Use accelerate for Conda on macos #40520

tobiasdiez · 2025-08-01T01:07:47Z

Replaces openblas by accelerate in the conda env.

Note: added virtual-packages.yml to define virtual package __osx with a recent macOS platform so that the accelerate package can be used.

📝 Checklist

The title is concise and informative.
The description explains in detail what this PR is about.
I have linked a relevant issue or discussion.
I have created tests covering the changes.
I have updated the documentation and checked the documentation preview.

⌛ Dependencies

Note: added `virtual-packages.yml` to define virtual package `__osx` with a recent macOS platform so that the accelerate package can be used.

tools/update-conda.py

Co-authored-by: Isuru Fernando <[email protected]>

tobiasdiez · 2025-08-01T03:09:13Z

@dimpase it fails to build now:

[615/1782] Compiling C++ object src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/meson-generated_src_sage_libs_linbox_linbox_flint_interface.pyx.cpp.o
  FAILED: [code=1] src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/meson-generated_src_sage_libs_linbox_linbox_flint_interface.pyx.cpp.o
  ccache arm64-apple-darwin20.0.0-clang++ -Isrc/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p -Isrc/sage/libs/linbox -I../src/sage/libs/linbox -Isrc/sage/libs/flint -I../src/sage/libs/flint -I/Users/runner/miniconda3/envs/sage-dev/include/python3.11 -I/Users/runner/miniconda3/envs/sage-dev/include -fvisibility=hidden -fvisibility-inlines-hidden -fdiagnostics-color=always -DNDEBUG -Wall -Winvalid-pch -std=c++11 -O3 -ftree-vectorize -fPIC -fstack-protector-strong -O2 -pipe -stdlib=libc++ -fvisibility-inlines-hidden -fmessage-length=0 -isystem /Users/runner/miniconda3/envs/sage-dev/include -D_FORTIFY_SOURCE=2 -isystem /Users/runner/miniconda3/envs/sage-dev/include -DDISABLE_COMMENTATOR -pthread -DFFLAS_COMPILED -DFFPACK_COMPILED -F/Applications/Xcode_15.4.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/System/Library/Frameworks/Accelerate.framework -idirafter/Applications/Xcode_15.4.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/System/Library/Frameworks/Accelerate.framework/Headers -MD -MQ src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/meson-generated_src_sage_libs_linbox_linbox_flint_interface.pyx.cpp.o -MF src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/meson-generated_src_sage_libs_linbox_linbox_flint_interface.pyx.cpp.o.d -o src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/meson-generated_src_sage_libs_linbox_linbox_flint_interface.pyx.cpp.o -c src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/src/sage/libs/linbox/linbox_flint_interface.pyx.cpp
  In file included from src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/src/sage/libs/linbox/linbox_flint_interface.pyx.cpp:1143:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/zring.h:160:
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givinteger.h:33:11: error: explicit specialization of 'Givaro::ZRing<Givaro::Integer>' after instantiation
     33 |     class ZRing<Integer> : public UnparametricZRing<Integer> {
        |           ^~~~~~~~~~~~~~
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/random-integer.h:90:51: note: implicit instantiation first required here
     90 |                 const_cast<Integer_Domain&>(_ring)=R._ring;
        |                                                   ^
  In file included from src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/src/sage/libs/linbox/linbox_flint_interface.pyx.cpp:1146:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1.h:41:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1denseops.inl:19:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1cyclo.inl:13:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.h:161:
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:269:23: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    269 |                 this->gcd(g,sub(t,y,x),n);
        |                 ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:280:23: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    280 |                 this->gcd(g,sub(t,y,x),n);
        |                 ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:480:15: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    480 |         this->gcd(g,u,v,two,n);
        |         ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:483:15: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    483 |         this->gcd(g,u,v,four,n);
        |         ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:486:15: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    486 |         this->gcd(g,u,v,sixt,n);
        |         ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:497:23: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    497 |                 this->gcd(kgg,kg,n);
        |                 ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:500:27: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    500 |                     this->gcd(g,u,v,kg,n); if (! this->isOne(g)) { delete [] A; delete [] X; delete [] Z; return g; }
        |                     ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:507:19: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    507 |             this->gcd(g,u,v,a,n); if (! this->isOne(g)) { delete [] A; delete [] X; delete [] Z; return g; }
        |             ~~~~  ^
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givintfactor.inl:534:19: error: no member named 'gcd' in 'IntFactorDom<MyRandIter>'
    534 |             this->gcd(ftm,f,n);
        |             ~~~~  ^
  In file included from src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/src/sage/libs/linbox/linbox_flint_interface.pyx.cpp:1146:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1.h:41:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1denseops.inl:19:
  /Users/runner/miniconda3/envs/sage-dev/include/givaro/givpoly1cyclo.inl:62:20: error: no member named 'divexact' in 'Givaro::IntFactorDom<>'
     62 |                 IF.divexact(Iq, n, IF.factor(If,n) );
        |                 ~~ ^
  In file included from src/sage/libs/linbox/linbox_flint_interface.cpython-311-darwin.so.p/src/sage/libs/linbox/linbox_flint_interface.pyx.cpp:1148:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/linbox/matrix/dense-matrix.h:85:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/linbox/matrix/densematrix/blas-matrix.h:42:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/linbox/vector/blas-vector.h:43:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/linbox/field/hom.h:33:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/modular.h:26:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/givaro/modular-ruint.h:24:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/recint/rint.h:53:
  In file included from /Users/runner/miniconda3/envs/sage-dev/include/recint/rdiv.h:48:
  /Users/runner/miniconda3/envs/sage-dev/include/recint/rfiddling.h:170:29: error: no member named 'Low' in 'rint<K>'
    170 |         set_highest_bit(max.Low.Value);
        |                         ~~~ ^

Do you also get this issue with #40397?

tobiasdiez · 2025-08-01T09:13:33Z

Thanks for looking into this Dima!

Some further investigation shows:

The offending code is removed/fixed in linbox-team/givaro@b0cf33e
This is released in givaro v4.2.1
which is already published on conda-forge
but the fflas in conda-forge still has a dependency on the older givaro
which is fixed by Rebuild for givaro 4.2.1 conda-forge/fflas-ffpack-feedstock#21

So we just have to wait for this last PR.

github-actions · 2025-10-28T17:08:22Z

Documentation preview for this PR (built with commit 6282ca5; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

tobiasdiez · 2025-10-29T05:56:17Z

This is now generally working, except that cython_alias fails because it still tries to find blas and not accelerate. @dimpase is it okay to remove get_cblas_pc_module_name in the cython alias handling?

dimpase · 2025-10-29T05:59:08Z

do you talk about a specific platform (conda on macOS) ?

tobiasdiez · 2025-10-29T06:33:13Z

do you talk about a specific platform (conda on macOS) ?

No in general. We don't really require blas - accelerate works as well.

dimpase · 2025-10-29T07:29:43Z

this should be harmonised with sage-distro

tobiasdiez · 2025-11-25T21:37:28Z

Tests are green, so this is now ready for review from my side.

cxzhong · 2025-11-26T06:59:24Z

Maybe you should after remove python 3.11 enviorment then merge this. Thank you very much. And it will cause conflict with remove python 3.11 PR.

tobiasdiez · 2025-11-26T11:12:18Z

Maybe you should after remove python 3.11 enviorment then merge this. Thank you very much. And it will cause conflict with remove python 3.11 PR.

You right. I'll deal with the merge conflicts once the other PR is merged into develop.

cxzhong · 2025-12-12T11:28:31Z

Looks good to me

dimpase

OK

dimpase · 2025-12-19T17:31:57Z

this needs an update of conda environments

tobiasdiez · 2025-12-20T16:43:34Z

this needs an update of conda environments

What do you mean? I've updated the conda env files a few weeks ago in this PR...

dimpase · 2025-12-20T18:18:26Z

I tried to rebase this over 10.8, and got lots of conflicts in environment-* (some packages are newer in 10.8 than in this branch)

dimpase · 2025-12-20T18:19:40Z

so you can merge, but then you still need to update environments, and push the update

tobiasdiez · 2025-12-21T13:02:41Z

so you can merge, but then you still need to update environments, and push the update

Okay, I've now updated the conda lock files.

cxzhong · 2025-12-21T13:24:13Z

run update-meson again. because meson 1.10.0 change the format. Or you pin the version of meson to 1.9.0

dimpase · 2025-12-21T18:05:59Z

update conda lock files needs a good instructon. I don't really know what this does conceptually.

Does it mean to pin the versions to the latest conda state? Then it can be done in bootstrap, no need to keep these environment* files.

Or does it mean to record the latest conda state the current branch is supposed to work in?
Then it needs to be checked on different platforms, not just in one local conda environment. Again, this seems to be some sort of labour intensive stuff best left to a computer.

Recording all the state of all packages is too much, it should suffice to record only the maximal elements state in the package poset.

tobiasdiez · 2025-12-22T10:23:19Z

run update-meson again. because meson 1.10.0 change the format. Or you pin the version of meson to 1.9.0

Done! Thanks

update conda lock files needs a good instructon. I don't really know what this does conceptually.
Or does it mean to record the latest conda state the current branch is supposed to work in? Then it needs to be checked on different platforms, not just in one local conda environment. Again, this seems to be some sort of labour intensive stuff best left to a computer.

Yes, those lock files record a set of versions of conda packages where we know that CI passes without errors. The purpose is two fold:

CI doesn't break randomly because someone pushes a new version of a package to the conda eco system
Devs can use the same lock files locally to setup a dev environment that is known to work

The lock files are automatically generated but need to be updated manually. In the future it would be nice to automate the process by something like dependabot/renovate.

antonio-rojas · 2026-01-13T07:55:18Z

The CBLAS_PC_MODULES variable is unused after this and can be dropped.

tobiasdiez · 2026-01-13T08:19:50Z

Thanks, done in #41439

It's no longer used after sagemath#40520. ### 📝 Checklist  - [ ] The title is concise and informative. - [ ] The description explains in detail what this PR is about. - [ ] I have linked a relevant issue or discussion. - [ ] I have created tests covering the changes. - [ ] I have updated the documentation and checked the documentation preview. ### ⌛ Dependencies    URL: sagemath#41439 Reported by: Tobias Diez Reviewer(s):

Use accelerate for Conda on macos

82437de

Note: added `virtual-packages.yml` to define virtual package `__osx` with a recent macOS platform so that the accelerate package can be used.

github-actions bot added the s: needs review label Aug 1, 2025

tobiasdiez mentioned this pull request Aug 1, 2025

macOS: allow to use Accelerate framework instead of OpenBLAS #40397

Draft

5 tasks

isuruf reviewed Aug 1, 2025

View reviewed changes

tools/update-conda.py Outdated Show resolved Hide resolved

Use libblas for consistency everywhere

de0b673

Co-authored-by: Isuru Fernando <[email protected]>

tobiasdiez marked this pull request as draft August 1, 2025 03:09

github-actions bot removed the s: needs review label Aug 1, 2025

Merge remote-tracking branch 'upstream/develop' into macos-conda-accel

532dfa8

tobiasdiez added 3 commits November 24, 2025 21:50

Merge remote-tracking branch 'upstream/develop' into macos-conda-accel

37c33bc

Remove get_cblas_pc_module_name

7505ec4

update conda lock files

01276d7

tobiasdiez force-pushed the macos-conda-accel branch from 4df387a to 01276d7 Compare November 25, 2025 07:26

tobiasdiez added 3 commits November 25, 2025 09:25

Readd OpenBLAS dependency in conda files

83a352e

fix tests

c8eaca6

Remove more CBLAS references

87847b5

tobiasdiez marked this pull request as ready for review November 25, 2025 21:37

github-actions bot added the s: needs review label Nov 25, 2025

tobiasdiez requested review from cxzhong and dimpase November 25, 2025 21:37

Merge remote-tracking branch 'upstream/develop' into macos-conda-accel

dedd29b

cxzhong added the s: positive review label Dec 12, 2025

dimpase reviewed Dec 12, 2025

View reviewed changes

dimpase approved these changes Dec 12, 2025

View reviewed changes

dimpase added s: needs work and removed s: positive review labels Dec 19, 2025

Merge branch 'develop' into macos-conda-accel

2330fad

Update conda lock files

7df00e7

github-actions bot added s: needs review and removed s: needs work labels Dec 21, 2025

tobiasdiez added 2 commits December 22, 2025 11:18

Merge remote-tracking branch 'upstream/develop' into macos-conda-accel

67489bf

Reformat meson files

6282ca5

cxzhong added s: positive review and removed s: needs review labels Dec 22, 2025

vbraun merged commit 174227b into sagemath:develop Dec 29, 2025
18 of 31 checks passed

github-actions bot removed the s: positive review label Dec 29, 2025

antonio-rojas mentioned this pull request Jan 13, 2026

Test suite regressions in the meson build #39872

Open

tobiasdiez deleted the macos-conda-accel branch January 13, 2026 08:16

tobiasdiez mentioned this pull request Jan 13, 2026

Remove CBLAS_PC_MODULES variable #41439

Merged

5 tasks

Uh oh!

Use accelerate for Conda on macos #40520

Use accelerate for Conda on macos #40520

Uh oh!

Conversation

tobiasdiez commented Aug 1, 2025

📝 Checklist

⌛ Dependencies

Uh oh!

Uh oh!

tobiasdiez commented Aug 1, 2025

Uh oh!

tobiasdiez commented Aug 1, 2025

Uh oh!

github-actions bot commented Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tobiasdiez commented Oct 29, 2025

Uh oh!

dimpase commented Oct 29, 2025

Uh oh!

tobiasdiez commented Oct 29, 2025

Uh oh!

dimpase commented Oct 29, 2025

Uh oh!

tobiasdiez commented Nov 25, 2025

Uh oh!

cxzhong commented Nov 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tobiasdiez commented Nov 26, 2025

Uh oh!

cxzhong commented Dec 12, 2025

Uh oh!

dimpase left a comment

Choose a reason for hiding this comment

Uh oh!

dimpase commented Dec 19, 2025

Uh oh!

tobiasdiez commented Dec 20, 2025

Uh oh!

dimpase commented Dec 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dimpase commented Dec 20, 2025

Uh oh!

tobiasdiez commented Dec 21, 2025

Uh oh!

cxzhong commented Dec 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dimpase commented Dec 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tobiasdiez commented Dec 22, 2025

Uh oh!

Uh oh!

antonio-rojas commented Jan 13, 2026

Uh oh!

tobiasdiez commented Jan 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

github-actions bot commented Oct 28, 2025 •

edited

Loading

cxzhong commented Nov 26, 2025 •

edited

Loading

dimpase commented Dec 20, 2025 •

edited

Loading

cxzhong commented Dec 21, 2025 •

edited

Loading

dimpase commented Dec 21, 2025 •

edited

Loading