:( I think my post was ignored....
I reiterate a few points. There is -NOT- the same amount of data testing in a 32x32 vs 32x32 images collide and a 32x32 vs 1024x768 test. The main reason for this is that pixels checked are only the ones that actually overlap. With your small images, I'm willing to bet that most of the time they don't overlap at all. With the large vs small image, they always overlap 100% (if the small image is on screen at all). Even with the collision layers stuff, this is true.
Reguardless of which collideimages type command you use, you -will not- get much faster with pixel perfect collision. It's just too many comparisons done on the CPU.
I suggest, once again, that you look into alternate collision methods.
I reiterate a few points. There is -NOT- the same amount of data testing in a 32x32 vs 32x32 images collide and a 32x32 vs 1024x768 test. The main reason for this is that pixels checked are only the ones that actually overlap. With your small images, I'm willing to bet that most of the time they don't overlap at all. With the large vs small image, they always overlap 100% (if the small image is on screen at all). Even with the collision layers stuff, this is true.
Reguardless of which collideimages type command you use, you -will not- get much faster with pixel perfect collision. It's just too many comparisons done on the CPU.
I suggest, once again, that you look into alternate collision methods.