Re: [AD] Clear MMX Improved !

[ Thread Index | Date Index | More lists.liballeg.org/allegro-developers Archives ]


varobert@xxxxxxxxxx writes:
> Ok, here it is. It's a majorly bug fixed version of MMX clear for both 
> 16-bit and 8-bit color depths ! Speed hasn't changed since the old 
> version, but this one should work correctly now.

Great.

> Replace iblit8.s and iblit16.s with these ones. These should be applied 
> *AFTER* JOSE ANTONIO LUQUE's patch (as he's using my old version).

This is why it's usually better to send changes in patch format: if you 
make this as a diff relative to whatever you last sent me, I'd be able to 
apply patches from both you and Jose in either order, even if you are 
both working on the same files! It really does make life easier (for me, 
but also for you since you don't need to work so hard at coordinating all 
the changes).

> I have 2 versions of the code: one with segment decode and one without 
> (as it's faster on PMMX and prolly K6 too because there's no decode to 
> do).

I thought you decided that this wasn't a big enough difference to care 
about? Not a huge deal, but with both versions included the code looks 
awfully large and complicated just for a memory clearing function :-) 
IMHO it would be better to stick with the simpler single-case version 
unless you have some actual tests that show this being a significant 
performance difference (remember that memory, and vram in particular, is 
an order of magnitude slower than the CPU, so it is quite possible that 
saving the odd cycle here and there won't actually affect the performance 
at all).


-- 
Shawn Hargreaves - shawn@xxxxxxxxxx - http://www.talula.demon.co.uk/
"A binary is barely software: it's more like hardware on a floppy disk."



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/