VP8: much faster DC transform handling

A lot of the time the DC block is empty: don't do the WHT in this case.
A lot of the rest of the time, there's only one coefficient: make a special
DC-only transform for that case.
When the block is empty, don't incorrectly mark luma DCT blocks as having DC
coefficients.

Originally committed as revision 24670 to svn://svn.ffmpeg.org/ffmpeg/trunk

1 files changed, 17 insertions, 4 deletions

diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c
index 5f5124803d..b8cf0b2f6c 100644
--- a/libavcodec/vp8dsp.c
+++ b/libavcodec/vp8dsp.c
@@ -51,13 +51,25 @@ static void vp8_luma_dc_wht_c(DCTELEM block[4][4][16], DCTELEM dc[16])
         dc[i*4+2] = 0;
         dc[i*4+3] = 0;
 
-        *block[i][0] = (t0 + t1) >> 3;
-        *block[i][1] = (t3 + t2) >> 3;
-        *block[i][2] = (t0 - t1) >> 3;
-        *block[i][3] = (t3 - t2) >> 3;
+        block[i][0][0] = (t0 + t1) >> 3;
+        block[i][1][0] = (t3 + t2) >> 3;
+        block[i][2][0] = (t0 - t1) >> 3;
+        block[i][3][0] = (t3 - t2) >> 3;
     }
 }
 
+static void vp8_luma_dc_wht_dc_c(DCTELEM block[4][4][16], DCTELEM dc[16])
+{
+    int i, val = (dc[0] + 3) >> 3;
+    dc[0] = 0;
+
+    for (i = 0; i < 4; i++) {
+        block[i][0][0] = val;
+        block[i][1][0] = val;
+        block[i][2][0] = val;
+        block[i][3][0] = val;
+    }
+}
 
 #define MUL_20091(a) ((((a)*20091) >> 16) + (a))
 #define MUL_35468(a)  (((a)*35468) >> 16)
@@ -480,6 +492,7 @@ VP8_BILINEAR(4)
 av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
 {
     dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;
+    dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;
     dsp->vp8_idct_add       = vp8_idct_add_c;
     dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;
     dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;