[fredroy]

Fix

• [LinearSystem] Add tests: the global matrix depends on the addresses of the components #6425
• MatrixLinearSystem assembles groups using an address-ordered std::map, causing run-to-run nondeterminism #6229

Before, the containers of the components (and of the pairs of mechanical states) of the assembly were ordered by addresses
Now they are ordered by the paths of the components in the scene graph, thus allowing deterministic results
(really poor results)

Now the contributions follow the order of the scene graph traversal:
- IndependentContributors stores vectors filled by visiting m_forceFields, m_masses and m_mechanicalMappings (collected in traversal order)
- the shared mapped matrices are created, and projected, in the order in which their pair of mechanical states is first encountered
- the duplicated mechanical states of a component are removed without sorting them by address

EDIT: add the benchmark scene

[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]

The only concern is the loss of performances. Otherwise 👍

[fredroy]

The only concern is the loss of performances. Otherwise 👍

You were right to be concerned, with a specially crafted scene (long graph depth and very long path name)

master: 200 iterations done in 7.45178 s ( 26.8392 FPS)
PR with paths: 200 iterations done in 438.29 s ( 0.456319 FPS).
😅

[fredroy]

The only concern is the loss of performances. Otherwise 👍

You were right to be concerned, with a specially crafted scene (long graph depth and very long path name)

master: 200 iterations done in 7.45178 s ( 26.8392 FPS) PR with paths: 200 iterations done in 438.29 s ( 0.456319 FPS). 😅

PR with new containers: 200 iterations done in 7.01615 s ( 28.5056 FPS).