Memory and performance improvements#358
Merged
daleroberts merged 1 commit into1.4.0from Apr 22, 2026
Merged
Conversation
daleroberts
force-pushed
the
dr-344
branch
2 times, most recently
from
28c9f2e to
3b6fdd9
Compare
- Matrix class: packed/symmetric storage, attach-to-mmap, calloc allocator - LM solver + Anderson acceleration CLI - Static build fixes (glibc 2.34+ pthread) - Matrix test suite
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 core of the change is a rework of the matrix_2d class to use packed and symmetric storage modes, a faster 3x3 copyelements, LAPACK bindings for the packed routines (dpptrf, dpptri, dspmv, dsymm, dpotrs), and bounds-check asserts on every element accessor.
There are also a way to tune threading with
--max-threadsas testing has shown that it's not always optimal to use all available threads (e.g., on NCI).Optimisations have also been made for
--stagedmode. For example, matrix allocator path now supports attach-to-mmap and calloc-backed buffers. Further, there is now a--stage-pathoption so the path location of where the mmap files are stored can be set. This allows, for example, to use the/iointensivepath on NCI.Finally, an optional Levenberg-Marquardt / trust-region solver and an Anderson-acceleration stage has been been included that mixes the last few iterates to speed convergence. This needs to be tested more and configured by various flags:
--lm-enabled,--lm-lambda-init,--lm-eta-good,--lm-eta-accept,--lm-gamma-up,--lm-gamma-down,--lm-max-rejects,--anderson-acceleration,--aa-depth.