changeset 8411:b749b637ecfa libavcodec

Optimize get_dct8x8_allowed(). 30 cpu cycles faster on pentium dual.
author michael
date Sat, 20 Dec 2008 23:11:30 +0000
parents b0ec9768b39b
children 2d48043ab521
files h264.c h264.h
diffstat 2 files changed, 5 insertions(+), 8 deletions(-) [+]
line wrap: on
line diff
--- a/h264.c	Sat Dec 20 18:06:01 2008 +0000
+++ b/h264.c	Sat Dec 20 23:11:30 2008 +0000
@@ -4039,13 +4039,10 @@
 }
 
 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));
 }
 
 /**
--- a/h264.h	Sat Dec 20 18:06:01 2008 +0000
+++ b/h264.h	Sat Dec 20 23:11:30 2008 +0000
@@ -339,7 +339,7 @@
     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;