Repository navigation
Use PyObject_CallOneArg where applicable - #3023
Conversation
|
This was actually part of my commit for 3020 before I pushed it up to GitHub. The reason I didn’t want to use this is because it’s not in the limited API, and I would still like to move towards compiling with the limited API. We could have a shim that converts this to the old thing in that theoretical case, but I just felt like this was a less clear win than NoArgs, also because NoArgs docs claim speed and OneArgs docs don’t mention it. |
|
I think this probably has the same level of "minor performance gain" as the other PR, but I didn't notice the limited API detail while making the PR |
9e8446a to
0020f72
Compare
Starbuck5
left a comment
Starbuck5
left a comment
There was a problem hiding this comment.
Alright, I'm okay with this since it's using the fancy vectorcall underneath.
We'll get some tiny unmeasurable speedups when these functions are utilized. But you're doing this more for code modernization, I get that.
I checked and the Rect performance increases I got from the vectorcall PR are preserved here.
MyreMylar
left a comment
MyreMylar
left a comment
There was a problem hiding this comment.
Alright LGTM 👍 Tests all pass locally, code replacement looks straightforward & correct to me
A PR that follows what #3020 did