Re: [xine-devel] Gone with the mm_support

[Miguel F. <mi...@ce...>]

Hi Darwin,

On Mon, 2002-11-25 at 17:31, Darwin O'Connor wrote:
> The mm_support functioned wasn't defined anywhere in xine-lib that I 
> could find. It is not readily available amoung the libraries I use to 
> compile unix programs under OS/2. Looking on Google it seems to be 
> defined in mmx.h as an inline function. There is an mmx.h that comes 
> with xine, but it only includes xineutils.h. I'm not sure where the 
> linker is finding mm_support for you people compiling under Linux.

Try file src/libffmpeg/libavcodec/i386/cputest.c

 
> To get things to compile I put back the
> "#define mm_support() xine_mm_accel()". I haven't tested this yet. Is 
> xine_mm_accel compatible with mm_support? If so should xine_mm_accel be 
> used for everyone so other people won't run into this problem?


yes, it is the same. the reason i've changed it is because i'm slowly
trying to make our ffmpeg copy closer to the ffmpeg repository. moving
this function to cputest.c means less patches to dsputil.h.

...and you if are wondering why can't just we use their cpu detection
function instead... yes, that's possible of course, but our cpu
detection code is more generic (works for more architectures) and has
some additional checks (like SSE support from OS).

regards,

Miguel