From c546280dafa11bc68deeb1cdfa3c5a636801d14b Mon Sep 17 00:00:00 2001 From: acki2 Date: Wed, 2 May 2001 23:21:27 +0000 Subject: [PATCH] - now we use fastmemcpy() for copying. Saves about 25% of copying time on K6-2+ git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@691 b3059339-0415-0410-9bf9-f77b7e298cf2 --- libvo/vo_dga.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) diff --git a/libvo/vo_dga.c b/libvo/vo_dga.c index f0dbe4e554..456cb81c3c 100644 --- a/libvo/vo_dga.c +++ b/libvo/vo_dga.c @@ -23,6 +23,9 @@ * - works only on x86 architectures * * $Log$ + * Revision 1.20 2001/05/02 23:21:27 acki2 + * - now we use fastmemcpy() for copying. Saves about 25% of copying time on K6-2+ + * * Revision 1.19 2001/05/01 22:37:37 acki2 * - now features 24->32 conversion (this is actually faster than letting the * codec produce depth 32 in the first place for avis :-))) ) @@ -113,6 +116,8 @@ LIBVO_EXTERN( dga ) #include "x11_common.h" +#include "fastmemcpy.h" + static vo_info_t vo_info = { #ifdef HAVE_DGA2 @@ -384,7 +389,17 @@ static uint32_t draw_frame( uint8_t *src[] ){ switch(SRC_MODE.vdm_conversion_func){ case VDM_CONV_NATIVE: - rep_movsl(d, s, lpl, vo_dga_vp_skip, numlines ); + {int i; + for(i=0; i< vo_dga_lines; i++){ + memcpy(d, s, vo_dga_bytes_per_line); + d+=vo_dga_vp_skip; + d+=vo_dga_bytes_per_line; + s+=vo_dga_bytes_per_line; + } + } + + + // rep_movsl(d, s, lpl, vo_dga_vp_skip, numlines ); break; case VDM_CONV_15TO16: printf("vo_dga: 15 to 16 not implemented yet!!!\n");