[LLVMbugs] [Bug 16834] New: Clang Generates Incorrect Code With Certain Arch and Optimization Combinations

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=16834

            Bug ID: 16834
           Summary: Clang Generates Incorrect Code With Certain Arch and
                    Optimization Combinations
           Product: clang
           Version: trunk
          Hardware: Macintosh
                OS: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: mathpup at gmail.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Created attachment 11007
  --> http://llvm.org/bugs/attachment.cgi?id=11007&action=edit
Test case

With the latest Clang/LLVM from SVN, the sample program abc.cpp produces
incorrect output when compiled with the optimization flags

-O3 -march=corei7-avx -fslp-vectorize -fslp-vectorize-aggressive

The problem goes away if -fslp-vectorize-aggressive is removed, and it also
goes away if -march=corei7-avx.

I am attaching sample C++ source for abc.cpp. I tried to narrow down this test
case, but this seems to be as simple as I can make it while sitting exhibiting
the   problem. The sample program is based on code used in Qt 5 for working
with Bezier curves.

Correct output:
[10, 20], [30, 40], [50, 60], [70, 80]
[10, 20], [20, 30], [30, 40], [40, 50]
[40, 50], [50, 60], [60, 70], [70, 80]

Incorrect output produced as described above:
[10, 20], [30, 40], [50, 60], [70, 80]
[10, 20], [20, 30], [30, 40], [22.5, 50]
[22.5, 50], [45, 60], [60, 70], [70, 80]

I did attempt to use bugpoint, but I could not figure out how to produce any
output that seemed useful.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20130808/3cccf65b/attachment.html>