Test: SIMD by ax3l · Pull Request #4938 · AMReX-Codes/amrex

ax3l · 2026-02-06T00:03:39Z

Summary

Needed to cover SIMD features and generic SIMD/non-SIMD patterns.
Bravely vibe coded, but now reviewed and improved for sensibility.

Additional background

#4924 #4607 #4600 #4520

Checklist

The proposed changes:

fix a bug or incorrect behavior in AMReX
add new capabilities to AMReX
changes answers in the test suite to more than roundoff level
are likely to significantly affect the results of downstream AMReX users
include documentation in the code and/or rst files, if appropriate

Test generic SIMD patterns with new unit tests.

Add SIMD runners to CI.

.github/workflows/gcc.yml

Tests/SIMD/GNUmakefile

Tests/SIMD/main.cpp

ax3l · 2026-02-06T01:14:09Z

/run-hpsf-gitlab-ci

github-actions · 2026-02-06T01:14:19Z

GitLab CI has started at https://gitlab.spack.io/amrex/amrex/-/pipelines/1415577.

amrex-gitlab-ci-reporter · 2026-02-06T01:35:50Z

GitLab CI 1415577 finished with status: failed. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1415577.

False-positive on `-Wnull-dereference` for ``` In member function ‘void amrex::ParticleContainer_impl<ParticleType, NArrayReal, NArrayInt, Allocator, CellAssignor>::WriteParticles(int, std::ofstream&, int, amrex::Vector<int>&, amrex::Vector<int>&, amrex::Vector<long int>&, const amrex::Vector<int>&, const amrex::Vector<int>&, const amrex::Vector<std::map<std::pair<int, int>, typename amrex::ParticleTile<T_ParticleType, NArrayReal, NArrayInt, Allocator>::SoA::IntVector> >&, bool) const [with T_ParticleType = amrex::Particle<3, 0>; int T_NArrayReal = 0; int T_NArrayInt = 0; Allocator = std::allocator; T_CellAssignor = amrex::DefaultAssignor]’: cc1plus: error: potential null pointer dereference [-Werror=null-dereference] cc1plus: error: potential null pointer dereference [-Werror=null-dereference] ```

ax3l · 2026-02-06T17:12:00Z

@WeiqunZhang can you please remove the [email protected] C++20 OMP [tests] test from required in the AMReX branch protection settings? I will need to modify this one to be named GNU@12 C++20 OMP SIMD [tests] for clarity if that is ok.

ax3l · 2026-02-06T17:12:07Z

/run-hpsf-gitlab-ci

github-actions · 2026-02-06T17:12:16Z

GitLab CI has started at https://gitlab.spack.io/amrex/amrex/-/pipelines/1416671.

ax3l · 2026-02-06T17:25:35Z

Oh wow, GCC has so many false positives. I might bump to 13...

False-positive on `-Wnull-dereference` for ``` inlined from ‘void amrex::ParticleContainer_impl<ParticleType, NArrayReal, NArrayInt, Allocator, CellAssignor>::WriteParticles(int, std::ofstream&, int, amrex::Vector<int>&, amrex::Vector<int>&, amrex::Vector<long int>&, const amrex::Vector<int>&, const amrex::Vector<int>&, const amrex::Vector<std::map<std::pair<int, int>, typename amrex::ParticleTile<T_ParticleType, NArrayReal, NArrayInt, Allocator>::SoA::IntVector> >&, bool) const [with T_ParticleType = amrex::Particle<3, 0>; int T_NArrayReal = 0; int T_NArrayInt = 0; Allocator = std::allocator; T_CellAssignor = amrex::DefaultAssignor]’ at /home/runner/work/amrex/amrex/Src/Particle/AMReX_ParticleIO.H:623:36: /usr/include/c++/12/bits/stl_construct.h:119:7: error: potential null pointer dereference [-Werror=null-dereference] 119 | ::new((void*)__p) _Tp(std::forward<_Args>(__args)...); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cc1plus: all warnings being treated as errors ```

WeiqunZhang · 2026-02-06T17:32:46Z

Sometimes, it's not entirely the compiler's fault. They always assume the worse case scenario. But if we also do that, it will affect the code's performance.

ax3l · 2026-02-06T17:34:00Z

Yeah it seems a bit unhappy in GCC 11 & 12 with the ParticleIO routines. I could disable that nullptr check, but since I try to build a first SIMD runner, I would like to not turn that one off to keep this covered for purely SIMD code paths.

Let me know if I shall add a purely new runner, there might be a reason to keep GCC 10 around as a runner as well for now.

amrex-gitlab-ci-reporter · 2026-02-06T17:45:38Z

GitLab CI 1416671 finished with status: success. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1416671.

Error: ``` inlined from ‘void amrex::particle_detail::packIOData(amrex::Vector<int>&, amrex::Vector<double>&, const PC&, int, int, const amrex::Vector<int>&, const amrex::Vector<int>&, const amrex::Vector<std::map<std::pair<int, int>, typename PC::IntVector> >&, const amrex::Vector<int>&, int, bool) [with PC = amrex::ParticleContainer_impl<amrex::Particle<3, 0>, 0, 0, std::allocator, amrex::DefaultAssignor>]’ at /home/runner/work/amrex/amrex/Src/Particle/AMReX_WriteBinaryParticleData.H:457:26, inlined from ‘void amrex::ParticleContainer_impl<ParticleType, NArrayReal, NArrayInt, Allocator, CellAssignor>::WriteParticles(int, std::ofstream&, int, amrex::Vector<int>&, amrex::Vector<int>&, amrex::Vector<long int>&, const amrex::Vector<int>&, const amrex::Vector<int>&, const amrex::Vector<std::map<std::pair<int, int>, typename amrex::ParticleTile<T_ParticleType, NArrayReal, NArrayInt, Allocator>::SoA::IntVector> >&, bool) const [with T_ParticleType = amrex::Particle<3, 0>; int T_NArrayReal = 0; int T_NArrayInt = 0; Allocator = std::allocator; T_CellAssignor = amrex::DefaultAssignor]’ at /home/runner/work/amrex/amrex/Src/Particle/AMReX_ParticleIO.H:623:36: /usr/include/c++/13/bits/stl_construct.h:119:7: error: potential null pointer dereference [-Werror=null-dereference] 119 | ::new((void*)__p) _Tp(std::forward<_Args>(__args)...); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ```

.github/workflows/gcc.yml

Tests/SIMD/main.cpp

WeiqunZhang · 2026-02-06T23:32:43Z

The null-dereference is due to this GCC bug. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108860 I think we have to add -Wno-null-dereference.

This reverts commit c5d1b5a.

ax3l · 2026-02-07T00:38:15Z

Src/Particle/AMReX_ParticleIO.H

        where[grid] = VisMF::FileOffset(ofs);

-        if (count[grid] == 0) { continue; }
+        if (count[grid] <= 0) { continue; }


🎉 Thank you, Weiqun!

ax3l · 2026-02-07T00:39:29Z

Thank you for your help, it works 🎉

And fixes a warning 🎉

Test: SIMD

63e9638

Test generic SIMD patterns with new unit tests.

ax3l requested a review from WeiqunZhang February 6, 2026 00:03

ax3l added the test label Feb 6, 2026

ax3l force-pushed the topic-test-simd branch from 23a422d to 4a6ebbb Compare February 6, 2026 00:06

CI: SIMD w/ vir-simd

73478fc

Add SIMD runners to CI.

ax3l force-pushed the topic-test-simd branch from 4a6ebbb to 73478fc Compare February 6, 2026 00:47

ax3l changed the title ~~Topic test simd~~ Test: SIMD Feb 6, 2026

ax3l mentioned this pull request Feb 6, 2026

Generalize SIMD Single Source Design #4924

Merged

5 tasks

ax3l force-pushed the topic-test-simd branch from 873dfc6 to 71b3bc3 Compare February 6, 2026 01:08

ax3l commented Feb 6, 2026

View reviewed changes

.github/workflows/gcc.yml Show resolved Hide resolved

ax3l commented Feb 6, 2026

View reviewed changes

Tests/SIMD/GNUmakefile Show resolved Hide resolved

ax3l commented Feb 6, 2026

View reviewed changes

Tests/SIMD/main.cpp Outdated Show resolved Hide resolved

ax3l changed the title ~~Test: SIMD~~ [WIP] Test: SIMD Feb 6, 2026

ax3l force-pushed the topic-test-simd branch from 71b3bc3 to ec95a2a Compare February 6, 2026 05:04

host device qualifiers, GPU sync, arenas

f5cfd25

ax3l force-pushed the topic-test-simd branch from ec95a2a to f5cfd25 Compare February 6, 2026 06:25

ax3l changed the title ~~[WIP] Test: SIMD~~ Test: SIMD Feb 6, 2026

CI: GCC 10->11 for SIMD

ab8b532

ax3l force-pushed the topic-test-simd branch from bd9ac93 to eef63f8 Compare February 6, 2026 06:59

ax3l force-pushed the topic-test-simd branch from eef63f8 to 9264891 Compare February 6, 2026 17:10

WeiqunZhang reviewed Feb 6, 2026

View reviewed changes

.github/workflows/gcc.yml Show resolved Hide resolved

Merge branch 'development' into topic-test-simd

b4b88c3

WeiqunZhang reviewed Feb 6, 2026

View reviewed changes

Tests/SIMD/main.cpp Show resolved Hide resolved

WeiqunZhang added 2 commits February 6, 2026 14:40

Maybe this can avoid GCC null-dereference warning

1acaa03

Try to work around GCC warning again

c5d1b5a

WeiqunZhang added 3 commits February 6, 2026 16:09

Revert "Try to work around GCC warning again"

9c902b9

This reverts commit c5d1b5a.

Remove Wnull-dereference

4b225dc

comment

dc00f2f

WeiqunZhang approved these changes Feb 7, 2026

View reviewed changes

ax3l commented Feb 7, 2026

View reviewed changes

ax3l added the warning label Feb 7, 2026

ax3l enabled auto-merge (squash) February 7, 2026 01:11

ax3l merged commit d7da504 into AMReX-Codes:development Feb 7, 2026
74 checks passed

ax3l mentioned this pull request Feb 26, 2026

SIMD: Add where and ternary operator #5095

Open

5 tasks

Conversation

ax3l commented Feb 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Additional background

Checklist

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ax3l commented Feb 6, 2026

Uh oh!

github-actions bot commented Feb 6, 2026

Uh oh!

amrex-gitlab-ci-reporter bot commented Feb 6, 2026

Uh oh!

ax3l commented Feb 6, 2026

Uh oh!

ax3l commented Feb 6, 2026

Uh oh!

github-actions bot commented Feb 6, 2026

Uh oh!

ax3l commented Feb 6, 2026

Uh oh!

WeiqunZhang commented Feb 6, 2026

Uh oh!

ax3l commented Feb 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

amrex-gitlab-ci-reporter bot commented Feb 6, 2026

Uh oh!

Uh oh!

Uh oh!

WeiqunZhang commented Feb 6, 2026

Uh oh!

ax3l Feb 7, 2026

Choose a reason for hiding this comment

Uh oh!

ax3l commented Feb 7, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ax3l commented Feb 6, 2026 •

edited

Loading

ax3l commented Feb 6, 2026 •

edited

Loading