[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
On Sun, 2005-01-30 at 11:45 +0100, Evert Glebbeek wrote:
> On Saturday 29 January 2005 00:27, Robert Ohannessian wrote:
> > Changing the vtable for some of the more common cases is the better
> > idea. We can start by using memcpy() for those cases, and then later on
> > add assembly versions.
> >
> > Better we do this as a multi-step process.
>
> Ok, so basically, post 4.2 material?
> I'm still tempted to want as many improvements in 4.2 as possible, but at
> some point it's going to be feature creep. Do we draw the line here and
> don't work on this to get it in 4.2?
> Much as I'd like to look at this now, I think we probably shouldn't...
>
We could probably apply the memset patch for now, it speeds up clearing
to color 0 in 15/16/24 bit color depth.
Then, for 4.2, anyone with asm skills (Bob or Angelo or AJ) can write a
non-line-oriented clear_to_color and blit and masked_blit. As I see it,
alignement just needs some padding bytes at the end of every line, so
the overhead of the current implementation which Bob pointed out could
still be mostly removed for such bitmaps.
--
Elias Pschernig