ENH: minimizer performance improvements #424
Merged
Conversation
This file contains 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
With profiling, this corresponds to roughly a factor of 2 performance improvement in a ternary benchmark system 1. Don't recomputed ALLOWED_MASS_RESIDUAL on the hot loop. We can compute it in the initializer if we assume the prescribed elemental amounts is constant 2. Replace np.sum() with manually computed phase composition sums in SystemState initializer and recompute() 3. Replace np.all call in consolidation check for remove_and_consolidate_phases The only thing left showing up in the profile inside the run_loop function is np.take, np.nonzero, and np.argsort, which only appear inside change_phases()
bocklund
changed the title
Codecov Report
@@ Coverage Diff @@
## develop #424 +/- ##
===========================================
+ Coverage 90.22% 90.27% +0.04%
===========================================
Files 50 50
Lines 7696 7723 +27
===========================================
+ Hits 6944 6972 +28
+ Misses 752 751 -1
📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more |
richardotis
reviewed
Jul 24, 2022
|
@richardotis I'm happy with this PR where it is if you want to take another quick pass before merging. I'm not ready to increase the scope for convergence related things yet, so this one can just be performance focused. |
richardotis
approved these changes
Aug 3, 2022
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Outdated suggestions cannot be applied.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.
Profiling-guided removal of (relatively) expensive NumPy calls
With a couple ternary benchmark systems, I'm seeing a consistent 2x speedup in the
run_looptime and almost all the profiled code that drops back into Python (from Cython) is gone.Summary of changes:
np.sumwith manually computed phase composition sums in SystemState initializer and recompute()np.allcall in consolidation check for remove_and_consolidate_phasesnp.take,np.nonzero,np.argsortwith loopy equivalents (this is less of a speedup, but we probably just hit this code path less in the benchmark and in practice)The intent was to preserve exactly the current functionality and just be pure performance improvements (any differences detected in review are not intentional).
Benchmarks
ebcfbdb (develop)
b3668f4