Skip to content

add patch for BLIS to fix auto-detection of POWER#15826

Merged
jfgrimm merged 1 commit intoeasybuilders:developfrom
Flamefire:20220707144323_new_pr_BLIS081
Jul 8, 2022
Merged

add patch for BLIS to fix auto-detection of POWER#15826
jfgrimm merged 1 commit intoeasybuilders:developfrom
Flamefire:20220707144323_new_pr_BLIS081

Conversation

@Flamefire
Copy link
Copy Markdown
Contributor

@Flamefire Flamefire commented Jul 7, 2022

(created using eb --new-pr)

@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 7, 2022

Test report by @jfgrimm
FAILED
Build succeeded for 4 out of 11 (11 easyconfigs in total)
node022.pri.viking.alces.network - Linux CentOS Linux 7.9.2009, x86_64, Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz (skylake_avx512), Python 3.6.8
See https://gist.github.com/55d27543e3a23b0d02b2d647bb0b6e0d for a full test report.

@boegelbot

This comment was marked as outdated.

Sign in to view
@Flamefire Flamefire force-pushed the 20220707144323_new_pr_BLIS081 branch 2 times, most recently from f61ed6a to 8a0c694 Compare July 7, 2022 14:46
@Flamefire Flamefire force-pushed the 20220707144323_new_pr_BLIS081 branch from 8a0c694 to 5e50cd1 Compare July 7, 2022 15:01
@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 7, 2022

Test report by @jfgrimm
FAILED
Build succeeded for 6 out of 10 (10 easyconfigs in total)
node022.pri.viking.alces.network - Linux CentOS Linux 7.9.2009, x86_64, Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz (skylake_avx512), Python 3.6.8
See https://gist.github.com/d3b672008e0c9fbc04c2c60da0b51b4c for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Flamefire commented Jul 7, 2022

Test report by @Flamefire
SUCCESS
Build succeeded for 10 out of 10 (10 easyconfigs in total)
taurusi8018 - Linux CentOS Linux 7.9.2009, x86_64, AMD EPYC 7352 24-Core Processor (zen2), 8 x NVIDIA NVIDIA A100-SXM4-40GB, 470.57.02, Python 2.7.5
See https://gist.github.com/ca6b0ac0bd3f22f4d795911c428571e9 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Flamefire commented Jul 7, 2022

Test report by @Flamefire
SUCCESS
Build succeeded for 10 out of 10 (10 easyconfigs in total)
taurusa11 - Linux CentOS Linux 7.7.1908, x86_64, Intel(R) Xeon(R) CPU E5-2603 v4 @ 1.70GHz (broadwell), 3 x NVIDIA GeForce GTX 1080 Ti, 460.32.03, Python 2.7.5
See https://gist.github.com/23f746f0845ef4d83a822b0fac88b40f for a full test report.

@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 7, 2022

@boegelbot please test @ jsc-zen2
CORE_CNT=16

@boegelbot
Copy link
Copy Markdown
Collaborator

boegelbot commented Jul 7, 2022

@jfgrimm: Request for testing this PR well received on jsczen2l1.int.jsc-zen2.easybuild-test.cluster

PR test command 'EB_PR=15826 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_15826 --ntasks="16" ~/boegelbot/eb_from_pr_upload_jsc-zen2.sh' executed!

  • exit code: 0
  • output:
Submitted batch job 1376

Test results coming soon (I hope)...

Details

- notification for comment with ID 1178221339 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

@boegelbot
Copy link
Copy Markdown
Collaborator

boegelbot commented Jul 7, 2022

Test report by @boegelbot
SUCCESS
Build succeeded for 10 out of 10 (10 easyconfigs in total)
jsczen2c1.int.jsc-zen2.easybuild-test.cluster - Linux Rocky Linux 8.5, x86_64, AMD EPYC 7742 64-Core Processor (zen2), Python 3.6.8
See https://gist.github.com/4c75c937ba0f2dd60a8a1f3c750193b7 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Flamefire commented Jul 8, 2022

@jfgrimm sorry for the failures, seemingly the versions are not really incremental so some patches I thought would still apply didn't. As can be seen by recent test reports this is now fixed.

@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 8, 2022

@boegelbot please test @ generoso

@boegelbot
Copy link
Copy Markdown
Collaborator

boegelbot commented Jul 8, 2022

@jfgrimm: Request for testing this PR well received on login1

PR test command 'EB_PR=15826 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_15826 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

  • exit code: 0
  • output:
Submitted batch job 8844

Test results coming soon (I hope)...

Details

- notification for comment with ID 1178658211 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 8, 2022

I'm still seeing test failures on my machine:

% blis_<dt><op>_<params>_<stor>      m     k   gflops   resid      result                                               
blis_sgemmt_lnn_rrr                100   100    37.48   5.39e-02   FAILURE                                              
blis_sgemmt_unn_rrr                100   100    40.96   6.05e-02   FAILURE

although I think that's unrelated to the changes in this PR, so shouldn't block this

@boegelbot
Copy link
Copy Markdown
Collaborator

boegelbot commented Jul 8, 2022

Test report by @boegelbot
SUCCESS
Build succeeded for 10 out of 10 (10 easyconfigs in total)
cns1 - Linux Rocky Linux 8.5, x86_64, Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz (haswell), Python 3.6.8
See https://gist.github.com/a1b1b3aedc25642e434d2f7266a838ac for a full test report.

@jfgrimm jfgrimm added this to the next release (4.6.1?) milestone Jul 8, 2022
jfgrimm
jfgrimm approved these changes Jul 8, 2022
View reviewed changes
Copy link
Copy Markdown
Member

@jfgrimm jfgrimm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@jfgrimm
Copy link
Copy Markdown
Member

jfgrimm commented Jul 8, 2022

Going in, thanks @Flamefire!

@jfgrimm jfgrimm merged commit cb610b5 into easybuilders:develop Jul 8, 2022
@Flamefire Flamefire deleted the 20220707144323_new_pr_BLIS081 branch July 8, 2022 14:11
@Flamefire Flamefire mentioned this pull request Jul 21, 2022
Merged
@Flamefire
Copy link
Copy Markdown
Contributor Author

Flamefire commented Jul 25, 2022

Seems like I introduced a bug here for 2.2 and 3.0: They don't support POWER10 so the code doesn't compile on any PPC. Fix included in #15889

This still doesn't fully solve the BLIS build on PPC due to segfaults in 0.9.0 (flame/blis#621) and timeouts during tests in 3.0.1

@boegel boegel changed the title Fix BLIS build on PPC add patch for BLIS to fix auto-detection of POWER Aug 3, 2022
@boegel boegel added bug fix and removed change labels Aug 3, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jfgrimm jfgrimm jfgrimm approved these changes

Assignees

No one assigned

Labels

bug fix enhancement

Projects

None yet

Milestone

4.6.1

Development

Successfully merging this pull request may close these issues.

4 participants

@Flamefire @jfgrimm @boegelbot @boegel