gh-132657: Add free-threading scaling benchmark for copy.copy and copy.deepcopy - #132658
gh-132657: Add free-threading scaling benchmark for copy.copy and copy.deepcopy#132658eendebakpt wants to merge 4 commits into
Conversation
Sorry, something went wrong.
colesbury
left a comment
There was a problem hiding this comment.
This looks good, but in general my preference is to add these benchmarks when we fix the scaling issue. I think of these like regression tests, and having a mix of "passing" and "failing" cases makes it harder to notice regressions.
Sorry, something went wrong.
Co-authored-by: Sam Gross <colesbury@gmail.com>
|
Thanks for the feedback. I suspect we will only need one of the two benchmarks as a regression test (as the reason for not scaling seems to be the same). Once we have a better scaling, I will remove one of the tests. |
Sorry, something went wrong.
This is taken from the PR pythonGH-132658.
|
@colesbury The deepcopy benchmark was already merged in #142843. The copy benchmark overlaps for a large part with the deepcopy benchmark (e.g. both have the same potential FT bottlenecks). Unless there are reasons to add the copy as well I suggest we close this PR. |
Sorry, something went wrong.
|
Sounds good to me |
Sorry, something went wrong.
In this branch two benchmarks for the free-threading scaling are added. They are not completely microbenchmarks (they test multiple aspects of the free-threaded performance).