1 //===- MVELaneInterleaving.cpp - Inverleave for MVE instructions ----------===// 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 // This pass interleaves around sext/zext/trunc instructions. MVE does not have 10 // a single sext/zext or trunc instruction that takes the bottom half of a 11 // vector and extends to a full width, like NEON has with MOVL. Instead it is 12 // expected that this happens through top/bottom instructions. So the MVE 13 // equivalent VMOVLT/B instructions take either the even or odd elements of the 14 // input and extend them to the larger type, producing a vector with half the 15 // number of elements each of double the bitwidth. As there is no simple 16 // instruction, we often have to turn sext/zext/trunc into a series of lane 17 // moves (or stack loads/stores, which we do not do yet). 18 // 19 // This pass takes vector code that starts at truncs, looks for interconnected 20 // blobs of operations that end with sext/zext (or constants/splats) of the 21 // form: 22 // %sa = sext v8i16 %a to v8i32 23 // %sb = sext v8i16 %b to v8i32 24 // %add = add v8i32 %sa, %sb 25 // %r = trunc %add to v8i16 26 // And adds shuffles to allow the use of VMOVL/VMOVN instrctions: 27 // %sha = shuffle v8i16 %a, undef, <0, 2, 4, 6, 1, 3, 5, 7> 28 // %sa = sext v8i16 %sha to v8i32 29 // %shb = shuffle v8i16 %b, undef, <0, 2, 4, 6, 1, 3, 5, 7> 30 // %sb = sext v8i16 %shb to v8i32 31 // %add = add v8i32 %sa, %sb 32 // %r = trunc %add to v8i16 33 // %shr = shuffle v8i16 %r, undef, <0, 4, 1, 5, 2, 6, 3, 7> 34 // Which can then be split and lowered to MVE instructions efficiently: 35 // %sa_b = VMOVLB.s16 %a 36 // %sa_t = VMOVLT.s16 %a 37 // %sb_b = VMOVLB.s16 %b 38 // %sb_t = VMOVLT.s16 %b 39 // %add_b = VADD.i32 %sa_b, %sb_b 40 // %add_t = VADD.i32 %sa_t, %sb_t 41 // %r = VMOVNT.i16 %add_b, %add_t 42 // 43 //===----------------------------------------------------------------------===// 44 45 #include "ARM.h" 46 #include "ARMBaseInstrInfo.h" 47 #include "ARMSubtarget.h" 48 #include "llvm/ADT/SetVector.h" 49 #include "llvm/Analysis/TargetTransformInfo.h" 50 #include "llvm/CodeGen/TargetLowering.h" 51 #include "llvm/CodeGen/TargetPassConfig.h" 52 #include "llvm/CodeGen/TargetSubtargetInfo.h" 53 #include "llvm/IR/BasicBlock.h" 54 #include "llvm/IR/Constant.h" 55 #include "llvm/IR/Constants.h" 56 #include "llvm/IR/DerivedTypes.h" 57 #include "llvm/IR/Function.h" 58 #include "llvm/IR/IRBuilder.h" 59 #include "llvm/IR/InstIterator.h" 60 #include "llvm/IR/InstrTypes.h" 61 #include "llvm/IR/Instruction.h" 62 #include "llvm/IR/Instructions.h" 63 #include "llvm/IR/IntrinsicInst.h" 64 #include "llvm/IR/Intrinsics.h" 65 #include "llvm/IR/IntrinsicsARM.h" 66 #include "llvm/IR/PatternMatch.h" 67 #include "llvm/IR/Type.h" 68 #include "llvm/IR/Value.h" 69 #include "llvm/InitializePasses.h" 70 #include "llvm/Pass.h" 71 #include "llvm/Support/Casting.h" 72 #include <algorithm> 73 #include <cassert> 74 75 using namespace llvm; 76 77 #define DEBUG_TYPE "mve-laneinterleave" 78 79 cl::opt<bool> EnableInterleave( 80 "enable-mve-interleave", cl::Hidden, cl::init(true), 81 cl::desc("Enable interleave MVE vector operation lowering")); 82 83 namespace { 84 85 class MVELaneInterleaving : public FunctionPass { 86 public: 87 static char ID; // Pass identification, replacement for typeid 88 89 explicit MVELaneInterleaving() : FunctionPass(ID) { 90 initializeMVELaneInterleavingPass(*PassRegistry::getPassRegistry()); 91 } 92 93 bool runOnFunction(Function &F) override; 94 95 StringRef getPassName() const override { return "MVE lane interleaving"; } 96 97 void getAnalysisUsage(AnalysisUsage &AU) const override { 98 AU.setPreservesCFG(); 99 AU.addRequired<TargetPassConfig>(); 100 FunctionPass::getAnalysisUsage(AU); 101 } 102 }; 103 104 } // end anonymous namespace 105 106 char MVELaneInterleaving::ID = 0; 107 108 INITIALIZE_PASS(MVELaneInterleaving, DEBUG_TYPE, "MVE lane interleaving", false, 109 false) 110 111 Pass *llvm::createMVELaneInterleavingPass() { 112 return new MVELaneInterleaving(); 113 } 114 115 static bool isProfitableToInterleave(SmallSetVector<Instruction *, 4> &Exts, 116 SmallSetVector<Instruction *, 4> &Truncs) { 117 // This is not always beneficial to transform. Exts can be incorporated into 118 // loads, Truncs can be folded into stores. 119 // Truncs are usually the same number of instructions, 120 // VSTRH.32(A);VSTRH.32(B) vs VSTRH.16(VMOVNT A, B) with interleaving 121 // Exts are unfortunately more instructions in the general case: 122 // A=VLDRH.32; B=VLDRH.32; 123 // vs with interleaving: 124 // T=VLDRH.16; A=VMOVNB T; B=VMOVNT T 125 // But those VMOVL may be folded into a VMULL. 126 127 // But expensive extends/truncs are always good to remove. FPExts always 128 // involve extra VCVT's so are always considered to be beneficial to convert. 129 for (auto *E : Exts) { 130 if (isa<FPExtInst>(E) || !isa<LoadInst>(E->getOperand(0))) { 131 LLVM_DEBUG(dbgs() << "Beneficial due to " << *E << "\n"); 132 return true; 133 } 134 } 135 for (auto *T : Truncs) { 136 if (T->hasOneUse() && !isa<StoreInst>(*T->user_begin())) { 137 LLVM_DEBUG(dbgs() << "Beneficial due to " << *T << "\n"); 138 return true; 139 } 140 } 141 142 // Otherwise, we know we have a load(ext), see if any of the Extends are a 143 // vmull. This is a simple heuristic and certainly not perfect. 144 for (auto *E : Exts) { 145 if (!E->hasOneUse() || 146 cast<Instruction>(*E->user_begin())->getOpcode() != Instruction::Mul) { 147 LLVM_DEBUG(dbgs() << "Not beneficial due to " << *E << "\n"); 148 return false; 149 } 150 } 151 return true; 152 } 153 154 static bool tryInterleave(Instruction *Start, 155 SmallPtrSetImpl<Instruction *> &Visited) { 156 LLVM_DEBUG(dbgs() << "tryInterleave from " << *Start << "\n"); 157 auto *VT = cast<FixedVectorType>(Start->getType()); 158 159 if (!isa<Instruction>(Start->getOperand(0))) 160 return false; 161 162 // Look for connected operations starting from Ext's, terminating at Truncs. 163 std::vector<Instruction *> Worklist; 164 Worklist.push_back(Start); 165 Worklist.push_back(cast<Instruction>(Start->getOperand(0))); 166 167 SmallSetVector<Instruction *, 4> Truncs; 168 SmallSetVector<Instruction *, 4> Exts; 169 SmallSetVector<Use *, 4> OtherLeafs; 170 SmallSetVector<Instruction *, 4> Ops; 171 172 while (!Worklist.empty()) { 173 Instruction *I = Worklist.back(); 174 Worklist.pop_back(); 175 176 switch (I->getOpcode()) { 177 // Truncs 178 case Instruction::Trunc: 179 case Instruction::FPTrunc: 180 if (Truncs.count(I)) 181 continue; 182 Truncs.insert(I); 183 Visited.insert(I); 184 break; 185 186 // Extend leafs 187 case Instruction::SExt: 188 case Instruction::ZExt: 189 case Instruction::FPExt: 190 if (Exts.count(I)) 191 continue; 192 for (auto *Use : I->users()) 193 Worklist.push_back(cast<Instruction>(Use)); 194 Exts.insert(I); 195 break; 196 197 case Instruction::Call: { 198 IntrinsicInst *II = dyn_cast<IntrinsicInst>(I); 199 if (!II) 200 return false; 201 202 switch (II->getIntrinsicID()) { 203 case Intrinsic::abs: 204 case Intrinsic::smin: 205 case Intrinsic::smax: 206 case Intrinsic::umin: 207 case Intrinsic::umax: 208 case Intrinsic::sadd_sat: 209 case Intrinsic::ssub_sat: 210 case Intrinsic::uadd_sat: 211 case Intrinsic::usub_sat: 212 case Intrinsic::minnum: 213 case Intrinsic::maxnum: 214 case Intrinsic::fabs: 215 case Intrinsic::fma: 216 case Intrinsic::ceil: 217 case Intrinsic::floor: 218 case Intrinsic::rint: 219 case Intrinsic::round: 220 case Intrinsic::trunc: 221 break; 222 default: 223 return false; 224 } 225 LLVM_FALLTHROUGH; // Fall through to treating these like an operator below. 226 } 227 // Binary/tertiary ops 228 case Instruction::Add: 229 case Instruction::Sub: 230 case Instruction::Mul: 231 case Instruction::AShr: 232 case Instruction::LShr: 233 case Instruction::Shl: 234 case Instruction::ICmp: 235 case Instruction::FCmp: 236 case Instruction::FAdd: 237 case Instruction::FMul: 238 case Instruction::Select: 239 if (Ops.count(I)) 240 continue; 241 Ops.insert(I); 242 243 for (Use &Op : I->operands()) { 244 if (!isa<FixedVectorType>(Op->getType())) 245 continue; 246 if (isa<Instruction>(Op)) 247 Worklist.push_back(cast<Instruction>(&Op)); 248 else 249 OtherLeafs.insert(&Op); 250 } 251 252 for (auto *Use : I->users()) 253 Worklist.push_back(cast<Instruction>(Use)); 254 break; 255 256 case Instruction::ShuffleVector: 257 // A shuffle of a splat is a splat. 258 if (cast<ShuffleVectorInst>(I)->isZeroEltSplat()) 259 continue; 260 LLVM_FALLTHROUGH; 261 262 default: 263 LLVM_DEBUG(dbgs() << " Unhandled instruction: " << *I << "\n"); 264 return false; 265 } 266 } 267 268 if (Exts.empty() && OtherLeafs.empty()) 269 return false; 270 271 LLVM_DEBUG({ 272 dbgs() << "Found group:\n Exts:"; 273 for (auto *I : Exts) 274 dbgs() << " " << *I << "\n"; 275 dbgs() << " Ops:"; 276 for (auto *I : Ops) 277 dbgs() << " " << *I << "\n"; 278 dbgs() << " OtherLeafs:"; 279 for (auto *I : OtherLeafs) 280 dbgs() << " " << *I->get() << " of " << *I->getUser() << "\n"; 281 dbgs() << "Truncs:"; 282 for (auto *I : Truncs) 283 dbgs() << " " << *I << "\n"; 284 }); 285 286 assert(!Truncs.empty() && "Expected some truncs"); 287 288 // Check types 289 unsigned NumElts = VT->getNumElements(); 290 unsigned BaseElts = VT->getScalarSizeInBits() == 16 291 ? 8 292 : (VT->getScalarSizeInBits() == 8 ? 16 : 0); 293 if (BaseElts == 0 || NumElts % BaseElts != 0) { 294 LLVM_DEBUG(dbgs() << " Type is unsupported\n"); 295 return false; 296 } 297 if (Start->getOperand(0)->getType()->getScalarSizeInBits() != 298 VT->getScalarSizeInBits() * 2) { 299 LLVM_DEBUG(dbgs() << " Type not double sized\n"); 300 return false; 301 } 302 for (Instruction *I : Exts) 303 if (I->getOperand(0)->getType() != VT) { 304 LLVM_DEBUG(dbgs() << " Wrong type on " << *I << "\n"); 305 return false; 306 } 307 for (Instruction *I : Truncs) 308 if (I->getType() != VT) { 309 LLVM_DEBUG(dbgs() << " Wrong type on " << *I << "\n"); 310 return false; 311 } 312 313 // Check that it looks beneficial 314 if (!isProfitableToInterleave(Exts, Truncs)) 315 return false; 316 317 // Create new shuffles around the extends / truncs / other leaves. 318 IRBuilder<> Builder(Start); 319 320 SmallVector<int, 16> LeafMask; 321 SmallVector<int, 16> TruncMask; 322 // LeafMask : 0, 2, 4, 6, 1, 3, 5, 7 8, 10, 12, 14, 9, 11, 13, 15 323 // TruncMask: 0, 4, 1, 5, 2, 6, 3, 7 8, 12, 9, 13, 10, 14, 11, 15 324 for (unsigned Base = 0; Base < NumElts; Base += BaseElts) { 325 for (unsigned i = 0; i < BaseElts / 2; i++) 326 LeafMask.push_back(Base + i * 2); 327 for (unsigned i = 0; i < BaseElts / 2; i++) 328 LeafMask.push_back(Base + i * 2 + 1); 329 } 330 for (unsigned Base = 0; Base < NumElts; Base += BaseElts) { 331 for (unsigned i = 0; i < BaseElts / 2; i++) { 332 TruncMask.push_back(Base + i); 333 TruncMask.push_back(Base + i + BaseElts / 2); 334 } 335 } 336 337 for (Instruction *I : Exts) { 338 LLVM_DEBUG(dbgs() << "Replacing ext " << *I << "\n"); 339 Builder.SetInsertPoint(I); 340 Value *Shuffle = Builder.CreateShuffleVector(I->getOperand(0), LeafMask); 341 bool FPext = isa<FPExtInst>(I); 342 bool Sext = isa<SExtInst>(I); 343 Value *Ext = FPext ? Builder.CreateFPExt(Shuffle, I->getType()) 344 : Sext ? Builder.CreateSExt(Shuffle, I->getType()) 345 : Builder.CreateZExt(Shuffle, I->getType()); 346 I->replaceAllUsesWith(Ext); 347 LLVM_DEBUG(dbgs() << " with " << *Shuffle << "\n"); 348 } 349 350 for (Use *I : OtherLeafs) { 351 LLVM_DEBUG(dbgs() << "Replacing leaf " << *I << "\n"); 352 Builder.SetInsertPoint(cast<Instruction>(I->getUser())); 353 Value *Shuffle = Builder.CreateShuffleVector(I->get(), LeafMask); 354 I->getUser()->setOperand(I->getOperandNo(), Shuffle); 355 LLVM_DEBUG(dbgs() << " with " << *Shuffle << "\n"); 356 } 357 358 for (Instruction *I : Truncs) { 359 LLVM_DEBUG(dbgs() << "Replacing trunc " << *I << "\n"); 360 361 Builder.SetInsertPoint(I->getParent(), ++I->getIterator()); 362 Value *Shuf = Builder.CreateShuffleVector(I, TruncMask); 363 I->replaceAllUsesWith(Shuf); 364 cast<Instruction>(Shuf)->setOperand(0, I); 365 366 LLVM_DEBUG(dbgs() << " with " << *Shuf << "\n"); 367 } 368 369 return true; 370 } 371 372 bool MVELaneInterleaving::runOnFunction(Function &F) { 373 if (!EnableInterleave) 374 return false; 375 auto &TPC = getAnalysis<TargetPassConfig>(); 376 auto &TM = TPC.getTM<TargetMachine>(); 377 auto *ST = &TM.getSubtarget<ARMSubtarget>(F); 378 if (!ST->hasMVEIntegerOps()) 379 return false; 380 381 bool Changed = false; 382 383 SmallPtrSet<Instruction *, 16> Visited; 384 for (Instruction &I : reverse(instructions(F))) { 385 if (I.getType()->isVectorTy() && 386 (isa<TruncInst>(I) || isa<FPTruncInst>(I)) && !Visited.count(&I)) 387 Changed |= tryInterleave(&I, Visited); 388 } 389 390 return Changed; 391 } 392