phat code If brute force doesn't solve your problems, then you aren't using enough.
Main

Projects

Downloads

Articles

Links

Forum

 

View Message

Back to Messages
Optimus Tue Jun 25 2024 at 4:36 pm
Oh hai
 
 
I think I tried with ffix or without and brun45 is still faster.

And yes, I realized I shouldn't def seg outside or inside function and expect in another one or outside to persist.

I am looking at more quirks now thoroughly, looking at the assembly. Like I had a rotozoomer doing 89fps and dropped at 56 because of something innocent. I used to have my texture as DIM texture%(0 to 31, 0 to 31), then I decided to become more "clever" and start using CONST texWidth% = 32, then I did DIM texture%(0 to texWidth%-1, 0 to texHeight% - 1) and suddenly I look at the produced assembly, it reads the width from some offset near the array, then does IMUL, instead of the usual left arithmetic shifts. I did learn before that I should have used CONST for my old demos when I didn't want to use magic numbers but define them to be more clear (and also it would tell the compiler to not read the variable but use directly the number, as I see in the asm output) but here is an exception for an array size.

Anyway,. maybe I look too hard. Meanwhile every morning before work I somehow like to work on this, I had a lot of progress, have now gouraud, pseudo-env mapped phong, at a good speed for Pentium 1. I posted some previews in discord. Maybe I should make a demo with it at some point.
 
 
 
 

Reply to this Message

Name
Subject
Message

No HTML is allowed, except for <code> <b> <i> <u> in the message only.
All URLs and email addresses will automatically be converted to hyperlinks.