{numlib}[GCC/12.2.0] FlexiBLAS v3.2.1, BLIS v0.9.0, OpenBLAS v0.3.21 #16602

boegel · 2022-11-09T20:57:48Z

(created using eb --new-pr)
requires:

…2.0.eb, OpenBLAS-0.3.21-GCC-12.2.0.eb

boegel · 2022-11-09T21:51:21Z

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
node3135.skitty.os - Linux RHEL 8.4, x86_64, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz (skylake_avx512), Python 3.6.8
See https://gist.github.com/ec71922a9a3838ee5b8110939c456201 for a full test report.

boegel · 2022-11-09T22:01:41Z

@boegelbot please test @ generoso

boegelbot · 2022-11-09T22:05:14Z

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

PR test command 'EB_PR=16602 EB_ARGS= EB_CONTAINER= /opt/software/slurm/bin/sbatch --job-name test_PR_16602 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

exit code: 0
output:

Submitted batch job 9486

Test results coming soon (I hope)...

- notification for comment with ID 1309434846 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 · 2022-11-09T22:17:18Z

Test report by @boegelbot
FAILED
Build succeeded for 1 out of 3 (3 easyconfigs in total)
cns2 - 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/cd2f7714872ba15083eec16163913b58 for a full test report.

			-->   LAPACK TESTING SUMMARY  <--
SUMMARY             	nb test run 	numerical error   	other error  
================   	===========	=================	================  
REAL             	1317867		0	(0.000%)	0	(0.000%)	
DOUBLE PRECISION	1318689		0	(0.000%)	0	(0.000%)	
COMPLEX          	769171		180	(0.023%)	0	(0.000%)	
COMPLEX16         	769982		99	(0.013%)	0	(0.000%)	

--> ALL PRECISIONS	4175709		279	(0.007%)	0	(0.000%)

branfosj · 2022-11-09T22:40:05Z

Test report by @branfosj
SUCCESS
Build succeeded for 4 out of 4 (3 easyconfigs in total)
bear-pg0105u36b.bear.cluster - Linux RHEL 8.6, x86_64, Intel(R) Xeon(R) Platinum 8360Y CPU @ 2.40GHz (icelake), Python 3.6.8
See https://gist.github.com/2a2bd588a730046103125c7240737c04 for a full test report.

            -->   LAPACK TESTING SUMMARY  <--
SUMMARY                 nb test run     numerical error     other error  
================    =========== =================   ================  
REAL                1317867     0   (0.000%)    0   (0.000%)    
DOUBLE PRECISION    1317597     10  (0.001%)    0   (0.000%)    
COMPLEX             774343      24  (0.003%)    0   (0.000%)    
COMPLEX16           776246      14  (0.002%)    0   (0.000%)    

--> ALL PRECISIONS  4186053     48  (0.001%)    0   (0.000%)

…asyconfigs into 20221109215743_new_pr_FlexiBLAS321

boegel · 2022-11-09T22:42:27Z

@boegelbot please test @ jsc-zen2

branfosj · 2022-11-09T22:42:52Z

Test report by @branfosj
SUCCESS
Build succeeded for 4 out of 4 (3 easyconfigs in total)
bear-pg0211u03a.bear.cluster - Linux Ubuntu 20.04.2 LTS (Focal Fossa), x86_64, Intel(R) Xeon(R) Gold 6248 CPU @ 2.50GHz (cascadelake), Python 3.8.5
See https://gist.github.com/5817753b642b309a3e064d66f9105d48 for a full test report.

            -->   LAPACK TESTING SUMMARY  <--
SUMMARY                 nb test run     numerical error     other error  
================    =========== =================   ================  
REAL                1317867     0   (0.000%)    0   (0.000%)    
DOUBLE PRECISION    1317597     9   (0.001%)    0   (0.000%)    
COMPLEX             774343      22  (0.003%)    0   (0.000%)    
COMPLEX16           776246      14  (0.002%)    0   (0.000%)    

--> ALL PRECISIONS  4186053     45  (0.001%)    0   (0.000%)

boegelbot · 2022-11-09T22:47:10Z

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

PR test command 'EB_PR=16602 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_16602 --ntasks=8 ~/boegelbot/eb_from_pr_upload_jsc-zen2.sh' executed!

exit code: 0
output:

Submitted batch job 1759

Test results coming soon (I hope)...

- notification for comment with ID 1309485571 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 · 2022-11-09T23:12:40Z

Test report by @boegelbot
FAILED
Build succeeded for 6 out of 8 (3 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/1f003936543cd9139d2053791e05ce93 for a full test report.

			-->   LAPACK TESTING SUMMARY  <--
SUMMARY             	nb test run 	numerical error   	other error  
================   	===========	=================	================  
REAL             	1317867		0	(0.000%)	0	(0.000%)	
DOUBLE PRECISION	1318689		0	(0.000%)	0	(0.000%)	
COMPLEX          	769171		180	(0.023%)	0	(0.000%)	
COMPLEX16         	769982		99	(0.013%)	0	(0.000%)	

--> ALL PRECISIONS	4175709		279	(0.007%)	0	(0.000%)

boegel · 2022-11-10T06:54:28Z

~~Hmm, I need to check whether the patch from #16406 is also required here...~~

edit:
All patches (from #16406 + #16510) are already used here, yet we're still seeing more failing tests than expected...

The GCCcore/12.2.0 installations on both generoso and jsc-zen2 were also rebuilt with the patch included in #16411, so that's fine too.

branfosj · 2022-11-10T09:09:22Z

Test report by @branfosj
FAILED
Build succeeded for 8 out of 10 (3 easyconfigs in total)
bear-pg0210u06a.bear.cluster - Linux RHEL 8.6, x86_64, Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz (broadwell), Python 3.6.8
See https://gist.github.com/5b24b879588401a2eb45da93a1930281 for a full test report.

            -->   LAPACK TESTING SUMMARY  <--
SUMMARY                 nb test run     numerical error     other error
================    =========== =================   ================
REAL                1317867     0   (0.000%)    0   (0.000%)
DOUBLE PRECISION    1318689     0   (0.000%)    0   (0.000%)
COMPLEX             769171      180 (0.023%)    0   (0.000%)
COMPLEX16           769982      99  (0.013%)    0   (0.000%)

--> ALL PRECISIONS  4175709     279 (0.007%)    0   (0.000%)

boegel · 2022-11-10T10:23:47Z

Test report by @boegel
FAILED
Build succeeded for 29 out of 31 (3 easyconfigs in total)
node2672.swalot.os - Linux RHEL 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/bf649421627bb232fd20e95beb08ef85 for a full test report.

bartoldeman · 2022-11-10T13:09:58Z

A few of these tests are failing due to picky LAPACK tests.
There are a few ways around this:

Increase threshold (duh)
Patch LAPACK a bit more with some brackets, see Many test failures when compiled with -march=haswell -ftree-vectorize Reference-LAPACK/lapack#732 (two of the files are already patched but a few more changes were needed with the updated LAPACK in OpenBLAS 0.3.21)
Disable FMA (-ffp-contract=off) or vectorization for LAPACK (FMA is cheaper to disable here, since vectorization can give more speedup).
I looked a bit more into the why and it's really about complex multiplication:
(a+bi)*(c+di)=(a*c-b*d)+i*(a*d+b*c)
without FMA this always gives the same answer numerically, and for complex conjugates (a+bi)*(a-bi) the resulting imaginary part is exactly zero.

But with FMAs there are 4 permutations possible (e.g. a*d+b*c -> FMA(a,d,b*c) or FMA(b,c,a*d)) and GCC sometimes uses one variation in the vectorized loop and another one in the unvectorized loop (used for the tail ends).

boegel · 2022-11-10T13:10:13Z

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
fair-mastodon-c6g-2xlarge-0001 - Linux Rocky Linux 8.5, AArch64, ARM UNKNOWN (graviton2), Python 3.6.8
See https://gist.github.com/2cedb0f54950219e8f9b1f87a0fa5353 for a full test report.

SebastianAchilles · 2022-11-10T13:51:55Z

Test report by @SebastianAchilles
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
jrc0871.jureca - Linux Rocky Linux 8.6, AArch64, ARM UNKNOWN (graviton2), 2 x NVIDIA NVIDIA A100-PCIE-40GB, 515.65.01, Python 3.6.8
See https://gist.github.com/f278dc857456da4b99886b89a9dc0636 for a full test report.

For some reason (GCC bug report pending) -ffp-contract=off is not sufficient for GCC 12, so we also need -mno-fma, implemented via a GCC pragma.

Adjust disable-fma-in-cscal-zscal.patch for GCC 12

boegel · 2022-11-11T17:04:42Z

@boegelbot please test @ generoso

boegelbot · 2022-11-11T17:05:08Z

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

PR test command 'EB_PR=16602 EB_ARGS= EB_CONTAINER= /opt/software/slurm/bin/sbatch --job-name test_PR_16602 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

exit code: 0
output:

Submitted batch job 9501

Test results coming soon (I hope)...

- notification for comment with ID 1311952862 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).

bartoldeman · 2022-11-11T17:25:20Z

The OpenBLAS FMA patch was adjusted because of a new issue with GCC 12 (not present in 11.3)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107647
https://godbolt.org/z/aWh6d1E4E

boegel · 2022-11-11T17:39:04Z

Test report by @boegel
SUCCESS
Build succeeded for 4 out of 4 (3 easyconfigs in total)
node3508.doduo.os - Linux RHEL 8.4, x86_64, AMD EPYC 7552 48-Core Processor (zen2), Python 3.6.8
See https://gist.github.com/87205938d1f05d7ad3f23c0b6980f20d for a full test report.

boegelbot · 2022-11-11T17:41:29Z

Test report by @boegelbot
SUCCESS
Build succeeded for 4 out of 4 (3 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/e4406121d3a11cc70b38a12f663624ad for a full test report.

boegel · 2022-11-11T17:43:21Z

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
fair-mastodon-c6g-2xlarge-0001 - Linux Rocky Linux 8.5, AArch64, ARM UNKNOWN (graviton2), Python 3.6.8
See https://gist.github.com/1d49696db11b04030f75b594729a52a9 for a full test report.

boegel · 2022-11-11T17:44:15Z

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
node2620.swalot.os - Linux RHEL 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/b4ecf824e906305dd0f59d9df1cf652c for a full test report.

boegel · 2022-11-11T17:53:07Z

@boegelbot please test @ jsc-zen2

boegelbot · 2022-11-11T17:57:08Z

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

PR test command 'EB_PR=16602 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_16602 --ntasks=8 ~/boegelbot/eb_from_pr_upload_jsc-zen2.sh' executed!

exit code: 0
output:

Submitted batch job 1765

Test results coming soon (I hope)...

- notification for comment with ID 1311996431 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 · 2022-11-11T18:09:17Z

Test report by @boegelbot
SUCCESS
Build succeeded for 3 out of 3 (3 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/d59278c88854b146105619e1c5f5607b for a full test report.

boegel · 2022-11-11T19:59:24Z

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
easybuild2.novalocal - Linux CentOS Stream 8, POWER, IBM pSeries (emulated by qemu) (power9le), Python 3.6.8
See https://gist.github.com/57c0340c1ecd73329326c973738b2858 for a full test report.

branfosj · 2022-11-12T07:44:19Z

Test report by @branfosj
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
bear-pg0105u36b.bear.cluster - Linux RHEL 8.7, x86_64, Intel(R) Xeon(R) Platinum 8360Y CPU @ 2.40GHz (icelake), Python 3.6.8
See https://gist.github.com/fe7d5404f42a1e6a318ee58ffb0f07d3 for a full test report.

branfosj

lgtm

branfosj · 2022-11-12T07:48:00Z

Test report by @branfosj
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in total)
bear-pg0211u08b.bear.cluster - Linux Ubuntu 20.04.2 LTS (Focal Fossa), x86_64, Intel(R) Xeon(R) Gold 6248 CPU @ 2.50GHz (cascadelake), Python 3.8.5
See https://gist.github.com/53b7cb28697ce3cfffff2bfcbee974d2 for a full test report.

branfosj · 2022-11-12T07:48:26Z

Going in, thanks @boegel!

adding easyconfigs: FlexiBLAS-3.2.1-GCC-12.2.0.eb, BLIS-0.9.0-GCC-12.…

b0f5a08

…2.0.eb, OpenBLAS-0.3.21-GCC-12.2.0.eb

boegel added the update label Nov 9, 2022

boegel added this to the next release (4.6.3?) milestone Nov 9, 2022

boegel mentioned this pull request Nov 9, 2022

{toolchain} foss/2022.10 #16607

Merged

Merge branch 'develop' of https://github.com/easybuilders/easybuild-e…

c40ccb9

…asyconfigs into 20221109215743_new_pr_FlexiBLAS321

easybuilders deleted a comment from boegelbot Nov 9, 2022

bartoldeman and others added 2 commits November 11, 2022 16:34

Adjust disable-fma-in-cscal-zscal.patch for GCC 12

94bb82b

For some reason (GCC bug report pending) -ffp-contract=off is not sufficient for GCC 12, so we also need -mno-fma, implemented via a GCC pragma.

Merge pull request #89 from ComputeCanada/openblas-lapack-fma-workaround

3fae9c7

Adjust disable-fma-in-cscal-zscal.patch for GCC 12

branfosj approved these changes Nov 12, 2022

View reviewed changes

branfosj merged commit d77ec0a into easybuilders:develop Nov 12, 2022

boegel deleted the 20221109215743_new_pr_FlexiBLAS321 branch November 12, 2022 17:12

{numlib}[GCC/12.2.0] FlexiBLAS v3.2.1, BLIS v0.9.0, OpenBLAS v0.3.21 #16602

{numlib}[GCC/12.2.0] FlexiBLAS v3.2.1, BLIS v0.9.0, OpenBLAS v0.3.21 #16602

Uh oh!

Conversation

boegel commented Nov 9, 2022

Uh oh!

boegel commented Nov 9, 2022

Uh oh!

boegel commented Nov 9, 2022

Uh oh!

boegelbot commented Nov 9, 2022

Uh oh!

boegelbot commented Nov 9, 2022 • edited by boegel Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

branfosj commented Nov 9, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

boegel commented Nov 9, 2022

Uh oh!

branfosj commented Nov 9, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

boegelbot commented Nov 9, 2022

Uh oh!

boegelbot commented Nov 9, 2022 • edited by boegel Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

boegel commented Nov 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

branfosj commented Nov 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

boegel commented Nov 10, 2022

Uh oh!

bartoldeman commented Nov 10, 2022

Uh oh!

boegel commented Nov 10, 2022

Uh oh!

SebastianAchilles commented Nov 10, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

boegelbot commented Nov 11, 2022

Uh oh!

bartoldeman commented Nov 11, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

boegelbot commented Nov 11, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

boegelbot commented Nov 11, 2022

Uh oh!

boegelbot commented Nov 11, 2022

Uh oh!

boegel commented Nov 11, 2022

Uh oh!

branfosj commented Nov 12, 2022

Uh oh!

branfosj left a comment

Choose a reason for hiding this comment

Uh oh!

branfosj commented Nov 12, 2022

Uh oh!

branfosj commented Nov 12, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

boegelbot commented Nov 9, 2022 •

edited by boegel

Loading

branfosj commented Nov 9, 2022 •

edited

Loading

branfosj commented Nov 9, 2022 •

edited

Loading

boegelbot commented Nov 9, 2022 •

edited by boegel

Loading

boegel commented Nov 10, 2022 •

edited

Loading

branfosj commented Nov 10, 2022 •

edited

Loading