Zachary Ware added the comment:
The buildbots appear to be happy with this, so I'm closing the issue. The
intel-ubuntu-icc Non-Debug builders each had their first green build after this
changeset. The OSX ICC builder still has some issues, but they seem to be OSX
specific (see issue25589).
Roundup Robot added the comment:
New changeset d7b5c2f99a99 by Zachary Ware in branch '2.7':
Issue #25827: Add support for ICC to configure
https://hg.python.org/cpython/rev/d7b5c2f99a99
New changeset c5e419464585 by Zachary Ware in branch '3.5':
Issue #25827: Add support for ICC to configure
ht
Zachary Ware added the comment:
After testing, I think I like '-fp-model strict' better in CFLAGS_NODIST. On
Ubuntu, `CC=gcc ./python -m test.test_distutils` passes when the flag is in
CFLAGS_NODIST, but fails when it's in BASECFLAGS. On OSX, it fails both ways,
due to deployment target issu
Alecsandru Patrascu added the comment:
Thank you David for including me in this issue.
On ICC, when executing the instrumented applications, it will generate dynamic
information file that has a unique name and .dyn suffix. From my previous
experience with GCC and CLANG PGO patch (issue24915),
R. David Murray added the comment:
This patch looks correct to me, but I don't have much experience with
configure. Are the .dyn files icc's profiling output?
I'm adding Alecsandru and Brett as nosy since they worked on the PGO stuff.
I'm also adding Stefan since he's shown interest in icc.
New submission from Zachary Ware:
Attached is a patch that adds support for ICC to configure, making it easier to
build with ICC on Unix and adjusting arguments to better fit ICC.
- Adds '--with-icc' argument to configure, which sets 'CC=icc' and 'CXX=icpc'
- Adds support for ICC PGO
- Prevents
