 | PA.preserveSet<CFGAnalyses>(); |
| 2181 | return PA; |
| 2182 | } |
| 2183 | |
| 2184 | INITIALIZE_PASS_BEGIN(AMDGPUCodeGenPrepare, DEBUG_TYPE, |
| 2185 | "AMDGPU IR optimizations" , false, false) |
| 2186 | INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker) |
| 2187 | INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass) |
| 2188 | INITIALIZE_PASS_DEPENDENCY(UniformityInfoWrapperPass) |
| 2189 | INITIALIZE_PASS_END(AMDGPUCodeGenPrepare, DEBUG_TYPE, "AMDGPU IR optimizations" , |
| 2190 | false, false) |
| 2191 | |
| 2192 | /// Create a workitem.id.x intrinsic call with range metadata. |
| 2193 | CallInst *AMDGPUCodeGenPrepareImpl::createWorkitemIdX(IRBuilder<> &B) const { |
| 2194 | CallInst *Tid = B.CreateIntrinsic(ID: Intrinsic::amdgcn_workitem_id_x, Args: {}); |
| 2195 | ST.makeLIDRangeMetadata(I: Tid); |
| 2196 | return Tid; |
| 2197 | } |
| 2198 | |
| 2199 | /// Replace the instruction with a direct workitem.id.x call. |
| 2200 | void AMDGPUCodeGenPrepareImpl::replaceWithWorkitemIdX(Instruction &I) const { |
| 2201 | IRBuilder<> B(&I); |
| 2202 | CallInst *Tid = createWorkitemIdX(B); |
| 2203 | BasicBlock::iterator BI(&I); |
| 2204 | ReplaceInstWithValue(BI, V: Tid); |
| 2205 | } |
| 2206 | |
| 2207 | /// Replace the instruction with (workitem.id.x & mask). |
| 2208 | void AMDGPUCodeGenPrepareImpl::replaceWithMaskedWorkitemIdX( |
| 2209 | Instruction &I, unsigned WaveSize) const { |
| 2210 | IRBuilder<> B(&I); |
| 2211 | CallInst *Tid = createWorkitemIdX(B); |
| 2212 | Constant *Mask = ConstantInt::get(Ty: Tid->getType(), V: WaveSize - 1); |
| 2213 | Value *AndInst = B.CreateAnd(LHS: Tid, RHS: Mask); |
| 2214 | BasicBlock::iterator BI(&I); |
| 2215 | ReplaceInstWithValue(BI, V: AndInst); |
| 2216 | } |
| 2217 | |
| 2218 | /// Try to optimize mbcnt instruction by replacing with workitem.id.x when |
| 2219 | /// work group size allows direct computation of lane ID. |
| 2220 | /// Returns true if optimization was applied, false otherwise. |
| 2221 | bool AMDGPUCodeGenPrepareImpl::tryReplaceWithWorkitemId(Instruction &I, |
| 2222 | unsigned Wave) const { |
| 2223 | std::optional<unsigned> MaybeX = ST.getReqdWorkGroupSize(F, Dim: 0); |
| 2224 | if (!MaybeX) |
| 2225 | return false; |
| 2226 | |
| 2227 | // When work group size == wave_size, each work group contains exactly one |
| 2228 | // wave, so the instruction can be replaced with workitem.id.x directly. |
| 2229 | if (*MaybeX == Wave) { |
| 2230 | replaceWithWorkitemIdX(I); |
| 2231 | return true; |
| 2232 | } |
| 2233 | |
| 2234 | // When work group evenly splits into waves, compute lane ID within wave |
| 2235 | // using bit masking: lane_id = workitem.id.x & (wave_size - 1). |
| 2236 | if (ST.hasWavefrontsEvenlySplittingXDim(F, /*RequiresUniformYZ=*/REquiresUniformYZ: true)) { |
| 2237 | replaceWithMaskedWorkitemIdX(I, WaveSize: Wave); |
| 2238 | return true; |
| 2239 | } |
| 2240 | |
| 2241 | return false; |
| 2242 | } |
| 2243 | |
| 2244 | /// Optimize mbcnt.lo calls on wave32 architectures for lane ID computation. |
| 2245 | bool AMDGPUCodeGenPrepareImpl::visitMbcntLo(IntrinsicInst &I) const { |
| 2246 | // This optimization only applies to wave32 targets where mbcnt.lo operates on |
| 2247 | // the full execution mask. |
| 2248 | if (!ST.isWave32()) |
| 2249 | return false; |
| 2250 | |
| 2251 | // Only optimize the pattern mbcnt.lo(~0, 0) which counts active lanes with |
| 2252 | // lower IDs. |
| 2253 | if (!match(V: &I, |
| 2254 | P: m_Intrinsic<Intrinsic::amdgcn_mbcnt_lo>(Op0: m_AllOnes(), Op1: m_Zero()))) |
| 2255 | return false; |
| 2256 | |
| 2257 | return tryReplaceWithWorkitemId(I, Wave: ST.getWavefrontSize()); |
| 2258 | } |
| 2259 | |
| 2260 | /// Optimize mbcnt.hi calls for lane ID computation. |
| 2261 | bool AMDGPUCodeGenPrepareImpl::visitMbcntHi(IntrinsicInst &I) const { |
| 2262 | // Abort if wave size is not known at compile time. |
| 2263 | if (!ST.isWaveSizeKnown()) |
| 2264 | return false; |
| 2265 | |
| 2266 | unsigned Wave = ST.getWavefrontSize(); |
| 2267 | |
| 2268 | // On wave32, the upper 32 bits of execution mask are always 0, so |
| 2269 | // mbcnt.hi(mask, val) always returns val unchanged. |
| 2270 | if (ST.isWave32()) { |
| 2271 | if (auto MaybeX = ST.getReqdWorkGroupSize(F, Dim: 0)) { |
| 2272 | // Replace mbcnt.hi(mask, val) with val only when work group size matches |
| 2273 | // wave size (single wave per work group). |
| 2274 | if (*MaybeX == Wave) { |
| 2275 | BasicBlock::iterator BI(&I); |
| 2276 | ReplaceInstWithValue(BI, V: I.getArgOperand(i: 1)); |
| 2277 | return true; |
| 2278 | } |
| 2279 | } |
| 2280 | } |
| 2281 | |
| 2282 | // Optimize the complete lane ID computation pattern: |
| 2283 | // mbcnt.hi(~0, mbcnt.lo(~0, 0)) which counts all active lanes with lower IDs |
| 2284 | // across the full execution mask. |
| 2285 | using namespace PatternMatch; |
| 2286 | |
| 2287 | // Check for pattern: mbcnt.hi(~0, mbcnt.lo(~0, 0)) |
| 2288 | if (!match(V: &I, P: m_Intrinsic<Intrinsic::amdgcn_mbcnt_hi>( |
| 2289 | Op0: m_AllOnes(), Op1: m_Intrinsic<Intrinsic::amdgcn_mbcnt_lo>( |
| 2290 | Op0: m_AllOnes(), Op1: m_Zero())))) |
| 2291 | return false; |
| 2292 | |
| 2293 | return tryReplaceWithWorkitemId(I, Wave); |
| 2294 | } |
| 2295 | |
| 2296 | char AMDGPUCodeGenPrepare::ID = 0; |
| 2297 | |
| 2298 | FunctionPass *llvm::createAMDGPUCodeGenPreparePass() { |
| 2299 | return new AMDGPUCodeGenPrepare(); |
| 2300 | } |
| 2301 | |