More efficient way of creating condensed sparsity pattern #436
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
Codecov ReportBase: 91.37% // Head: 92.00% // Increases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## master #436 +/- ##
==========================================
+ Coverage 91.37% 92.00% +0.63%
==========================================
Files 22 22
Lines 3258 3703 +445
==========================================
+ Hits 2977 3407 +430
- Misses 281 296 +15
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
fredrikekre
reviewed
Mar 8, 2022
fredrikekre
reviewed
Oct 3, 2022
KristofferC
reviewed
Oct 5, 2022
| # Store linear constraint index for each constrained dof | ||
| distribute = Dict{Int,Int}(acs[c].constrained_dof => c for c in 1:length(acs)) | ||
|
|
||
| #Adding new entries to K is extremely slow, so create a new sparsity triplet for the condensed sparsity pattern | ||
| N = length(acs)*2 # TODO: Better size estimate for additional condensed sparsity pattern. | ||
| I = Int[]; resize!(I, N) |
There was a problem hiding this comment.
Could use sizehint! here instead and then use push! instead of setindex!.
There was a problem hiding this comment.
Yes, and perhaps in _create_sparsity_pattern() aswell? How much does push! Grow the vector if it grows beyond the size hint?
fredrikekre
added a commit
that referenced
this pull request
Nov 25, 2022
This patch removes some unnecessary allocations of a sparse matrix column which was needed before when the sparse matrix changed in-place. Since #436 we create a new matrix for the new entries and then add them in the end instead.
fredrikekre
added a commit
that referenced
this pull request
Dec 10, 2022
After #436 the condensation of the pattern is done by creating a new matrix which is added to the original matrix. Before this patch there is still a check for whether each new entry already exist in the original matrix before adding it to the new matrix. With the new approach implemented in #436 this seems unnecessary and this patch removes the check. This also removes some extra complexity from the code. A datapoint, which support this, is the Stoke's flow example in the documentation. In that problem, the condensation adds 32k new entries, of which 30k are new, and 2k exist in the original matrix. Checking whether the 32k elements exist is much more expensive than simply including the extra 2k entries in the new matrix. The new approach reduces the time for creating the combined matrix from 4ms to 3.2ms. Matrix creation isn't a bottleneck by any means, but it is nice to see that we with simpler code also get better performance.
fredrikekre
added a commit
that referenced
this pull request
Dec 11, 2022
After #436 the condensation of the pattern is done by creating a new matrix which is added to the original matrix. Before this patch there is still a check for whether each new entry already exist in the original matrix before adding it to the new matrix. With the new approach implemented in #436 this seems unnecessary and this patch removes the check. This also removes some extra complexity from the code. A datapoint, which support this, is the Stoke's flow example in the documentation. In that problem, the condensation adds 32k new entries, of which 30k are new, and 2k exist in the original matrix. Checking whether the 32k elements exist is much more expensive than simply including the extra 2k entries in the new matrix. The new approach reduces the time for creating the combined matrix from 4ms to 3.2ms. Matrix creation isn't a bottleneck by any means, but it is nice to see that we with simpler code also get better performance.
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.
Creating the sparsity pattern with affine constraints was extremely slow.
A lot of %gc time still which I dont understand