Repository navigation
Wrap raw Vec handles with PetscVec and split the MPI launch - #332
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CI on main is failing again on the Julia 1 jobs. PETSc.jl 0.5.5 deprecates
VecPtrentirely, so the keyword form #330 switched to also prints a one-time warning ("VecPtr goes in v0.6: wrap a raw Vec with PetscVec(petsclib, ptr; own)"), and that record shows up in every test that captures logs or counts warnings. CI now resolves PETSc.jl 0.5.6.PETScCompat.VecPtrnow callsPetscVec(pl, ptr; own = false)on PETSc.jl 0.5.4 and later. That method exists from 0.5.4 and returns the sameVecPtrwrapper as before, so nothing else changes. None of the other names deprecated in 0.5.5 and 0.5.6 are called from src or ext.The 3-rank MPI job on Julia 1 was also being killed by the one-hour guard around each
mpiexeclaunch: the one-thread launch ran seven scripts and took just over an hour on current runners, on unchanged main as well. That launch is now two,explicit/implicit/adjointanddm/types/second_order/exit, each with its own guard.all.jlalready loads each script in its own module, andexit.jlstill runs last in its launch. No test content changes.With PETSc.jl 0.5.6, solves log no warnings (the same check on unchanged main captures the warning above), and the solves still pass with PETSc.jl 0.4.14, which the Julia 1.10 jobs resolve. All 17 jobs pass on the fork, with the 3-rank Julia 1 MPI job at 40 minutes: https://github.com/singhharsh1708/PETScDiffEq.jl/actions/runs/41192091938
AI Disclosure: Claude assisted with the code. Testing and review by @singhharsh1708.