H.264: eliminate non-transposed scantable support.

It was an ugly hack to begin with and didn't give any performance.

NOTE: this patch opens up some future simplifications to be made (such as
removing some of the scantables from H264Context) but doesn't take advantage
of them yet.

Originally committed as revision 26329 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/h264.c

@@ -687,7 +687,6 @@ static void free_tables(H264Context *h){
 
 static void init_dequant8_coeff_table(H264Context *h){
     int i,q,x;
-    const int transpose = (h->h264dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly
     h->dequant8_coeff[0] = h->dequant8_buffer[0];
     h->dequant8_coeff[1] = h->dequant8_buffer[1];
 
@@ -701,7 +700,7 @@ static void init_dequant8_coeff_table(H264Context *h){
             int shift = div6[q];
             int idx = rem6[q];
             for(x=0; x<64; x++)
-                h->dequant8_coeff[i][q][transpose ? (x>>3)|((x&7)<<3) : x] =
+                h->dequant8_coeff[i][q][(x>>3)|((x&7)<<3)] =
                     ((uint32_t)dequant8_coeff_init[idx][ dequant8_coeff_init_scan[((x>>1)&12) | (x&3)] ] *
                     h->pps.scaling_matrix8[i][x]) << shift;
         }
@@ -710,7 +709,6 @@ static void init_dequant8_coeff_table(H264Context *h){
 
 static void init_dequant4_coeff_table(H264Context *h){
     int i,j,q,x;
-    const int transpose = (h->h264dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
     for(i=0; i<6; i++ ){
         h->dequant4_coeff[i] = h->dequant4_buffer[i];
         for(j=0; j<i; j++){
@@ -726,7 +724,7 @@ static void init_dequant4_coeff_table(H264Context *h){
             int shift = div6[q] + 2;
             int idx = rem6[q];
             for(x=0; x<16; x++)
-                h->dequant4_coeff[i][q][transpose ? (x>>2)|((x<<2)&0xF) : x] =
+                h->dequant4_coeff[i][q][(x>>2)|((x<<2)&0xF)] =
                     ((uint32_t)dequant4_coeff_init[idx][(x&1) + ((x>>2)&1)] *
                     h->pps.scaling_matrix4[i][x]) << shift;
         }
@@ -1597,31 +1595,19 @@ static int init_poc(H264Context *h){
  */
 static void init_scan_tables(H264Context *h){
     int i;
-    if(h->h264dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
-        memcpy(h->zigzag_scan, zigzag_scan, 16*sizeof(uint8_t));
-        memcpy(h-> field_scan,  field_scan, 16*sizeof(uint8_t));
-    }else{
-        for(i=0; i<16; i++){
+    for(i=0; i<16; i++){
 #define T(x) (x>>2) | ((x<<2) & 0xF)
-            h->zigzag_scan[i] = T(zigzag_scan[i]);
-            h-> field_scan[i] = T( field_scan[i]);
+        h->zigzag_scan[i] = T(zigzag_scan[i]);
+        h-> field_scan[i] = T( field_scan[i]);
 #undef T
-        }
     }
-    if(h->h264dsp.h264_idct8_add == ff_h264_idct8_add_c){
-        memcpy(h->zigzag_scan8x8,       ff_zigzag_direct,     64*sizeof(uint8_t));
-        memcpy(h->zigzag_scan8x8_cavlc, zigzag_scan8x8_cavlc, 64*sizeof(uint8_t));
-        memcpy(h->field_scan8x8,        field_scan8x8,        64*sizeof(uint8_t));
-        memcpy(h->field_scan8x8_cavlc,  field_scan8x8_cavlc,  64*sizeof(uint8_t));
-    }else{
-        for(i=0; i<64; i++){
+    for(i=0; i<64; i++){
 #define T(x) (x>>3) | ((x&7)<<3)
-            h->zigzag_scan8x8[i]       = T(ff_zigzag_direct[i]);
-            h->zigzag_scan8x8_cavlc[i] = T(zigzag_scan8x8_cavlc[i]);
-            h->field_scan8x8[i]        = T(field_scan8x8[i]);
-            h->field_scan8x8_cavlc[i]  = T(field_scan8x8_cavlc[i]);
+        h->zigzag_scan8x8[i]       = T(ff_zigzag_direct[i]);
+        h->zigzag_scan8x8_cavlc[i] = T(zigzag_scan8x8_cavlc[i]);
+        h->field_scan8x8[i]        = T(field_scan8x8[i]);
+        h->field_scan8x8_cavlc[i]  = T(field_scan8x8_cavlc[i]);
 #undef T
-        }
     }
     if(h->sps.transform_bypass){ //FIXME same ugly
         h->zigzag_scan_q0          = zigzag_scan;

libavcodec/h264idct.c

@@ -33,24 +33,24 @@ static av_always_inline void idct_internal(uint8_t *dst, DCTELEM *block, int str
 
     block[0] += 1<<(shift-1);
 
-    for(i=0; i<4; i++){
-        const int z0=  block[0 + block_stride*i]     +  block[2 + block_stride*i];
-        const int z1=  block[0 + block_stride*i]     -  block[2 + block_stride*i];
-        const int z2= (block[1 + block_stride*i]>>1) -  block[3 + block_stride*i];
-        const int z3=  block[1 + block_stride*i]     + (block[3 + block_stride*i]>>1);
-
-        block[0 + block_stride*i]= z0 + z3;
-        block[1 + block_stride*i]= z1 + z2;
-        block[2 + block_stride*i]= z1 - z2;
-        block[3 + block_stride*i]= z0 - z3;
-    }
-
     for(i=0; i<4; i++){
         const int z0=  block[i + block_stride*0]     +  block[i + block_stride*2];
         const int z1=  block[i + block_stride*0]     -  block[i + block_stride*2];
         const int z2= (block[i + block_stride*1]>>1) -  block[i + block_stride*3];
         const int z3=  block[i + block_stride*1]     + (block[i + block_stride*3]>>1);
 
+        block[i + block_stride*0]= z0 + z3;
+        block[i + block_stride*1]= z1 + z2;
+        block[i + block_stride*2]= z1 - z2;
+        block[i + block_stride*3]= z0 - z3;
+    }
+
+    for(i=0; i<4; i++){
+        const int z0=  block[0 + block_stride*i]     +  block[2 + block_stride*i];
+        const int z1=  block[0 + block_stride*i]     -  block[2 + block_stride*i];
+        const int z2= (block[1 + block_stride*i]>>1) -  block[3 + block_stride*i];
+        const int z3=  block[1 + block_stride*i]     + (block[3 + block_stride*i]>>1);
+
         dst[i + 0*stride]= cm[ add*dst[i + 0*stride] + ((z0 + z3) >> shift) ];
         dst[i + 1*stride]= cm[ add*dst[i + 1*stride] + ((z1 + z2) >> shift) ];
         dst[i + 2*stride]= cm[ add*dst[i + 2*stride] + ((z1 - z2) >> shift) ];
@@ -78,51 +78,51 @@ void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride){
 
     for( i = 0; i < 8; i++ )
     {
-        const int a0 =  block[0+i*8] + block[4+i*8];
-        const int a2 =  block[0+i*8] - block[4+i*8];
-        const int a4 = (block[2+i*8]>>1) - block[6+i*8];
-        const int a6 = (block[6+i*8]>>1) + block[2+i*8];
+        const int a0 =  block[i+0*8] + block[i+4*8];
+        const int a2 =  block[i+0*8] - block[i+4*8];
+        const int a4 = (block[i+2*8]>>1) - block[i+6*8];
+        const int a6 = (block[i+6*8]>>1) + block[i+2*8];
 
         const int b0 = a0 + a6;
         const int b2 = a2 + a4;
         const int b4 = a2 - a4;
         const int b6 = a0 - a6;
 
-        const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
-        const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
-        const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
-        const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
+        const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
+        const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
+        const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
+        const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
 
         const int b1 = (a7>>2) + a1;
         const int b3 =  a3 + (a5>>2);
         const int b5 = (a3>>2) - a5;
         const int b7 =  a7 - (a1>>2);
 
-        block[0+i*8] = b0 + b7;
-        block[7+i*8] = b0 - b7;
-        block[1+i*8] = b2 + b5;
-        block[6+i*8] = b2 - b5;
-        block[2+i*8] = b4 + b3;
-        block[5+i*8] = b4 - b3;
-        block[3+i*8] = b6 + b1;
-        block[4+i*8] = b6 - b1;
+        block[i+0*8] = b0 + b7;
+        block[i+7*8] = b0 - b7;
+        block[i+1*8] = b2 + b5;
+        block[i+6*8] = b2 - b5;
+        block[i+2*8] = b4 + b3;
+        block[i+5*8] = b4 - b3;
+        block[i+3*8] = b6 + b1;
+        block[i+4*8] = b6 - b1;
     }
     for( i = 0; i < 8; i++ )
     {
-        const int a0 =  block[i+0*8] + block[i+4*8];
-        const int a2 =  block[i+0*8] - block[i+4*8];
-        const int a4 = (block[i+2*8]>>1) - block[i+6*8];
-        const int a6 = (block[i+6*8]>>1) + block[i+2*8];
+        const int a0 =  block[0+i*8] + block[4+i*8];
+        const int a2 =  block[0+i*8] - block[4+i*8];
+        const int a4 = (block[2+i*8]>>1) - block[6+i*8];
+        const int a6 = (block[6+i*8]>>1) + block[2+i*8];
 
         const int b0 = a0 + a6;
         const int b2 = a2 + a4;
         const int b4 = a2 - a4;
         const int b6 = a0 - a6;
 
-        const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
-        const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
-        const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
-        const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
+        const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
+        const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
+        const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
+        const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
 
         const int b1 = (a7>>2) + a1;
         const int b3 =  a3 + (a5>>2);