[Scummvm-devel] Re: SDL Altivec blitters

[Bob Ippolito]

If you take a look at the svn repository, some 565 blitters were 
implemented, but nobody provided me with good tests for them so I 
really can't say if they're faster or not.  1555 would be easy to 
implement, but it's hard to justify without a good way to test it.

-bob

On Mar 5, 2005, at 7:59 PM, Max Horn wrote:

> Hi folks,
>
> I just stumbled by chance over the SDL Altivec work you've been doing 
> (because I found out on veriontracker.com that a new version of Frozen 
> Bubble had been released by Bob, which I like a lot, then read about 
> this "Altivec SDL", which prompted me to google for it, after which I 
> read the whole thread on the SDL list and to download the various 
> patches). Looks very cool, thanks for your efforts!
>
> Anyway, as somebody else mentioned on the SDL list, a popular app 
> using 16 bit mode only is ScummVM, of which I am project lead. It 
> certainly would benefit quite some from an accelerated 1555 mode in 
> SDL; while we currently use a 565 surface on OSX, the only reason for 
> that is that this is what SDL gives us when we ask for an "optimal" 16 
> bit surface. I can easily change it to specifically request a (1)555 
> surface from SDL, as our frontend code is specifically written to 
> support both 555 and 565 modes.
>
> I did spend some time speeding up our Scale2x/Advmame2x, Hq2x/Hq3x 
> etc. scalers, even using some AltiVec, but stopped at some point 
> because more time was spent in SDL than in our own code anyway...
>
> Anyway: The reason of my mail is mostly meant as a "heads up", as I 
> have no time to look into writing 16<->32 blit code myself; but in 
> case anybody else is interested in doing that, I'll be happy to do 
> testing with ScummVM, and to modify ScummVM to use the (1)555 mode for 
> this.
>
> Please keep up the good work,
>
> Max
>