SOFA dev reports (2022)

Join us every Thursday 9:30 am CET online: www.sofa-framework.org/sofa-dev-meeting

14/12/2022

News

SOFA v22.12 release is postponed due to sick leave of the release manager Sorry for the delay!

Technical discussions

CI with macos-3 is down again Future investigation on why the comm macos-3 ↔ jenkins fails is needed
CGAL was disactivated see commit Consider reactivating CGAL plugin on the CI when the CI will be fixed

SOFA PR review

#3526 [All] Fix example scenes showing warnings or errors Has not been reviewed yet. Need more time. Will probably not be ready for v22.12 → keep to review
#3535 [all] Lifecycle v22.06 Changes requested → wip
#3534 [all] Lifecycle v22.12 Changes requested → wip
#3465 [Sofa.Helper] Use DataTypeInfo in NameDecoder Wait for CI then ready but wait for v23.06 → ready / postponed

BeamAdapter PR review

#84 Remove deprecation warnings Approved → ready
#80 Restore BeamLengthMapping Restored for now, but don't know what this component does. To deprecate later if needed. → keep to review
#50 [Mapping] Clean and optimize AdaptiveBeamMapping Still decisions to take on the support of std parallelism → keep to review

SofaPython3 PR review

#326 Remove v22.06 for the CI → ready and merged
#325 Fix: mass binding is defined to have multiple inheritance → ready and merged

PR merged within the week

07/12/2022

News

SOFA v22.12 freeze of branch today, release process will start right after this branch creation
Next week is the last 2022 meeting! We are now preparing the 2023 agenda (exact dates are being validated)
HackaTech in Grenoble in January

Technical discussions

Integration of matrix assembly branch: further to STC, the conclusion was that the interaction between several objects is a feature which is interesting to keep. There is therefore still work to do and this branch will not get into the release however it should be merged in master as soon as possible (January is targeted).
PRs respecting the lifecycle are open: #3534 & #3535
Label "pr: test" is added

SOFA PR review

#3529 [LinearSolver.Direct] Segfault in SparseLDLSolver due to empty system matrix Damned let's rebase it and re-run the CI. After doing it: fix are legit and must be fixed. → back to wip
#3526 [All] Fix example scenes showing warnings or errors Not ready yet, let's see if it has to be in v22.12.00 or to be later backported in v22.12.01. Comments done. → keep to review
#3520 [SofaPhysicsAPI] Add several methods using output parameters To be rebased, commented → wip
#3519 [SofaPhysicsAPI] Some small cleaning and add access to VisualModel* →ready
#3491 [Simulation] Worker threads are no longer static Rebase made the diff way clearer. Wait for CI then ready → keep to review
#3482 [Simulation] Task scheduler can accept callable Wait for CI then ready → keep to review

SofaPython3 PR review

#316 Add a counter to memcach so that we can detect data change Fix needed when we use the cache of SOFA data loaded into python. The unit test had an issue but not emitting a proper test error. A benchmark comparing the use and no use of this cache in term of performance would be really nice. → ready
#315 [SofaConstraintSolver] Add bindings to set W, lambda and dfree PR is a nice hack which is being tested by various users. API of SOFA needs to be updated and would therefore need more work. Let's keep it → to review + experimental
#314 [Sofa.Core] Add spelling suggestion to addObject & addChild → ready

PR merged within the week

30/11/2022

News

Preparation of the 2023 agenda: make your 2023 events known! Share them with us: contact@sofa-framework.org

Technical discussions

Discuss the creation of the v22.12 branch and freeze for release
- Branch will be created and frozen
- Notify us for any upcoming contributions to be integrated
- Only pending contribution we are aware of: matrix assembly
Contribution process for large contribution (breaking / incompatible with SOFA)
- A scheme presenting the process for integration of such contributions could be formalized
Erik PR for v22.06: the very same branch will be created and merged in the SOFA repository

SOFA PR review

#3508 [FEM.Elastic] Clean draw method → ready
#3505 [VectorTypeInfo] Change default handling of data buffer → ready
#3491 [Simulation] Worker threads are no longer static Sounds good but PR needs a rebase, it will ease the review → keep to review
#3482 [Simulation] Task scheduler can accept callable Sounds good but PR needs a rebase, it will ease the review → keep to review
#3480 [Simulation] Separate factory code from TaskScheduler → ready
#3465 [Sofa.Helper] Use DataTypeInfo in NameDecoder Test failing →back to wip

SofaPython3 PR review

#318 [DataHelper] Change default access of vector data buffer → ready
#316 FIX Issue 313 Set a proper title, CI jobs restarted → keep to reveiw
#315 [SofaConstraintSolver] Add bindings to set W, lambda and dfree Awaiting reply to comment → keep to review
#314 [Sofa.Core] Add spelling suggestion to addObject & addChild. Update of pybind11 required, to be done but not urgent (no time currently, sorry) → keep to review

PR merged within the week

23/11/2022

News

Update further to the SOFA Week 2022: 231 registered participants!!!
- Training working smoothly
- Symposium
  - Hard to motivate people to contribute → does not produce a real contribution
  - Consortium : could present the community activity worldwide
  - Nice event showing the activity of international partners, mostly users
  - Inria : would be nice to have general presentation of the team
- STC
  - No real dynamics, hard to embark new people
  - Coding-sprints : should we skip it or make it separated?
  - Coding-sprints : onsite event should help
  - Coding-sprints : STC tasks too large / complicated – coding sprint more focused !
  - Coding-sprints : e.g. a week with a scrum organization
Preparation of the 2023 agenda: make your 2023 events known!
Accompany new users and make them become developers
- Weekly meetings : open to all (including Inria teams)
- Support + companionship till the contribution
- At the SOFA dev → presentation of technical topics, one topic per month, upvoted by the students/engineers

SOFA PR review

#3476 [MultiThreading] Clean file format → ready
#3474 [Lagrangian.Solver] Merge duplicated code into small but expressive functions → ready
#3471 [Visual] Introduce TrailRenderer → ready
#3465 [Sofa.Helper] Use DataTypeInfo in NameDecoder Unit and scene tests are failing due to Ogl → wip
#3357 [all] Remove in-class typedefs pointing to common sofa::type Test is failing. GItHub diff is unclear. [ci-build][force-full-build] to see if the test is related to the PR. → keep to review

PR merged within the week

09/11/2022

News

SOFA Week 2022 : next week!!! Already +185 registered participants and program all available online
Release manager needed! Share all our job offers
STC #14: add the topics you would like to address in discussions

Technical discussions

Timer for debug only has been increased by Fred
A new test is failing : RemovingBilateralInteractionConstraint.scn on MacOS

SOFA PR review

#3465 Use DataTypeInfo in NameDecoder Refactoring could be done in NameDecoder to keep only string resolution from the compiler, but for class template, the code could be moved in BaseClass using defaulttype. Damien will work on it → wip
#3464 Revert "[Sofa.Core] Linear time getRoot() method in BaseNode and Node" Breaks some scenes with a diamond (node with multiple parents). Issue reported in #3469 https://github.com/sofa-framework/sofa/issues/3469 → ready
#3462 [DefaultType] fix rigidcoord compilation Follows #3424 → ready
#3461 [DefaultType] Remove template aliases Might be closed since #3465 has been opened. → closed
#3456 [GUI.Qt] Restore Stats (Charts) and DocBrowser for Qt6 We know fully support Qt6, we should update the doc accordingly. → ready
#3454 [GUI.Qt] Restore QGLViewer for Qt6 → ready
#3452 [image, Component.FEM, Component.LinearSolver] fix Eigen3 assertion with SVD → ready
#3424 [Helper] Add mutex in IdFactory Not sure it solves any problem and it might degrade performances. A full refactoring of AdvancedTimer should be the objective to make it simple, thread-safe and shiny! → close
#3422 [Helper] Portable thread local storage duration Go for it! AdvancedTimer may not itself be thread-safe (and a pretty complex class) → ready
#3360 [DefaultType] fix rigidcoord compilation It would be nice to have an associated test. Translation in 3,3 case is not supported (toMatrix gives only a rotation matrix). This PR fixes the compilation → ready
#3357 [all] Remove in-class typedefs to Vector3 Types with new naming are proposed. Should be moved in another PR → wip
#3042 [Core, Helper] Add spelling suggestion in ObjectFactory Checking spelling is nice, a single header only lib has been added. A test is added as well. → ready

PR merged within the week

02/11/2022

News

SOFA Week 2022 in two weeks, time to RE-GI-STER
Release manager needed! Share all our job offers

Technical discussions

Timeout seems to occur in debug, Fred suggested to increase the timer for debug only
Backport of PRs should be done and a tag v22.06.01 should be created
A discussion about a "SOFA 2.0" is scheduled on day 2 of the STC (Thursday 17th). Do not hesitate to join and create your SOFA wish-list
Types in SOFA: internal type in class or SOFA type::
- When to use internal types
- Define a naming convention for internal types (through a PR in doc)
- Issue of manpower: how to propagate this change among many other

SOFA PR review

#3448 [Simulation.Core] Clean Multithreading code → ready
#3439 [plugins] Fix warnings in SofaSphFluid and image Message should not be addressed to users, the drawing should just be invalidated → wip
#3436 [Sofa.Config] CMake: Remove "both" as a choice for SOFA_FLOATING_POINT_TYPE parameter → ready
#3435 [All] Fix compilation with float as floating_point_type Nice! Compilation and running the caduceus scene with floats is fixed. FYI no performance gain. Use of signature using SReal → ready
#3428 [core.topology] Add mechanism to check checkTopologyInputTypes Reviewed and approved → ready and merge
#3424 [Helper] Add mutex in IdFactory Linked to #3422 → keep to review
#3422 [Helper] Portable thread local storage duration A priori (Francois J.) since C+11 a thread-safe instantiation is done when initializing a static variable While here using thread_local, it creates one static variable per thread. AdvancedTimer may not itself be thread-safe. Error might come from a concurrent writing of new events in a map in AdvancedTimer. → keep to review
#3413 [solidmechanics.fem] Clean and optimise TriangularFEMForceField draw method → ready
#3404 [Core] Convert warning to error in object factory Conflicts to be solved. → wip
#3369 [Topology] Fix topologyHandler removal New mechanism with a set storing the registered topologies → ready
#3368 [Topology] Add fonction in BaseMeshTopology to compute all topology containers Two tests are failing and a comment to be added → wip
#3357 [all] Remove in-class typedefs to Vector3 Compilation issue in tests and bindings. Moreover, types with new naming are proposed. Should be discussed with all. If so, PR title should be updated. → wip
#3339 [Mapping] Add checks at init in TopologicalMapping CanCreate of TopologicalMapping calls findLinkDest() which checks the link at creation and if not found, the search continues in parent nodes. Issue #3449 created. To be rebased once #3428 is merged, then if CI agrees ready → keep to review
#3329 [Constraint.Lagrangian] Update BilateralInteractionConstraint to support topological changes Test has been fixed. New force full build (QuaterAngle test strangely failing). → keep to review

PR merged within the week

26/10/2022

News

SOFA Week 2022: 20 days to register! 14-18th November
Jobs: our community recruits massively
We should take part in NAFEMS 2022 on November 24th

Technical discussions

Topics of discussion must be added in preparation of the STC#14

SOFA PR review

#3427 [SofaNewmat] Fix modules → ready
#3426 [helper.accessor] Support more std vector methods in WriteAccessorVector → ready
#3423 [SofaMatrix] Update plugin CMakeLists to avoid strong dependency on Qt → ready
#3421 [image] Remove example scenes requiring Flexible plugin → ready
#3420 [FEM.Elastic] Speedup hexa drawing in force field → ready
#3418 [Lagrangian.Solver] Add events in the constraint solver pipeline Code in ConstraintSolver should really be separated between core code (in a renamed BaseConstraintSolver class) and the implementation ConstraintSolver. → ready
#3413 [solidmechanics.fem] Clean and optimise TriangularFEMForceField draw method Review : maybe to create a dedicated function called in addForce. To be discussed in PR. → keep to review
#3412 [TriangularFEMForceField] Avoid double write access to the TriangleInfo Data in Review done, but not 7 day old. → ready
#3408 [Topology.mapping] Types cleaning in Edge2QuadTopologicalMapping → ready
#3407 [SofaCarving] Update and add more tests in SofaCarving_test → ready
#3404 [Core] Convert warning to error in object factory Scenes to be ignored : error is due to the fact that now CI detects the not-compiled templates. However, the compilation time with these templates increases significantly. → wip
#3401 [Sofa.Core] VecId: move definition of VecTypeLabels in its own Translation Unit → ready
#3398 [Engine.Analyze] ClusteringEngine: set correct values in load() → ready
#3393 [Lagrangian.Solver] Deprecate MechanicalAccumulateConstraint Restart force full build → wait for CI
#3369 [Topology] Fix topologyHandler removal → wip
#3368 [Topology] Add fonction in BaseMeshTopology to compute all topology containers Scene test is failing and function naming to be improved → back to wip
#3339 [Mapping] Add checks at init in TopologicalMapping New approach proposed in the TopologicalMapping to make the checks on input/output topologies. Some missing implementation and find a solution to get the topological element name from enum class. New review needed → wip
#3329 [Constraint.Lagrangian] Update BilateralInteractionConstraint to support topological changes Test on MacOS failing → back to wip, must be
#3299 [all] Replace all Vector2, Vector3, Vector4 by their short name alias Vec2, Vec3 branch is protected. Please open it so that we can update and your PR → wip
#3242 [Core] Make some functions non-virtual Await for regression PR to be merged → keep to review

Registration PR review

#2 [CMake] Fix NG compil without compat → closed
#4 getTemplateName is no longer virtual merge master and rebuild → keep to review
#5 Restore tests and add test on getTemplateName → merged
#7 Remove example scenes requiring Flexible plugin → merged

PR merged within the week

19/10/2022

News

[SOFA Week 2022] More than 130 participants registered already, spread the word!
First PR from Leo Hunout on the SofaGym project!

Technical discussions

CI issues on MacOS, solved by reboot
STC #14: add the topics you would like to address in discussions
#3288 [all] Removal of datatype info from Data.h (step 2) → Damien presented the PR and no objection was raised, only a few minor suggestions. Need to keep working on it to complete the PR.

SOFA PR review

#3386 [all] Implement getModuleComponentList for most modules → ready
#3376 [Simulation] Change LoadFromMemory signature to remove unused parameter Size → ready
#3373 [Helper] Deprecate error-prone resize and add emplace_back → ready
#3369 [Topology] Fix topologyHandler removal Test are failing → back to wip
#3368 [Topology] Add fonction in BaseMeshTopology to compute all topology containers Scene test is failing and function naming to be improved → back to wip
#3360 [DefaultType] fix rigidcoord compilation Changes requested → back to wip
#3358 [Core] STL parallelism support Postponed until support of Ubuntu 20.04 is dropped → postponed
#3356 [All] Reduce calls to getValue Conflicts to solve → If conflicts are solve: ready
#3345 [Config] Option to enable interprocedural optimization → ready
#3335 [LinearAlgebra] Remove unused operations on BaseMatrix Postponed because code could be used in private plugins → postponed
#3329 [Constraint.Lagrangian] Update BilateralInteractionConstraint to support topological changes one last tiny change and ready → to review
#3299 [all] Replace all Vector2, Vector3, Vector4 by their short name alias Vec2, Vec3 A unit test to fix and conflicts to solve. Otherwise ready → to review
#3294 [Constraint.Lagrangian.Solver] Add GetConstraintForce component Support of multiple objects is validated. Force norm is validated. Unit test is required. → wip
#3271 [Topology] Improve TopologyHandler registration → ready
#3242 [Core] Make some functions non-virtual wait more PR related to unit testing to validate. One unit test fails → wip

PR merged within the week

12/10/2022

News

SOFA recruits, find out more about these opportunities
[SOFA Week 2022] More than 95 participants registered already, spread the word!
Test the Matrix branch : #2777 waiting your reviews!
Another SOFA Training session occurred on Monday!

Technical discussions

CI issues on MacOS, machines rebooted
Upcoming discussion about the progress of the Matrix project and objectives: naming, blocking points, specific cases
STC #14: add the topics you would like to address in discussions
#3299 [all] Replace all Vector2, Vector3, Vector4 by their short name alias Vec2, Vec3 This change should not have a strong impact on regression. Only issues should come when compiling in float. This raises a question: do we really want to support float (required for several API, CUDA, OpenGL) and how to keep the simplest and most readable code. In many codes, the member types are generic while in fact the implementation only works for a specific template (eg. Vec3). Now: check compilation and regression. Then ready!
Good first issues are available, give it a try!

SOFA PR review

#3365 [CImgPlugin] Replace new keyword with creating unique pointers to properly clean up allocated memory. → ready
#3353 [sceneutility] Deprecate MakeAlias and MakeDataAlias → closed
#3351 [examples] Another pass of plugin finder One crash on scene test (error 139). Name of some classes are really not good : see issue #3320 https://github.com/sofa-framework/sofa/issues/3320 → back to wip
#3345 [Config] Option to enable interprocedural optimization New test to be conducted to check whether it activates properly the interprocedural optimization on Windows → back to wip
#3344 [git] Add various temporary files to .gitignore Other folder should be also ignored (extlibs, license, include) → ready
#3329 [Constraint.Lagrangian] Update BilateralInteractionConstraint to support topological changes Test failing and review comments to take into account → back to wip
#3328 [Constraint.Lagrangian] Remove merge option from BilateralInteractionConstraint → ready
#3327 [Constraint.lagrangian] Fix BilateralInteractionConstraint double init and clean some Data Activating constraint at a specific time, should now be done through a script or a controller → ready
#3315 [S.C.MechanicalLoad] PlaneForcefield: Fix "potential divide by zero" warnings in msvc → ready
#3278 [Sofa.Core] minor refactoring for Data::read() to move into BaseData the reading code Tests are failing → back to wip
#3271 [Topology] Improve TopologyHandler registration One test added in the PR is failing → back to wip
#3242 [Core] Make some functions non-virtual Add comment in the code referring to the PR → to review, then ready
#2777 [LinearSystem] Refactor matrix assembly Unit tests added! To be reviewed @all. Performances using the benchmark scenes seem similar to the current ones. → keep to review and meeting upcoming

PR merged within the week

06/10/2022

News

SOFA recruits, find out more about these opportunities
SOFA Week 2022: First participants arelisted here

Technical discussions

Program STC #14 for days 2 & 3 (November 17-18th): feel free to add any topic to discuss

SOFA PR review

#3237: ready. Needed to be rebased due to conflicts.
#3242: Add comments with link to the PR with details. After that => Ready
#3245: Ready
#3269: Changes requested
#3278: Failing unit tests. A description is requested.
#3287: Change requested, otherwise ready.
#3288: FAiling tests on Windows. This PR will be discussed next week.
#3289: Compilation failing
#3290: postponed. The need to repeat using Inherit1::initData is not understood
#3294: WIP
#3299: Compilation failing => WIP
#3301: Failing scenes => WIP. Wait for https://github.com/sofa-framework/sofa/pull/3351
#3304: Ready
#3305: Ready. To present next dev meeting so everyone is aware of this PR
#3314: Ready. To present next dev meeting so everyone is aware of this PR
#3315: Ready. But an issue has been written.
#3327: Deprecation requested. Await answer from M. Talbot.
#3331: Ready
#3332: Ready
#3334: Keep to review. Need more feedback. Deprecation requested.
#3338: Ready
#3339: Answer requested. Need more investigation => WIP
#3344: Keep to review
#3345: Need more tests

BeamAdapter PR review

#50: wait to review https://github.com/sofa-framework/sofa/pull/3358. An alternative is to support only compilers with parallel algorithms support.
#57: Ready. Vector of gravity vectors has been replaced by a single gravity vector from the current Node
#59: Ready
#62: Keep to review
#64: Need regression tests. Keep to review

PR merged within the week

28/09/2022

News

SOFA recruits, find out more about these opportunities
Servers back online, take the opportunity to register for the SOFA Week 2022
First participants are listed here

Technical discussions

New gravity API proposed in #2988 Presentation done, but to be further discussed with other physics developers. Main changes are for users, therefore scene description must be clearly explained to users. → to review
Performance issue coming from #3280 VisualStudio seems to struggle with the inlining, RealeaseWithDebugInfo should be slower (which is the case on the CI). Inlining could be forced but this is not the good way. A deeper investigation is needed but Alex has no time for now. Any help welcome! Reported in #3343 https://github.com/sofa-framework/sofa/issues/3343 → to review
Memory leak fixed in SOFA in #3337 Nice PR fixing the leak while unloading SOFA. Few questions posted and now waiting the 7 days of review, then should be ready → keep to review
Program STC #14 for days 2 & 3 (November 17-18th): feel free to add any topic to discuss
Issue of css on the online documentation

SOFA PR review

#3344 [git] Add various temporary files to .gitignore → ready
#3339 [Mapping] Add checks at init in TopologicalMapping Reviews added, improve message for user. → back to wip
#3335 [LinearAlgebra] Remove unused operations on BaseMatrix Request again review awaiting → keep to review
#3294 [Constraint.Lagrangian.Solver] Add GetConstraintForce component Test to be conducted to validate the value returned. Also should make sure that indices are well managed when several objects are in the scene. → keep to review
#3290 [All] Compilation with "/fpermissive-" (MSVC) Need to be discussed, investigated by more → keep to review
#3287 [Collision] Cleaning pass on types and remove unused code Extract the changes on MeshDiscreteIntersection in another PR, else ready. → keep to review, then ready
#3282 [Sofa.Defaulttype] Move Rigid{Coord, Deriv, Mass} from RigidTypes into their own headers → ready
#3279 [Sofa.Core] Finally disable deprecated code in BaseData and ExectParam's Aspects. → ready
#3271 [Topology] Improve TopologyHandler registration Few questions and typos to fix. → wip
#3270 [Sofa.Core] Add some const qualifier to collision methods & use SReal & nodiscard. → ready
#3268 [SolidMechanics] Fix updating the internal spring indices of SpringForceField Need review of @epernod → keep to review
#3245 [Topology_test] Add tests in EdgeSetTopology_test to check topological changes Need review of @hugtalbot → keep to review
#3237 [Sofa.DefaultType] Constexpr VecTypes and RigidTypes These classes were fully inline. Need a rebase. Need to test whether it can really be constexpr-ed! → keep to review, after CI ready !
#2777 [LinearSystem] Refactor matrix assembly Boilerplate code added to get the compatibility layer if the new API is not implemented. Tests remain to be added, the work will then be widely presented → keep to review

PR merged within the week

14/09/2022

News

Due to new malicious act on Inria Lille infrastructure, our website http://sofa-framework.org is down for now. We'll keep you posted when to register again for the #SOFAWeek2022
The CI is down, investigation shows this is related to a problem in the Inria infrastructure a ticket was open
Presentation of the work to be done by Leo Hunout
Work on parallel linear solver performed by Stephane & Alex is now public: https://github.com/SofaDefrost/SofaCUDALinearSolver
SOFA recruits, find out more about these opportunities
Guillaume Paran as guest star in the SOFA dev meeting!

SOFA PR review

#3216 [Topology] Fix initTopology call chains for mesh without topology (ready, postpone... )
#3221 [Type] Constexpr Quaternion (ready, a future PR will come to implement constexpr validation in unittest)
#3231 [Constraint.Lagrangian.Correction] LinearSolverCC: use FullVector API if detected (ready, future discussion planned on more general refactoring to have dynamic_cast less architecture if possible)
#3239 [GUI.performer] Add remove elements function for LineCollisionModel (ready, wait the CI is back)
#3249 [SolidMechanics.Spring] RestShapeSpringForceField: Fix addKToMatrix (ready, opened a discussion about how we should handle the "conditionale" sizing of the array)
#3283 [Core][Type] Minor warning fixes (ready)
#3284 [StateContainer] Optimize vector operations (ready)
#3265 [Config] Remove deprecated macros (back to wip to only handle the case with Quat, BaseData will be in #3279)

SofaPython3 PR review

#302 [bindings.SofaTypes] Follow changes in Mat (ready, wait the CI)
#301 [all] Fix memory leak (validated by Paul, ready wait the CI)
#300 [examples] Update all plugins and remove DefaultVisualManagerLoop (back to wip)
#298 [Core] Improve some tests related to the binding of getClassName() (ready, wait the CI)
#293 Fixed Memory Leak in DataContainer.array() and DataContainer.writeable() (closed integrated in #301)

PR merged within the week

07/09/2022

News

SOFA recruits, find out more about these job opportunities
Do not miss the annual conference for SOFA: Symposium day on 15th November
SOFA v22.06 is out, get it here: https://www.sofa-framework.org/download/

Technical discussions

SofaCUDALinearSolver : GPU-based direct linear solver: further to Stephane’s work, a direct GPU linear solver based on cusolver library (available in CUDA toolkit). However, inveistation and limitations remain :
- the library is opaque (API does not provide access to factors/submatrices, and therefore it does not allow a CPU-GPU hybrid factorization-resolution)
- a limitation is that it cannot take advantage of the constant pattern of the system matrix: another matrix is available for this.cusolver_rf (only a LU factorization) → an investigation to do
[Type] Constexpr Quaternion #3221
- Constexpr requires to define all in the header file (making compilation time higher)
- Let's test to define it constexpr while defining it in the inl file (including it in the header) and see if it works as constexpr

SOFA PR review

#3239 [GUI.performer] Add remove elements function for LineCollisionModel A comment to be added. Missing a check on the found topology. → wip
#3237 [Sofa.DefaultType] Constexpr VecTypes and RigidTypes Wait for #3221 → keep to review
#3231 [Constraint.Lagrangian.Correction] LinearSolverCC: use FullVector API if detected Loss of genericity and complexity with the lambda function. Todo: make the lambda include the two for loops. → wip
#3221 [Type] Constexpr Quaternion Let's test to define it constexpr while defining it in the inl file (including it in the header) and see if it works as constexpr → keep to review
#3216 [Topology] Fix initTopology call chains for mesh without topology Rebuild taking into account regression PR and then ready if CI is ok → keep to review
#3126 [Lagrangian.Model] Remove use of LateSpecialization pattern Try to find a design pattern to properly treat the specialization cases in SOFA. Vec3Type and Rigid3Type are working so far (but not taking CudaType). Check if the proposed pattern does suit a sfinae design. → keep to review

SofaPython3 PR review

#301 [all] Fix memory leak More see #293 → will be closed after investigation
#293 Fixed Memory Leak in DataContainer.array() and DataContainer.writeable() The PR #301 opened suggests a new solution. Should be merged into #293. Remain todo: solve the new case (solved by the std::make_unique) to have something generic whatever the version of pybind. Damien will have a look at it → keep to review

PR merged within the week

31/08/2022

News

Save your seat for the 14-18th November: SOFA Week 2022
SOFA v22.06 is out!
- Discover all features online
- Give us feedback and share your enthusiasm
End of Stephane’s internship :
- CUDA code incoming in a dedicated plugin
- Benchmarking code remains to do on several PR

Technical discussions

Update about the SofaMatrix project
- Refactoring is over
- 10% last percent (which are taking 90% of the time)
- A doc has been written
- New feature: able to check column and row value given when filling the matrix (data in the matrix assembler): no check, check system size, check submatrix, topology-aware check
- Maybe looking at a compatibility layer
- Later (after merge in master)
  - Modify all components with the new API (ongoing) Geometric stiffness : non-linear mapping which is itself non-mapped → need to differentiate linear and nonlinear mappings
  - Speed up of matrix mapping
  - Python bindings
  - Multithreading
  - CUDA compat
  - Test
PRs to be backported in v22.06 have been sorted, ready for backport v22.06.01

SOFA PR review

#3216 [Topology] Fix initTopology call chains for mesh without topology Test to fix → wip
#3176 [Core] Minor cleaning → ready
#3151 [Core] Make src an explicit link and support late update Still tests failing → wip
#3126 [Lagrangian.Model] Remove use of LateSpecialization pattern Try to find a design pattern to treat properly the specialization cases in SOFA. Vec3Type and Rigid3Type are working so far (but not taking CudaType). → keep to review
#3059 [Sofa.Core] Linear time getRoot() method in BaseNode and Node Test with a SPtr but not convincing and no way to guarantee the fact that there is no circular dependency → ready

SofaPython3 PR review

#293 Fixed Memory Leak in DataContainer.array() and DataContainer.writeable() Not sure about the fix, we may need to find another design avoid the new, discussion ongoing → keep to review

PR merged within the week

24/08/2022

News

SOFA Week is coming: request a place on stage and present your work
SOFA v22.06 is out!
- Download it here: https://www.sofa-framework.org/download/
- Discover all features online: https://www.sofa-framework.org/about/sofa-v22-06/
- Give us feedback and share your enthusiasm: https://github.com/sofa-framework/sofa/discussions/3215

SOFA PR review

#3230 [SolidMechanics] Remove useless data indice attribute in SpringFF Code removed is actually used → closed
#3227 [all] Set of warning fixes → ready
#3226 [LinearSolver.Direct] BTDLinearSolver: Clean debug informations and rename data Additional refactoring of message should be done later → ready
#3225 [CMake] FIX parent modules library consistency → ready
#3224 [core] remove shadow variables Ok to keep it as is, but naming should be improved and the for loop should be modernized. To be done in a following PR. → keep to review until next PR
#3221 [Sofa.Type] Constexpr Quaternion Feedback expected from @damienmarchal else ready. Move Quat::writeOpenGlMatrix in another PR. → keep to review
#3216 [Topology.c.dynamic] Fix initTopology call chains for mesh without topology CI says no, fix tests → wip
#3188 [objectmodel] Avoid extra copy in DataFileNameVector → ready
#3183 [objectmodel] Remove memory leaks → ready
#3151 [Core] Make src an explicit link and support late update Warnings and test failure to investigate. PR will be separated in two, keep only the src as a data → wip
#3063 [metis] Upgrade metis and GKlib PR done to check a metis-related issue. Potential source of error, mark it as postponed if needed later → closed

SofaPython3 PR review

#297 Fix warnings Review merged. Wait for CI then merged → ready
#294 Get info to try to fix the failing test Function "PythonFactory::toPython" finding the downcast where an error must occur. To be investigated → wip

BeamAdapter PR review

#50 [Mapping] Clean and optimize AdaptiveBeamMapping Configuration code could be set in SOFA, but we should agree on the way to use multithreading. → ready

PR merged within the week

17/08/2022

News

SOFA v22.06 is out!
- Download it here
- Discover all features online
- Give us feedback and share your enthusiasm

Technical discussions

Opening codes foster the robustness of these! Never hesitate to share your contributions to benefit from our dev engineering community!

SOFA PR review

#3224 [core] remove shadow variables Ask for more modern C++ changes → wip
#3223 [SofaImplicitField] Fix CMake for new NG architecture → ready
#3221 [Sofa.Type] Constexpr Quaternion → ready (wait for the 7 days)
#3217 [Type] Add NoInit constructor for Quat class → ready (wait for the 7 days)
#3216 [Topology.c.dynamic] Fix initTopology call chains for mesh without topology → ready
#3183 [objectmodel] Remove memory leaks PR looks really nice, wait for devs to take a look → keep to review (to be merged next week)
#3175 [Spring] RestSpringsForceField: Unify Vec/Rigid implementation Code for Rigid was in cpp, therefore when instantiating in CUDA you need to include the cpp. Factorization was needed. It would also ensure a better coherency of the code. The check for the DataType isRigid could be moved in sofa::defaulttype (in another PR) → ready
#3164 [SofaCUDA] Make CudaConstantForceField compatible with CudaRigid types → ready
#3155 [linearsolver] Create data links for preconditioners and ShewchukPCGLinearSolver → ready
#3152 [constraint.lagrangian] Create data links in ConstraintCorrection for linear solver → ready

SofaPython3 PR review

#296 [Plugin][CMake] Explicit RELOCATABLE parameter → ready and merged
#294 Get info to try to fix the failing test It is a PR for debug → wip
#293 Fixed Memory Leak in DataContainer.array() and DataContainer.writeable() Need review of @damienmarchal → keep to review

PR merged within the week

10/08/2022

News

SOFA Week 2022: preparation started and registration is open!
Release will be out by Friday, stay tuned!

Technical discussions

Check dashboard: strange macOS problem with access to matrices with SofaPython3. Apparently a binding to LCPConstraintSolver is needed while the binding to ConstraintSolverImpl does not suffice. To be further investigated
#3203 Topology changes for SphereCollisionModel on EdgeSetTopologyContainer The SphereCollisionModel works historically without a topology (relying on the mstate). Tests are failing due to the absence of a PointSetTopology, an alternative could be to change the warning into an info message that without a topology, no topological change will be supported. Another change in order to make the component topologically robust would be to make the data radius as a PointData

SOFA PR review

#3190 [objectmodel] Remove redefinition of initData → ready
#3188 [objectmodel] Avoid extra copy in DataFileNameVector Conflict on PR, when fixed PR is ready → keep to review
#3183 [objectmodel] Remove memory leaks PR looks really nice, wait for devs to take a look → keep to review
#3176 [objectmodel] Minor cleaning Test failing → wip
#3173 [MappedMatrix] MechanicalMatrixMapper: adds option → ready
#3164 [SofaCUDA] Make CudaConstantForceField compatible with CudaRigid types Restart CI #3163 [all] Remove code disabled in v21.06 → ready
#3162 [Component.Setting] Introduce PrintLog component to set printLog to all the graph components Agreed for a Node based behavior even if this would mean to pollute Node.h → suggestion to close
#3155 [linearsolver] Create data links for preconditioners and ShewchukPCGLinearSolver Use links directly instead of internal point variable → wip
#3152 [constraint.lagrangian] Create data links in ConstraintCorrection for linear solver Use links directly instead of internal point variable → wip
#3124 [Sofa.Type] Deprecate unsafe Vec constructor from pointer Crash on Image::getCorners. Check if related to the PR → wip
#3053 [Constraint.Lagrangian] Add the NNCG as NLCP solver in GenericCS Wait for CI to compile and then merge → keep to review
#3004 [SofaCUDA] Add ConstraintCorrection instantiation in CudaVec3f/CudaVec3f1 → ready

SofaPython3 PR review

#293 Fixed Memory Leak in DataContainer.array() and DataContainer.writeable() Need review of @damienmarchal → stay to review
#291 Fix minor warning → ready

PR merged within the week

27/07/2022

News

SOFA Week 2022: preparation started and registration is open!

SOFA PR review

#3167 [all] Remove pointer to itself in initData Poke @damienmarchal @JeremieA → wip
#3154 [SofaSparseSolver] Parallelize the assembly of the compliance matrix for the LDL solver Choice should be offered to the user to use a multithreaded approach, through a data e.g. Evaluate the overhead due to use of Task with only 1 thread. Same for #3093 → wip
#3151 [Core] Make src an explicit link and support late update Warnings and test failure to investigate → wip
#3142 [Core.Topology] Fix info message when Topology given to topologyHandler is not dynamic → ready
#3141 [SolidMechanics.FEM.HyperElastic] Reformat and clean TetrahedronHyperelasticityFEMForceField → ready
#3138 [Sofa.Type] Quat: Optimize rotate() (and inverseRotate()) → ready
#3124 [Sofa.Type] Deprecate unsafe Vec constructor from pointer Strange scene crash on Windows (image plugin), rebuild and see → keep to review and then merge (CI)
#3120 [CMake] Use standard macros + fix out-of-tree builds Reply needed on : #3120 (review) → keep to review
#3053 [Constraint.Lagrangian] Add the NNCG as NLCP solver in GenericCS Create separate functions in buildSystem. Avoid using else function with the d_resolutionMethod. Add a regression scene with unbuilt. → wip

SofaPython3 PR review

#289 Binding to access to the compliance matrix in constraint space W compliance matrix can be accessed, only MacOS tests are strangely failing. → ready (and investigate compilation on MacOS on dashboard)

PR merged within the week

20/07/2022

News

Twinical startup project got the 1st Prize of the ILab challenge
SOFA v22.06 release: postponed to August when Guillaume comes back. TODO before:
- Merge all v22.06-tagged PR
- Test a full build (custom job on CI) as for the Summer School
Open-source is a philosophy, get involved in your own way!

SOFA PR review

#3123 [GUI] Remove the reset button, replaced by reload scene Only the button reset is removed, used by no one. For the future, the work on improving the SOFA API has to be done to ensure a good initialization / update of components. A good test would be to run sequentially: storeResetState() and reset() → ready
#3120 [CMake] Use standard macros + fix out-of-tree builds This PR might be breaking for plugin compilation (including move of files), this PR changes the way we organize / name the code (which is good!). It would be therefore nice to rediscuss it with @guparan and @fredroy to all agree on the organization and merge it, but a question comes: should the PR be integrated in the release. → keep to review
#3115 [GL.Rendering3D] Clean up GL components → ready
#3109 [Helper] Make clearer from where plugins are loaded UX is not perfect : some plugins loaded are in blue in the GUI (the one loaded by the GUI itself). Even with a tooltip, the user can be a bit confused but this is a dev feature. An issue was created on the double mechanism of plugin loading #3147 → ready
#3105 [GUI.Qt] Move default format setup before application creation → ready
#3090 [All] Lifecycle update before v22.06 → ready and we will wait before backporting it
#3077 [SofaCUDA] Fix CMake to compile plugin without legacy headers option Remove BoxStiffSpringForceField (unavailable since not anymore in the compiled packages) → wip
#3059 [Sofa.Core] Linear time getRoot() method in BaseNode and Node Test using a smart pointer will be done and tested on the CI → wip
#3053 [Constraint.Lagrangian] Add the NNCG as NLCP solver in GenericCS Rename data as solverMethod Issue suggesting the factorization of ConstraintSolvers was created in #3148 → wip
#2988 [All] Propose a new gravity API Propose this new mechanism in a new issue to be discussed with all users / devs, see #3149 → back to wip

SofaPython3 PR review

#265 [SofaPython3] Add a LinkPath object in both the binding and plugin. Interesting proposal to have an object LinkPath for the path in the graph of objects and associated binding in SofaPython3. Unify the different ways to access this path. Simply the scene writing. → ready

PR merged within the week

13/07/2022

News

SOFA Website is down since Sunday due to an act of vandalism where our servers are located, see question posted on GitHub Discussions #3127
SOFA v22.06 release: due to the above point, the release process is delayed because of unavailability of CI infrastructure
Summer school "Deformation in robotics" was a success with more than 40 participants presenting their work in soft-robotics! Feedback very positive!
This Monday tool place the SOFA workshop at EMBC 2022 in Glasgow. Attendees were motivated, interested and many questions came up!

Technical discussions

LICENSE at SOFA root: we should find a way to display it without changing the file
Several PR were to be discussed before v22.06

SOFA PR review

#3119 [GUI.Qt] Save scene graph lock state persistently Save the state of the graph lock and remember it → ready
#3116 [Mass] Make warnings become info_msg in masses Make msg homogeneous with other l_topology, for now we accept the search in the graph and it removes the warnings in all scenes. For later, a SceneChecker for explicit link could be created instead of checking this component per component → ready
#3120 [CMake] Use standard macros + fix out-of-tree builds Init in Sofa.GUI is fine but else it should not be a plugin. Add a new macro sofa_add_library when it is a library not containing components (not loadable as RequiredPlugin) → wip
#3090 [All] Lifecycle update before v22.06 Change CMake to move GLOBAL loc → wip

SofaPython3 PR review

#284 [SofaPython3] Separates the pre-loaded python modules out of the python (Fix issue #283) → ready and issue created to add GHA test for SOFA use in a python interpreter #287

PR merged within the week

06/07/2022

News

Summer school "Deformation in robotics" was a success with more than 40 participants presenting their work in soft-robotics!
Next week
- SOFA at EMBC 2022 in Glasgow July 11th
- SOFA v22.06 should be released

Technical discussions

Issue of rendering in QtViewer brought on GitHub discussions #3100, branch being tested and seems to work

SOFA PR review

#3101 [GUI.Qt] Redirect Qt messages to the SOFA output stream Ready but not 7 day old → keep to review -#3093 [SofaLinearSolver] Parallelize the assembly of compliance matrix Cleaning of the code, solve the threadsafetyness of getValue() and work on SparseLDLSolver. Temporary vector used in a loop also not threadsafe. → wip
#3090 [All] Lifecycle update before v22.06 66 regressions failing on Ubuntu, rebuild ongoing → keep to review
#3084 [IO.Mesh] Fix binary loading in case of degenerated triangles Ready but on last change needed on #2999 → ready
#3080 [Core] Dynamic control of the computation of the bounding box Doc must be updated → ready
#3077 [SofaCUDA] Fix CMake to compile plugin without legacy headers option The CUDA version of BoxStiffSpringForceField will be removed and add warning when trying to instantiate removed component → wip
#3063 [metis] Upgrade metis and GKlib Ready but we keep it to review on purpose to postpone its merge after stable release v22.06 → keep to review
#3059 [Sofa.Core] Linear time getRoot() method in BaseNode and Node Ready only one pointer could be set as unique instead → keep temporarily to review
#3055 [Constraint.Lagrangian.Correction] Searching for Direct Linear Solver → keep to review
#3026 [runSofa] Add a button and a mechanism to activate the scenegraph updates → ready

SofaPython3 PR review

#268 Compilation without Compatibility Layer from SOFA Seems perfect! (poked guparan) → ready
#265 [SofaPython3] Add a LinkPath object in both the binding and plugin PR has some conflicts, issue created #282 → back to wip

BeamAdapter PR

Issue of rendering in QtViewer brought on GitHub discussions #3100, branch being tested and seems to work

PR merged within the week

Home

SOFA dev meeting reports
- 2026
- 2025
- 2024
- 2023
- 2022
Documentation
Devs