Alberto
Alberto
New performance metrics with the added _pg_do_notnormalized_rects_intersect() function : (notice how the 4 args not normalized case went from 0.9083898999997473s down to 0.742086800000834s and the total time from 9.983s down...
New stats after latest commit: ``` --------- Not normalized Colliding rects --------- Rect: 0.40994370000044 4 args: 0.6392197999994096 2 args: 1.141009800001484 1 tuple arg: 1.0302646000018285 --------- Normalized Colliding rects --------- Rect:...
Latest test ``` --------- Not normalized Colliding rects --------- Rect: 0.40383030000157305 4 args: 0.7056063999989419 2 args: 1.1588473999981943 1 tuple arg: 1.0563916999999492 --------- Normalized Colliding rects --------- Rect: 0.40069299999959185 4 args:...
Latest test: ``` --------- Not normalized Colliding rects --------- Rect: 0.38273279999884835 4 args: 0.6658422000000428 2 args: 1.1407941999987088 1 tuple arg: 0.9429583000001003 --------- Normalized Colliding rects --------- Rect: 0.4201164999994944 4 args:...
Super cool! Yeah Sure i can take a look with my "fancy graphs" lol. Also i've never used premultiplied as a blend... how does it work? is it some sort...
Oh ok now i'm interested, so for now i can't use premultiplied alpha for images i've loaded with pygame.image.load? how does it interact with `.convert_apha()`? Could we implement something like...
I'll post some results here so if people want to examine them they easily can. PS. My setup: Windows 11, Ryzen 5 3600X 3.8-4.2GHz AVX2, SSE2 capable, NVIDIA RTX 2070...
Oh god what did i start here... And it's curious to see that the alpha blit (in the last graph) goes so much up and down... is there a problem...
An update on this. If we want to optimize many functions inside math.c, i will need new API functions that are more performant. That is because using new API reduces...
I personally dig the idea. Always nice to have frequently used functions in pygame C api