[LLVMbugs] [Bug 6833] New: Combination of GVN, ScalarRepl, and other passes misoptimizes code

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=6833

           Summary: Combination of GVN, ScalarRepl, and other passes
                    misoptimizes code
           Product: libraries
           Version: trunk
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: Interprocedural Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: criswell at uiuc.edu
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=4677)
 --> (http://llvm.org/bugs/attachment.cgi?id=4677)
Input assembly which triggers the bug

This bug was found in the release_27 branch of LLVM using Mac OS X (Snow
Leopard).

The following command reproduces the problem:

opt -preverify -domtree -verify -scalarrepl -indvars -internalize -inline -gvn 
bugpoint-reduced-simplified.bc -f -o /tmp/f.bc

The resulting bitcode file creates a constant expression with the wrong index
(the index is extremely large).

Attached are LLVM assembly files of the input and output of the above command;
notice that the GEP constant expression for plus8dir is incorrect.

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.