[Ffmpeg-devel] [PATCH] put_mpeg4_qpel16_h_lowpass altivec, take 2
Brian Foley
bfoley
Sun Nov 26 18:44:47 CET 2006
On Sun, Nov 26, 2006 at 05:23:35PM +0100, Luca Barbato wrote:
> Brian Foley wrote:
> > On Mon, Nov 20, 2006 at 02:43:17AM +0100, Michael Niedermayer wrote:
> >> Hi
> >>
> >> On Sun, Nov 19, 2006 at 11:20:14PM +0000, Brian Foley wrote:
> >>> +static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
> >> code duplication, move copy_block17 to a common header like dsputil.h or
> >> dsputil_internal.h or whatever, dont copy and paste it
> >
> > OK, done. I've moved all the copy_block* stuff into dsputil.h. sh4 was
> > using a copy of this too as it turns out.
>
> make it a separate patch alone.
Here you go. That, at least is easy :)
Cheers,
Brian.
-------------- next part --------------
Index: dsputil.c
===================================================================
--- dsputil.c (revision 7171)
+++ dsputil.c (working copy)
@@ -1513,83 +1513,6 @@
}
}
-static inline void copy_block2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST16(dst , LD16(src ));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- ST32(dst+8 , LD32(src+8 ));
- ST32(dst+12, LD32(src+12));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- ST32(dst+8 , LD32(src+8 ));
- ST32(dst+12, LD32(src+12));
- dst[16]= src[16];
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- dst[8]= src[8];
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-
#define QPEL_MC(r, OPNAME, RND, OP) \
static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
Index: dsputil.h
===================================================================
--- dsputil.h (revision 7171)
+++ dsputil.h (working copy)
@@ -698,4 +698,81 @@
return score;\
}
+
+static inline void copy_block2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST16(dst , LD16(src ));
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
+static inline void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST32(dst , LD32(src ));
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
+static inline void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST32(dst , LD32(src ));
+ ST32(dst+4 , LD32(src+4 ));
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
+static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST32(dst , LD32(src ));
+ ST32(dst+4 , LD32(src+4 ));
+ dst[8]= src[8];
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
+static inline void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST32(dst , LD32(src ));
+ ST32(dst+4 , LD32(src+4 ));
+ ST32(dst+8 , LD32(src+8 ));
+ ST32(dst+12, LD32(src+12));
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
+static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+{
+ int i;
+ for(i=0; i<h; i++)
+ {
+ ST32(dst , LD32(src ));
+ ST32(dst+4 , LD32(src+4 ));
+ ST32(dst+8 , LD32(src+8 ));
+ ST32(dst+12, LD32(src+12));
+ dst[16]= src[16];
+ dst+=dstStride;
+ src+=srcStride;
+ }
+}
+
#endif
Index: sh4/qpel.c
===================================================================
--- sh4/qpel.c (revision 7171)
+++ sh4/qpel.c (working copy)
@@ -564,73 +564,6 @@
#undef op_avg
#undef op_put
-/* not yet optimized */
-static inline void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- ST32(dst+8 , LD32(src+8 ));
- ST32(dst+12, LD32(src+12));
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- ST32(dst+8 , LD32(src+8 ));
- ST32(dst+12, LD32(src+12));
- dst[16]= src[16];
- dst+=dstStride;
- src+=srcStride;
- }
-}
-
-static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
-{
- int i;
- for(i=0; i<h; i++)
- {
- ST32(dst , LD32(src ));
- ST32(dst+4 , LD32(src+4 ));
- dst[8]= src[8];
- dst+=dstStride;
- src+=srcStride;
- }
-}
-/* end not optimized */
-
#define QPEL_MC(r, OPNAME, RND, OP) \
static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
More information about the ffmpeg-devel
mailing list