MMX for H.264 deblocking filter

Originally committed as revision 4158 to svn://svn.ffmpeg.org/ffmpeg/trunk

1 files changed, 91 insertions, 0 deletions

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 926832ff16..29e1b0d3c2 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2640,6 +2640,92 @@ static void h261_loop_filter_c(uint8_t *src, int stride){
     }
 }
 
+static inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int *tc0)
+{
+    int i, d;
+    for( i = 0; i < 4; i++ ) {
+        if( tc0[i] < 0 ) {
+            pix += 4*ystride;
+            continue;
+        }
+        for( d = 0; d < 4; d++ ) {
+            const int p0 = pix[-1*xstride];
+            const int p1 = pix[-2*xstride];
+            const int p2 = pix[-3*xstride];
+            const int q0 = pix[0];
+            const int q1 = pix[1*xstride];
+            const int q2 = pix[2*xstride];
+    
+            if( ABS( p0 - q0 ) < alpha &&
+                ABS( p1 - p0 ) < beta &&
+                ABS( q1 - q0 ) < beta ) {
+    
+                int tc = tc0[i];
+                int i_delta;
+    
+                if( ABS( p2 - p0 ) < beta ) {
+                    pix[-2*xstride] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0[i], tc0[i] );
+                    tc++;
+                }
+                if( ABS( q2 - q0 ) < beta ) {
+                    pix[xstride] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0[i], tc0[i] );
+                    tc++;
+                }
+    
+                i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+                pix[-xstride] = clip_uint8( p0 + i_delta );    /* p0' */
+                pix[0]        = clip_uint8( q0 - i_delta );    /* q0' */
+            }
+            pix += ystride;
+        }
+    }
+}
+static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int *tc0)
+{
+    h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
+}
+static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int *tc0)
+{
+    h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
+}
+
+static inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int *tc0)
+{
+    int i, d;
+    for( i = 0; i < 4; i++ ) {
+        const int tc = tc0[i];
+        if( tc <= 0 ) {
+            pix += 2*ystride;
+            continue;
+        }
+        for( d = 0; d < 2; d++ ) {
+            const int p0 = pix[-1*xstride];
+            const int p1 = pix[-2*xstride];
+            const int q0 = pix[0];
+            const int q1 = pix[1*xstride];
+
+            if( ABS( p0 - q0 ) < alpha &&
+                ABS( p1 - p0 ) < beta &&
+                ABS( q1 - q0 ) < beta ) {
+
+                int delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+
+                pix[-xstride] = clip_uint8( p0 + delta );    /* p0' */
+                pix[0]        = clip_uint8( q0 - delta );    /* q0' */
+            }
+            pix += ystride;
+        }
+    }
+}
+static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int *tc0)
+{
+    h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
+}
+static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int *tc0)
+{
+    h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
+}
+
 static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
 {
     int s, i;
@@ -3739,6 +3825,11 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->diff_bytes= diff_bytes_c;
     c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
     c->bswap_buf= bswap_buf;
+
+    c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
+    c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
+    c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
+    c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
     
     c->h263_h_loop_filter= h263_h_loop_filter_c;
     c->h263_v_loop_filter= h263_v_loop_filter_c;