From 2c5f9584bacd1f038c59478e3a1096f218357dfb Mon Sep 17 00:00:00 2001 From: Hal Finkel Date: Fri, 20 Feb 2015 05:08:21 +0000 Subject: [PATCH] [PowerPC] Loop Data Prefetching for the BG/Q The IBM BG/Q supercomputer's A2 cores have a hardware prefetching unit, the L1P, but it does not prefetch directly into the A2's L1 cache. Instead, it prefetches into its own L1P buffer, and the latency to access that buffer is significantly higher than that to the L1 cache (although smaller than the latency to the L2 cache). As a result, especially when multiple hardware threads are not actively busy, explicitly prefetching data into the L1 cache is advantageous. I've been using this pass out-of-tree for data prefetching on the BG/Q for well over a year, and it has worked quite well. It is enabled by default only for the BG/Q, but can be enabled for other cores as well via a command-line option. Eventually, we might want to add some TTI interfaces and move this into Transforms/Scalar (there is nothing particularly target dependent about it, although only machines like the BG/Q will benefit from its simplistic strategy). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229966 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/PowerPC/CMakeLists.txt | 1 + lib/Target/PowerPC/PPC.h | 1 + lib/Target/PowerPC/PPCLoopDataPrefetch.cpp | 231 +++++++++++++++++++++ lib/Target/PowerPC/PPCTargetMachine.cpp | 15 ++ test/CodeGen/PowerPC/loop-data-prefetch.ll | 29 +++ 5 files changed, 277 insertions(+) create mode 100644 lib/Target/PowerPC/PPCLoopDataPrefetch.cpp create mode 100644 test/CodeGen/PowerPC/loop-data-prefetch.ll diff --git a/lib/Target/PowerPC/CMakeLists.txt b/lib/Target/PowerPC/CMakeLists.txt index ea988b1093d..936ed7f2b1b 100644 --- a/lib/Target/PowerPC/CMakeLists.txt +++ b/lib/Target/PowerPC/CMakeLists.txt @@ -23,6 +23,7 @@ add_llvm_target(PowerPCCodeGen PPCEarlyReturn.cpp PPCFastISel.cpp PPCFrameLowering.cpp + PPCLoopDataPrefetch.cpp PPCLoopPreIncPrep.cpp PPCMCInstLower.cpp PPCMachineFunctionInfo.cpp diff --git a/lib/Target/PowerPC/PPC.h b/lib/Target/PowerPC/PPC.h index bc5fc97b4f1..5e5a9b1999d 100644 --- a/lib/Target/PowerPC/PPC.h +++ b/lib/Target/PowerPC/PPC.h @@ -34,6 +34,7 @@ namespace llvm { #ifndef NDEBUG FunctionPass *createPPCCTRLoopsVerify(); #endif + FunctionPass *createPPCLoopDataPrefetchPass(); FunctionPass *createPPCLoopPreIncPrepPass(PPCTargetMachine &TM); FunctionPass *createPPCEarlyReturnPass(); FunctionPass *createPPCVSXCopyPass(); diff --git a/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp b/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp new file mode 100644 index 00000000000..efd2d92b928 --- /dev/null +++ b/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp @@ -0,0 +1,231 @@ +//===-------- PPCLoopDataPrefetch.cpp - Loop Data Prefetching Pass --------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file implements a Loop Data Prefetching Pass. +// +//===----------------------------------------------------------------------===// + +#define DEBUG_TYPE "ppc-loop-data-prefetch" +#include "PPC.h" +#include "llvm/Transforms/Scalar.h" +#include "llvm/ADT/Statistic.h" +#include "llvm/Analysis/AssumptionCache.h" +#include "llvm/Analysis/CodeMetrics.h" +#include "llvm/Analysis/InstructionSimplify.h" +#include "llvm/Analysis/LoopInfo.h" +#include "llvm/Analysis/ScalarEvolution.h" +#include "llvm/Analysis/ScalarEvolutionExpander.h" +#include "llvm/Analysis/ScalarEvolutionExpressions.h" +#include "llvm/Analysis/TargetTransformInfo.h" +#include "llvm/Analysis/ValueTracking.h" +#include "llvm/IR/CFG.h" +#include "llvm/IR/Dominators.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/IntrinsicInst.h" +#include "llvm/IR/Module.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/Debug.h" +#include "llvm/Transforms/Utils/BasicBlockUtils.h" +#include "llvm/Transforms/Utils/Local.h" +#include "llvm/Transforms/Utils/ValueMapper.h" +using namespace llvm; + +// By default, we limit this to creating 16 PHIs (which is a little over half +// of the allocatable register set). +static cl::opt +PrefetchWrites("ppc-loop-prefetch-writes", cl::Hidden, cl::init(false), + cl::desc("Prefetch write addresses")); + +// This seems like a reasonable default for the BG/Q (this pass is enabled, by +// default, only on the BG/Q). +static cl::opt +PrefDist("ppc-loop-prefetch-distance", cl::Hidden, cl::init(300), + cl::desc("The loop prefetch distance")); + +static cl::opt +CacheLineSize("ppc-loop-prefetch-cache-line", cl::Hidden, cl::init(64), + cl::desc("The loop prefetch cache line size")); + +namespace llvm { + void initializePPCLoopDataPrefetchPass(PassRegistry&); +} + +namespace { + + class PPCLoopDataPrefetch : public FunctionPass { + public: + static char ID; // Pass ID, replacement for typeid + PPCLoopDataPrefetch() : FunctionPass(ID) { + initializePPCLoopDataPrefetchPass(*PassRegistry::getPassRegistry()); + } + + void getAnalysisUsage(AnalysisUsage &AU) const override { + AU.addRequired(); + AU.addPreserved(); + AU.addRequired(); + AU.addPreserved(); + AU.addRequired(); + // FIXME: For some reason, preserving SE here breaks LSR (even if + // this pass changes nothing). + // AU.addPreserved(); + AU.addRequired(); + } + + bool runOnFunction(Function &F) override; + bool runOnLoop(Loop *L); + + private: + AssumptionCache *AC; + LoopInfo *LI; + ScalarEvolution *SE; + const TargetTransformInfo *TTI; + const DataLayout *DL; + }; +} + +char PPCLoopDataPrefetch::ID = 0; +INITIALIZE_PASS_BEGIN(PPCLoopDataPrefetch, "ppc-loop-data-prefetch", + "PPC Loop Data Prefetch", false, false) +INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker) +INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass) +INITIALIZE_PASS_DEPENDENCY(LoopInfoWrapperPass) +INITIALIZE_PASS_DEPENDENCY(ScalarEvolution) +INITIALIZE_PASS_END(PPCLoopDataPrefetch, "ppc-loop-data-prefetch", + "PPC Loop Data Prefetch", false, false) + +FunctionPass *llvm::createPPCLoopDataPrefetchPass() { return new PPCLoopDataPrefetch(); } + +bool PPCLoopDataPrefetch::runOnFunction(Function &F) { + LI = &getAnalysis().getLoopInfo(); + SE = &getAnalysis(); + DL = F.getParent()->getDataLayout(); + AC = &getAnalysis().getAssumptionCache(F); + TTI = &getAnalysis().getTTI(F); + + bool MadeChange = false; + + for (LoopInfo::iterator I = LI->begin(), E = LI->end(); + I != E; ++I) { + Loop *L = *I; + MadeChange |= runOnLoop(L); + } + + return MadeChange; +} + +bool PPCLoopDataPrefetch::runOnLoop(Loop *L) { + bool MadeChange = false; + + // Only prefetch in the inner-most loop + if (!L->empty()) + return MadeChange; + + SmallPtrSet EphValues; + CodeMetrics::collectEphemeralValues(L, AC, EphValues); + + // Calculate the number of iterations ahead to prefetch + CodeMetrics Metrics; + for (Loop::block_iterator I = L->block_begin(), IE = L->block_end(); + I != IE; ++I) { + + // If the loop already has prefetches, then assume that the user knows + // what he or she is doing and don't add any more. + for (BasicBlock::iterator J = (*I)->begin(), JE = (*I)->end(); + J != JE; ++J) + if (CallInst *CI = dyn_cast(J)) + if (Function *F = CI->getCalledFunction()) + if (F->getIntrinsicID() == Intrinsic::prefetch) + return MadeChange; + + Metrics.analyzeBasicBlock(*I, *TTI, EphValues); + } + unsigned LoopSize = Metrics.NumInsts; + if (!LoopSize) + LoopSize = 1; + + unsigned ItersAhead = PrefDist/LoopSize; + if (!ItersAhead) + ItersAhead = 1; + + SmallVector, 16> PrefLoads; + for (Loop::block_iterator I = L->block_begin(), IE = L->block_end(); + I != IE; ++I) { + for (BasicBlock::iterator J = (*I)->begin(), JE = (*I)->end(); + J != JE; ++J) { + Value *PtrValue; + Instruction *MemI; + + if (LoadInst *LMemI = dyn_cast(J)) { + MemI = LMemI; + PtrValue = LMemI->getPointerOperand(); + } else if (StoreInst *SMemI = dyn_cast(J)) { + if (!PrefetchWrites) continue; + MemI = SMemI; + PtrValue = SMemI->getPointerOperand(); + } else continue; + + unsigned PtrAddrSpace = PtrValue->getType()->getPointerAddressSpace(); + if (PtrAddrSpace) + continue; + + if (L->isLoopInvariant(PtrValue)) + continue; + + const SCEV *LSCEV = SE->getSCEV(PtrValue); + const SCEVAddRecExpr *LSCEVAddRec = dyn_cast(LSCEV); + if (!LSCEVAddRec) + continue; + + // We don't want to double prefetch individual cache lines. If this load + // is known to be within one cache line of some other load that has + // already been prefetched, then don't prefetch this one as well. + bool DupPref = false; + for (SmallVector, + 16>::iterator K = PrefLoads.begin(), KE = PrefLoads.end(); + K != KE; ++K) { + const SCEV *PtrDiff = SE->getMinusSCEV(LSCEVAddRec, K->second); + if (const SCEVConstant *ConstPtrDiff = + dyn_cast(PtrDiff)) { + int64_t PD = abs64(ConstPtrDiff->getValue()->getSExtValue()); + if (PD < (int64_t) CacheLineSize) { + DupPref = true; + break; + } + } + } + if (DupPref) + continue; + + const SCEV *NextLSCEV = SE->getAddExpr(LSCEVAddRec, SE->getMulExpr( + SE->getConstant(LSCEVAddRec->getType(), ItersAhead), + LSCEVAddRec->getStepRecurrence(*SE))); + if (!isSafeToExpand(NextLSCEV, *SE)) + continue; + + PrefLoads.push_back(std::make_pair(MemI, LSCEVAddRec)); + + Type *I8Ptr = Type::getInt8PtrTy((*I)->getContext(), PtrAddrSpace); + SCEVExpander SCEVE(*SE, "prefaddr"); + Value *PrefPtrValue = SCEVE.expandCodeFor(NextLSCEV, I8Ptr, MemI); + + IRBuilder<> Builder(MemI); + Module *M = (*I)->getParent()->getParent(); + Type *I32 = Type::getInt32Ty((*I)->getContext()); + Value *PrefetchFunc = Intrinsic::getDeclaration(M, Intrinsic::prefetch); + Builder.CreateCall4(PrefetchFunc, PrefPtrValue, + ConstantInt::get(I32, MemI->mayReadFromMemory() ? 0 : 1), + ConstantInt::get(I32, 3), ConstantInt::get(I32, 1)); + + MadeChange = true; + } + } + + return MadeChange; +} + diff --git a/lib/Target/PowerPC/PPCTargetMachine.cpp b/lib/Target/PowerPC/PPCTargetMachine.cpp index a81f7fbfc1c..b219e93dabe 100644 --- a/lib/Target/PowerPC/PPCTargetMachine.cpp +++ b/lib/Target/PowerPC/PPCTargetMachine.cpp @@ -43,6 +43,11 @@ EnableGEPOpt("ppc-gep-opt", cl::Hidden, cl::desc("Enable optimizations on complex GEPs"), cl::init(true)); +static cl::opt +EnablePrefetch("enable-ppc-prefetching", + cl::desc("disable software prefetching on PPC"), + cl::init(false), cl::Hidden); + extern "C" void LLVMInitializePowerPCTarget() { // Register the targets RegisterTargetMachine A(ThePPC32Target); @@ -240,6 +245,16 @@ TargetPassConfig *PPCTargetMachine::createPassConfig(PassManagerBase &PM) { void PPCPassConfig::addIRPasses() { addPass(createAtomicExpandPass(&getPPCTargetMachine())); + // For the BG/Q (or if explicitly requested), add explicit data prefetch + // intrinsics. + bool UsePrefetching = + Triple(TM->getTargetTriple()).getVendor() == Triple::BGQ && + getOptLevel() != CodeGenOpt::None; + if (EnablePrefetch.getNumOccurrences() > 0) + UsePrefetching = EnablePrefetch; + if (UsePrefetching) + addPass(createPPCLoopDataPrefetchPass()); + if (TM->getOptLevel() == CodeGenOpt::Aggressive && EnableGEPOpt) { // Call SeparateConstOffsetFromGEP pass to extract constants within indices // and lower a GEP with multiple indices to either arithmetic operations or diff --git a/test/CodeGen/PowerPC/loop-data-prefetch.ll b/test/CodeGen/PowerPC/loop-data-prefetch.ll new file mode 100644 index 00000000000..88714819bd7 --- /dev/null +++ b/test/CodeGen/PowerPC/loop-data-prefetch.ll @@ -0,0 +1,29 @@ +; RUN: llc -mcpu=a2 < %s | FileCheck %s +target datalayout = "E-m:e-i64:64-n32:64" +target triple = "powerpc64-bgq-linux" + +; Function Attrs: nounwind +define void @foo(double* nocapture %a, double* nocapture readonly %b) #0 { +entry: + br label %for.body + +for.body: ; preds = %for.body, %entry + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] + %arrayidx = getelementptr inbounds double* %b, i64 %indvars.iv + %0 = load double* %arrayidx, align 8 + %add = fadd double %0, 1.000000e+00 + %arrayidx2 = getelementptr inbounds double* %a, i64 %indvars.iv + store double %add, double* %arrayidx2, align 8 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp eq i64 %indvars.iv.next, 1600 + br i1 %exitcond, label %for.end, label %for.body + +for.end: ; preds = %for.body + ret void + +; CHECK-LABEL: @foo +; CHECK: dcbt +} + +attributes #0 = { nounwind } +