[llvm-bugs] [Bug 36883] New: EarlyCSE MemorySSA breaks vpx_idct32x32_1024_add_neon from libvpx when targeting armv7

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=36883

            Bug ID: 36883
           Summary: EarlyCSE MemorySSA breaks vpx_idct32x32_1024_add_neon
                    from libvpx when targeting armv7
           Product: libraries
           Version: 5.0
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: martin at martin.st
                CC: llvm-bugs at lists.llvm.org

Since SVN r306477 of LLVM ([EarlyCSE][MemorySSA] Enable MemorySSA in
function-simplification pass of EarlyCSE.; present in the 5.0 and 6.0
releases), the vpx_idct32x32_1024_add_neon (intrinsics based) function of
libvpx gets broken/misoptimized. Not yet sure if this is an LLVM bug or some
other wrong with the function that allows the optimizer to break it.

To work around the issue, one can build with "-mllvm
-enable-npm-earlycse-memssa=false -mllvm -enable-earlycse-memssa=false".

The same issue is reported for libvpx at
https://bugs.chromium.org/p/webm/issues/detail?id=1510.

-- 
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/20180323/7b0fc4f9/attachment.html>