Commit 78ff3248 authored by dionoea's avatar dionoea

Optimze nearest neighboor scaling implementation. This is used by the sub...

Optimze nearest neighboor scaling implementation. This is used by the sub pictures core. CPU decrease compared to previous version with my test filter was: 66% for YUVA, 65% for I420 and 71% for RGBA.
parent b40e1114
/*****************************************************************************
* resize.c: video scaling module for YUVP/A pictures
* resize.c: video scaling module for YUVP/A, I420 and RGBA pictures
* Uses the low quality "nearest neighbour" algorithm.
*****************************************************************************
* Copyright (C) 2003 the VideoLAN team
* Copyright (C) 2003-2007 the VideoLAN team
* $Id$
*
* Author: Gildas Bazin <gbazin@videolan.org>
* Authors: Gildas Bazin <gbazin@videolan.org>
* Antoine Cellerier <dionoea @t videolan dot org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
......@@ -109,7 +110,7 @@ static void CloseFilter( vlc_object_t *p_this )
static picture_t *Filter( filter_t *p_filter, picture_t *p_pic )
{
picture_t *p_pic_dst;
int i_plane, i, j, k, l;
int i_plane;
if( !p_pic ) return NULL;
......@@ -127,56 +128,83 @@ static picture_t *Filter( filter_t *p_filter, picture_t *p_pic )
{
for( i_plane = 0; i_plane < p_pic_dst->i_planes; i_plane++ )
{
const int i_src_pitch = p_pic->p[i_plane].i_pitch;
const int i_dst_pitch = p_pic_dst->p[i_plane].i_pitch;
const int i_src_height = p_filter->fmt_in.video.i_height;
const int i_src_width = p_filter->fmt_in.video.i_width;
const int i_dst_height = p_filter->fmt_out.video.i_height;
const int i_dst_width = p_filter->fmt_out.video.i_width;
const int i_dst_visible_lines =
p_pic_dst->p[i_plane].i_visible_lines;
const int i_dst_visible_pitch =
p_pic_dst->p[i_plane].i_visible_pitch;
const int i_dst_hidden_pitch = i_dst_pitch - i_dst_visible_pitch;
#define SHIFT_SIZE 16
const int i_height_coef = ( i_src_height << SHIFT_SIZE )
/ i_dst_height;
const int i_width_coef = ( i_src_width << SHIFT_SIZE )
/ i_dst_width;
const int i_src_height_1 = i_src_height - 1;
const int i_src_width_1 = i_src_width - 1;
uint8_t *p_src = p_pic->p[i_plane].p_pixels;
uint8_t *p_dst = p_pic_dst->p[i_plane].p_pixels;
int i_src_pitch = p_pic->p[i_plane].i_pitch;
int i_dst_pitch = p_pic_dst->p[i_plane].i_pitch;
uint8_t *p_dstendline = p_dst + i_dst_visible_pitch;
const uint8_t *p_dstend = p_dst + i_dst_visible_lines*i_dst_pitch;
for( i = 0; i < p_pic_dst->p[i_plane].i_visible_lines; i++ )
int l = 1<<(SHIFT_SIZE-1);
for( ; p_dst < p_dstend;
p_dst += i_dst_hidden_pitch,
p_dstendline += i_dst_pitch, l += i_height_coef )
{
l = ( p_filter->fmt_in.video.i_height * i +
p_filter->fmt_out.video.i_height / 2 ) /
p_filter->fmt_out.video.i_height;
l = __MIN( (int)p_filter->fmt_in.video.i_height - 1, l );
int k = 1<<(SHIFT_SIZE-1);
uint8_t *p_srcl = p_src
+ (__MIN( i_src_height_1, l >> SHIFT_SIZE )*i_src_pitch);
for( j = 0; j < p_pic_dst->p[i_plane].i_visible_pitch; j++ )
for( ; p_dst < p_dstendline; p_dst++, k += i_width_coef )
{
k = ( p_filter->fmt_in.video.i_width * j +
p_filter->fmt_out.video.i_width / 2 ) /
p_filter->fmt_out.video.i_width;
k = __MIN( (int)p_filter->fmt_in.video.i_width - 1, k );
p_dst[i * i_dst_pitch + j] = p_src[l * i_src_pitch + k];
*p_dst = p_srcl[__MIN( i_src_width_1, k >> SHIFT_SIZE )];
}
}
}
}
else /* RGBA */
{
uint8_t *p_src = p_pic->p->p_pixels;
uint8_t *p_dst = p_pic_dst->p->p_pixels;
int i_src_pitch = p_pic->p->i_pitch;
int i_dst_pitch = p_pic_dst->p->i_pitch;
for( i = 0; i < p_pic_dst->p->i_visible_lines; i++ )
const int i_src_pitch = p_pic->p->i_pitch;
const int i_dst_pitch = p_pic_dst->p->i_pitch;
const int i_src_height = p_filter->fmt_in.video.i_height;
const int i_src_width = p_filter->fmt_in.video.i_width;
const int i_dst_height = p_filter->fmt_out.video.i_height;
const int i_dst_width = p_filter->fmt_out.video.i_width;
const int i_dst_visible_lines =
p_pic_dst->p->i_visible_lines;
const int i_dst_visible_pitch =
p_pic_dst->p->i_visible_pitch;
const int i_dst_hidden_pitch = i_dst_pitch - i_dst_visible_pitch;
const int i_height_coef = ( i_src_height << SHIFT_SIZE )
/ i_dst_height;
const int i_width_coef = ( i_src_width << SHIFT_SIZE )
/ i_dst_width;
const int i_src_height_1 = i_src_height - 1;
const int i_src_width_1 = i_src_width - 1;
uint32_t *p_src = (uint32_t*)p_pic->p->p_pixels;
uint32_t *p_dst = (uint32_t*)p_pic_dst->p->p_pixels;
uint32_t *p_dstendline = p_dst + (i_dst_visible_pitch>>2);
const uint32_t *p_dstend = p_dst + i_dst_visible_lines*(i_dst_pitch>>2);
int l = 1<<(SHIFT_SIZE-1);
for( ; p_dst < p_dstend;
p_dst += (i_dst_hidden_pitch>>2),
p_dstendline += (i_dst_pitch>>2),
l += i_height_coef )
{
l = ( p_filter->fmt_in.video.i_height * i +
p_filter->fmt_out.video.i_height / 2 ) /
p_filter->fmt_out.video.i_height;
l = __MIN( (int)p_filter->fmt_in.video.i_height - 1, l );
for( j = 0; j < p_pic_dst->p->i_visible_pitch/4; j++ )
int k = 1<<(SHIFT_SIZE-1);
uint32_t *p_srcl = p_src
+ (__MIN( i_src_height_1, l >> SHIFT_SIZE )*(i_src_pitch>>2));
for( ; p_dst < p_dstendline; p_dst++, k += i_width_coef )
{
k = ( p_filter->fmt_in.video.i_width * j +
p_filter->fmt_out.video.i_width / 2 ) /
p_filter->fmt_out.video.i_width;
k = __MIN( (int)p_filter->fmt_in.video.i_width - 1, k );
*(uint32_t*)(&p_dst[i * i_dst_pitch + 4*j]) =
*(uint32_t*)(&p_src[l * i_src_pitch + 4*k]);
*p_dst = p_srcl[__MIN( i_src_width_1, k >> SHIFT_SIZE )];
}
}
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment