Use vectorcall for all-positional-argument calls #5896
Conversation
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
If a handle or object is called with only positional arguments, it is
straightforward to use PyObject_Vectorcall instead of
PyObject_CallObject.
Benchmarked by adding a trivial function to pybind11_benchmark:
```
m.def("call_func_with_int", [](py::object func) {
return func(py::cast(1));
});
```
and then running `python -m timeit --setup 'from pybind11_benchmark import call_func_with_int; f = lambda x: x + 1' 'call_func_with_int(f)'`.
Before on M4 mac: 57.6 nsec per loop
After on M4 mac: 48.4 nsec per loop
For comparison, the included collatz benchmark takes 33.1 nsec per
loop, just calling `f(1)` directly takes 17.8 nec per loop, and simply
running `pass` takes 4.19 nsec per loop.
swolchok
commented
Nov 14, 2025
include/pybind11/cast.h
Outdated
| // Disable warnings about useless comparisons when N == 0. | ||
| PYBIND11_WARNING_PUSH | ||
| PYBIND11_WARNING_DISABLE_GCC("-Wtype-limits") | ||
| PYBIND11_WARNING_DISABLE_INTEL(186) |
Contributor
Author
There was a problem hiding this comment.
not sure why suppressing the icc warning didn't work :(
Contributor
Author
|
FWIW, I attempted to extend this to use vectorcall for the unpacking/kwargs/etc. cases, but my naive/straightforward attempt ended up adding too much cost for kwarg handling to end up actually improving performance. (the tuple unpacking case in particular is bottlenecked on args_proxy having to use slow PyIter-based iteration instead of PyTuple_GET_ITEM as currently written, so we couldn't help there either even though it doesn't use kwargs.) |
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.
If a handle or object is called with only positional arguments, it is straightforward to use PyObject_Vectorcall instead of PyObject_CallObject.
Benchmarked by adding a trivial function to pybind11_benchmark:
and then running
python -m timeit --setup 'from pybind11_benchmark import call_func_with_int; f = lambda x: x + 1' 'call_func_with_int(f)'.Before on M4 mac: 57.6 nsec per loop
After on M4 mac: 48.4 nsec per loop
For comparison, the included collatz benchmark takes 33.1 nsec per loop, just calling
f(1)directly takes 17.8 nec per loop, and simply runningpasstakes 4.19 nsec per loop.Suggested changelog entry: