New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tensorflow build fails with -mavx512f #10689

Closed
Zehvogel opened this Issue Jun 13, 2017 · 5 comments

Comments

Projects
None yet
7 participants
@Zehvogel

Zehvogel commented Jun 13, 2017

System information

  • Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Fedora 25
  • TensorFlow installed from (source or binary): source
  • TensorFlow version (use command below): latest commit f58e6ce , happens with older versions as well (23caaa5, f48673b, 9a15e0a, 3c4cb08 to only name a few)
  • Bazel version (if compiling from source): 0.4.5
  • CUDA/cuDNN version: none
  • CPU model and memory: XEON PHI 7250, 96GiB ram
  • Exact command to reproduce:
    ./configure
    specify optimization flags: -mavx512f
    bazel build --config=opt --verbose_failures //tensorflow/tools/pip_package:build_pip_package

Describe the problem

Building Tensorflow from source fails when avx512 instructions are activated.
After managing to compile most of the source the build fails somewhere in the eigen part of the code.

builderror.txt

Don't let yourself get thrown off by the "MKL" in some path names, its just the name of the virtualenv, for this report MKL was turned off in the configure.
See also similar Issue #9849

@aselle

This comment has been minimized.

Show comment
Hide comment
@aselle

aselle Jun 13, 2017

Member

Also seems to be like this #7530.
@benoitsteiner, could you take a look?

Member

aselle commented Jun 13, 2017

Also seems to be like this #7530.
@benoitsteiner, could you take a look?

@cYclonEK

This comment has been minimized.

Show comment
Hide comment
@cYclonEK

cYclonEK Jun 19, 2017

I guess my issue is the same.
Build Log.txt
On 16.04.2 with GCC 7.1.0
I did not use -mavx512f but -march=native (which enables those kind of architecture optimizations I guess).
And same issue with and without MKL.

cYclonEK commented Jun 19, 2017

I guess my issue is the same.
Build Log.txt
On 16.04.2 with GCC 7.1.0
I did not use -mavx512f but -march=native (which enables those kind of architecture optimizations I guess).
And same issue with and without MKL.

@yvdriess

This comment has been minimized.

Show comment
Hide comment
@yvdriess

yvdriess Sep 4, 2017

I can confirm the same issue with latest master ( 1209491 ) on a Skylake Xeon CPU, using either -march=native or -mavx512f, with either GCC 6 or 7.

yvdriess commented Sep 4, 2017

I can confirm the same issue with latest master ( 1209491 ) on a Skylake Xeon CPU, using either -march=native or -mavx512f, with either GCC 6 or 7.

@quaeler

This comment has been minimized.

Show comment
Hide comment
@quaeler

quaeler Sep 5, 2017

Contributor

Seems possibly related to #12781 also.

Contributor

quaeler commented Sep 5, 2017

Seems possibly related to #12781 also.

@tensorflowbutler

This comment has been minimized.

Show comment
Hide comment
@tensorflowbutler

tensorflowbutler Dec 4, 2017

Member

This issue is automatically closed due to lack of activity. Please re-open if this is still an issue for you. Thanks!

Member

tensorflowbutler commented Dec 4, 2017

This issue is automatically closed due to lack of activity. Please re-open if this is still an issue for you. Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment