[MemCpyOpt] Do move the memset, but look at its dest's dependencies.

In effect a partial revert of r237858, which was a dumb shortcut.
Looking at the dependencies of the destination should be the proper
fix: if the new memset would depend on anything other than itself,
the transformation isn't correct.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@237874 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index a90204d26bd874d619b71323940c4bc5d30b915d..66d6ac6f3a0990ab37cfbd52601b2da434ef1033 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -859,6 +859,13 @@ bool MemCpyOpt::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
   if (MemSet->getDest() != MemCpy->getDest())
     return false;
 
+  // Check that there are no other dependencies on the memset destination.
+  MemDepResult DstDepInfo =
+      MD->getPointerDependencyFrom(AliasAnalysis::getLocationForDest(MemSet),
+                                   false, MemCpy, MemCpy->getParent());
+  if (DstDepInfo.getInst() != MemSet)
+    return false;
+
   // Use the same i8* dest as the memcpy, killing the memset dest if different.
   Value *Dest = MemCpy->getRawDest();
   Value *DestSize = MemSet->getLength();
@@ -874,7 +881,7 @@ bool MemCpyOpt::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
     if (ConstantInt *SrcSizeC = dyn_cast<ConstantInt>(SrcSize))
       Align = MinAlign(SrcSizeC->getZExtValue(), DestAlign);
 
-  IRBuilder<> Builder(MemSet);
+  IRBuilder<> Builder(MemCpy);
 
   // If the sizes have different types, zext the smaller one.
   if (DestSize->getType() != SrcSize->getType()) {

diff --git a/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll b/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
index 25411f2698278684bbee8b2c2ba0e155a95a5c1d..993c8a1bb2d046d9f325efc1573965c0ed4f64d8 100644 (file)
--- a/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
+++ b/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
@@ -1,4 +1,4 @@
-; RUN: opt -memcpyopt -S %s | FileCheck %s
+; RUN: opt -basicaa -memcpyopt -S %s | FileCheck %s
 
 target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
 
@@ -140,6 +140,26 @@ define i8 @test_intermediate_read(i8* %a, i8* %b) #0 {
   ret i8 %r
 }
 
+%struct = type { [8 x i8], [8 x i8] }
+
+; CHECK-LABEL: define void @test_intermediate_write
+; CHECK-NEXT: %a = alloca %struct
+; CHECK-NEXT: %a0 = getelementptr %struct, %struct* %a, i32 0, i32 0, i32 0
+; CHECK-NEXT: %a1 = getelementptr %struct, %struct* %a, i32 0, i32 1, i32 0
+; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %a0, i8 0, i64 16, i32 1, i1 false)
+; CHECK-NEXT: store i8 1, i8* %a1
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a0, i8* %b, i64 8, i32 1, i1 false)
+; CHECK-NEXT: ret void
+define void @test_intermediate_write(i8* %b) #0 {
+  %a = alloca %struct
+  %a0 = getelementptr %struct, %struct* %a, i32 0, i32 0, i32 0
+  %a1 = getelementptr %struct, %struct* %a, i32 0, i32 1, i32 0
+  call void @llvm.memset.p0i8.i64(i8* %a0, i8 0, i64 16, i32 1, i1 false)
+  store i8 1, i8* %a1
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a0, i8* %b, i64 8, i32 1, i1 false)
+  ret void
+}
+
 declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i32, i1)
 declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1)