Optimize get_dct8x8_allowed().

30 cpu cycles faster on pentium dual.

Originally committed as revision 16248 to svn://svn.ffmpeg.org/ffmpeg/trunk

diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 5a2bb7f..202ff91 100644 (file)
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -4039,13 +4039,10 @@ static inline int get_level_prefix(GetBitContext *gb){
 }
 
 static inline int get_dct8x8_allowed(H264Context *h){
-    int i;
-    for(i=0; i<4; i++){
-        if(!IS_SUB_8X8(h->sub_mb_type[i])
-           || (!h->sps.direct_8x8_inference_flag && IS_DIRECT(h->sub_mb_type[i])))
-            return 0;
-    }
-    return 1;
+    if(h->sps.direct_8x8_inference_flag)
+        return !(*(uint64_t*)h->sub_mb_type & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8                )*0x0001000100010001ULL));
+    else
+        return !(*(uint64_t*)h->sub_mb_type & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8|MB_TYPE_DIRECT2)*0x0001000100010001ULL));
 }
 
 /**

diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index 7708195..40a94a0 100644 (file)
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -339,7 +339,7 @@ typedef struct H264Context{
     int mb_field_decoding_flag;
     int mb_mbaff;              ///< mb_aff_frame && mb_field_decoding_flag
 
-    unsigned int sub_mb_type[4];
+    uint16_t sub_mb_type[4];
 
     //POC stuff
     int poc_lsb;