Skip to content

[FEM.Elastic] Remove complex unproductive stiffness matrix data structure#6133

Merged
hugtalbot merged 4 commits into
sofa-framework:masterfrom
alxbilger:removestiffness
Jun 5, 2026
Merged

[FEM.Elastic] Remove complex unproductive stiffness matrix data structure#6133
hugtalbot merged 4 commits into
sofa-framework:masterfrom
alxbilger:removestiffness

Conversation

@alxbilger

@alxbilger alxbilger commented May 28, 2026

Copy link
Copy Markdown
Contributor

Since #6069, it is suspected that the ElementStiffnessMatrix data structure did not achieve its performance objectives. It was suggested to replace it by a simple matrix, and contain them in a vector.

This new PR applies the suggestion, and removes totally the complex data structure, that is no longer used.

[with-all-tests]


By submitting this pull request, I acknowledge that
I have read, understand, and agree SOFA Developer Certificate of Origin (DCO).


Reviewers will merge this pull-request only if

  • it builds with SUCCESS for all platforms on the CI.
  • it does not generate new warnings.
  • it does not generate new unit test failures.
  • it does not generate new scene test failures.
  • it does not break API compatibility.
  • it is more than 1 week old (or has fast-merge label).

@alxbilger alxbilger added pr: enhancement About a possible enhancement pr: status to review To notify reviewers to review this pull-request labels May 28, 2026
@alxbilger alxbilger requested a review from fredroy May 28, 2026 12:41
@fredroy

fredroy commented May 29, 2026

Copy link
Copy Markdown
Contributor

@alxbilger did you run benchmarks on this branch vs master ?
I can do it on macOS and linux if you want. (to compare with #6069)

@fredroy

fredroy commented May 29, 2026

Copy link
Copy Markdown
Contributor

[ci-build][with-all-tests]

@alxbilger

Copy link
Copy Markdown
Contributor Author

@fredroy fredroy force-pushed the removestiffness branch from eacad69 to 52c475c Compare May 31, 2026 22:38
@fredroy

fredroy commented Jun 1, 2026

Copy link
Copy Markdown
Contributor

on macOS on M3Pro ( 12 P + 4E cores)

Configuration Master FPS RemoveStiffness FPS Speedup
Tetra Linear / Parallel / MatrixFree 4.203 5.935 +41.2%
Tetra Linear / Parallel / Assembled 2.346 2.359 +0.5%
Tetra Linear / Sequential / MatrixFree 2.553 4.176 +63.6%
Tetra Linear / Sequential / Assembled 2.339 2.358 +0.8%
Tetra Corotational / Parallel / MatrixFree 4.083 5.826 +42.7%
Tetra Corotational / Parallel / Assembled 2.344 2.356 +0.5%
Tetra Corotational / Sequential / MatrixFree 1.586 3.585 +126.1%
Tetra Corotational / Sequential / Assembled 2.337 2.348 +0.5%
Hexa Linear / Parallel / MatrixFree 5.114 8.540 +67.0%
Hexa Linear / Parallel / Assembled 2.349 2.350 +0.1%
Hexa Linear / Sequential / MatrixFree 3.391 5.286 +55.9%
Hexa Linear / Sequential / Assembled 2.347 2.343 -0.2%
Hexa Corotational / Parallel / MatrixFree 5.130 8.397 +63.7%
Hexa Corotational / Parallel / Assembled 2.347 2.353 +0.2%
Hexa Corotational / Sequential / MatrixFree 3.039 4.836 +59.1%
Hexa Corotational / Sequential / Assembled 2.335 2.344 +0.4%

Linux on i7 13700K ( 8 P + 8E cores)

Configuration Master FPS RemoveStiffness FPS Speedup
Tetra Linear / Parallel / MatrixFree 7.749 7.082 -8.6%
Tetra Linear / Parallel / Assembled 2.643 2.653 +0.4%
Tetra Linear / Sequential / MatrixFree 2.159 2.111 -2.2%
Tetra Linear / Sequential / Assembled 2.626 2.641 +0.6%
Tetra Corotational / Parallel / MatrixFree 7.290 6.616 -9.3%
Tetra Corotational / Parallel / Assembled 2.622 2.639 +0.6%
Tetra Corotational / Sequential / MatrixFree 1.921 1.866 -2.9%
Tetra Corotational / Sequential / Assembled 2.607 2.616 +0.4%
Hexa Linear / Parallel / MatrixFree 11.344 14.027 +23.7%
Hexa Linear / Parallel / Assembled 2.636 2.651 +0.6%
Hexa Linear / Sequential / MatrixFree 2.692 4.563 +69.5%
Hexa Linear / Sequential / Assembled 2.631 2.635 +0.1%
Hexa Corotational / Parallel / MatrixFree 11.314 13.679 +20.9%
Hexa Corotational / Parallel / Assembled 2.640 2.637 -0.1%
Hexa Corotational / Sequential / MatrixFree 2.517 4.283 +70.2%
Hexa Corotational / Sequential / Assembled 2.616 2.626 +0.4%

Those numbers are virtually identical to those in #6069 so I guess we can close it

@fredroy

fredroy commented Jun 1, 2026

Copy link
Copy Markdown
Contributor

By the way: Windows/MSVC 2026 (Core i7-11800H @ 2.30GHz)

Configuration Master FPS RemoveStiffness FPS Speedup
Tetra Linear / Parallel / MatrixFree 2.585 2.952 +14.2%
Tetra Linear / Parallel / Assembled 1.363 1.287 -5.6%
Tetra Linear / Sequential / MatrixFree 1.018 1.161 +14.1%
Tetra Linear / Sequential / Assembled 1.349 1.282 -5.0%
Tetra Corotational / Parallel / MatrixFree 2.294 2.988 +30.3%
Tetra Corotational / Parallel / Assembled 1.240 1.296 +4.6%
Tetra Corotational / Sequential / MatrixFree 0.617 0.781 +26.5%
Tetra Corotational / Sequential / Assembled 1.185 1.262 +6.5%
Hexa Linear / Parallel / MatrixFree 5.113 7.301 +42.8%
Hexa Linear / Parallel / Assembled 1.350 1.394 +3.3%
Hexa Linear / Sequential / MatrixFree 1.713 2.102 +22.7%
Hexa Linear / Sequential / Assembled 1.342 1.385 +3.2%
Hexa Corotational / Parallel / MatrixFree 4.439 6.035 +35.9%
Hexa Corotational / Parallel / Assembled 1.315 1.354 +3.0%
Hexa Corotational / Sequential / MatrixFree 1.315 1.517 +15.4%
Hexa Corotational / Sequential / Assembled 1.298 1.334 +2.8%

@fredroy

fredroy commented Jun 1, 2026

Copy link
Copy Markdown
Contributor

I rebased the CUDA implementation (#6071) on this branch

@bakpaul bakpaul added pr: status ready Approved a pull-request, ready to be squashed and removed pr: status to review To notify reviewers to review this pull-request labels Jun 4, 2026
@hugtalbot hugtalbot merged commit f82e2e0 into sofa-framework:master Jun 5, 2026
13 checks passed
@hugtalbot hugtalbot added this to the v26.06 milestone Jun 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

pr: enhancement About a possible enhancement pr: status ready Approved a pull-request, ready to be squashed

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants