drawspans optimizations in c?

[qbism]

[leileilol]

I tried earlier tonight. No luck, and things weren't even faster when I removed all of the stepping.

Isn't DrawSpans16 in assembler?

[mh]

[qbism]

[Spike]

[mh]

[qbism]

[qbism]

[leileilol]

[qbism]

17fps vs 12.8fps: That's a lot of FPS still on the table. I wonder if any typical compiler settings could be screwing up the unroll benefit? Possible that unoptimized could be faster in this loop. Or maybe related to 486 instruction set?

[mh]

I think it's just the 486; I've run Quake on one of those before and it's a miracle to be getting FPS above single digits.

My benchmarks came from a timedemo demo1 by the way.

It's not really Duff's either as it doesn't have a do...while embedded in the switch, but then again since we already know there's an absolute maximum on spancount it doesn't need to.

[frag.machine]

[qbism]

Tute!

[frag.machine]

[mh]

There's some floating point division further up in the loop that I'd like to have a pop at doing something about too. I suspect that might be causing more performance issues.