Skip to content

F
ffmpeg
Commit 359f98de authored by Michael Niedermayer's avatar Michael Niedermayer
Browse files
Options

h263_v_loop_filter_mmx 

Originally committed as revision 2552 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent e1f03acc
Hide whitespace changes
Inline Side-by-side
Showing with 92 additions and 0 deletions
libavcodec/i386/dsputil_mmx.c
View file @ 359f98de
...@@ -22,6 +22,8 @@ ...@@ -22,6 +22,8 @@
#include "../dsputil.h" #include "../dsputil.h"
#include "../simple_idct.h" #include "../simple_idct.h"
extern const uint8_t ff_h263_loop_filter_strength[32];
int mm_flags; /* multimedia extension flags */ int mm_flags; /* multimedia extension flags */
/* pixel operations */ /* pixel operations */
...@@ -34,6 +36,8 @@ static const uint64_t ff_pw_3 __attribute__ ((aligned(8))) = 0x0003000300030003 ...@@ -34,6 +36,8 @@ static const uint64_t ff_pw_3 __attribute__ ((aligned(8))) = 0x0003000300030003
static const uint64_t ff_pw_16 __attribute__ ((aligned(8))) = 0x0010001000100010ULL; static const uint64_t ff_pw_16 __attribute__ ((aligned(8))) = 0x0010001000100010ULL;
static const uint64_t ff_pw_15 __attribute__ ((aligned(8))) = 0x000F000F000F000FULL; static const uint64_t ff_pw_15 __attribute__ ((aligned(8))) = 0x000F000F000F000FULL;
static const uint64_t ff_pb_FC __attribute__ ((aligned(8))) = 0xFCFCFCFCFCFCFCFCULL;
#define JUMPALIGN() __asm __volatile (".balign 8"::) #define JUMPALIGN() __asm __volatile (".balign 8"::)
#define MOVQ_ZERO(regd) __asm __volatile ("pxor %%" #regd ", %%" #regd ::) #define MOVQ_ZERO(regd) __asm __volatile ("pxor %%" #regd ", %%" #regd ::)
...@@ -465,6 +469,92 @@ static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){ ...@@ -465,6 +469,92 @@ static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
dst[i+0] += src[i+0]; dst[i+0] += src[i+0];
} }
static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale){
const int strength= ff_h263_loop_filter_strength[qscale];
asm volatile(
"pxor %%mm7, %%mm7 \n\t"
"movq %0, %%mm0 \n\t"
"movq %0, %%mm1 \n\t"
"movq %3, %%mm2 \n\t"
"movq %3, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"punpckhbw %%mm7, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
"punpckhbw %%mm7, %%mm3 \n\t"
"psubw %%mm2, %%mm0 \n\t"
"psubw %%mm3, %%mm1 \n\t"
"movq %1, %%mm2 \n\t"
"movq %1, %%mm3 \n\t"
"movq %2, %%mm4 \n\t"
"movq %2, %%mm5 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
"punpckhbw %%mm7, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm4 \n\t"
"punpckhbw %%mm7, %%mm5 \n\t"
"psubw %%mm2, %%mm4 \n\t"
"psubw %%mm3, %%mm5 \n\t"
"psllw $2, %%mm4 \n\t"
"psllw $2, %%mm5 \n\t"
"paddw %%mm0, %%mm4 \n\t"
"paddw %%mm1, %%mm5 \n\t"
"pxor %%mm6, %%mm6 \n\t"
"pcmpgtw %%mm4, %%mm6 \n\t"
"pcmpgtw %%mm5, %%mm7 \n\t"
"pxor %%mm6, %%mm4 \n\t"
"pxor %%mm7, %%mm5 \n\t"
"psubw %%mm6, %%mm4 \n\t"
"psubw %%mm7, %%mm5 \n\t"
"psrlw $3, %%mm4 \n\t"
"psrlw $3, %%mm5 \n\t"
"packuswb %%mm5, %%mm4 \n\t" //abs(d)
"packsswb %%mm7, %%mm6 \n\t" //sign(d)
"pxor %%mm7, %%mm7 \n\t"
"movd %4, %%mm2 \n\t"
"punpcklbw %%mm2, %%mm2 \n\t"
"punpcklbw %%mm2, %%mm2 \n\t"
"punpcklbw %%mm2, %%mm2 \n\t" //2*strength
"psubusb %%mm4, %%mm2 \n\t" // S(2*strength - abs(d))
"movq %%mm2, %%mm3 \n\t" // S(2*strength - abs(d))
"psubusb %%mm4, %%mm3 \n\t" // S(S(2*strength - abs(d)) - abs(d))
"psubb %%mm3, %%mm2 \n\t" // MIN(abs(d), S(2*strength - abs(d)))
"movq %1, %%mm3 \n\t"
"movq %2, %%mm4 \n\t"
"pxor %%mm6, %%mm3 \n\t"
"pxor %%mm6, %%mm4 \n\t"
"paddusb %%mm2, %%mm3 \n\t"
"psubusb %%mm2, %%mm4 \n\t"
"pxor %%mm6, %%mm3 \n\t"
"pxor %%mm6, %%mm4 \n\t"
"movq %%mm3, %1 \n\t"
"movq %%mm4, %2 \n\t"
"paddusb %%mm2, %%mm2 \n\t"
"packsswb %%mm1, %%mm0 \n\t"
"pcmpgtb %%mm0, %%mm7 \n\t"
"pxor %%mm7, %%mm0 \n\t"
"psubb %%mm7, %%mm0 \n\t"
"movq %%mm0, %%mm1 \n\t"
"psubusb %%mm2, %%mm0 \n\t"
"psubb %%mm0, %%mm1 \n\t"
"pand %5, %%mm1 \n\t"
"psrlw $2, %%mm1 \n\t"
"pxor %%mm7, %%mm1 \n\t"
"psubb %%mm7, %%mm1 \n\t"
"movq %0, %%mm3 \n\t"
"movq %3, %%mm4 \n\t"
"psubb %%mm1, %%mm3 \n\t"
"paddb %%mm1, %%mm4 \n\t"
"movq %%mm3, %0 \n\t"
"movq %%mm4, %3 \n\t"
: "+m" (*(uint64_t*)(src - 2*stride)),
"+m" (*(uint64_t*)(src - 1*stride)),
"+m" (*(uint64_t*)(src + 0*stride)),
"+m" (*(uint64_t*)(src + 1*stride))
: "g" (2*strength), "m"(ff_pb_FC)
);
}
#ifdef CONFIG_ENCODERS #ifdef CONFIG_ENCODERS
static int pix_norm1_mmx(uint8_t *pix, int line_size) { static int pix_norm1_mmx(uint8_t *pix, int line_size) {
int tmp; int tmp;
...@@ -1691,6 +1781,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) ...@@ -1691,6 +1781,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->pix_norm1 = pix_norm1_mmx; c->pix_norm1 = pix_norm1_mmx;
c->sse[0] = sse16_mmx; c->sse[0] = sse16_mmx;
#endif //CONFIG_ENCODERS #endif //CONFIG_ENCODERS
c->h263_v_loop_filter= h263_v_loop_filter_mmx;
if (mm_flags & MM_MMXEXT) { if (mm_flags & MM_MMXEXT) {
c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2; c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment