[Ffmpeg-devel] [PATCH] snow mmx + sse2 part 3
Mon Mar 20 23:29:32 CET 2006
On Mon, 2006-03-20 at 20:47 +0200, Oded Shimon wrote:
> On Mon, Mar 20, 2006 at 11:54:18AM -0500, Robert Edele wrote:
> > Here's part three of my snow asm patch, which covers the mmx and sse2
> > implementations of ff_snow_vertical_compose().
> You might as well send all your mmx, a patch is only seperable when the
> different parts really cover different things or are logical as seperate.
> this might as well be just a single patch.
I disagree. The three functions can work independently and this patch is
large enough as is.
> > Index: libavcodec/i386/dsputil_mmx.c
> > ===================================================================
> > RCS file: /cvsroot/ffmpeg/ffmpeg/libavcodec/i386/dsputil_mmx.c,v
> > retrieving revision 1.113
> > diff -u -r1.113 dsputil_mmx.c
> > --- libavcodec/i386/dsputil_mmx.c 7 Mar 2006 22:45:56 -0000 1.113
> > +++ libavcodec/i386/dsputil_mmx.c 20 Mar 2006 16:45:48 -0000
> > @@ -2564,6 +2564,9 @@
> > }
> > #endif
> > +extern void ff_snow_vertical_compose97i_sse2(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
> > +extern void ff_snow_vertical_compose97i_mmx(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
> I still HIGHLY dislike these declerations. A much better approach would be
> to either use static functions and #include the .c file directly, or to use
> some common header. I preffer the former as an added bonus there is no
> (additional) namespace bloat.
It does seem to be the safe thing to do. I've revised the patch to
#include snowdsp_mmx.c directly from dsputil_mmx.c.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 13320 bytes
Desc: not available
More information about the ffmpeg-devel