Port SVQ3 to use the new mb_luma_dc method of storing luma DC coefficients.

Doesn't help speed as there isn't an asm implementation yet, but consistency
is a good thing.

Originally committed as revision 26348 to svn://svn.ffmpeg.org/ffmpeg/trunk

3 files changed, 23 insertions, 25 deletions

diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index e6b50e9c12..0efbad918a 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -65,7 +65,7 @@ void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block,
 void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
 
 void ff_h264_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qmul);
-void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
+void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
 void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
 
 void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 85aa3a8198..40dc276dd6 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -1215,7 +1215,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                         }
                     }
                 }else
-                    ff_svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
+                    ff_svq3_luma_dc_dequant_idct_c(h->mb, h->mb_luma_dc, s->qscale);
             }
             if(h->deblocking_filter)
                 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, simple);
diff --git a/libavcodec/svq3.c b/libavcodec/svq3.c
index 4a4a1c52cd..a52744228e 100644
--- a/libavcodec/svq3.c
+++ b/libavcodec/svq3.c
@@ -125,22 +125,18 @@ static const uint32_t svq3_dequant_coeff[32] = {
     61694, 68745, 77615, 89113,100253,109366,126635,141533
 };
 
-
-void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp)
-{
+void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp){
     const int qmul = svq3_dequant_coeff[qp];
 #define stride 16
     int i;
     int temp[16];
-    static const int x_offset[4] = {0, 1*stride, 4* stride,  5*stride};
-    static const int y_offset[4] = {0, 2*stride, 8* stride, 10*stride};
+    static const uint8_t x_offset[4]={0, 1*stride, 4*stride, 5*stride};
 
-    for (i = 0; i < 4; i++){
-        const int offset = y_offset[i];
-        const int z0 = 13*(block[offset+stride*0] +    block[offset+stride*4]);
-        const int z1 = 13*(block[offset+stride*0] -    block[offset+stride*4]);
-        const int z2 =  7* block[offset+stride*1] - 17*block[offset+stride*5];
-        const int z3 = 17* block[offset+stride*1] +  7*block[offset+stride*5];
+    for(i=0; i<4; i++){
+        const int z0 = 13*(input[4*i+0] +    input[4*i+2]);
+        const int z1 = 13*(input[4*i+0] -    input[4*i+2]);
+        const int z2 =  7* input[4*i+1] - 17*input[4*i+3];
+        const int z3 = 17* input[4*i+1] +  7*input[4*i+3];
 
         temp[4*i+0] = z0+z3;
         temp[4*i+1] = z1+z2;
@@ -148,17 +144,17 @@ void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp)
         temp[4*i+3] = z0-z3;
     }
 
-    for (i = 0; i < 4; i++){
-        const int offset = x_offset[i];
-        const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]);
-        const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]);
-        const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
-        const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
-
-        block[stride*0 +offset] = ((z0 + z3)*qmul + 0x80000) >> 20;
-        block[stride*2 +offset] = ((z1 + z2)*qmul + 0x80000) >> 20;
-        block[stride*8 +offset] = ((z1 - z2)*qmul + 0x80000) >> 20;
-        block[stride*10+offset] = ((z0 - z3)*qmul + 0x80000) >> 20;
+    for(i=0; i<4; i++){
+        const int offset= x_offset[i];
+        const int z0= 13*(temp[4*0+i] +    temp[4*2+i]);
+        const int z1= 13*(temp[4*0+i] -    temp[4*2+i]);
+        const int z2=  7* temp[4*1+i] - 17*temp[4*3+i];
+        const int z3= 17* temp[4*1+i] +  7*temp[4*3+i];
+
+        output[stride* 0+offset] = ((z0 + z3)*qmul + 0x80000) >> 20;
+        output[stride* 2+offset] = ((z1 + z2)*qmul + 0x80000) >> 20;
+        output[stride* 8+offset] = ((z1 - z2)*qmul + 0x80000) >> 20;
+        output[stride*10+offset] = ((z0 - z3)*qmul + 0x80000) >> 20;
     }
 }
 #undef stride
@@ -648,7 +644,9 @@ static int svq3_decode_mb(H264Context *h, unsigned int mb_type)
         }
     }
     if (IS_INTRA16x16(mb_type)) {
-        if (svq3_decode_block(&s->gb, h->mb, 0, 0)){
+        AV_ZERO128(h->mb_luma_dc+0);
+        AV_ZERO128(h->mb_luma_dc+8);
+        if (svq3_decode_block(&s->gb, h->mb_luma_dc, 0, 1)){
             av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding intra luma dc\n");
             return -1;
         }