Blitting BOBs line by line

[aros-sg]

Has anyone ever tested if this would make sense in some cases (maybe many planes, maybe AGA, maybe CPU >= 020). For example for 32x32 bobs. Image data still interleaved, but only single plane mask so saving memory. For each line cpu would read mask into bltafwm, bltadat, bltalwm. Only once per line. Not once per plane-line as blitter would need to do for a standard BOB blit.

[roondar]

Interesting idea. I'd expect that to be a bit slower than blitting as normal due to the large number of blits you're doing, but if it works you can certainly save some memory over standard interleaved blitting. The only real question is whether or not going for straight non-interleaved blitting might not be faster if the goal is to simply save on mask memory.

[DanScott]

Originally Posted by aros-sg

Has anyone ever tested if this would make sense in some cases (maybe many planes, maybe AGA, maybe CPU >= 020). For example for 32x32 bobs. Image data still interleaved, but only single plane mask so saving memory. For each line cpu would read mask into bltafwm, bltadat, bltalwm. Only once per line. Not once per plane-line as blitter would need to do for a standard BOB blit.

Why don't you try it out and report back with your findings ?

[aros-sg]

Originally Posted by DanScott

Why don't you try it out and report back with your findings ?

I'm not really coding much of anything at all these days.

Memory saving -> mask also would not need to be in chip ram. That would be advantage over std plane by plane blit, if otherwise speed would end up being similar. no idea. Cpu also can read mask faster from memory than blitter I guess (in addition to needing to read it only once per line, not once per planeline as said), so maybe there is a slight chance that in some cases it would be faster than std interleaved (with interleaved=multiplane mask) bob blit.

[roondar]

Hmm, using Fast Memory to store the mask might improve things. And yes, you'd only need to set the mask once per line so you do save some memory reads by the Blitter this way. Still, repeatedly calling and waiting on the Blitter to do the line-by-line blits does take some time in itself, so yeah - I don't honestly know what the result would be here. I guess only experiments can really show how this ends up working one way or the other.

[sandruzzo]

If you have a bob that fit perfectly into bitmap, you can skip sources setting since blitter, after each blit will point next line