SIMD surface.fill

[MightyJosip]

This should improve the performance of simple surface.fill

Test code:

from timeit import repeat
import pygame

pygame.init()
surf = pygame.Surface((800, 600))
G = globals()

teststr = "surf.fill((24, 24, 24))"
l = [min(repeat(teststr, globals=G, number=1000, repeat=10)) for _ in range(5)]
print(f"fill: {sum(l) / len(l)}")

My results

pygame-ce 2.5.0.dev1 (SDL 2.28.5, Python 3.11.0)
fill: 0.03824264000868425

pygame-ce 2.4.0 (SDL 2.28.5, Python 3.11.0)
fill: 0.08401216000784188

[itzpr3d4t0r]

Cool to see you venture into SIMD! Unfortunately I think you just found out something I've discovered a while ago, that is going for an AVX/SSE naive strategy for optimizing the fill algorithm (see some of my findings here: #2390). That strategy is fine for small blits but it doesn't scale whell with surface size due to cache pollution. In the graph below the current best SDL version is the SDL (SSE) one that uses stream instructions to minimize cache pollutions and what you are going to is the AVX2 naive one. This should be investigated further but I'm pretty sure this is the case.

[MightyJosip]

oh yea you are right