Split VC1 loop filter into separate functions for h/v and size

Originally committed as revision 18521 to svn://svn.ffmpeg.org/ffmpeg/trunk

3 files changed, 64 insertions, 24 deletions

diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 3599259f8e..60b2f22991 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -482,7 +482,12 @@ typedef struct DSPContext {
     void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
     void (*vc1_v_overlap)(uint8_t* src, int stride);
     void (*vc1_h_overlap)(uint8_t* src, int stride);
-    void (*vc1_loop_filter)(uint8_t *src, int step, int stride, int len, int pq);
+    void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq);
+    void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq);
+    void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq);
+    void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq);
+    void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq);
+    void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
     /* put 8x8 block with bicubic interpolation and quarterpel precision
      * last argument is actually round value instead of height
      */
diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c
index 604ab0a4a0..0add2c7f55 100644
--- a/libavcodec/vc1.c
+++ b/libavcodec/vc1.c
@@ -314,15 +314,15 @@ static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
 {
     int i, j;
     if(!s->first_slice_line)
-        s->dsp.vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
-    s->dsp.vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
+        s->dsp.vc1_v_loop_filter16(s->dest[0], s->linesize, pq);
+    s->dsp.vc1_v_loop_filter16(s->dest[0] + 8*s->linesize, s->linesize, pq);
     for(i = !s->mb_x*8; i < 16; i += 8)
-        s->dsp.vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
+        s->dsp.vc1_h_loop_filter16(s->dest[0] + i, s->linesize, pq);
     for(j = 0; j < 2; j++){
         if(!s->first_slice_line)
-            s->dsp.vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
+            s->dsp.vc1_v_loop_filter8(s->dest[j+1], s->uvlinesize, pq);
         if(s->mb_x)
-            s->dsp.vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
+            s->dsp.vc1_h_loop_filter8(s->dest[j+1], s->uvlinesize, pq);
     }
 }
 
@@ -2982,9 +2982,9 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
             s->dsp.vc1_inv_trans_8x8(block);
             s->dsp.add_pixels_clamped(block, dst, linesize);
             if(apply_filter && cbp_top  & 0xC)
-                s->dsp.vc1_loop_filter(dst, 1, linesize, 8, mquant);
+                s->dsp.vc1_v_loop_filter8(dst, linesize, mquant);
             if(apply_filter && cbp_left & 0xA)
-                s->dsp.vc1_loop_filter(dst, linesize, 1, 8, mquant);
+                s->dsp.vc1_h_loop_filter8(dst, linesize, mquant);
         }
         break;
     case TT_4X4:
@@ -3006,9 +3006,9 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
             if(!(subblkpat & (1 << (3 - j))) && !skip_block){
                 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
                 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
-                    s->dsp.vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
+                    s->dsp.vc1_v_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, mquant);
                 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
-                    s->dsp.vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
+                    s->dsp.vc1_h_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, mquant);
             }
         }
         break;
@@ -3031,9 +3031,9 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
             if(!(subblkpat & (1 << (1 - j))) && !skip_block){
                 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
                 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
-                    s->dsp.vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
+                    s->dsp.vc1_v_loop_filter8(dst + j*4*linesize, linesize, mquant);
                 if(apply_filter && cbp_left & (2 << j))
-                    s->dsp.vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
+                    s->dsp.vc1_h_loop_filter4(dst + j*4*linesize, linesize, mquant);
             }
         }
         break;
@@ -3056,9 +3056,9 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
             if(!(subblkpat & (1 << (1 - j))) && !skip_block){
                 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
                 if(apply_filter && cbp_top & (2 << j))
-                    s->dsp.vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
+                    s->dsp.vc1_v_loop_filter4(dst + j*4, linesize, mquant);
                 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
-                    s->dsp.vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
+                    s->dsp.vc1_h_loop_filter8(dst + j*4, linesize, mquant);
             }
         }
         break;
@@ -3181,9 +3181,9 @@ static int vc1_decode_p_mb(VC1Context *v)
                             top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
                         }
                         if(left_cbp & 0xC)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
+                            s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                         if(top_cbp  & 0xA)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
+                            s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                     }
                     block_cbp |= 0xF << (i << 2);
                 } else if(val) {
@@ -3198,9 +3198,9 @@ static int vc1_decode_p_mb(VC1Context *v)
                             top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
                         }
                         if(left_cbp & 0xC)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
+                            s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                         if(top_cbp  & 0xA)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
+                            s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                     }
                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
                     block_cbp |= pat << (i << 2);
@@ -3312,9 +3312,9 @@ static int vc1_decode_p_mb(VC1Context *v)
                             top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
                         }
                         if(left_cbp & 0xC)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
+                            s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                         if(top_cbp  & 0xA)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
+                            s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                     }
                     block_cbp |= 0xF << (i << 2);
                 } else if(is_coded[i]) {
@@ -3329,9 +3329,9 @@ static int vc1_decode_p_mb(VC1Context *v)
                             top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
                         }
                         if(left_cbp & 0xC)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
+                            s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                         if(top_cbp  & 0xA)
-                            s->dsp.vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
+                            s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, mquant);
                     }
                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
                     block_cbp |= pat << (i << 2);
diff --git a/libavcodec/vc1dsp.c b/libavcodec/vc1dsp.c
index 5420e7e260..5773ab1e3c 100644
--- a/libavcodec/vc1dsp.c
+++ b/libavcodec/vc1dsp.c
@@ -130,7 +130,7 @@ static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
  * @param pq block quantizer
  * @see 8.6
  */
-static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
+static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
 {
     int i;
     int filt3;
@@ -146,6 +146,36 @@ static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
     }
 }
 
+static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, 1, stride, 4, pq);
+}
+
+static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, stride, 1, 4, pq);
+}
+
+static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, 1, stride, 8, pq);
+}
+
+static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, stride, 1, 8, pq);
+}
+
+static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, 1, stride, 16, pq);
+}
+
+static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)
+{
+    vc1_loop_filter(src, stride, 1, 16, pq);
+}
+
 /** Do inverse transform on 8x8 block
 */
 static void vc1_inv_trans_8x8_c(DCTELEM block[64])
@@ -517,7 +547,12 @@ void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
     dsp->vc1_h_overlap = vc1_h_overlap_c;
     dsp->vc1_v_overlap = vc1_v_overlap_c;
-    dsp->vc1_loop_filter = vc1_loop_filter;
+    dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
+    dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
+    dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;
+    dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;
+    dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;
+    dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;
 
     dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;