Need to further optimize render routine to hit 64-sprite SAT about jcap HOT 5 CLOSED

from jcap.

@konimaru Sounds good; hope you're having a great time!

The code in question is vga_render.spin on the wip/video/sprites branch.

I think if time can be shaved off anywhere it's in either the sprite visibility check section or the sprite tile line parsing section.

Like I said above, bumping up the number of render cogs from 5 to 6 (thus overwriting the SPIN interpreter in Cog 0) lets me hit my goal as well as give me time for some extra effects like fine and parallax scrolling.

So unless you see a really critical optimization, to the tune of hundreds of instructions per line, I think the code is in its almost final state.

Thanks!

from jcap.

OK, I'm back to my daily grind. Is that vga_render.spin still current, i.e. can I work with/on it?

from jcap.

@konimaru welcome back! You'll want to look at HEAD~1 of wip/video/sprites for the last working version of render. In the most recent commit I started working on getting fine sprite scrolling down. Take a look here for specifics on that line of effort: https://forums.parallax.com/discussion/168164/shifting-consecutive-longs

If you look at the most recent commit, all the commented out code will certainly be removed in place of whatever algorithm I implement to populate the buffer w/ fine scrolling.

from jcap.

Tentatively satisfied by commit 7de0fe5

from jcap.