llvm-mirror/lib/CodeGen/AtomicExpandPass.cpp
Robin Morisset 98b0bed638 Lower idempotent RMWs to fence+load
Summary:
I originally tried doing this specifically for X86 in the backend in D5091,
but it was rather brittle and generally running too late to be general.
Furthermore, other targets may want to implement similar optimizations.
So I reimplemented it at the IR-level, fitting it into AtomicExpandPass
as it interacts with that pass (which could not be cleanly done before
at the backend level).

This optimization relies on a new target hook, which is only used by X86
for now, as the correctness of the optimization on other targets remains
an open question. If it is found correct on other targets, it should be
trivial to enable for them.

Details of the optimization are discussed in D5091.

Test Plan: make check-all + a new test

Reviewers: jfb

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5422

llvm-svn: 218455
2014-09-25 17:27:43 +00:00

564 lines
20 KiB
C++

//===-- AtomicExpandPass.cpp - Expand atomic instructions -------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains a pass (at IR level) to replace atomic instructions with
// either (intrinsic-based) load-linked/store-conditional loops or AtomicCmpXchg.
//
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/Passes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetSubtargetInfo.h"
using namespace llvm;
#define DEBUG_TYPE "atomic-expand"
namespace {
class AtomicExpand: public FunctionPass {
const TargetMachine *TM;
public:
static char ID; // Pass identification, replacement for typeid
explicit AtomicExpand(const TargetMachine *TM = nullptr)
: FunctionPass(ID), TM(TM) {
initializeAtomicExpandPass(*PassRegistry::getPassRegistry());
}
bool runOnFunction(Function &F) override;
private:
bool bracketInstWithFences(Instruction *I, AtomicOrdering Order,
bool IsStore, bool IsLoad);
bool expandAtomicLoad(LoadInst *LI);
bool expandAtomicLoadToLL(LoadInst *LI);
bool expandAtomicLoadToCmpXchg(LoadInst *LI);
bool expandAtomicStore(StoreInst *SI);
bool expandAtomicRMW(AtomicRMWInst *AI);
bool expandAtomicRMWToLLSC(AtomicRMWInst *AI);
bool expandAtomicRMWToCmpXchg(AtomicRMWInst *AI);
bool expandAtomicCmpXchg(AtomicCmpXchgInst *CI);
bool isIdempotentRMW(AtomicRMWInst *AI);
bool simplifyIdempotentRMW(AtomicRMWInst *AI);
};
}
char AtomicExpand::ID = 0;
char &llvm::AtomicExpandID = AtomicExpand::ID;
INITIALIZE_TM_PASS(AtomicExpand, "atomic-expand",
"Expand Atomic calls in terms of either load-linked & store-conditional or cmpxchg",
false, false)
FunctionPass *llvm::createAtomicExpandPass(const TargetMachine *TM) {
return new AtomicExpand(TM);
}
bool AtomicExpand::runOnFunction(Function &F) {
if (!TM || !TM->getSubtargetImpl()->enableAtomicExpand())
return false;
auto TargetLowering = TM->getSubtargetImpl()->getTargetLowering();
SmallVector<Instruction *, 1> AtomicInsts;
// Changing control-flow while iterating through it is a bad idea, so gather a
// list of all atomic instructions before we start.
for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
if (I->isAtomic())
AtomicInsts.push_back(&*I);
}
bool MadeChange = false;
for (auto I : AtomicInsts) {
auto LI = dyn_cast<LoadInst>(I);
auto SI = dyn_cast<StoreInst>(I);
auto RMWI = dyn_cast<AtomicRMWInst>(I);
auto CASI = dyn_cast<AtomicCmpXchgInst>(I);
assert((LI || SI || RMWI || CASI || isa<FenceInst>(I)) &&
"Unknown atomic instruction");
auto FenceOrdering = Monotonic;
bool IsStore, IsLoad;
if (TargetLowering->getInsertFencesForAtomic()) {
if (LI && isAtLeastAcquire(LI->getOrdering())) {
FenceOrdering = LI->getOrdering();
LI->setOrdering(Monotonic);
IsStore = false;
IsLoad = true;
} else if (SI && isAtLeastRelease(SI->getOrdering())) {
FenceOrdering = SI->getOrdering();
SI->setOrdering(Monotonic);
IsStore = true;
IsLoad = false;
} else if (RMWI && (isAtLeastRelease(RMWI->getOrdering()) ||
isAtLeastAcquire(RMWI->getOrdering()))) {
FenceOrdering = RMWI->getOrdering();
RMWI->setOrdering(Monotonic);
IsStore = IsLoad = true;
} else if (CASI && !TargetLowering->hasLoadLinkedStoreConditional() &&
(isAtLeastRelease(CASI->getSuccessOrdering()) ||
isAtLeastAcquire(CASI->getSuccessOrdering()))) {
// If a compare and swap is lowered to LL/SC, we can do smarter fence
// insertion, with a stronger one on the success path than on the
// failure path. As a result, fence insertion is directly done by
// expandAtomicCmpXchg in that case.
FenceOrdering = CASI->getSuccessOrdering();
CASI->setSuccessOrdering(Monotonic);
CASI->setFailureOrdering(Monotonic);
IsStore = IsLoad = true;
}
if (FenceOrdering != Monotonic) {
MadeChange |= bracketInstWithFences(I, FenceOrdering, IsStore, IsLoad);
}
}
if (LI && TargetLowering->shouldExpandAtomicLoadInIR(LI)) {
MadeChange |= expandAtomicLoad(LI);
} else if (SI && TargetLowering->shouldExpandAtomicStoreInIR(SI)) {
MadeChange |= expandAtomicStore(SI);
} else if (RMWI) {
// There are two different ways of expanding RMW instructions:
// - into a load if it is idempotent
// - into a Cmpxchg/LL-SC loop otherwise
// we try them in that order.
MadeChange |= (isIdempotentRMW(RMWI) &&
simplifyIdempotentRMW(RMWI)) ||
(TargetLowering->shouldExpandAtomicRMWInIR(RMWI) &&
expandAtomicRMW(RMWI));
} else if (CASI && TargetLowering->hasLoadLinkedStoreConditional()) {
MadeChange |= expandAtomicCmpXchg(CASI);
}
}
return MadeChange;
}
bool AtomicExpand::bracketInstWithFences(Instruction *I, AtomicOrdering Order,
bool IsStore, bool IsLoad) {
IRBuilder<> Builder(I);
auto LeadingFence =
TM->getSubtargetImpl()->getTargetLowering()->emitLeadingFence(
Builder, Order, IsStore, IsLoad);
auto TrailingFence =
TM->getSubtargetImpl()->getTargetLowering()->emitTrailingFence(
Builder, Order, IsStore, IsLoad);
// The trailing fence is emitted before the instruction instead of after
// because there is no easy way of setting Builder insertion point after
// an instruction. So we must erase it from the BB, and insert it back
// in the right place.
// We have a guard here because not every atomic operation generates a
// trailing fence.
if (TrailingFence) {
TrailingFence->removeFromParent();
TrailingFence->insertAfter(I);
}
return (LeadingFence || TrailingFence);
}
bool AtomicExpand::expandAtomicLoad(LoadInst *LI) {
if (TM->getSubtargetImpl()
->getTargetLowering()
->hasLoadLinkedStoreConditional())
return expandAtomicLoadToLL(LI);
else
return expandAtomicLoadToCmpXchg(LI);
}
bool AtomicExpand::expandAtomicLoadToLL(LoadInst *LI) {
auto TLI = TM->getSubtargetImpl()->getTargetLowering();
IRBuilder<> Builder(LI);
// On some architectures, load-linked instructions are atomic for larger
// sizes than normal loads. For example, the only 64-bit load guaranteed
// to be single-copy atomic by ARM is an ldrexd (A3.5.3).
Value *Val =
TLI->emitLoadLinked(Builder, LI->getPointerOperand(), LI->getOrdering());
LI->replaceAllUsesWith(Val);
LI->eraseFromParent();
return true;
}
bool AtomicExpand::expandAtomicLoadToCmpXchg(LoadInst *LI) {
IRBuilder<> Builder(LI);
AtomicOrdering Order = LI->getOrdering();
Value *Addr = LI->getPointerOperand();
Type *Ty = cast<PointerType>(Addr->getType())->getElementType();
Constant *DummyVal = Constant::getNullValue(Ty);
Value *Pair = Builder.CreateAtomicCmpXchg(
Addr, DummyVal, DummyVal, Order,
AtomicCmpXchgInst::getStrongestFailureOrdering(Order));
Value *Loaded = Builder.CreateExtractValue(Pair, 0, "loaded");
LI->replaceAllUsesWith(Loaded);
LI->eraseFromParent();
return true;
}
bool AtomicExpand::expandAtomicStore(StoreInst *SI) {
// This function is only called on atomic stores that are too large to be
// atomic if implemented as a native store. So we replace them by an
// atomic swap, that can be implemented for example as a ldrex/strex on ARM
// or lock cmpxchg8/16b on X86, as these are atomic for larger sizes.
// It is the responsibility of the target to only return true in
// shouldExpandAtomicRMW in cases where this is required and possible.
IRBuilder<> Builder(SI);
AtomicRMWInst *AI =
Builder.CreateAtomicRMW(AtomicRMWInst::Xchg, SI->getPointerOperand(),
SI->getValueOperand(), SI->getOrdering());
SI->eraseFromParent();
// Now we have an appropriate swap instruction, lower it as usual.
return expandAtomicRMW(AI);
}
bool AtomicExpand::expandAtomicRMW(AtomicRMWInst *AI) {
if (TM->getSubtargetImpl()
->getTargetLowering()
->hasLoadLinkedStoreConditional())
return expandAtomicRMWToLLSC(AI);
else
return expandAtomicRMWToCmpXchg(AI);
}
/// Emit IR to implement the given atomicrmw operation on values in registers,
/// returning the new value.
static Value *performAtomicOp(AtomicRMWInst::BinOp Op, IRBuilder<> &Builder,
Value *Loaded, Value *Inc) {
Value *NewVal;
switch (Op) {
case AtomicRMWInst::Xchg:
return Inc;
case AtomicRMWInst::Add:
return Builder.CreateAdd(Loaded, Inc, "new");
case AtomicRMWInst::Sub:
return Builder.CreateSub(Loaded, Inc, "new");
case AtomicRMWInst::And:
return Builder.CreateAnd(Loaded, Inc, "new");
case AtomicRMWInst::Nand:
return Builder.CreateNot(Builder.CreateAnd(Loaded, Inc), "new");
case AtomicRMWInst::Or:
return Builder.CreateOr(Loaded, Inc, "new");
case AtomicRMWInst::Xor:
return Builder.CreateXor(Loaded, Inc, "new");
case AtomicRMWInst::Max:
NewVal = Builder.CreateICmpSGT(Loaded, Inc);
return Builder.CreateSelect(NewVal, Loaded, Inc, "new");
case AtomicRMWInst::Min:
NewVal = Builder.CreateICmpSLE(Loaded, Inc);
return Builder.CreateSelect(NewVal, Loaded, Inc, "new");
case AtomicRMWInst::UMax:
NewVal = Builder.CreateICmpUGT(Loaded, Inc);
return Builder.CreateSelect(NewVal, Loaded, Inc, "new");
case AtomicRMWInst::UMin:
NewVal = Builder.CreateICmpULE(Loaded, Inc);
return Builder.CreateSelect(NewVal, Loaded, Inc, "new");
default:
llvm_unreachable("Unknown atomic op");
}
}
bool AtomicExpand::expandAtomicRMWToLLSC(AtomicRMWInst *AI) {
auto TLI = TM->getSubtargetImpl()->getTargetLowering();
AtomicOrdering MemOpOrder = AI->getOrdering();
Value *Addr = AI->getPointerOperand();
BasicBlock *BB = AI->getParent();
Function *F = BB->getParent();
LLVMContext &Ctx = F->getContext();
// Given: atomicrmw some_op iN* %addr, iN %incr ordering
//
// The standard expansion we produce is:
// [...]
// fence?
// atomicrmw.start:
// %loaded = @load.linked(%addr)
// %new = some_op iN %loaded, %incr
// %stored = @store_conditional(%new, %addr)
// %try_again = icmp i32 ne %stored, 0
// br i1 %try_again, label %loop, label %atomicrmw.end
// atomicrmw.end:
// fence?
// [...]
BasicBlock *ExitBB = BB->splitBasicBlock(AI, "atomicrmw.end");
BasicBlock *LoopBB = BasicBlock::Create(Ctx, "atomicrmw.start", F, ExitBB);
// This grabs the DebugLoc from AI.
IRBuilder<> Builder(AI);
// The split call above "helpfully" added a branch at the end of BB (to the
// wrong place), but we might want a fence too. It's easiest to just remove
// the branch entirely.
std::prev(BB->end())->eraseFromParent();
Builder.SetInsertPoint(BB);
Builder.CreateBr(LoopBB);
// Start the main loop block now that we've taken care of the preliminaries.
Builder.SetInsertPoint(LoopBB);
Value *Loaded = TLI->emitLoadLinked(Builder, Addr, MemOpOrder);
Value *NewVal =
performAtomicOp(AI->getOperation(), Builder, Loaded, AI->getValOperand());
Value *StoreSuccess =
TLI->emitStoreConditional(Builder, NewVal, Addr, MemOpOrder);
Value *TryAgain = Builder.CreateICmpNE(
StoreSuccess, ConstantInt::get(IntegerType::get(Ctx, 32), 0), "tryagain");
Builder.CreateCondBr(TryAgain, LoopBB, ExitBB);
Builder.SetInsertPoint(ExitBB, ExitBB->begin());
AI->replaceAllUsesWith(Loaded);
AI->eraseFromParent();
return true;
}
bool AtomicExpand::expandAtomicRMWToCmpXchg(AtomicRMWInst *AI) {
AtomicOrdering MemOpOrder =
AI->getOrdering() == Unordered ? Monotonic : AI->getOrdering();
Value *Addr = AI->getPointerOperand();
BasicBlock *BB = AI->getParent();
Function *F = BB->getParent();
LLVMContext &Ctx = F->getContext();
// Given: atomicrmw some_op iN* %addr, iN %incr ordering
//
// The standard expansion we produce is:
// [...]
// %init_loaded = load atomic iN* %addr
// br label %loop
// loop:
// %loaded = phi iN [ %init_loaded, %entry ], [ %new_loaded, %loop ]
// %new = some_op iN %loaded, %incr
// %pair = cmpxchg iN* %addr, iN %loaded, iN %new
// %new_loaded = extractvalue { iN, i1 } %pair, 0
// %success = extractvalue { iN, i1 } %pair, 1
// br i1 %success, label %atomicrmw.end, label %loop
// atomicrmw.end:
// [...]
BasicBlock *ExitBB = BB->splitBasicBlock(AI, "atomicrmw.end");
BasicBlock *LoopBB = BasicBlock::Create(Ctx, "atomicrmw.start", F, ExitBB);
// This grabs the DebugLoc from AI.
IRBuilder<> Builder(AI);
// The split call above "helpfully" added a branch at the end of BB (to the
// wrong place), but we want a load. It's easiest to just remove
// the branch entirely.
std::prev(BB->end())->eraseFromParent();
Builder.SetInsertPoint(BB);
LoadInst *InitLoaded = Builder.CreateLoad(Addr);
// Atomics require at least natural alignment.
InitLoaded->setAlignment(AI->getType()->getPrimitiveSizeInBits());
Builder.CreateBr(LoopBB);
// Start the main loop block now that we've taken care of the preliminaries.
Builder.SetInsertPoint(LoopBB);
PHINode *Loaded = Builder.CreatePHI(AI->getType(), 2, "loaded");
Loaded->addIncoming(InitLoaded, BB);
Value *NewVal =
performAtomicOp(AI->getOperation(), Builder, Loaded, AI->getValOperand());
Value *Pair = Builder.CreateAtomicCmpXchg(
Addr, Loaded, NewVal, MemOpOrder,
AtomicCmpXchgInst::getStrongestFailureOrdering(MemOpOrder));
Value *NewLoaded = Builder.CreateExtractValue(Pair, 0, "newloaded");
Loaded->addIncoming(NewLoaded, LoopBB);
Value *Success = Builder.CreateExtractValue(Pair, 1, "success");
Builder.CreateCondBr(Success, ExitBB, LoopBB);
Builder.SetInsertPoint(ExitBB, ExitBB->begin());
AI->replaceAllUsesWith(NewLoaded);
AI->eraseFromParent();
return true;
}
bool AtomicExpand::expandAtomicCmpXchg(AtomicCmpXchgInst *CI) {
auto TLI = TM->getSubtargetImpl()->getTargetLowering();
AtomicOrdering SuccessOrder = CI->getSuccessOrdering();
AtomicOrdering FailureOrder = CI->getFailureOrdering();
Value *Addr = CI->getPointerOperand();
BasicBlock *BB = CI->getParent();
Function *F = BB->getParent();
LLVMContext &Ctx = F->getContext();
// If getInsertFencesForAtomic() returns true, then the target does not want
// to deal with memory orders, and emitLeading/TrailingFence should take care
// of everything. Otherwise, emitLeading/TrailingFence are no-op and we
// should preserve the ordering.
AtomicOrdering MemOpOrder =
TLI->getInsertFencesForAtomic() ? Monotonic : SuccessOrder;
// Given: cmpxchg some_op iN* %addr, iN %desired, iN %new success_ord fail_ord
//
// The full expansion we produce is:
// [...]
// fence?
// cmpxchg.start:
// %loaded = @load.linked(%addr)
// %should_store = icmp eq %loaded, %desired
// br i1 %should_store, label %cmpxchg.trystore,
// label %cmpxchg.failure
// cmpxchg.trystore:
// %stored = @store_conditional(%new, %addr)
// %success = icmp eq i32 %stored, 0
// br i1 %success, label %cmpxchg.success, label %loop/%cmpxchg.failure
// cmpxchg.success:
// fence?
// br label %cmpxchg.end
// cmpxchg.failure:
// fence?
// br label %cmpxchg.end
// cmpxchg.end:
// %success = phi i1 [true, %cmpxchg.success], [false, %cmpxchg.failure]
// %restmp = insertvalue { iN, i1 } undef, iN %loaded, 0
// %res = insertvalue { iN, i1 } %restmp, i1 %success, 1
// [...]
BasicBlock *ExitBB = BB->splitBasicBlock(CI, "cmpxchg.end");
auto FailureBB = BasicBlock::Create(Ctx, "cmpxchg.failure", F, ExitBB);
auto SuccessBB = BasicBlock::Create(Ctx, "cmpxchg.success", F, FailureBB);
auto TryStoreBB = BasicBlock::Create(Ctx, "cmpxchg.trystore", F, SuccessBB);
auto LoopBB = BasicBlock::Create(Ctx, "cmpxchg.start", F, TryStoreBB);
// This grabs the DebugLoc from CI
IRBuilder<> Builder(CI);
// The split call above "helpfully" added a branch at the end of BB (to the
// wrong place), but we might want a fence too. It's easiest to just remove
// the branch entirely.
std::prev(BB->end())->eraseFromParent();
Builder.SetInsertPoint(BB);
TLI->emitLeadingFence(Builder, SuccessOrder, /*IsStore=*/true,
/*IsLoad=*/true);
Builder.CreateBr(LoopBB);
// Start the main loop block now that we've taken care of the preliminaries.
Builder.SetInsertPoint(LoopBB);
Value *Loaded = TLI->emitLoadLinked(Builder, Addr, MemOpOrder);
Value *ShouldStore =
Builder.CreateICmpEQ(Loaded, CI->getCompareOperand(), "should_store");
// If the the cmpxchg doesn't actually need any ordering when it fails, we can
// jump straight past that fence instruction (if it exists).
Builder.CreateCondBr(ShouldStore, TryStoreBB, FailureBB);
Builder.SetInsertPoint(TryStoreBB);
Value *StoreSuccess = TLI->emitStoreConditional(
Builder, CI->getNewValOperand(), Addr, MemOpOrder);
StoreSuccess = Builder.CreateICmpEQ(
StoreSuccess, ConstantInt::get(Type::getInt32Ty(Ctx), 0), "success");
Builder.CreateCondBr(StoreSuccess, SuccessBB,
CI->isWeak() ? FailureBB : LoopBB);
// Make sure later instructions don't get reordered with a fence if necessary.
Builder.SetInsertPoint(SuccessBB);
TLI->emitTrailingFence(Builder, SuccessOrder, /*IsStore=*/true,
/*IsLoad=*/true);
Builder.CreateBr(ExitBB);
Builder.SetInsertPoint(FailureBB);
TLI->emitTrailingFence(Builder, FailureOrder, /*IsStore=*/true,
/*IsLoad=*/true);
Builder.CreateBr(ExitBB);
// Finally, we have control-flow based knowledge of whether the cmpxchg
// succeeded or not. We expose this to later passes by converting any
// subsequent "icmp eq/ne %loaded, %oldval" into a use of an appropriate PHI.
// Setup the builder so we can create any PHIs we need.
Builder.SetInsertPoint(ExitBB, ExitBB->begin());
PHINode *Success = Builder.CreatePHI(Type::getInt1Ty(Ctx), 2);
Success->addIncoming(ConstantInt::getTrue(Ctx), SuccessBB);
Success->addIncoming(ConstantInt::getFalse(Ctx), FailureBB);
// Look for any users of the cmpxchg that are just comparing the loaded value
// against the desired one, and replace them with the CFG-derived version.
SmallVector<ExtractValueInst *, 2> PrunedInsts;
for (auto User : CI->users()) {
ExtractValueInst *EV = dyn_cast<ExtractValueInst>(User);
if (!EV)
continue;
assert(EV->getNumIndices() == 1 && EV->getIndices()[0] <= 1 &&
"weird extraction from { iN, i1 }");
if (EV->getIndices()[0] == 0)
EV->replaceAllUsesWith(Loaded);
else
EV->replaceAllUsesWith(Success);
PrunedInsts.push_back(EV);
}
// We can remove the instructions now we're no longer iterating through them.
for (auto EV : PrunedInsts)
EV->eraseFromParent();
if (!CI->use_empty()) {
// Some use of the full struct return that we don't understand has happened,
// so we've got to reconstruct it properly.
Value *Res;
Res = Builder.CreateInsertValue(UndefValue::get(CI->getType()), Loaded, 0);
Res = Builder.CreateInsertValue(Res, Success, 1);
CI->replaceAllUsesWith(Res);
}
CI->eraseFromParent();
return true;
}
bool AtomicExpand::isIdempotentRMW(AtomicRMWInst* RMWI) {
auto C = dyn_cast<ConstantInt>(RMWI->getValOperand());
if(!C)
return false;
AtomicRMWInst::BinOp Op = RMWI->getOperation();
switch(Op) {
case AtomicRMWInst::Add:
case AtomicRMWInst::Sub:
case AtomicRMWInst::Or:
case AtomicRMWInst::Xor:
return C->isZero();
case AtomicRMWInst::And:
return C->isMinusOne();
// FIXME: we could also treat Min/Max/UMin/UMax by the INT_MIN/INT_MAX/...
default:
return false;
}
}
bool AtomicExpand::simplifyIdempotentRMW(AtomicRMWInst* RMWI) {
auto TLI = TM->getSubtargetImpl()->getTargetLowering();
if (auto ResultingLoad = TLI->lowerIdempotentRMWIntoFencedLoad(RMWI)) {
if (TLI->shouldExpandAtomicLoadInIR(ResultingLoad))
expandAtomicLoad(ResultingLoad);
return true;
}
return false;
}