[llvm-dev] Maintaining TBAA information in stack coloring pass

Hal Finkel via llvm-dev llvm-dev at lists.llvm.org
Thu Jul 27 05:12:44 PDT 2017

On 07/27/2017 02:18 AM, Hiroshi 7 Inoue via llvm-dev wrote:
> Dear all,
> While investigating recent failures in bootstrap on ppc64le buildbot, 
> I found that the current stack coloring pass does not regard TBAA 
> information when merging two allocas of different types (Actually 
> there is a FIXME comment on handling of TBAA information in 
> StackColoring.cpp)
> Although the current failures are observed only on ppc64le, the cause 
> of failures seems to be a platform-neutral issue and other platforms 
> can suffer from similar problems.
> I submitted a work-in-progress patch in Phabricator to resolve this 
> issue. https://reviews.llvm.org/D35907
> Any suggestions or comments would be truly appreciated.

I'll take a look. Stack coloring already has a hack that updates the 
(otherwise frozen) IR so that using BasicAA in the backend continues to 
work afterward. Dealing with this issue is also certainly necessary.


> Regards,
> Hiroshi
> -----
> Hiroshi Inoue <inouehrs at jp.ibm.com>
> IBM Research - Tokyo
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170727/d821d58f/attachment.html>

More information about the llvm-dev mailing list