[llvm-commits] [llvm] r50444 - in /llvm/trunk: lib/Transforms/Scalar/MemCpyOptimizer.cpp test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll

[Owen Anderson]

Author: resistor
Date: Tue Apr 29 16:26:06 2008
New Revision: 50444

URL: http://llvm.org/viewvc/llvm-project?rev=50444&view=rev
Log:
Fix a bug in memcpyopt where the memcpy-memcpy transform was never being applied because 
we were checking for it in the wrong order.  This caused a miscompilation because the
return slot optimization assumes that the call it is dealing with is NOT a memcpy.

Added:
    llvm/trunk/test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll
Modified:
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=50444&r1=50443&r2=50444&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Tue Apr 29 16:26:06 2008
@@ -615,10 +615,12 @@
   if (dep == MemoryDependenceAnalysis::None ||
       dep == MemoryDependenceAnalysis::NonLocal)
     return false;
-  else if (CallInst* C = dyn_cast<CallInst>(dep))
-    return performCallSlotOptzn(M, C);
-  else if (!isa<MemCpyInst>(dep))
-    return false;
+  else if (!isa<MemCpyInst>(dep)) {
+    if (CallInst* C = dyn_cast<CallInst>(dep))
+      return performCallSlotOptzn(M, C);
+    else
+      return false;
+  }
   
   MemCpyInst* MDep = cast<MemCpyInst>(dep);
   

Added: llvm/trunk/test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll?rev=50444&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll (added)
+++ llvm/trunk/test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll Tue Apr 29 16:26:06 2008
@@ -0,0 +1,17 @@
+; RUN: llvm-as < %s | opt -memcpyopt | llvm-dis | grep {call.*memcpy.*agg.result}
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target triple = "i386-apple-darwin8"
+ at x = external global { x86_fp80, x86_fp80 }		; <{ x86_fp80, x86_fp80 }*> [#uses=1]
+
+define void @foo({ x86_fp80, x86_fp80 }* noalias sret %agg.result) nounwind  {
+entry:
+	%x.0 = alloca { x86_fp80, x86_fp80 }		; <{ x86_fp80, x86_fp80 }*> [#uses=1]
+	%x.01 = bitcast { x86_fp80, x86_fp80 }* %x.0 to i8*		; <i8*> [#uses=2]
+	call void @llvm.memcpy.i32( i8* %x.01, i8* bitcast ({ x86_fp80, x86_fp80 }* @x to i8*), i32 32, i32 16 )
+	%agg.result2 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8*		; <i8*> [#uses=1]
+	call void @llvm.memcpy.i32( i8* %agg.result2, i8* %x.01, i32 32, i32 16 )
+	ret void
+}
+
+declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind