Skip to content
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

nvofbf/2022.07 toolchain, incl. NVHPC v22.7, FFTW v3.3.10, FlexiBLAS 3.0.4, OpenBLAS v0.3.20, ... #16724

Merged
merged 11 commits into from
Jan 5, 2023

Conversation

Micket
Copy link
Contributor

@Micket Micket commented Nov 23, 2022

(created using eb --new-pr)

Having some issues with BLIS still, didn't have time to really investigate it at all yet.

…-CUDA-11.7.0.eb, OpenBLAS-0.3.20-NVHPC-22.7-CUDA-11.7.0.eb, BLIS-0.9.0-NVHPC-22.7-CUDA-11.7.0.eb, FlexiBLAS-3.2.0-NVHPC-22.7-CUDA-11.7.0.eb, ScaLAPACK-2.2.0-nvompi-2022.07.eb, nvofbf-2022.07.eb
@Micket Micket marked this pull request as draft November 23, 2022 18:49
@boegelbot

This comment was marked as resolved.

@Micket

This comment was marked as outdated.

@bartoldeman
Copy link
Contributor

Note that for BLIS, I had a quick look if it can be compiled with nvc, and the answer is no, since its inline asm kernels use the rbp register, which can be used by GCC using -fomit-frame-pointer, but the equivalent -Mnoframe doesn't allow it.


"kernels/haswell/3/sup/bli_gemmsup_rd_haswell_asm_d6x8m.c", line 166: error: register "bp" has a fixed purpose and may not be clobbered in an asm statement
        begin_asm()

@Micket Micket marked this pull request as ready for review November 27, 2022 00:07
@Micket
Copy link
Contributor Author

Micket commented Nov 27, 2022

Test report by @Micket
FAILED
Build succeeded for 6 out of 9 (9 easyconfigs in total)
alvis-c1 - Linux Rocky Linux 8.5, x86_64, Intel Xeon Processor (Skylake), Python 3.6.8
See https://gist.github.com/02aae05585cdf7909373a578a13033a2 for a full test report.

@Micket
Copy link
Contributor Author

Micket commented Nov 27, 2022

@bartoldeman did you actually manage to build flexiblas with Intel ABI?

The error i'm seeing doesn't just look like some missing compiler flag.
E.g.

"/dev/shm/FlexiBLAS/3.2.0/NVHPC-22.7-CUDA-11.7.0/flexiblas-3.2.0/src/wrapper_blas_intel.c", line 244: error: identifier "hook_pos_caxpy" is undefined
        hook_pos_caxpy = 0;

Looking at the source code, yeah; hook_pos_caxpy really isn't defined anywhere in wrapper_blas_intel.c (like it is in the wrapper_blas_gnu).

flexiblas-3.2.0]$ grep -R hook_pos_caxpy
src/wrapper_blas_intel.c:       hook_pos_caxpy = 0;
src/wrapper_blas_gnu.c:static TLS_STORE uint8_t hook_pos_caxpy = 0;
src/wrapper_blas_gnu.c: hook_pos_caxpy = 0;
src/wrapper_blas_gnu.c:    hook_pos_caxpy++;
src/wrapper_blas_gnu.c:    if ( hook_pos_caxpy < __flexiblas_hooks->caxpy.nhook ) {
src/wrapper_blas_gnu.c:        fn = __flexiblas_hooks->caxpy.f77_hook_function[hook_pos_caxpy];
src/wrapper_blas_gnu.c:        hook_pos_caxpy = 0;

I'm seeing a ton of errors due to all these undefined hook_pos_xxx variables, and many more broken casts

"/dev/shm/FlexiBLAS/3.2.0/NVHPC-22.7-CUDA-11.7.0/flexiblas-3.2.0/src/wrapper_blas_intel.c", line 242: warning: a value of type "void *" cannot be assigned to an entity of type "void (*)(void *, void *, void *, void *, void *, void *)"
        fn = current_backend->blas.caxpy.f77_blas_function;
           ^

@bartoldeman
Copy link
Contributor

I did, but may have worked around the above somehow. I'll double check later.

@Micket
Copy link
Contributor Author

Micket commented Dec 1, 2022

Yeah I stole all your patches already. I've just been to busy to update this PR with the new stuff. I think i'll just patch out the renaming of flexiblas_intel vs flexiblas since things will get rather painful otherwise.

@bartoldeman
Copy link
Contributor

NVHPC 22.11 no longer complains about the bp register in BLIS, so at least that issue is solved (BLIS still needs a patch to for correct compilation flags however)

@easybuilders easybuilders deleted a comment from boegelbot Dec 9, 2022
@easybuilders easybuilders deleted a comment from boegelbot Dec 9, 2022
@easybuilders easybuilders deleted a comment from boegelbot Dec 9, 2022
@Micket
Copy link
Contributor Author

Micket commented Dec 9, 2022

Test report by @Micket
SUCCESS
Build succeeded for 9 out of 9 (9 easyconfigs in total)
vera-c2 - Linux Rocky Linux 8.5, x86_64, Intel Xeon Processor (Skylake), Python 3.6.8
See https://gist.github.com/ae5c15b96ece76ffbb2135383e9cffed for a full test report.

@boegel boegel added this to the 4.x milestone Jan 4, 2023
@boegel
Copy link
Member

boegel commented Jan 4, 2023

@boegelbot please test @ generoso

@boegelbot
Copy link
Collaborator

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

PR test command 'EB_PR=16724 EB_ARGS= EB_CONTAINER= /opt/software/slurm/bin/sbatch --job-name test_PR_16724 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

  • exit code: 0
  • output:
Submitted batch job 9934

Test results coming soon (I hope)...

- notification for comment with ID 1370807134 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
Collaborator

Test report by @boegelbot
SUCCESS
Build succeeded for 9 out of 9 (9 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/c04609a7acde2b3347db4e03adcf4008 for a full test report.

@boegel boegel removed this from the 4.x milestone Jan 4, 2023
add missing patches to OpenBLAS 0.3.20 with NVHPC/22.7 + retain -fb versionsuffix for ScalaPACK on top of FlexiBLAS
Copy link
Member

@boegel boegel left a comment

Choose a reason for hiding this comment

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

lgtm

@boegel
Copy link
Member

boegel commented Jan 5, 2023

Test report by @boegel
SUCCESS
Build succeeded for 9 out of 9 (9 easyconfigs in total)
node3130.skitty.os - Linux RHEL 8.6, x86_64, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz (skylake_avx512), Python 3.6.8
See https://gist.github.com/129ca78c2a186a83a7f0773f8c7f77f0 for a full test report.

@boegel
Copy link
Member

boegel commented Jan 5, 2023

Going in, thanks @Micket!

@boegel boegel merged commit c803ac3 into easybuilders:develop Jan 5, 2023
@espenfl
Copy link

espenfl commented Jan 17, 2023

Thanks a lot @Micket. Greatly appreciated.

@Micket Micket deleted the 20221123184901_new_pr_NVHPC227 branch January 17, 2023 09:47
@boegel boegel changed the title {numlib}[NVHPC/22.7-CUDA-11.7.0] NVHPC v22.7, FFTW v3.3.10, OpenBLAS v0.3.20, ... nvofbf/2022.07 toolchain, incl. NVHPC v22.7, FFTW v3.3.10, OpenBLAS v0.3.20, ... Jan 18, 2023
@boegel boegel changed the title nvofbf/2022.07 toolchain, incl. NVHPC v22.7, FFTW v3.3.10, OpenBLAS v0.3.20, ... nvofbf/2022.07 toolchain, incl. NVHPC v22.7, FFTW v3.3.10, FlexiBLAS 3.0.4, OpenBLAS v0.3.20, ... Jan 18, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants