ffmpeg/libavcodec/i386/dsputil_mmx_avg.h
Fabrice Bellard de6d9b6404 Initial revision
Originally committed as revision 5 to svn://svn.ffmpeg.org/ffmpeg/trunk
2001-07-22 14:18:56 +00:00

353 lines
7.8 KiB
C

/*
* DSP utils : average functions are compiled twice for 3dnow/mmx2
* Copyright (c) 2000, 2001 Gerard Lantau.
*
* 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
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*
* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
*/
static void DEF(put_pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
int dh, hh;
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
hh=h>>2;
dh=h&3;
while(hh--) {
__asm __volatile(
"movq %4, %%mm0\n\t"
"movq 1%4, %%mm1\n\t"
"movq %5, %%mm2\n\t"
"movq 1%5, %%mm3\n\t"
"movq %6, %%mm4\n\t"
"movq 1%6, %%mm5\n\t"
"movq %7, %%mm6\n\t"
"movq 1%7, %%mm7\n\t"
PAVGB" %%mm1, %%mm0\n\t"
PAVGB" %%mm3, %%mm2\n\t"
PAVGB" %%mm5, %%mm4\n\t"
PAVGB" %%mm7, %%mm6\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm2, %1\n\t"
"movq %%mm4, %2\n\t"
"movq %%mm6, %3\n\t"
:"=m"(*p), "=m"(*(p+line_size)), "=m"(*(p+line_size*2)), "=m"(*(p+line_size*3))
:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2)), "m"(*(pix+line_size*3))
:"memory");
pix += line_size*4; p += line_size*4;
}
while(dh--) {
__asm __volatile(
"movq %1, %%mm0\n\t"
"movq 1%1, %%mm1\n\t"
PAVGB" %%mm1, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix)
:"memory");
pix += line_size; p += line_size;
}
emms();
}
static void DEF(put_pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
int dh, hh;
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
hh=h>>1;
dh=h&1;
while(hh--) {
__asm __volatile(
"movq %2, %%mm0\n\t"
"movq %3, %%mm1\n\t"
"movq %4, %%mm2\n\t"
PAVGB" %%mm1, %%mm0\n\t"
PAVGB" %%mm2, %%mm1\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm1, %1\n\t"
:"=m"(*p), "=m"(*(p+line_size))
:"m"(*pix), "m"(*(pix+line_size)),
"m"(*(pix+line_size*2))
:"memory");
pix += line_size*2;
p += line_size*2;
}
if(dh) {
__asm __volatile(
"movq %1, %%mm0\n\t"
"movq %2, %%mm1\n\t"
PAVGB" %%mm1, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix),
"m"(*(pix+line_size))
:"memory");
}
emms();
}
static void DEF(avg_pixels)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
int dh, hh;
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
hh=h>>2;
dh=h&3;
while(hh--) {
__asm __volatile(
"movq %0, %%mm0\n\t"
"movq %4, %%mm1\n\t"
"movq %1, %%mm2\n\t"
"movq %5, %%mm3\n\t"
"movq %2, %%mm4\n\t"
"movq %6, %%mm5\n\t"
"movq %3, %%mm6\n\t"
"movq %7, %%mm7\n\t"
PAVGB" %%mm1, %%mm0\n\t"
PAVGB" %%mm3, %%mm2\n\t"
PAVGB" %%mm5, %%mm4\n\t"
PAVGB" %%mm7, %%mm6\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm2, %1\n\t"
"movq %%mm4, %2\n\t"
"movq %%mm6, %3\n\t"
:"=m"(*p), "=m"(*(p+line_size)), "=m"(*(p+line_size*2)), "=m"(*(p+line_size*3))
:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2)), "m"(*(pix+line_size*3))
:"memory");
pix += line_size*4; p += line_size*4;
}
while(dh--) {
__asm __volatile(
"movq %0, %%mm0\n\t"
"movq %1, %%mm1\n\t"
PAVGB" %%mm1, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix)
:"memory");
pix += line_size; p += line_size;
}
emms();
}
static void DEF(avg_pixels_x2)( UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
int dh, hh;
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
hh=h>>1;
dh=h&1;
while(hh--) {
__asm __volatile(
"movq %2, %%mm2\n\t"
"movq 1%2, %%mm3\n\t"
"movq %3, %%mm4\n\t"
"movq 1%3, %%mm5\n\t"
"movq %0, %%mm0\n\t"
"movq %1, %%mm1\n\t"
PAVGB" %%mm3, %%mm2\n\t"
PAVGB" %%mm2, %%mm0\n\t"
PAVGB" %%mm5, %%mm4\n\t"
PAVGB" %%mm4, %%mm1\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm1, %1\n\t"
:"=m"(*p), "=m"(*(p+line_size))
:"m"(*pix), "m"(*(pix+line_size))
:"memory");
pix += line_size*2;
p += line_size*2;
}
if(dh) {
__asm __volatile(
"movq %1, %%mm1\n\t"
"movq 1%1, %%mm2\n\t"
"movq %0, %%mm0\n\t"
PAVGB" %%mm2, %%mm1\n\t"
PAVGB" %%mm1, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix)
:"memory");
}
emms();
}
static void DEF(avg_pixels_y2)( UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
int dh, hh;
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
hh=h>>1;
dh=h&1;
while(hh--) {
__asm __volatile(
"movq %2, %%mm2\n\t"
"movq %3, %%mm3\n\t"
"movq %3, %%mm4\n\t"
"movq %4, %%mm5\n\t"
"movq %0, %%mm0\n\t"
"movq %1, %%mm1\n\t"
PAVGB" %%mm3, %%mm2\n\t"
PAVGB" %%mm2, %%mm0\n\t"
PAVGB" %%mm5, %%mm4\n\t"
PAVGB" %%mm4, %%mm1\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm1, %1\n\t"
:"=m"(*p), "=m"(*(p+line_size))
:"m"(*pix), "m"(*(pix+line_size)), "m"(*(pix+line_size*2))
:"memory");
pix += line_size*2;
p += line_size*2;
}
if(dh) {
__asm __volatile(
"movq %1, %%mm1\n\t"
"movq %2, %%mm2\n\t"
"movq %0, %%mm0\n\t"
PAVGB" %%mm2, %%mm1\n\t"
PAVGB" %%mm1, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix), "m"(*(pix+line_size))
:"memory");
}
emms();
}
static void DEF(avg_pixels_xy2)( UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
UINT8 *p;
const UINT8 *pix;
p = block;
pix = pixels;
__asm __volatile(
"pxor %%mm7, %%mm7\n\t"
"movq %0, %%mm6\n\t"
::"m"(mm_wtwo[0]):"memory");
do {
__asm __volatile(
"movq %1, %%mm0\n\t"
"movq %2, %%mm1\n\t"
"movq 1%1, %%mm4\n\t"
"movq 1%2, %%mm5\n\t"
"movq %%mm0, %%mm2\n\t"
"movq %%mm1, %%mm3\n\t"
"punpcklbw %%mm7, %%mm0\n\t"
"punpcklbw %%mm7, %%mm1\n\t"
"punpckhbw %%mm7, %%mm2\n\t"
"punpckhbw %%mm7, %%mm3\n\t"
"paddusw %%mm1, %%mm0\n\t"
"paddusw %%mm3, %%mm2\n\t"
"movq %%mm4, %%mm1\n\t"
"movq %%mm5, %%mm3\n\t"
"punpcklbw %%mm7, %%mm4\n\t"
"punpcklbw %%mm7, %%mm5\n\t"
"punpckhbw %%mm7, %%mm1\n\t"
"punpckhbw %%mm7, %%mm3\n\t"
"paddusw %%mm5, %%mm4\n\t"
"paddusw %%mm3, %%mm1\n\t"
"paddusw %%mm6, %%mm4\n\t"
"paddusw %%mm6, %%mm1\n\t"
"paddusw %%mm4, %%mm0\n\t"
"paddusw %%mm1, %%mm2\n\t"
"psrlw $2, %%mm0\n\t"
"psrlw $2, %%mm2\n\t"
"packuswb %%mm2, %%mm0\n\t"
PAVGB" %0, %%mm0\n\t"
"movq %%mm0, %0\n\t"
:"=m"(*p)
:"m"(*pix),
"m"(*(pix+line_size))
:"memory");
pix += line_size;
p += line_size ;
} while(--h);
emms();
}
static void DEF(sub_pixels_x2)( DCTELEM *block, const UINT8 *pixels, int line_size, int h)
{
DCTELEM *p;
const UINT8 *pix;
p = block;
pix = pixels;
__asm __volatile(
"pxor %%mm7, %%mm7":::"memory");
do {
__asm __volatile(
"movq 1%1, %%mm2\n\t"
"movq %0, %%mm0\n\t"
PAVGB" %1, %%mm2\n\t"
"movq 8%0, %%mm1\n\t"
"movq %%mm2, %%mm3\n\t"
"punpcklbw %%mm7, %%mm2\n\t"
"punpckhbw %%mm7, %%mm3\n\t"
"psubsw %%mm2, %%mm0\n\t"
"psubsw %%mm3, %%mm1\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm1, 8%0\n\t"
:"=m"(*p)
:"m"(*pix)
:"memory");
pix += line_size;
p += 8;
} while (--h);
emms();
}
static void DEF(sub_pixels_y2)( DCTELEM *block, const UINT8 *pixels, int line_size, int h)
{
DCTELEM *p;
const UINT8 *pix;
p = block;
pix = pixels;
__asm __volatile(
"pxor %%mm7, %%mm7":::"memory");
do {
__asm __volatile(
"movq %2, %%mm2\n\t"
"movq %0, %%mm0\n\t"
PAVGB" %1, %%mm2\n\t"
"movq 8%0, %%mm1\n\t"
"movq %%mm2, %%mm3\n\t"
"punpcklbw %%mm7, %%mm2\n\t"
"punpckhbw %%mm7, %%mm3\n\t"
"psubsw %%mm2, %%mm0\n\t"
"psubsw %%mm3, %%mm1\n\t"
"movq %%mm0, %0\n\t"
"movq %%mm1, 8%0\n\t"
:"=m"(*p)
:"m"(*pix), "m"(*(pix+line_size))
:"memory");
pix += line_size;
p += 8;
} while (--h);
emms();
}