Draft
Conversation
When algorithm operates by modifying the input array in place, it is good manners to make a copy first so that the function does not mutate user input. I am guessing that that was the motivation for the use of `copy()` here. However, as far as I can tell, this algorithm does not modify `signal` (or its alias `a`) in place. It reassigns the variable `a` in a loop, but does not actually modify the input array object. Therefore, we can remove this copying and increase the speed. Also, note that the call to `np.array` explicitly cast the input to a literal numpy array. By removing that from the code, we accept dask arrays and cupy arrays and allow them to flow through the algorithm via the NEP-18 numpy dispatch mechanism.
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This builds on the commits from #12, demonstrating support for sparse arrays.
In this initial draft, I resort to a copy-paste of
multitau.pyto make some sparse-specific changes inmultitau_sparse.py. Using current (and also possible future) numpy protocols, it ought to be possible to support bothnumpy.ndarrayandsparse.COOarrays in the same codepath.The memory usage is of the sparse implementation about 1/3 of the dense implementation for a 2048x2048 image with 1% density.
In exchange for that memory efficiency, the sparse version runs roughly half as fast. We may see greater speed gains for choices of lags and levels that require more multiplication.