[llvm-commits] [llvm] r74792 - in /llvm/trunk: lib/Analysis/ScalarEvolution.cpp test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll

Nick Lewycky nicholas at mxc.ca
Sat Jul 4 10:24:53 PDT 2009


Author: nicholas
Date: Sat Jul  4 12:24:52 2009
New Revision: 74792

URL: http://llvm.org/viewvc/llvm-project?rev=74792&view=rev
Log:
When comparing constants, consider a less wide constant to be "less complex"
than a wider one, before trying to compare their contents which will crash
if their sizes are different.

Added:
    llvm/trunk/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll
Modified:
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=74792&r1=74791&r2=74792&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Sat Jul  4 12:24:52 2009
@@ -495,6 +495,8 @@
       // Compare constant values.
       if (const SCEVConstant *LC = dyn_cast<SCEVConstant>(LHS)) {
         const SCEVConstant *RC = cast<SCEVConstant>(RHS);
+        if (LC->getValue()->getBitWidth() != RC->getValue()->getBitWidth())
+          return LC->getValue()->getBitWidth() < RC->getValue()->getBitWidth();
         return LC->getValue()->getValue().ult(RC->getValue()->getValue());
       }
 

Added: llvm/trunk/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll?rev=74792&view=auto

==============================================================================
--- llvm/trunk/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll (added)
+++ llvm/trunk/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll Sat Jul  4 12:24:52 2009
@@ -0,0 +1,16 @@
+; RUN: llvm-as < %s | opt -analyze -scalar-evolution
+; PR4501
+
+define void @test() {
+entry:
+        %0 = load i16* undef, align 1
+        %1 = lshr i16 %0, 8
+        %2 = and i16 %1, 3
+        %3 = zext i16 %2 to i32
+        %4 = load i8* undef, align 1
+        %5 = lshr i8 %4, 4
+        %6 = and i8 %5, 1
+        %7 = zext i8 %6 to i32
+        %t1 = add i32 %3, %7
+        ret void
+}





More information about the llvm-commits mailing list