Move filter_luma_intra into dsputil for later addition of asm.

Originally committed as revision 16228 to svn://svn.ffmpeg.org/ffmpeg/trunk

3 files changed, 63 insertions, 91 deletions

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index b3b54d9d6f..6cd7131c57 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2987,6 +2987,63 @@ static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int b
     h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
 }
 
+static inline void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
+{
+    int d;
+    for( d = 0; d < 16; d++ ) {
+        const int p2 = pix[-3*xstride];
+        const int p1 = pix[-2*xstride];
+        const int p0 = pix[-1*xstride];
+
+        const int q0 = pix[ 0*xstride];
+        const int q1 = pix[ 1*xstride];
+        const int q2 = pix[ 2*xstride];
+
+        if( FFABS( p0 - q0 ) < alpha &&
+            FFABS( p1 - p0 ) < beta &&
+            FFABS( q1 - q0 ) < beta ) {
+
+            if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
+                if( FFABS( p2 - p0 ) < beta)
+                {
+                    const int p3 = pix[-4*xstride];
+                    /* p0', p1', p2' */
+                    pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
+                    pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
+                    pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
+                } else {
+                    /* p0' */
+                    pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+                }
+                if( FFABS( q2 - q0 ) < beta)
+                {
+                    const int q3 = pix[3*xstride];
+                    /* q0', q1', q2' */
+                    pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
+                    pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
+                    pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
+                } else {
+                    /* q0' */
+                    pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+                }
+            }else{
+                /* p0', q0' */
+                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+            }
+        }
+        pix += ystride;
+    }
+}
+static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
+}
+static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
+}
+
 static inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
 {
     int i, d;
@@ -4500,6 +4557,8 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
 
     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_luma_intra= h264_v_loop_filter_luma_intra_c;
+    c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_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->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index aeb467f105..d4df99bafe 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -351,6 +351,8 @@ typedef struct DSPContext {
     void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
     void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
     /* v/h_loop_filter_luma_intra: align 16 */
+    void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
+    void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
     void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
     void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
     void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 64824a43e4..1558721015 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -5854,53 +5854,7 @@ static void filter_mb_edgev( H264Context *h, uint8_t *pix, int stride, int16_t b
             tc[i] = bS[i] ? (tc0_table+52)[index_a][bS[i] - 1] : -1;
         h->s.dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
     } else {
-        /* 16px edge length, because bS=4 is triggered by being at
-         * the edge of an intra MB, so all 4 bS are the same */
-            for( d = 0; d < 16; d++ ) {
-                const int p0 = pix[-1];
-                const int p1 = pix[-2];
-                const int p2 = pix[-3];
-
-                const int q0 = pix[0];
-                const int q1 = pix[1];
-                const int q2 = pix[2];
-
-                if( FFABS( p0 - q0 ) < alpha &&
-                    FFABS( p1 - p0 ) < beta &&
-                    FFABS( q1 - q0 ) < beta ) {
-
-                    if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
-                        if( FFABS( p2 - p0 ) < beta)
-                        {
-                            const int p3 = pix[-4];
-                            /* p0', p1', p2' */
-                            pix[-1] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
-                            pix[-2] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
-                            pix[-3] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
-                        } else {
-                            /* p0' */
-                            pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                        }
-                        if( FFABS( q2 - q0 ) < beta)
-                        {
-                            const int q3 = pix[3];
-                            /* q0', q1', q2' */
-                            pix[0] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
-                            pix[1] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
-                            pix[2] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
-                        } else {
-                            /* q0' */
-                            pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-                        }
-                    }else{
-                        /* p0', q0' */
-                        pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                        pix[ 0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-                    }
-                    tprintf(h->s.avctx, "filter_mb_edgev i:%d d:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, p2, p1, p0, q0, q1, q2, pix[-2], pix[-1], pix[0], pix[1]);
-                }
-                pix += stride;
-            }
+        h->s.dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
     }
 }
 static void filter_mb_edgecv( H264Context *h, uint8_t *pix, int stride, int16_t bS[4], int qp ) {
@@ -6083,50 +6037,7 @@ static void filter_mb_edgeh( H264Context *h, uint8_t *pix, int stride, int16_t b
             tc[i] = bS[i] ? (tc0_table+52)[index_a][bS[i] - 1] : -1;
         h->s.dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
     } else {
-        /* 16px edge length, see filter_mb_edgev */
-            for( d = 0; d < 16; d++ ) {
-                const int p0 = pix[-1*pix_next];
-                const int p1 = pix[-2*pix_next];
-                const int p2 = pix[-3*pix_next];
-                const int q0 = pix[0];
-                const int q1 = pix[1*pix_next];
-                const int q2 = pix[2*pix_next];
-
-                if( FFABS( p0 - q0 ) < alpha &&
-                    FFABS( p1 - p0 ) < beta &&
-                    FFABS( q1 - q0 ) < beta ) {
-
-                    const int p3 = pix[-4*pix_next];
-                    const int q3 = pix[ 3*pix_next];
-
-                    if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
-                        if( FFABS( p2 - p0 ) < beta) {
-                            /* p0', p1', p2' */
-                            pix[-1*pix_next] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
-                            pix[-2*pix_next] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
-                            pix[-3*pix_next] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
-                        } else {
-                            /* p0' */
-                            pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                        }
-                        if( FFABS( q2 - q0 ) < beta) {
-                            /* q0', q1', q2' */
-                            pix[0*pix_next] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
-                            pix[1*pix_next] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
-                            pix[2*pix_next] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
-                        } else {
-                            /* q0' */
-                            pix[0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-                        }
-                    }else{
-                        /* p0', q0' */
-                        pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                        pix[ 0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-                    }
-                    tprintf(h->s.avctx, "filter_mb_edgeh i:%d d:%d, qp:%d, indexA:%d, alpha:%d, beta:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, qp, index_a, alpha, beta, bS[i], p2, p1, p0, q0, q1, q2, pix[-2*pix_next], pix[-pix_next], pix[0], pix[pix_next]);
-                }
-                pix++;
-            }
+        h->s.dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
     }
 }