Fix a potential bug in r183584.

  r183584 tries to derive some info from the code *AFTER* a call and apply
these derived info to the code *BEFORE* the call, which is not always safe
as the call in question may never return, and in this case, the derived
info is invalid.

  Thank Duncan for pointing out this potential bug.

rdar://14073661

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183606 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 8600c9ebf7931dcfb75bb9ad7f58000da9396543..c3259254f8b4b2c75ea2ff39024e102e189f1827 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -626,10 +626,14 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
       return false;
 
     Type *StructTy = cast<PointerType>(A->getType())->getElementType();
-    // If StructTy is an opaque type, it should have at least <cpyLen> bytes,
-    // as implified by the copy-instruction.
-    uint64_t destSize = StructTy->isSized() ?
-                        TD->getTypeAllocSize(StructTy) : cpyLen;
+    if (!StructTy->isSized()) {
+      // The call may never return and hence the copy-instruction may never
+      // be executed, and therefore it's not safe to say "the destination
+      // has at least <cpyLen> bytes, as implied by the copy-instruction",
+      return false;
+    }
+
+    uint64_t destSize = TD->getTypeAllocSize(StructTy);
     if (destSize < srcSize)
       return false;
   } else {

diff --git a/test/Transforms/MemCpyOpt/memcpy.ll b/test/Transforms/MemCpyOpt/memcpy.ll
index fa768712ec937c2ce9475941a22b946f19db30f2..148623a6ce71d6f8dcc36f9ddea5e31c90ce9c0c 100644 (file)
--- a/test/Transforms/MemCpyOpt/memcpy.ll
+++ b/test/Transforms/MemCpyOpt/memcpy.ll
@@ -185,28 +185,6 @@ define void @test10(%opaque* noalias nocapture sret %x, i32 %y) {
   ret void
 }
 
-; Test11 is similar to test10 except that the instruction "store i32 %y, i32* %a"
-; before the call-site is deleted. MemCopyOpt is able to optimize this snippet into
-;
-;  %x1 = bitcast %opaque* %x to i32*
-;  call void @foo(i32* noalias nocapture %x1)
-;  ret void
-; 
-
-define void @test11(%opaque* noalias nocapture sret %x, i32 %y) {
-; CHECK: test11
-; CHECK: %x1 = bitcast %opaque* %x to i32*
-; CHECK: call void @foo(i32* noalias nocapture %x1)
-; CHECK: ret void
-
-  %a = alloca i32, align 4
-  call void @foo(i32* noalias nocapture %a)
-  %c = load i32* %a
-  %d = bitcast %opaque* %x to i32*
-  store i32 %c, i32* %d
-  ret void
-}
-
 declare void @f1(%struct.big* sret)
 declare void @f2(%struct.big*)