[llvm-commits] [llvm] r53932 - in /llvm/trunk: lib/CodeGen/SelectionDAG/DAGCombiner.cpp test/CodeGen/X86/2008-07-22-CombinerCrash.ll

Evan Cheng evan.cheng at apple.com
Tue Jul 22 13:42:56 PDT 2008


Author: evancheng
Date: Tue Jul 22 15:42:56 2008
New Revision: 53932

URL: http://llvm.org/viewvc/llvm-project?rev=53932&view=rev
Log:
Fix pr2566: incorrect assumption about bit_convert. It doesn't not have to output a vector value. Patch by Nicolas Capens!

Added:
    llvm/trunk/test/CodeGen/X86/2008-07-22-CombinerCrash.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=53932&r1=53931&r2=53932&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Tue Jul 22 15:42:56 2008
@@ -4953,7 +4953,8 @@
     // look though conversions that change things like v4f32 to v2f64.
     if (V->getOpcode() == ISD::BIT_CONVERT) {
       SDOperand ConvInput = V->getOperand(0);
-      if (ConvInput.getValueType().getVectorNumElements() == NumElts)
+      if (ConvInput.getValueType().isVector() &&
+          ConvInput.getValueType().getVectorNumElements() == NumElts)
         V = ConvInput.Val;
     }
 

Added: llvm/trunk/test/CodeGen/X86/2008-07-22-CombinerCrash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2008-07-22-CombinerCrash.ll?rev=53932&view=auto

==============================================================================
--- llvm/trunk/test/CodeGen/X86/2008-07-22-CombinerCrash.ll (added)
+++ llvm/trunk/test/CodeGen/X86/2008-07-22-CombinerCrash.ll Tue Jul 22 15:42:56 2008
@@ -0,0 +1,16 @@
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2
+; PR2566
+
+external global i16		; <i16*>:0 [#uses=1]
+external global <4 x i16>		; <<4 x i16>*>:1 [#uses=1]
+
+declare void @abort()
+
+define void @t() nounwind {
+	load i16* @0		; <i16>:1 [#uses=1]
+	zext i16 %1 to i64		; <i64>:2 [#uses=1]
+	bitcast i64 %2 to <4 x i16>		; <<4 x i16>>:3 [#uses=1]
+	shufflevector <4 x i16> %3, <4 x i16> undef, <4 x i32> zeroinitializer		; <<4 x i16>>:4 [#uses=1]
+	store <4 x i16> %4, <4 x i16>* @1
+	ret void
+}





More information about the llvm-commits mailing list