From 8e2b8ae3b11b9778da5cb54d330e984b33979bf9 Mon Sep 17 00:00:00 2001 From: Rafael Espindola Date: Wed, 14 Nov 2012 05:08:56 +0000 Subject: [PATCH] Handle DAG CSE adding new uses during ReplaceAllUsesWith. Fixes PR14333. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167912 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 13 ++++++++++++- test/CodeGen/X86/pr14333.ll | 12 ++++++++++++ 2 files changed, 24 insertions(+), 1 deletion(-) create mode 100644 test/CodeGen/X86/pr14333.ll diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp index 5416a5c1d7a..37d7731aa15 100644 --- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp +++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -7728,7 +7728,18 @@ bool DAGCombiner::MergeConsecutiveStores(StoreSDNode* St) { if (StoreNodes[i].MemNode == EarliestOp) continue; StoreSDNode *St = cast(StoreNodes[i].MemNode); - DAG.ReplaceAllUsesOfValueWith(SDValue(St, 0), St->getChain()); + // ReplaceAllUsesWith will replace all uses that existed when it was + // called, but graph optimizations may cause new ones to appear. For + // example, the case in pr14333 looks like + // + // St's chain -> St -> another store -> X + // + // And the only difference from St to the other store is the chain. + // When we change it's chain to be St's chain they become identical, + // get CSEed and the net result is that X is now a use of St. + // Since we know that St is redundant, just iterate. + while (!St->use_empty()) + DAG.ReplaceAllUsesWith(SDValue(St, 0), St->getChain()); removeFromWorkList(St); DAG.DeleteNode(St); } diff --git a/test/CodeGen/X86/pr14333.ll b/test/CodeGen/X86/pr14333.ll new file mode 100644 index 00000000000..86c12ef6b54 --- /dev/null +++ b/test/CodeGen/X86/pr14333.ll @@ -0,0 +1,12 @@ +; RUN: llc -mtriple=x86_64-unknown-unknown < %s +%foo = type { i64, i64 } +define void @bar(%foo* %zed) { + %tmp = getelementptr inbounds %foo* %zed, i64 0, i32 0 + store i64 0, i64* %tmp, align 8 + %tmp2 = getelementptr inbounds %foo* %zed, i64 0, i32 1 + store i64 0, i64* %tmp2, align 8 + %tmp3 = bitcast %foo* %zed to i8* + call void @llvm.memset.p0i8.i64(i8* %tmp3, i8 0, i64 16, i32 8, i1 false) + ret void +} +declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind