Poisson reconstruction - experiments and speedup using parallel Mesh_3 #7891
Conversation
It is threadsafe, but it still isn't optimal within each thread: when computing a refinement point for Mesh_3, the oracle will be successively asked about points that are close by, but when we switch from one dichotomy to the next, the hint is likely terrible.
With a static hint and re-initialization in the constructor, we could still get issues if someone did something like: Poisson poisson_1(...); Poisson poisson_2(...) mesh(poisson_1); mesh(poisson_2); // problem: the static hint points to a cell of T1
|
If might be interesting to look again at #4996 (in term of runtime + still crashing) |
...e_reconstruction_3/examples/Poisson_surface_reconstruction_3/poisson_and_parallel_mesh_3.cpp
Show resolved
Hide resolved
|
Successfully tested in CGAL-6.0-Ic-129 |
|
Is this branch still experimental or fixes done already a good addition to have? |
|
We have another potential improvement in the pipeline, and I would rather keep it all in the same branch. |
|
Would that invalidate any of the fixes here? |
|
Potentially yes, it's still related to hints. |
…ois/cgal into Poisson-use_parallel_mesh_3-GF
|
Successfully tested in CGAL-6.0-Ic-282 |
lrineau
added
rm only: ready for master
|
This pull-request was previously marked with the label |
Any update? |
|
@janetournois could you please update CHANGES.md so that it can be merged for 6.0.1? |
|
Successfully tested in CGAL-6.0.1-Ic-346 |
sloriot
added
Tested
CHANGES.md not updated
and removed
Under discussion
Under Testing
labels
|
This pull-request was previously marked with the label |
Summary of Changes
Describe what your pull request changes to CGAL (this can be skipped if it solves an issue already in the tracker or if it is a Feature or Small Feature submitted to the CGAL Wiki).
Release Management