From 98d6d2372feb2f8bb4c3941b278027499cbfc00b Mon Sep 17 00:00:00 2001 From: Benjamin Kramer Date: Fri, 26 Aug 2011 01:22:29 +0000 Subject: [PATCH] SimplifyCFG: If we have a PHI node that can evaluate to NULL and do a load or store to the address returned by the PHI node then we can consider this incoming value as dead and remove the edge pointing there, unless there are instructions that can affect control flow executed in between. In theory this could be extended to other instructions, eg. division by zero, but it's likely that it will "miscompile" some code because people depend on div by zero not trapping. NULL pointer dereference usually leads to a crash so we should be on the safe side. This shrinks the size of a Release clang by 16k on x86_64. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138618 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Transforms/Utils/SimplifyCFG.cpp | 69 +++++++++++++++ .../SimplifyCFG/phi-undef-loadstore.ll | 87 +++++++++++++++++++ 2 files changed, 156 insertions(+) create mode 100644 test/Transforms/SimplifyCFG/phi-undef-loadstore.ll diff --git a/lib/Transforms/Utils/SimplifyCFG.cpp b/lib/Transforms/Utils/SimplifyCFG.cpp index 2d60c35afb1..dce4e7abab1 100644 --- a/lib/Transforms/Utils/SimplifyCFG.cpp +++ b/lib/Transforms/Utils/SimplifyCFG.cpp @@ -2723,6 +2723,72 @@ bool SimplifyCFGOpt::SimplifyCondBranch(BranchInst *BI, IRBuilder<> &Builder) { return false; } +/// Check if passing a value to an instruction will cause undefined behavior. +static bool passingValueIsAlwaysUndefined(Value *V, Instruction *I) { + Constant *C = dyn_cast(V); + if (!C) + return false; + + if (!I->hasOneUse()) // FIXME: There is no reason to limit this to one use. + return false; + + if (C->isNullValue()) { + Instruction *Use = I->use_back(); + + // Now make sure that there are no instructions in between that can alter + // control flow (eg. calls) + for (BasicBlock::iterator i = ++BasicBlock::iterator(I); &*i != Use; ++i) + if (i == I->getParent()->end() || + !i->isSafeToSpeculativelyExecute()) + return false; + + // Look through GEPs. A load from a GEP derived from NULL is still undefined + if (GetElementPtrInst *GEP = dyn_cast(Use)) + if (GEP->getPointerOperand() == I) + return passingValueIsAlwaysUndefined(V, GEP); + + // Look through bitcasts. + if (BitCastInst *BC = dyn_cast(Use)) + return passingValueIsAlwaysUndefined(V, BC); + + // load from null is undefined + if (isa(Use)) + return true; + + // store to null is undef + if (isa(Use) && Use->getOperand(1) == I) + return true; + } + return false; +} + +/// If BB has an incoming value that will always trigger undefined behavior +/// (eg. null pointer derefence), remove the branch leading here. +static bool removeUndefIntroducingPredecessor(BasicBlock *BB) { + for (BasicBlock::iterator i = BB->begin(); + PHINode *PHI = dyn_cast(i); ++i) + for (unsigned i = 0, e = PHI->getNumIncomingValues(); i != e; ++i) + if (passingValueIsAlwaysUndefined(PHI->getIncomingValue(i), PHI)) { + TerminatorInst *T = PHI->getIncomingBlock(i)->getTerminator(); + IRBuilder<> Builder(T); + if (BranchInst *BI = dyn_cast(T)) { + BB->removePredecessor(PHI->getIncomingBlock(i)); + // Turn uncoditional branches into unreachables and remove the dead + // destination from conditional branches. + if (BI->isUnconditional()) + Builder.CreateUnreachable(); + else + Builder.CreateBr(BI->getSuccessor(0) == BB ? BI->getSuccessor(1) : + BI->getSuccessor(0)); + BI->eraseFromParent(); + return true; + } + // TODO: SwitchInst. + } + + return false; +} + bool SimplifyCFGOpt::run(BasicBlock *BB) { bool Changed = false; @@ -2746,6 +2812,9 @@ bool SimplifyCFGOpt::run(BasicBlock *BB) { // Check for and eliminate duplicate PHI nodes in this block. Changed |= EliminateDuplicatePHINodes(BB); + // Check for and remove branches that will always cause undefined behavior. + Changed |= removeUndefIntroducingPredecessor(BB); + // Merge basic blocks into their predecessor if there is only one distinct // pred, and if there is only one distinct successor of the predecessor, and // if there are no PHI nodes. diff --git a/test/Transforms/SimplifyCFG/phi-undef-loadstore.ll b/test/Transforms/SimplifyCFG/phi-undef-loadstore.ll new file mode 100644 index 00000000000..65d888ea01e --- /dev/null +++ b/test/Transforms/SimplifyCFG/phi-undef-loadstore.ll @@ -0,0 +1,87 @@ +; RUN: opt -simplifycfg -S < %s | FileCheck %s + +declare void @bar() nounwind + +define i32 @test1(i32* %a, i32 %b, i32* %c, i32 %d) nounwind { +entry: + %tobool = icmp eq i32 %b, 0 + br i1 %tobool, label %if.else, label %if.then + +if.then: ; preds = %entry + tail call void @bar() nounwind + br label %if.end7 + +if.else: ; preds = %entry + %tobool3 = icmp eq i32 %d, 0 + br i1 %tobool3, label %if.end7, label %if.then4 + +if.then4: ; preds = %if.else + tail call void @bar() nounwind + br label %if.end7 + +if.end7: ; preds = %if.else, %if.then4, %if.then + %x.0 = phi i32* [ %a, %if.then ], [ %c, %if.then4 ], [ null, %if.else ] + %tmp9 = load i32* %x.0 + ret i32 %tmp9 + +; CHECK: @test1 +; CHECK: if.else: +; CHECK: br label %if.end7 + +; CHECK: phi i32* [ %a, %if.then ], [ %c, %if.else ] +} + +define i32 @test2(i32* %a, i32 %b, i32* %c, i32 %d) nounwind { +entry: + %tobool = icmp eq i32 %b, 0 + br i1 %tobool, label %if.else, label %if.then + +if.then: ; preds = %entry + tail call void @bar() nounwind + br label %if.end7 + +if.else: ; preds = %entry + %tobool3 = icmp eq i32 %d, 0 + br i1 %tobool3, label %if.end7, label %if.then4 + +if.then4: ; preds = %if.else + tail call void @bar() nounwind + br label %if.end7 + +if.end7: ; preds = %if.else, %if.then4, %if.then + %x.0 = phi i32* [ %a, %if.then ], [ null, %if.then4 ], [ null, %if.else ] + %tmp9 = load i32* %x.0 + ret i32 %tmp9 +; CHECK: @test2 +; CHECK: if.else: +; CHECK: unreachable + +; CHECK-NOT: phi +} + +define i32 @test3(i32* %a, i32 %b, i32* %c, i32 %d) nounwind { +entry: + %tobool = icmp eq i32 %b, 0 + br i1 %tobool, label %if.else, label %if.then + +if.then: ; preds = %entry + tail call void @bar() nounwind + br label %if.end7 + +if.else: ; preds = %entry + %tobool3 = icmp eq i32 %d, 0 + br i1 %tobool3, label %if.end7, label %if.then4 + +if.then4: ; preds = %if.else + tail call void @bar() nounwind + br label %if.end7 + +if.end7: ; preds = %if.else, %if.then4, %if.then + %x.0 = phi i32* [ %a, %if.then ], [ null, %if.then4 ], [ null, %if.else ] + tail call void @bar() nounwind + %tmp9 = load i32* %x.0 + ret i32 %tmp9 +; CHECK: @test3 +; CHECK: if.end7: +; CHECK: phi i32* [ %a, %if.then ], [ null, %if.then4 ], [ null, %if.else ] +}