Made Surface.set_at() fastcall

[itzpr3d4t0r]

This PR aims to get a quick and easy perf improvement with fastcall as well as removing a redundant check.

Results:
TLDR, about 34-40% improvement on my machine.

OLD
mean: 0.226506186666514
mean: 0.8316974133334346
mean: 3.5231991799999682
NEW
mean: 0.16023285333330933
mean: 0.6178482066666523
mean: 2.6118607199999437

Code Used:

from statistics import mean
from timeit import repeat

import pygame

pygame.init()

screen = pygame.display.set_mode((10, 10))

c = (255, 23, 1)
G = globals()
reps = 15
positions = [
    (0, 0), (1, 0), (0, 1), (1, 1),
    (0, 0), (1, 0), (0, 1), (1, 1),
    (0, 0), (1, 0), (0, 1), (1, 1),
    (0, 0), (1, 0), (0, 1), (1, 1),
]

color = pygame.Color(255, 23, 1)


def test(stmt: str):
    data = repeat(stmt, globals=G, repeat=reps)
    print(f"mean: {mean(data)}")


test("screen.set_at((0, 0), color)")
test_str = """
screen.set_at((0, 0), color)
screen.set_at((1, 0), color)
screen.set_at((0, 1), color)
screen.set_at((1, 1), color)
"""
test(test_str)
test_str = """
for pos in positions:
    screen.set_at(pos, color)
"""
test(test_str)

pygame.quit()

[ankith26]

set_at is typically called in tight loops, so I think this optimisation is pretty good.

The CI is failing on a warning, which I'm sure you figured by now, just needs an explicit cast.

[ankith26]

LGTM now, thanks! 👍

[Starbuck5]

This conflicts with Scriptline's PR #2111.

Lets try to get that in first, then you can rebase this against that?

Edit: resolved, earlier PR merged.

[zoldalma999]

I'll convert it draft for now, feel free to open it back when the mentioned pr is merged

Resolved.

[ankith26]

To the person merging this: squash and merge this PR so that it doesn't conflict with my other PR. It would probably add me as a co-author of this PR, which you can remove if you want

[zoldalma999]

Looks good to me, thanks! Nice speed ups!