Fix slow parallel checkpoint read (COMM_SELF + O(N^2) metadata loop) - #229
Merged
Merged
Conversation
The restart reader opened the BP file on MPI_COMM_SELF, so every rank independently parsed the full global metadata instead of a single rank-0 read + broadcast. Phase 1 also had every rank loop over all g_ndomains reading each subdomain's extent/ncells via per-element Mode::Sync Gets. Both costs scale ~O(N_ranks^2) and made restart reads dominate wall time at large rank counts (~3 h at 65536 ranks vs a 2-3 min write). - Open the reader on MPI_COMM_WORLD so ADIOS2 reads/broadcasts metadata collectively and can aggregate reads. - In Phase 1, each rank now reads only its own subdomain entry and MPI_Allgathers the ncells/xmin/xmax arrays; global_extent and the reconstruction check are computed from the gathered layout. Non-MPI path unchanged. Verified ~50x faster on Frontier (read 209 s vs ~3 h) in the sibling entity_bh tree. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This was referenced Oct 5, 2026
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The restart reader opened the BP file on MPI_COMM_SELF, so every rank independently parsed the full global metadata instead of a single rank-0 read + broadcast. Phase 1 also had every rank loop over all g_ndomains reading each subdomain's extent/ncells via per-element Mode::Sync Gets. Both costs scale ~O(N_ranks^2) and made restart reads dominate wall time at large rank counts (~3 h at 65536 ranks vs a 2-3 min write).
Verified ~50x faster on Frontier (read 209 s vs ~3 h).