Skip to content

Reuse Gpu tags in MLCellLinOp::applyBC()#4899

Merged
WeiqunZhang merged 3 commits intoAMReX-Codes:developmentfrom
ankithadas:MLCellLinOp-applyBC-reuse-tags
Jan 19, 2026
Merged

Reuse Gpu tags in MLCellLinOp::applyBC()#4899
WeiqunZhang merged 3 commits intoAMReX-Codes:developmentfrom
ankithadas:MLCellLinOp-applyBC-reuse-tags

Conversation

@ankithadas
Copy link
Copy Markdown
Contributor

@ankithadas ankithadas commented Jan 16, 2026

Summary

Same optimization as implemented in #4882.

Additional background

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

@ankithadas
Copy link
Copy Markdown
Contributor Author

ankithadas commented Jan 16, 2026 •
edited
Loading

Small performance improvement when MLCellLinOp::applyBC() is repeatedly called.

Test case: LinearSolvers/ABecLaplacian_C/3d

For n_cell=320 and using amrex.use_gpu_aware_mpi=1 tiny_profiler.device_synchronize_around_region = 1

After:

MLMG: Final Iter. 6 resid, resid/resid0 = 1.941288332e-05, 7.152966126e-11
MLMG: Timers: Solve = 0.597130692 Iter = 0.585457216 Bottom = 0.109930995
Level 0 max-norm error: 4.496977726e-05 1-norm error: 1.176279099e-05 2-norm error: 1.591549109e-05
Level 1 max-norm error: 3.196255245e-05 1-norm error: 6.879450408e-07 2-norm error: 3.154006476e-06
Level 2 max-norm error: 2.148941828e-06 1-norm error: 1.121850939e-08 2-norm error: 1.178754315e-07


TinyProfiler total time across processes [min...avg...max]: 2.185 ... 2.185 ... 2.185

--------------------------------------------------------------------------------------------
Name                                         NCalls  Excl. Min  Excl. Avg  Excl. Max   Max %
--------------------------------------------------------------------------------------------
MLPoisson::Fsmooth()                           2904     0.9028     0.9028     0.9028  41.31%
MLCellLinOp::applyBC()                         5878     0.2398     0.2398     0.2398  10.98%
MLPoisson::Fapply()                            2974     0.1544     0.1544     0.1544   7.06%
amrex::Dot()                                   4948     0.1241     0.1241     0.1241   5.68%

Before

MLMG: Final Iter. 6 resid, resid/resid0 = 1.941288332e-05, 7.152966126e-11
MLMG: Timers: Solve = 0.627737748 Iter = 0.616312779 Bottom = 0.118744412
Level 0 max-norm error: 4.496977726e-05 1-norm error: 1.176279099e-05 2-norm error: 1.591549109e-05
Level 1 max-norm error: 3.196255245e-05 1-norm error: 6.879450408e-07 2-norm error: 3.154006476e-06
Level 2 max-norm error: 2.148941828e-06 1-norm error: 1.121850939e-08 2-norm error: 1.178754315e-07


TinyProfiler total time across processes [min...avg...max]: 2.294 ... 2.294 ... 2.294

--------------------------------------------------------------------------------------------
Name                                         NCalls  Excl. Min  Excl. Avg  Excl. Max   Max %
--------------------------------------------------------------------------------------------
MLPoisson::Fsmooth()                           2904     0.9031     0.9031     0.9031  39.37%
MLCellLinOp::applyBC()                         5878     0.3404     0.3404     0.3404  14.84%
MLPoisson::Fapply()                            2974     0.1543     0.1543     0.1543   6.73%
amrex::Dot()                                   4948     0.1263     0.1263     0.1263   5.51%

@ankithadas ankithadas marked this pull request as ready for review January 16, 2026 11:27
@WeiqunZhang
Copy link
Copy Markdown
Member

WeiqunZhang commented Jan 19, 2026

/run-hpsf-gitlab-ci

@github-actions
Copy link
Copy Markdown

github-actions bot commented Jan 19, 2026

GitLab CI has started at https://gitlab.spack.io/amrex/amrex/-/pipelines/1392856.

@amrex-gitlab-ci-reporter
Copy link
Copy Markdown

amrex-gitlab-ci-reporter bot commented Jan 19, 2026

GitLab CI 1392856 finished with status: success. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1392856.

@WeiqunZhang WeiqunZhang merged commit e585140 into AMReX-Codes:development Jan 19, 2026
74 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@WeiqunZhang WeiqunZhang WeiqunZhang approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@ankithadas @WeiqunZhang