This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Many vectorization failures on my x86_64 lto bootstrap.
- From: Uros Bizjak <ubizjak at gmail dot com>
- To: Toon Moene <toon at moene dot org>
- Cc: GCC Development <gcc at gcc dot gnu dot org>
- Date: Sat, 10 Nov 2012 13:08:11 +0100
- Subject: Re: Many vectorization failures on my x86_64 lto bootstrap.
Hello!
> Compare the log file of the test runs from my lto bootstrap:
>
> http://gcc.gnu.org/ml/gcc-testresults/2012-11/msg00832.html
Platform: x86_64-unknown-linux-gnu
configure flags: ... --with-arch=native --with-tune=native
> to that of a "normal" x86_64 bootstrap:
>
> http://gcc.gnu.org/ml/gcc-testresults/2012-11/msg00790.html
Platform: x86_64-unknown-linux-gnu
configure flags: ... --with-arch=core2 --with-cpu=core2
> What log files, if any, would be helpful to hunt this down ?
You have AVX capable machine, and --with-arch=native enables 256 bit
vectorization behind your back.
The vectorizer testcases are not prepared for an internal rerun of
vectorization and - since each internal report merges it own dump to
the report -- scan-tree-dump-times fail.
So, testsuite issue.
Uros.