Add a new IDCT permutation, used in xvid_sse2 and possibly future similar IDCTs.

Originally committed as revision 12842 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/dsputil.c

@@ -151,6 +151,8 @@ static const uint8_t simple_mmx_permutation[64]={
         0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
 };
 
+static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7};
+
 void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_scantable){
     int i;
     int end;
@@ -4476,6 +4478,10 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
         for(i=0; i<64; i++)
             c->idct_permutation[i]= (i&0x24) | ((i&3)<<3) | ((i>>3)&3);
         break;
+    case FF_SSE2_IDCT_PERM:
+        for(i=0; i<64; i++)
+            c->idct_permutation[i]= (i&0x38) | idct_sse2_row_perm[i&7];
+        break;
     default:
         av_log(avctx, AV_LOG_ERROR, "Internal error, IDCT permutation not set\n");
     }

libavcodec/dsputil.h

@@ -408,6 +408,7 @@ typedef struct DSPContext {
 #define FF_SIMPLE_IDCT_PERM 3
 #define FF_TRANSPOSE_IDCT_PERM 4
 #define FF_PARTTRANS_IDCT_PERM 5
+#define FF_SSE2_IDCT_PERM 6
 
     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);