[mypyc] Use faster METH_FASTCALL wrapper functions on Python 3.7+ #9894
Conversation
|
LGTM, just one suggestion. |
| # This is because CPyArg_ParseStackAndKeywords format string requires | ||
| # them grouped in that way. | ||
| groups = make_arg_groups(real_args) | ||
| reordered_args = reorder_arg_groups(groups) |
ilevkivskyi
Jan 18, 2021
Collaborator
These preparatory steps here are very similar to the legacy wrapper below. Would it make sense to factor them out in a helper?
JukkaL
Jan 23, 2021
Author
Collaborator
Refactored some of the shared code. I didn't share everything, since I'm planning further changes that may only be relevant for new-style wrapper functions.
JukkaL
added a commit
that referenced
this pull request
Jan 23, 2021
Allocate a vectorcall function pointer as a struct field for native classes that include `__call__`, including nested functions. This lets us use METH_FASTCALL wrapper functions with `__call__` methods. See https://www.python.org/dev/peps/pep-0590/ for details of why we jump through these hoops. This makes the `nested_func` microbenchmark about 1.5x faster. Follow-up to #9894.
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Outdated suggestions cannot be applied.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.
Implement faster argument parsing based on METH_FASTCALL on supported
Python versions.
Use
vgetargskeywordsfastextracted from Python 3.9 with some modifications:*argsand**kwargsThe modifications are very similar to what we have in the old-style
argument parsing logic.
The legacy calling convention is still used for
__init__and__call__. I'll add__call__support in a separate PR. I haven't looked into supporting__init__yet.
Here are some benchmark results (on Python 3.8)
However, the above benchmarks are still slower when compiled. I'll continue
working on further improvements after this PR.
Fixes mypyc/mypyc#578.