Re: [AD] Bug in Allegro's color convertors?

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


Eric Botcazou wrote:
Yes, but that bug will *never* show up unless someone runs a 32-bpp
Allegro prog in a window with a desktop of 24-bpp. I'd rather be 'correct'
when doing color conversion at that level.


It's a loss performance-wise to do it at that level, because we do it for
each refresh, whereas the user could do it only when necessary.

We can write back into the screen a 0 value for the alpha channel. Or we can change makecol to detect this particular setup. I'd rather not have users insert MID(0, x, 255) for every parameter of makecol!

What if they're explicitly using makeacol? Then the alpha channel will be filled with some info, which we don't want to show, especially not on *other* pixels!

If you can, I'd also like for you to benchmark old vs new convertor on a non-MMX system. Although the non-MMX code is 2-2.5 cycles longer, it doesn't have partial loads from memory and uses one less memory operation, so that may give some performence improvements.

[snip]

--
- Robert J Ohannessian
"Microsoft code is probably O(n^20)" (my CS prof)
http://pages.infinit.net/voidstar/



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