[PGO] Profile guided code size optimization.
Summary:
Enable some of the existing size optimizations for cold code under PGO.
A ~5% code size saving in big internal app under PGO.
The way it gets BFI/PSI is discussed in the RFC thread
http://lists.llvm.org/pipermail/llvm-dev/2019-March/130894.html
Note it doesn't currently touch loop passes.
Reviewers: davidxl, eraman
Reviewed By: eraman
Subscribers: mgorny, javed.absar, smeenai, mehdi_amini, eraman, zzheng, steven_wu, dexonsmith, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59514
llvm-svn: 358422
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 2cc2e82..d8b188a 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -4178,7 +4178,7 @@
auto InstCombineErase = [this](Instruction *I) {
eraseInstFromFunction(*I);
};
- LibCallSimplifier Simplifier(DL, &TLI, ORE, InstCombineRAUW,
+ LibCallSimplifier Simplifier(DL, &TLI, ORE, BFI, PSI, InstCombineRAUW,
InstCombineErase);
if (Value *With = Simplifier.optimizeCall(CI)) {
++NumSimplified;
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
index ee18536..a42cf95 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
+++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
@@ -52,12 +52,14 @@
class APInt;
class AssumptionCache;
+class BlockFrequencyInfo;
class DataLayout;
class DominatorTree;
class GEPOperator;
class GlobalVariable;
class LoopInfo;
class OptimizationRemarkEmitter;
+class ProfileSummaryInfo;
class TargetLibraryInfo;
class User;
@@ -304,6 +306,8 @@
const DataLayout &DL;
const SimplifyQuery SQ;
OptimizationRemarkEmitter &ORE;
+ BlockFrequencyInfo *BFI;
+ ProfileSummaryInfo *PSI;
// Optional analyses. When non-null, these can both be used to do better
// combining and will be updated to reflect any changes.
@@ -315,11 +319,11 @@
InstCombiner(InstCombineWorklist &Worklist, BuilderTy &Builder,
bool MinimizeSize, bool ExpensiveCombines, AliasAnalysis *AA,
AssumptionCache &AC, TargetLibraryInfo &TLI, DominatorTree &DT,
- OptimizationRemarkEmitter &ORE, const DataLayout &DL,
- LoopInfo *LI)
+ OptimizationRemarkEmitter &ORE, BlockFrequencyInfo *BFI,
+ ProfileSummaryInfo *PSI, const DataLayout &DL, LoopInfo *LI)
: Worklist(Worklist), Builder(Builder), MinimizeSize(MinimizeSize),
ExpensiveCombines(ExpensiveCombines), AA(AA), AC(AC), TLI(TLI), DT(DT),
- DL(DL), SQ(DL, &TLI, &DT, &AC), ORE(ORE), LI(LI) {}
+ DL(DL), SQ(DL, &TLI, &DT, &AC), ORE(ORE), BFI(BFI), PSI(PSI), LI(LI) {}
/// Run the combiner over the entire worklist until it is empty.
///
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 6ee94a9..7e26dee 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -46,14 +46,17 @@
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/BasicAliasAnalysis.h"
+#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/CFG.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/EHPersonalities.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/InstructionSimplify.h"
+#include "llvm/Analysis/LazyBlockFrequencyInfo.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/TargetFolder.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/ValueTracking.h"
@@ -3478,7 +3481,8 @@
static bool combineInstructionsOverFunction(
Function &F, InstCombineWorklist &Worklist, AliasAnalysis *AA,
AssumptionCache &AC, TargetLibraryInfo &TLI, DominatorTree &DT,
- OptimizationRemarkEmitter &ORE, bool ExpensiveCombines = true,
+ OptimizationRemarkEmitter &ORE, BlockFrequencyInfo *BFI,
+ ProfileSummaryInfo *PSI, bool ExpensiveCombines = true,
LoopInfo *LI = nullptr) {
auto &DL = F.getParent()->getDataLayout();
ExpensiveCombines |= EnableExpensiveCombines;
@@ -3509,7 +3513,7 @@
MadeIRChange |= prepareICWorklistFromFunction(F, DL, &TLI, Worklist);
InstCombiner IC(Worklist, Builder, F.hasMinSize(), ExpensiveCombines, AA,
- AC, TLI, DT, ORE, DL, LI);
+ AC, TLI, DT, ORE, BFI, PSI, DL, LI);
IC.MaxArraySizeForCombine = MaxArraySize;
if (!IC.run())
@@ -3529,8 +3533,15 @@
auto *LI = AM.getCachedResult<LoopAnalysis>(F);
auto *AA = &AM.getResult<AAManager>(F);
+ const ModuleAnalysisManager &MAM =
+ AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
+ ProfileSummaryInfo *PSI =
+ MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+ auto *BFI = (PSI && PSI->hasProfileSummary()) ?
+ &AM.getResult<BlockFrequencyAnalysis>(F) : nullptr;
+
if (!combineInstructionsOverFunction(F, Worklist, AA, AC, TLI, DT, ORE,
- ExpensiveCombines, LI))
+ BFI, PSI, ExpensiveCombines, LI))
// No changes, all analyses are preserved.
return PreservedAnalyses::all();
@@ -3554,6 +3565,8 @@
AU.addPreserved<AAResultsWrapperPass>();
AU.addPreserved<BasicAAWrapperPass>();
AU.addPreserved<GlobalsAAWrapperPass>();
+ AU.addRequired<ProfileSummaryInfoWrapperPass>();
+ LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU);
}
bool InstructionCombiningPass::runOnFunction(Function &F) {
@@ -3570,9 +3583,15 @@
// Optional analyses.
auto *LIWP = getAnalysisIfAvailable<LoopInfoWrapperPass>();
auto *LI = LIWP ? &LIWP->getLoopInfo() : nullptr;
+ ProfileSummaryInfo *PSI =
+ &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
+ BlockFrequencyInfo *BFI =
+ (PSI && PSI->hasProfileSummary()) ?
+ &getAnalysis<LazyBlockFrequencyInfoPass>().getBFI() :
+ nullptr;
return combineInstructionsOverFunction(F, Worklist, AA, AC, TLI, DT, ORE,
- ExpensiveCombines, LI);
+ BFI, PSI, ExpensiveCombines, LI);
}
char InstructionCombiningPass::ID = 0;
@@ -3585,6 +3604,8 @@
INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass)
INITIALIZE_PASS_DEPENDENCY(GlobalsAAWrapperPass)
INITIALIZE_PASS_DEPENDENCY(OptimizationRemarkEmitterWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(LazyBlockFrequencyInfoPass)
+INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass)
INITIALIZE_PASS_END(InstructionCombiningPass, "instcombine",
"Combine redundant instructions", false, false)
diff --git a/llvm/lib/Transforms/Scalar/ConstantHoisting.cpp b/llvm/lib/Transforms/Scalar/ConstantHoisting.cpp
index f1cc5e4..683cf32 100644
--- a/llvm/lib/Transforms/Scalar/ConstantHoisting.cpp
+++ b/llvm/lib/Transforms/Scalar/ConstantHoisting.cpp
@@ -41,6 +41,7 @@
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/BlockFrequencyInfo.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/IR/BasicBlock.h"
@@ -60,6 +61,7 @@
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
@@ -111,6 +113,7 @@
if (ConstHoistWithBlockFrequency)
AU.addRequired<BlockFrequencyInfoWrapperPass>();
AU.addRequired<DominatorTreeWrapperPass>();
+ AU.addRequired<ProfileSummaryInfoWrapperPass>();
AU.addRequired<TargetTransformInfoWrapperPass>();
}
@@ -126,6 +129,7 @@
"Constant Hoisting", false, false)
INITIALIZE_PASS_DEPENDENCY(BlockFrequencyInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_PASS_END(ConstantHoistingLegacyPass, "consthoist",
"Constant Hoisting", false, false)
@@ -148,7 +152,8 @@
ConstHoistWithBlockFrequency
? &getAnalysis<BlockFrequencyInfoWrapperPass>().getBFI()
: nullptr,
- Fn.getEntryBlock());
+ Fn.getEntryBlock(),
+ &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI());
if (MadeChange) {
LLVM_DEBUG(dbgs() << "********** Function after Constant Hoisting: "
@@ -548,7 +553,9 @@
ConstCandVecType::iterator &MaxCostItr) {
unsigned NumUses = 0;
- if(!Entry->getParent()->hasOptSize() || std::distance(S,E) > 100) {
+ bool OptForSize = Entry->getParent()->hasOptSize() ||
+ llvm::shouldOptimizeForSize(Entry->getParent(), PSI, BFI);
+ if (!OptForSize || std::distance(S,E) > 100) {
for (auto ConstCand = S; ConstCand != E; ++ConstCand) {
NumUses += ConstCand->Uses.size();
if (ConstCand->CumulativeCost > MaxCostItr->CumulativeCost)
@@ -919,13 +926,14 @@
/// Optimize expensive integer constants in the given function.
bool ConstantHoistingPass::runImpl(Function &Fn, TargetTransformInfo &TTI,
DominatorTree &DT, BlockFrequencyInfo *BFI,
- BasicBlock &Entry) {
+ BasicBlock &Entry, ProfileSummaryInfo *PSI) {
this->TTI = &TTI;
this->DT = &DT;
this->BFI = BFI;
this->DL = &Fn.getParent()->getDataLayout();
this->Ctx = &Fn.getContext();
this->Entry = &Entry;
+ this->PSI = PSI;
// Collect all constant candidates.
collectConstantCandidates(Fn);
@@ -962,7 +970,9 @@
auto BFI = ConstHoistWithBlockFrequency
? &AM.getResult<BlockFrequencyAnalysis>(F)
: nullptr;
- if (!runImpl(F, TTI, DT, BFI, F.getEntryBlock()))
+ auto &MAM = AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
+ auto *PSI = MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+ if (!runImpl(F, TTI, DT, BFI, F.getEntryBlock(), PSI))
return PreservedAnalyses::all();
PreservedAnalyses PA;
diff --git a/llvm/lib/Transforms/Scalar/LoopLoadElimination.cpp b/llvm/lib/Transforms/Scalar/LoopLoadElimination.cpp
index 7526610..68bb28d 100644
--- a/llvm/lib/Transforms/Scalar/LoopLoadElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopLoadElimination.cpp
@@ -29,11 +29,14 @@
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
+#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/GlobalsModRef.h"
+#include "llvm/Analysis/LazyBlockFrequencyInfo.h"
#include "llvm/Analysis/LoopAccessAnalysis.h"
#include "llvm/Analysis/LoopAnalysisManager.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemorySSA.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
@@ -54,6 +57,7 @@
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils.h"
#include "llvm/Transforms/Utils/LoopVersioning.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
#include <algorithm>
#include <cassert>
#include <forward_list>
@@ -159,8 +163,9 @@
class LoadEliminationForLoop {
public:
LoadEliminationForLoop(Loop *L, LoopInfo *LI, const LoopAccessInfo &LAI,
- DominatorTree *DT)
- : L(L), LI(LI), LAI(LAI), DT(DT), PSE(LAI.getPSE()) {}
+ DominatorTree *DT, BlockFrequencyInfo *BFI,
+ ProfileSummaryInfo* PSI)
+ : L(L), LI(LI), LAI(LAI), DT(DT), BFI(BFI), PSI(PSI), PSE(LAI.getPSE()) {}
/// Look through the loop-carried and loop-independent dependences in
/// this loop and find store->load dependences.
@@ -529,7 +534,11 @@
}
if (!Checks.empty() || !LAI.getPSE().getUnionPredicate().isAlwaysTrue()) {
- if (L->getHeader()->getParent()->hasOptSize()) {
+ auto *HeaderBB = L->getHeader();
+ auto *F = HeaderBB->getParent();
+ bool OptForSize = F->hasOptSize() ||
+ llvm::shouldOptimizeForSize(HeaderBB, PSI, BFI);
+ if (OptForSize) {
LLVM_DEBUG(
dbgs() << "Versioning is needed but not allowed when optimizing "
"for size.\n");
@@ -572,6 +581,8 @@
LoopInfo *LI;
const LoopAccessInfo &LAI;
DominatorTree *DT;
+ BlockFrequencyInfo *BFI;
+ ProfileSummaryInfo *PSI;
PredicatedScalarEvolution PSE;
};
@@ -579,6 +590,7 @@
static bool
eliminateLoadsAcrossLoops(Function &F, LoopInfo &LI, DominatorTree &DT,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI,
function_ref<const LoopAccessInfo &(Loop &)> GetLAI) {
// Build up a worklist of inner-loops to transform to avoid iterator
// invalidation.
@@ -597,7 +609,7 @@
bool Changed = false;
for (Loop *L : Worklist) {
// The actual work is performed by LoadEliminationForLoop.
- LoadEliminationForLoop LEL(L, &LI, GetLAI(*L), &DT);
+ LoadEliminationForLoop LEL(L, &LI, GetLAI(*L), &DT, BFI, PSI);
Changed |= LEL.processLoop();
}
return Changed;
@@ -622,10 +634,14 @@
auto &LI = getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
auto &LAA = getAnalysis<LoopAccessLegacyAnalysis>();
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+ auto *PSI = &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
+ auto *BFI = (PSI && PSI->hasProfileSummary()) ?
+ &getAnalysis<LazyBlockFrequencyInfoPass>().getBFI() :
+ nullptr;
// Process each loop nest in the function.
return eliminateLoadsAcrossLoops(
- F, LI, DT,
+ F, LI, DT, BFI, PSI,
[&LAA](Loop &L) -> const LoopAccessInfo & { return LAA.getInfo(&L); });
}
@@ -638,6 +654,8 @@
AU.addRequired<DominatorTreeWrapperPass>();
AU.addPreserved<DominatorTreeWrapperPass>();
AU.addPreserved<GlobalsAAWrapperPass>();
+ AU.addRequired<ProfileSummaryInfoWrapperPass>();
+ LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU);
}
};
@@ -653,6 +671,8 @@
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(ScalarEvolutionWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
+INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(LazyBlockFrequencyInfoPass)
INITIALIZE_PASS_END(LoopLoadElimination, LLE_OPTION, LLE_name, false, false)
FunctionPass *llvm::createLoopLoadEliminationPass() {
@@ -668,13 +688,17 @@
auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
auto &AA = AM.getResult<AAManager>(F);
auto &AC = AM.getResult<AssumptionAnalysis>(F);
+ auto &MAM = AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
+ auto *PSI = MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+ auto *BFI = (PSI && PSI->hasProfileSummary()) ?
+ &AM.getResult<BlockFrequencyAnalysis>(F) : nullptr;
MemorySSA *MSSA = EnableMSSALoopDependency
? &AM.getResult<MemorySSAAnalysis>(F).getMSSA()
: nullptr;
auto &LAM = AM.getResult<LoopAnalysisManagerFunctionProxy>(F).getManager();
bool Changed = eliminateLoadsAcrossLoops(
- F, LI, DT, [&](Loop &L) -> const LoopAccessInfo & {
+ F, LI, DT, BFI, PSI, [&](Loop &L) -> const LoopAccessInfo & {
LoopStandardAnalysisResults AR = {AA, AC, DT, LI, SE, TLI, TTI, MSSA};
return LAM.getResult<LoopAccessAnalysis>(L, AR);
});
diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
index 71e9953..86891eb 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
@@ -294,7 +294,8 @@
return LoopUnrollResult::Unmodified;
TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences(
- L, SE, TTI, OptLevel, None, None, None, None, None, None);
+ L, SE, TTI, nullptr, nullptr, OptLevel,
+ None, None, None, None, None, None);
if (AllowUnrollAndJam.getNumOccurrences() > 0)
UP.UnrollAndJam = AllowUnrollAndJam;
if (UnrollAndJamThreshold.getNumOccurrences() > 0)
diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
index 53d4ae8..c113e4d 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -23,7 +23,9 @@
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Analysis/AssumptionCache.h"
+#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/CodeMetrics.h"
+#include "llvm/Analysis/LazyBlockFrequencyInfo.h"
#include "llvm/Analysis/LoopAnalysisManager.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopPass.h"
@@ -55,6 +57,7 @@
#include "llvm/Transforms/Utils.h"
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
#include "llvm/Transforms/Utils/UnrollLoop.h"
#include <algorithm>
#include <cassert>
@@ -165,7 +168,8 @@
/// Gather the various unrolling parameters based on the defaults, compiler
/// flags, TTI overrides and user specified parameters.
TargetTransformInfo::UnrollingPreferences llvm::gatherUnrollingPreferences(
- Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI, int OptLevel,
+ Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI, int OptLevel,
Optional<unsigned> UserThreshold, Optional<unsigned> UserCount,
Optional<bool> UserAllowPartial, Optional<bool> UserRuntime,
Optional<bool> UserUpperBound, Optional<bool> UserAllowPeeling) {
@@ -198,7 +202,9 @@
TTI.getUnrollingPreferences(L, SE, UP);
// Apply size attributes
- if (L->getHeader()->getParent()->hasOptSize()) {
+ bool OptForSize = L->getHeader()->getParent()->hasOptSize() ||
+ llvm::shouldOptimizeForSize(L->getHeader(), PSI, BFI);
+ if (OptForSize) {
UP.Threshold = UP.OptSizeThreshold;
UP.PartialThreshold = UP.PartialOptSizeThreshold;
}
@@ -963,7 +969,9 @@
static LoopUnrollResult tryToUnrollLoop(
Loop *L, DominatorTree &DT, LoopInfo *LI, ScalarEvolution &SE,
const TargetTransformInfo &TTI, AssumptionCache &AC,
- OptimizationRemarkEmitter &ORE, bool PreserveLCSSA, int OptLevel,
+ OptimizationRemarkEmitter &ORE,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI,
+ bool PreserveLCSSA, int OptLevel,
bool OnlyWhenForced, bool ForgetAllSCEV, Optional<unsigned> ProvidedCount,
Optional<unsigned> ProvidedThreshold, Optional<bool> ProvidedAllowPartial,
Optional<bool> ProvidedRuntime, Optional<bool> ProvidedUpperBound,
@@ -989,7 +997,7 @@
bool NotDuplicatable;
bool Convergent;
TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences(
- L, SE, TTI, OptLevel, ProvidedThreshold, ProvidedCount,
+ L, SE, TTI, BFI, PSI, OptLevel, ProvidedThreshold, ProvidedCount,
ProvidedAllowPartial, ProvidedRuntime, ProvidedUpperBound,
ProvidedAllowPeeling);
// Exit early if unrolling is disabled.
@@ -1176,7 +1184,8 @@
bool PreserveLCSSA = mustPreserveAnalysisID(LCSSAID);
LoopUnrollResult Result = tryToUnrollLoop(
- L, DT, LI, SE, TTI, AC, ORE, PreserveLCSSA, OptLevel, OnlyWhenForced,
+ L, DT, LI, SE, TTI, AC, ORE, nullptr, nullptr,
+ PreserveLCSSA, OptLevel, OnlyWhenForced,
ForgetAllSCEV, ProvidedCount, ProvidedThreshold, ProvidedAllowPartial,
ProvidedRuntime, ProvidedUpperBound, ProvidedAllowPeeling);
@@ -1257,6 +1266,7 @@
bool Changed =
tryToUnrollLoop(&L, AR.DT, &AR.LI, AR.SE, AR.TTI, AR.AC, *ORE,
+ /*BFI*/ nullptr, /*PSI*/ nullptr,
/*PreserveLCSSA*/ true, OptLevel, OnlyWhenForced,
/*ForgetAllSCEV*/ false, /*Count*/ None,
/*Threshold*/ None, /*AllowPartial*/ false,
@@ -1359,6 +1369,8 @@
AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
ProfileSummaryInfo *PSI =
MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+ auto *BFI = (PSI && PSI->hasProfileSummary()) ?
+ &AM.getResult<BlockFrequencyAnalysis>(F) : nullptr;
bool Changed = false;
@@ -1394,7 +1406,7 @@
// The API here is quite complex to call and we allow to select some
// flavors of unrolling during construction time (by setting UnrollOpts).
LoopUnrollResult Result = tryToUnrollLoop(
- &L, DT, &LI, SE, TTI, AC, ORE,
+ &L, DT, &LI, SE, TTI, AC, ORE, BFI, PSI,
/*PreserveLCSSA*/ true, UnrollOpts.OptLevel, UnrollOpts.OnlyWhenForced,
/*ForgetAllSCEV*/ false, /*Count*/ None,
/*Threshold*/ None, UnrollOpts.AllowPartial, UnrollOpts.AllowRuntime,
diff --git a/llvm/lib/Transforms/Utils/CMakeLists.txt b/llvm/lib/Transforms/Utils/CMakeLists.txt
index cb3dc17..c232aa6 100644
--- a/llvm/lib/Transforms/Utils/CMakeLists.txt
+++ b/llvm/lib/Transforms/Utils/CMakeLists.txt
@@ -51,6 +51,7 @@
SimplifyCFG.cpp
SimplifyIndVar.cpp
SimplifyLibCalls.cpp
+ SizeOpts.cpp
SplitModule.cpp
StripNonLineTableDebugInfo.cpp
SymbolRewriter.cpp
diff --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
index f35328ca..23b88db 100644
--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -16,8 +16,10 @@
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/Triple.h"
+#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Analysis/ValueTracking.h"
@@ -34,6 +36,7 @@
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
using namespace llvm;
using namespace PatternMatch;
@@ -2375,7 +2378,9 @@
// Don't rewrite fputs to fwrite when optimising for size because fwrite
// requires more arguments and thus extra MOVs are required.
- if (CI->getFunction()->hasOptSize())
+ bool OptForSize = CI->getFunction()->hasOptSize() ||
+ llvm::shouldOptimizeForSize(CI->getParent(), PSI, BFI);
+ if (OptForSize)
return nullptr;
// Check if has any use
@@ -2750,9 +2755,10 @@
LibCallSimplifier::LibCallSimplifier(
const DataLayout &DL, const TargetLibraryInfo *TLI,
OptimizationRemarkEmitter &ORE,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI,
function_ref<void(Instruction *, Value *)> Replacer,
function_ref<void(Instruction *)> Eraser)
- : FortifiedSimplifier(TLI), DL(DL), TLI(TLI), ORE(ORE),
+ : FortifiedSimplifier(TLI), DL(DL), TLI(TLI), ORE(ORE), BFI(BFI), PSI(PSI),
UnsafeFPShrink(false), Replacer(Replacer), Eraser(Eraser) {}
void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) {
diff --git a/llvm/lib/Transforms/Utils/SizeOpts.cpp b/llvm/lib/Transforms/Utils/SizeOpts.cpp
new file mode 100644
index 0000000..1519751
--- /dev/null
+++ b/llvm/lib/Transforms/Utils/SizeOpts.cpp
@@ -0,0 +1,37 @@
+//===-- SizeOpts.cpp - code size optimization related code ----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains some shared code size optimization related code.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Analysis/BlockFrequencyInfo.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
+using namespace llvm;
+
+static cl::opt<bool> ProfileGuidedSizeOpt(
+ "pgso", cl::Hidden, cl::init(true),
+ cl::desc("Enable the profile guided size optimization. "));
+
+bool llvm::shouldOptimizeForSize(Function *F, ProfileSummaryInfo *PSI,
+ BlockFrequencyInfo *BFI) {
+ assert(F);
+ if (!PSI || !BFI || !PSI->hasProfileSummary())
+ return false;
+ return ProfileGuidedSizeOpt && PSI->isFunctionColdInCallGraph(F, *BFI);
+}
+
+bool llvm::shouldOptimizeForSize(BasicBlock *BB, ProfileSummaryInfo *PSI,
+ BlockFrequencyInfo *BFI) {
+ assert(BB);
+ if (!PSI || !BFI || !PSI->hasProfileSummary())
+ return false;
+ return ProfileGuidedSizeOpt && PSI->isColdBlock(BB, BFI);
+}
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 5612e95..06b70b3 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -88,6 +88,7 @@
#include "llvm/Analysis/LoopIterator.h"
#include "llvm/Analysis/MemorySSA.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
@@ -134,6 +135,7 @@
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
#include "llvm/Transforms/Utils/LoopVersioning.h"
+#include "llvm/Transforms/Utils/SizeOpts.h"
#include "llvm/Transforms/Vectorize/LoopVectorizationLegality.h"
#include <algorithm>
#include <cassert>
@@ -1452,12 +1454,13 @@
auto *LAA = &getAnalysis<LoopAccessLegacyAnalysis>();
auto *DB = &getAnalysis<DemandedBitsWrapperPass>().getDemandedBits();
auto *ORE = &getAnalysis<OptimizationRemarkEmitterWrapperPass>().getORE();
+ auto *PSI = &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
std::function<const LoopAccessInfo &(Loop &)> GetLAA =
[&](Loop &L) -> const LoopAccessInfo & { return LAA->getInfo(&L); };
return Impl.runImpl(F, *SE, *LI, *TTI, *DT, *BFI, TLI, *DB, *AA, *AC,
- GetLAA, *ORE);
+ GetLAA, *ORE, PSI);
}
void getAnalysisUsage(AnalysisUsage &AU) const override {
@@ -1483,6 +1486,7 @@
AU.addPreserved<BasicAAWrapperPass>();
AU.addPreserved<GlobalsAAWrapperPass>();
+ AU.addRequired<ProfileSummaryInfoWrapperPass>();
}
};
@@ -6054,6 +6058,7 @@
INITIALIZE_PASS_DEPENDENCY(LoopAccessLegacyAnalysis)
INITIALIZE_PASS_DEPENDENCY(DemandedBitsWrapperPass)
INITIALIZE_PASS_DEPENDENCY(OptimizationRemarkEmitterWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass)
INITIALIZE_PASS_END(LoopVectorize, LV_NAME, lv_name, false, false)
namespace llvm {
@@ -7147,7 +7152,8 @@
Loop *L, PredicatedScalarEvolution &PSE, LoopInfo *LI, DominatorTree *DT,
LoopVectorizationLegality *LVL, TargetTransformInfo *TTI,
TargetLibraryInfo *TLI, DemandedBits *DB, AssumptionCache *AC,
- OptimizationRemarkEmitter *ORE, LoopVectorizeHints &Hints) {
+ OptimizationRemarkEmitter *ORE, BlockFrequencyInfo *BFI,
+ ProfileSummaryInfo *PSI, LoopVectorizeHints &Hints) {
assert(EnableVPlanNativePath && "VPlan-native path is disabled.");
Function *F = L->getHeader()->getParent();
@@ -7162,10 +7168,12 @@
// Get user vectorization factor.
const unsigned UserVF = Hints.getWidth();
- // Check the function attributes to find out if this function should be
- // optimized for size.
+ // Check the function attributes and profiles to find out if this function
+ // should be optimized for size.
bool OptForSize =
- Hints.getForce() != LoopVectorizeHints::FK_Enabled && F->hasOptSize();
+ Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
+ (F->hasOptSize() ||
+ llvm::shouldOptimizeForSize(L->getHeader(), PSI, BFI));
// Plan how to best vectorize, return the best VF and its cost.
const VectorizationFactor VF = LVP.planInVPlanNativePath(OptForSize, UserVF);
@@ -7245,10 +7253,12 @@
return false;
}
- // Check the function attributes to find out if this function should be
- // optimized for size.
+ // Check the function attributes and profiles to find out if this function
+ // should be optimized for size.
bool OptForSize =
- Hints.getForce() != LoopVectorizeHints::FK_Enabled && F->hasOptSize();
+ Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
+ (F->hasOptSize() ||
+ llvm::shouldOptimizeForSize(L->getHeader(), PSI, BFI));
// Entrance to the VPlan-native vectorization path. Outer loops are processed
// here. They may require CFG and instruction level transformations before
@@ -7257,7 +7267,7 @@
// pipeline.
if (!L->empty())
return processLoopInVPlanNativePath(L, PSE, LI, DT, &LVL, TTI, TLI, DB, AC,
- ORE, Hints);
+ ORE, BFI, PSI, Hints);
assert(L->empty() && "Inner loop expected.");
// Check the loop for a trip count threshold: vectorize loops with a tiny trip
@@ -7523,7 +7533,7 @@
DominatorTree &DT_, BlockFrequencyInfo &BFI_, TargetLibraryInfo *TLI_,
DemandedBits &DB_, AliasAnalysis &AA_, AssumptionCache &AC_,
std::function<const LoopAccessInfo &(Loop &)> &GetLAA_,
- OptimizationRemarkEmitter &ORE_) {
+ OptimizationRemarkEmitter &ORE_, ProfileSummaryInfo *PSI_) {
SE = &SE_;
LI = &LI_;
TTI = &TTI_;
@@ -7535,6 +7545,7 @@
GetLAA = &GetLAA_;
DB = &DB_;
ORE = &ORE_;
+ PSI = PSI_;
// Don't attempt if
// 1. the target claims to have no vector registers, and
@@ -7603,8 +7614,12 @@
LoopStandardAnalysisResults AR = {AA, AC, DT, LI, SE, TLI, TTI, MSSA};
return LAM.getResult<LoopAccessAnalysis>(L, AR);
};
+ const ModuleAnalysisManager &MAM =
+ AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
+ ProfileSummaryInfo *PSI =
+ MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
bool Changed =
- runImpl(F, SE, LI, TTI, DT, BFI, &TLI, DB, AA, AC, GetLAA, ORE);
+ runImpl(F, SE, LI, TTI, DT, BFI, &TLI, DB, AA, AC, GetLAA, ORE, PSI);
if (!Changed)
return PreservedAnalyses::all();
PreservedAnalyses PA;