VP8 bilinear filter

Originally committed as revision 23813 to svn://svn.ffmpeg.org/ffmpeg/trunk

3 files changed, 85 insertions, 8 deletions

diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c
index f7163f4103..0000706ed2 100644
--- a/libavcodec/vp8.c
+++ b/libavcodec/vp8.c
@@ -45,6 +45,7 @@ typedef struct {
     DSPContext dsp;
     VP8DSPContext vp8dsp;
     H264PredContext hpc;
+    vp8_mc_func put_pixels_tab[3][3][3];
     AVFrame frames[4];
     AVFrame *framep[4];
     uint8_t *edge_emu_buffer;
@@ -379,8 +380,13 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
     buf      += 3;
     buf_size -= 3;
 
-    if (s->profile)
-        av_log(s->avctx, AV_LOG_WARNING, "Profile %d not fully handled\n", s->profile);
+    if (s->profile > 3)
+        av_log(s->avctx, AV_LOG_WARNING, "Unknown profile %d\n", s->profile);
+
+    if (!s->profile)
+        memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_epel_pixels_tab, sizeof(s->put_pixels_tab));
+    else    // profile 1-3 use bilinear, 4+ aren't defined so whatever
+        memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_bilinear_pixels_tab, sizeof(s->put_pixels_tab));
 
     if (header_size > buf_size - 7*s->keyframe) {
         av_log(s->avctx, AV_LOG_ERROR, "Header size larger than data provided\n");
@@ -951,7 +957,7 @@ static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
         /* Y */
         vp8_mc(s, 1, dst[0], s->framep[mb->ref_frame]->data[0], &mb->mv,
                x_off, y_off, 16, 16, width, height, s->linesize,
-               s->vp8dsp.put_vp8_epel_pixels_tab[0]);
+               s->put_pixels_tab[0]);
 
         /* U/V */
         uvmv = mb->mv;
@@ -962,10 +968,10 @@ static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
         x_off >>= 1; y_off >>= 1; width >>= 1; height >>= 1;
         vp8_mc(s, 0, dst[1], s->framep[mb->ref_frame]->data[1], &uvmv,
                x_off, y_off, 8, 8, width, height, s->uvlinesize,
-               s->vp8dsp.put_vp8_epel_pixels_tab[1]);
+               s->put_pixels_tab[1]);
         vp8_mc(s, 0, dst[2], s->framep[mb->ref_frame]->data[2], &uvmv,
                x_off, y_off, 8, 8, width, height, s->uvlinesize,
-               s->vp8dsp.put_vp8_epel_pixels_tab[1]);
+               s->put_pixels_tab[1]);
     } else {
         int x, y;
 
@@ -976,7 +982,7 @@ static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
                        s->framep[mb->ref_frame]->data[0], &mb->bmv[4*y + x],
                        4*x + x_off, 4*y + y_off, 4, 4,
                        width, height, s->linesize,
-                       s->vp8dsp.put_vp8_epel_pixels_tab[2]);
+                       s->put_pixels_tab[2]);
             }
         }
 
@@ -1002,12 +1008,12 @@ static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
                        s->framep[mb->ref_frame]->data[1], &uvmv,
                        4*x + x_off, 4*y + y_off, 4, 4,
                        width, height, s->uvlinesize,
-                       s->vp8dsp.put_vp8_epel_pixels_tab[2]);
+                       s->put_pixels_tab[2]);
                 vp8_mc(s, 0, dst[2] + 4*y*s->uvlinesize + x*4,
                        s->framep[mb->ref_frame]->data[2], &uvmv,
                        4*x + x_off, 4*y + y_off, 4, 4,
                        width, height, s->uvlinesize,
-                       s->vp8dsp.put_vp8_epel_pixels_tab[2]);
+                       s->put_pixels_tab[2]);
             }
         }
     }
diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c
index ada2c9d67d..78ca9099d8 100644
--- a/libavcodec/vp8dsp.c
+++ b/libavcodec/vp8dsp.c
@@ -348,6 +348,61 @@ VP8_EPEL_HV(16, FILTER_6TAP, FILTER_6TAP, h6v6)
 VP8_EPEL_HV(8,  FILTER_6TAP, FILTER_6TAP, h6v6)
 VP8_EPEL_HV(4,  FILTER_6TAP, FILTER_6TAP, h6v6)
 
+#define VP8_BILINEAR(SIZE) \
+static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
+{ \
+    int a = 8-mx, b = mx; \
+    int x, y; \
+\
+    for (y = 0; y < h; y++) { \
+        for (x = 0; x < SIZE; x++) \
+            dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
+        dst += stride; \
+        src += stride; \
+    } \
+} \
+static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
+{ \
+    int c = 8-my, d = my; \
+    int x, y; \
+\
+    for (y = 0; y < h; y++) { \
+        for (x = 0; x < SIZE; x++) \
+            dst[x] = (c*src[x] + d*src[x+stride] + 4) >> 3; \
+        dst += stride; \
+        src += stride; \
+    } \
+} \
+\
+static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
+{ \
+    int a = 8-mx, b = mx; \
+    int c = 8-my, d = my; \
+    int x, y; \
+    uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
+    uint8_t *tmp = tmp_array; \
+\
+    for (y = 0; y < h+1; y++) { \
+        for (x = 0; x < SIZE; x++) \
+            tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
+        tmp += SIZE; \
+        src += stride; \
+    } \
+\
+    tmp = tmp_array; \
+\
+    for (y = 0; y < h; y++) { \
+        for (x = 0; x < SIZE; x++) \
+            dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
+        dst += stride; \
+        tmp += SIZE; \
+    } \
+}
+
+VP8_BILINEAR(16)
+VP8_BILINEAR(8)
+VP8_BILINEAR(4)
+
 #define VP8_MC_FUNC(IDX, SIZE) \
     dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
     dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
@@ -359,6 +414,17 @@ VP8_EPEL_HV(4,  FILTER_6TAP, FILTER_6TAP, h6v6)
     dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
 
+#define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
+    dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
+
 av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
 {
     dsp->vp8_luma_dc_wht = vp8_luma_dc_wht_c;
@@ -381,4 +447,8 @@ av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
     VP8_MC_FUNC(0, 16);
     VP8_MC_FUNC(1, 8);
     VP8_MC_FUNC(2, 4);
+
+    VP8_BILINEAR_MC_FUNC(0, 16);
+    VP8_BILINEAR_MC_FUNC(1, 8);
+    VP8_BILINEAR_MC_FUNC(2, 4);
 }
diff --git a/libavcodec/vp8dsp.h b/libavcodec/vp8dsp.h
index 2cd9e7a10b..18d6916985 100644
--- a/libavcodec/vp8dsp.h
+++ b/libavcodec/vp8dsp.h
@@ -58,6 +58,7 @@ typedef struct VP8DSPContext {
      * so something like put_vp8_epel_pixels_tab[width>>3][2*!!my-(my&1)][2*!!mx-(mx&1)](..., mx, my)
      */
     vp8_mc_func put_vp8_epel_pixels_tab[3][3][3];
+    vp8_mc_func put_vp8_bilinear_pixels_tab[3][3][3];
 } VP8DSPContext;
 
 void ff_put_vp8_pixels16_c(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y);