2006-07-27 19:35:30 +00:00
|
|
|
/*
|
2006-10-07 15:33:14 +00:00
|
|
|
* AltiVec-enhanced yuv2yuvX
|
|
|
|
*
|
|
|
|
* Copyright (C) 2004 Romain Dolbeau <romain@dolbeau.org>
|
2008-02-25 09:43:52 +00:00
|
|
|
* based on the equivalent C code in swscale.c
|
2006-10-07 15:33:14 +00:00
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* This file is part of Libav.
|
2006-10-07 15:33:14 +00:00
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* Libav is free software; you can redistribute it and/or
|
2009-03-29 13:47:30 +00:00
|
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
|
|
* License as published by the Free Software Foundation; either
|
|
|
|
* version 2.1 of the License, or (at your option) any later version.
|
2006-10-07 15:33:14 +00:00
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* Libav is distributed in the hope that it will be useful,
|
2006-10-07 15:33:14 +00:00
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
2009-03-29 13:47:30 +00:00
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* Lesser General Public License for more details.
|
2006-10-07 15:33:14 +00:00
|
|
|
*
|
2009-03-29 13:47:30 +00:00
|
|
|
* You should have received a copy of the GNU Lesser General Public
|
2011-03-18 17:35:10 +00:00
|
|
|
* License along with Libav; if not, write to the Free Software
|
2007-07-05 10:18:58 +00:00
|
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
2006-10-07 15:33:14 +00:00
|
|
|
*/
|
2004-03-11 17:58:11 +00:00
|
|
|
|
2011-06-03 03:04:04 +00:00
|
|
|
#include <inttypes.h>
|
|
|
|
#include "config.h"
|
|
|
|
#include "libswscale/swscale.h"
|
|
|
|
#include "libswscale/swscale_internal.h"
|
|
|
|
#include "libavutil/cpu.h"
|
|
|
|
#include "yuv2rgb_altivec.h"
|
|
|
|
|
2006-04-06 10:08:54 +00:00
|
|
|
#define vzero vec_splat_s32(0)
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2004-03-11 17:58:11 +00:00
|
|
|
static inline void
|
2009-08-16 21:11:28 +00:00
|
|
|
altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
|
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
register int i;
|
|
|
|
vector unsigned int altivec_vectorShiftInt19 =
|
2008-01-17 10:24:14 +00:00
|
|
|
vec_add(vec_splat_u32(10), vec_splat_u32(9));
|
2011-05-26 16:57:26 +00:00
|
|
|
if ((unsigned int)dest % 16) {
|
2007-12-02 14:07:01 +00:00
|
|
|
/* badly aligned store, we force store alignment */
|
|
|
|
/* and will handle load misalignment on val w/ vec_perm */
|
2007-04-28 11:44:49 +00:00
|
|
|
vector unsigned char perm1;
|
|
|
|
vector signed int v1;
|
|
|
|
for (i = 0 ; (i < dstW) &&
|
2011-05-26 16:57:26 +00:00
|
|
|
(((unsigned int)dest + i) % 16) ; i++) {
|
2007-04-28 11:44:49 +00:00
|
|
|
int t = val[i] >> 19;
|
|
|
|
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
|
|
|
|
}
|
|
|
|
perm1 = vec_lvsl(i << 2, val);
|
|
|
|
v1 = vec_ld(i << 2, val);
|
|
|
|
for ( ; i < (dstW - 15); i+=16) {
|
|
|
|
int offset = i << 2;
|
|
|
|
vector signed int v2 = vec_ld(offset + 16, val);
|
|
|
|
vector signed int v3 = vec_ld(offset + 32, val);
|
|
|
|
vector signed int v4 = vec_ld(offset + 48, val);
|
|
|
|
vector signed int v5 = vec_ld(offset + 64, val);
|
2008-01-17 10:24:14 +00:00
|
|
|
vector signed int v12 = vec_perm(v1, v2, perm1);
|
|
|
|
vector signed int v23 = vec_perm(v2, v3, perm1);
|
|
|
|
vector signed int v34 = vec_perm(v3, v4, perm1);
|
|
|
|
vector signed int v45 = vec_perm(v4, v5, perm1);
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
vector signed int vA = vec_sra(v12, altivec_vectorShiftInt19);
|
|
|
|
vector signed int vB = vec_sra(v23, altivec_vectorShiftInt19);
|
|
|
|
vector signed int vC = vec_sra(v34, altivec_vectorShiftInt19);
|
|
|
|
vector signed int vD = vec_sra(v45, altivec_vectorShiftInt19);
|
|
|
|
vector unsigned short vs1 = vec_packsu(vA, vB);
|
|
|
|
vector unsigned short vs2 = vec_packsu(vC, vD);
|
|
|
|
vector unsigned char vf = vec_packsu(vs1, vs2);
|
|
|
|
vec_st(vf, i, dest);
|
|
|
|
v1 = v5;
|
|
|
|
}
|
|
|
|
} else { // dest is properly aligned, great
|
|
|
|
for (i = 0; i < (dstW - 15); i+=16) {
|
|
|
|
int offset = i << 2;
|
|
|
|
vector signed int v1 = vec_ld(offset, val);
|
|
|
|
vector signed int v2 = vec_ld(offset + 16, val);
|
|
|
|
vector signed int v3 = vec_ld(offset + 32, val);
|
|
|
|
vector signed int v4 = vec_ld(offset + 48, val);
|
|
|
|
vector signed int v5 = vec_sra(v1, altivec_vectorShiftInt19);
|
|
|
|
vector signed int v6 = vec_sra(v2, altivec_vectorShiftInt19);
|
|
|
|
vector signed int v7 = vec_sra(v3, altivec_vectorShiftInt19);
|
|
|
|
vector signed int v8 = vec_sra(v4, altivec_vectorShiftInt19);
|
|
|
|
vector unsigned short vs1 = vec_packsu(v5, v6);
|
|
|
|
vector unsigned short vs2 = vec_packsu(v7, v8);
|
|
|
|
vector unsigned char vf = vec_packsu(vs1, vs2);
|
|
|
|
vec_st(vf, i, dest);
|
|
|
|
}
|
2004-03-11 17:58:11 +00:00
|
|
|
}
|
2007-04-28 11:44:49 +00:00
|
|
|
for ( ; i < dstW ; i++) {
|
|
|
|
int t = val[i] >> 19;
|
|
|
|
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
|
2004-03-11 17:58:11 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-06-03 02:28:56 +00:00
|
|
|
static void
|
|
|
|
yuv2yuvX_altivec_real(SwsContext *c,
|
|
|
|
const int16_t *lumFilter, const int16_t **lumSrc,
|
2011-05-26 15:32:32 +00:00
|
|
|
int lumFilterSize, const int16_t *chrFilter,
|
|
|
|
const int16_t **chrUSrc, const int16_t **chrVSrc,
|
2011-06-03 02:28:56 +00:00
|
|
|
int chrFilterSize, const int16_t **alpSrc,
|
|
|
|
uint8_t *dest, uint8_t *uDest,
|
|
|
|
uint8_t *vDest, uint8_t *aDest,
|
|
|
|
int dstW, int chrDstW)
|
2004-03-11 17:58:11 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
const vector signed int vini = {(1 << 18), (1 << 18), (1 << 18), (1 << 18)};
|
|
|
|
register int i, j;
|
|
|
|
{
|
2010-01-22 03:26:30 +00:00
|
|
|
DECLARE_ALIGNED(16, int, val)[dstW];
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i = 0; i < (dstW -7); i+=4) {
|
|
|
|
vec_st(vini, i << 2, val);
|
|
|
|
}
|
|
|
|
for (; i < dstW; i++) {
|
|
|
|
val[i] = (1 << 18);
|
|
|
|
}
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
for (j = 0; j < lumFilterSize; j++) {
|
|
|
|
vector signed short l1, vLumFilter = vec_ld(j << 1, lumFilter);
|
|
|
|
vector unsigned char perm, perm0 = vec_lvsl(j << 1, lumFilter);
|
|
|
|
vLumFilter = vec_perm(vLumFilter, vLumFilter, perm0);
|
|
|
|
vLumFilter = vec_splat(vLumFilter, 0); // lumFilter[j] is loaded 8 times in vLumFilter
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
perm = vec_lvsl(0, lumSrc[j]);
|
|
|
|
l1 = vec_ld(0, lumSrc[j]);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i = 0; i < (dstW - 7); i+=8) {
|
|
|
|
int offset = i << 2;
|
|
|
|
vector signed short l2 = vec_ld((i << 1) + 16, lumSrc[j]);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int v1 = vec_ld(offset, val);
|
|
|
|
vector signed int v2 = vec_ld(offset + 16, val);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short ls = vec_perm(l1, l2, perm); // lumSrc[j][i] ... lumSrc[j][i+7]
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int i1 = vec_mule(vLumFilter, ls);
|
|
|
|
vector signed int i2 = vec_mulo(vLumFilter, ls);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int vf1 = vec_mergeh(i1, i2);
|
|
|
|
vector signed int vf2 = vec_mergel(i1, i2); // lumSrc[j][i] * lumFilter[j] ... lumSrc[j][i+7] * lumFilter[j]
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int vo1 = vec_add(v1, vf1);
|
|
|
|
vector signed int vo2 = vec_add(v2, vf2);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vec_st(vo1, offset, val);
|
|
|
|
vec_st(vo2, offset + 16, val);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
l1 = l2;
|
|
|
|
}
|
|
|
|
for ( ; i < dstW; i++) {
|
|
|
|
val[i] += lumSrc[j][i] * lumFilter[j];
|
|
|
|
}
|
|
|
|
}
|
2008-01-17 10:24:14 +00:00
|
|
|
altivec_packIntArrayToCharArray(val, dest, dstW);
|
2004-03-11 17:58:11 +00:00
|
|
|
}
|
2007-04-28 11:44:49 +00:00
|
|
|
if (uDest != 0) {
|
2010-01-22 03:26:30 +00:00
|
|
|
DECLARE_ALIGNED(16, int, u)[chrDstW];
|
|
|
|
DECLARE_ALIGNED(16, int, v)[chrDstW];
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
for (i = 0; i < (chrDstW -7); i+=4) {
|
|
|
|
vec_st(vini, i << 2, u);
|
|
|
|
vec_st(vini, i << 2, v);
|
|
|
|
}
|
|
|
|
for (; i < chrDstW; i++) {
|
|
|
|
u[i] = (1 << 18);
|
|
|
|
v[i] = (1 << 18);
|
|
|
|
}
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
for (j = 0; j < chrFilterSize; j++) {
|
|
|
|
vector signed short l1, l1_V, vChrFilter = vec_ld(j << 1, chrFilter);
|
|
|
|
vector unsigned char perm, perm0 = vec_lvsl(j << 1, chrFilter);
|
|
|
|
vChrFilter = vec_perm(vChrFilter, vChrFilter, perm0);
|
|
|
|
vChrFilter = vec_splat(vChrFilter, 0); // chrFilter[j] is loaded 8 times in vChrFilter
|
|
|
|
|
2011-05-26 15:32:32 +00:00
|
|
|
perm = vec_lvsl(0, chrUSrc[j]);
|
|
|
|
l1 = vec_ld(0, chrUSrc[j]);
|
|
|
|
l1_V = vec_ld(0, chrVSrc[j]);
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
for (i = 0; i < (chrDstW - 7); i+=8) {
|
|
|
|
int offset = i << 2;
|
2011-05-26 15:32:32 +00:00
|
|
|
vector signed short l2 = vec_ld((i << 1) + 16, chrUSrc[j]);
|
|
|
|
vector signed short l2_V = vec_ld((i << 1) + 16, chrVSrc[j]);
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
vector signed int v1 = vec_ld(offset, u);
|
|
|
|
vector signed int v2 = vec_ld(offset + 16, u);
|
|
|
|
vector signed int v1_V = vec_ld(offset, v);
|
|
|
|
vector signed int v2_V = vec_ld(offset + 16, v);
|
|
|
|
|
2011-05-26 15:32:32 +00:00
|
|
|
vector signed short ls = vec_perm(l1, l2, perm); // chrUSrc[j][i] ... chrUSrc[j][i+7]
|
|
|
|
vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrVSrc[j][i] ... chrVSrc[j][i]
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
vector signed int i1 = vec_mule(vChrFilter, ls);
|
|
|
|
vector signed int i2 = vec_mulo(vChrFilter, ls);
|
|
|
|
vector signed int i1_V = vec_mule(vChrFilter, ls_V);
|
|
|
|
vector signed int i2_V = vec_mulo(vChrFilter, ls_V);
|
|
|
|
|
|
|
|
vector signed int vf1 = vec_mergeh(i1, i2);
|
2011-05-26 15:32:32 +00:00
|
|
|
vector signed int vf2 = vec_mergel(i1, i2); // chrUSrc[j][i] * chrFilter[j] ... chrUSrc[j][i+7] * chrFilter[j]
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int vf1_V = vec_mergeh(i1_V, i2_V);
|
2011-05-26 15:32:32 +00:00
|
|
|
vector signed int vf2_V = vec_mergel(i1_V, i2_V); // chrVSrc[j][i] * chrFilter[j] ... chrVSrc[j][i+7] * chrFilter[j]
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
vector signed int vo1 = vec_add(v1, vf1);
|
|
|
|
vector signed int vo2 = vec_add(v2, vf2);
|
|
|
|
vector signed int vo1_V = vec_add(v1_V, vf1_V);
|
|
|
|
vector signed int vo2_V = vec_add(v2_V, vf2_V);
|
|
|
|
|
|
|
|
vec_st(vo1, offset, u);
|
|
|
|
vec_st(vo2, offset + 16, u);
|
|
|
|
vec_st(vo1_V, offset, v);
|
|
|
|
vec_st(vo2_V, offset + 16, v);
|
|
|
|
|
|
|
|
l1 = l2;
|
|
|
|
l1_V = l2_V;
|
|
|
|
}
|
|
|
|
for ( ; i < chrDstW; i++) {
|
2011-05-26 15:32:32 +00:00
|
|
|
u[i] += chrUSrc[j][i] * chrFilter[j];
|
|
|
|
v[i] += chrVSrc[j][i] * chrFilter[j];
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
|
|
|
}
|
2008-01-17 10:24:14 +00:00
|
|
|
altivec_packIntArrayToCharArray(u, uDest, chrDstW);
|
|
|
|
altivec_packIntArrayToCharArray(v, vDest, chrDstW);
|
2004-03-11 17:58:11 +00:00
|
|
|
}
|
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2011-06-03 03:04:04 +00:00
|
|
|
static void hScale_altivec_real(int16_t *dst, int dstW,
|
|
|
|
const uint8_t *src, int srcW,
|
|
|
|
int xInc, const int16_t *filter,
|
|
|
|
const int16_t *filterPos, int filterSize)
|
2009-05-09 10:56:37 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
register int i;
|
2010-01-22 03:26:30 +00:00
|
|
|
DECLARE_ALIGNED(16, int, tempo)[4];
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
if (filterSize % 4) {
|
|
|
|
for (i=0; i<dstW; i++) {
|
|
|
|
register int j;
|
|
|
|
register int srcPos = filterPos[i];
|
|
|
|
register int val = 0;
|
|
|
|
for (j=0; j<filterSize; j++) {
|
|
|
|
val += ((int)src[srcPos + j])*filter[filterSize*i + j];
|
|
|
|
}
|
2009-02-14 16:39:15 +00:00
|
|
|
dst[i] = FFMIN(val>>7, (1<<15)-1);
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
2007-04-28 11:44:49 +00:00
|
|
|
else
|
|
|
|
switch (filterSize) {
|
|
|
|
case 4:
|
2004-04-06 00:25:47 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i=0; i<dstW; i++) {
|
|
|
|
register int srcPos = filterPos[i];
|
|
|
|
|
|
|
|
vector unsigned char src_v0 = vec_ld(srcPos, src);
|
|
|
|
vector unsigned char src_v1, src_vF;
|
|
|
|
vector signed short src_v, filter_v;
|
|
|
|
vector signed int val_vEven, val_s;
|
|
|
|
if ((((int)src + srcPos)% 16) > 12) {
|
|
|
|
src_v1 = vec_ld(srcPos + 16, src);
|
|
|
|
}
|
|
|
|
src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
|
|
|
|
|
|
|
|
src_v = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
|
|
|
|
// now put our elements in the even slots
|
|
|
|
src_v = vec_mergeh(src_v, (vector signed short)vzero);
|
|
|
|
|
|
|
|
filter_v = vec_ld(i << 3, filter);
|
2008-07-04 13:49:45 +00:00
|
|
|
// The 3 above is 2 (filterSize == 4) + 1 (sizeof(short) == 2).
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2008-07-04 13:49:45 +00:00
|
|
|
// The neat trick: We only care for half the elements,
|
2004-04-06 00:25:47 +00:00
|
|
|
// high or low depending on (i<<3)%16 (it's 0 or 8 here),
|
2008-07-04 13:49:45 +00:00
|
|
|
// and we're going to use vec_mule, so we choose
|
|
|
|
// carefully how to "unpack" the elements into the even slots.
|
2007-04-28 11:44:49 +00:00
|
|
|
if ((i << 3) % 16)
|
2008-01-17 10:24:14 +00:00
|
|
|
filter_v = vec_mergel(filter_v, (vector signed short)vzero);
|
2007-04-28 11:44:49 +00:00
|
|
|
else
|
2008-01-17 10:24:14 +00:00
|
|
|
filter_v = vec_mergeh(filter_v, (vector signed short)vzero);
|
2007-04-28 11:44:49 +00:00
|
|
|
|
|
|
|
val_vEven = vec_mule(src_v, filter_v);
|
|
|
|
val_s = vec_sums(val_vEven, vzero);
|
|
|
|
vec_st(val_s, 0, tempo);
|
2009-02-14 16:39:15 +00:00
|
|
|
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
case 8:
|
2004-04-06 00:25:47 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i=0; i<dstW; i++) {
|
|
|
|
register int srcPos = filterPos[i];
|
|
|
|
|
|
|
|
vector unsigned char src_v0 = vec_ld(srcPos, src);
|
|
|
|
vector unsigned char src_v1, src_vF;
|
|
|
|
vector signed short src_v, filter_v;
|
|
|
|
vector signed int val_v, val_s;
|
|
|
|
if ((((int)src + srcPos)% 16) > 8) {
|
|
|
|
src_v1 = vec_ld(srcPos + 16, src);
|
|
|
|
}
|
|
|
|
src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
|
|
|
|
|
|
|
|
src_v = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
|
|
|
|
filter_v = vec_ld(i << 4, filter);
|
2004-04-06 00:25:47 +00:00
|
|
|
// the 4 above is 3 (filterSize == 8) + 1 (sizeof(short) == 2)
|
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
val_v = vec_msums(src_v, filter_v, (vector signed int)vzero);
|
|
|
|
val_s = vec_sums(val_v, vzero);
|
|
|
|
vec_st(val_s, 0, tempo);
|
2009-02-14 16:39:15 +00:00
|
|
|
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
case 16:
|
2004-04-06 00:25:47 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i=0; i<dstW; i++) {
|
|
|
|
register int srcPos = filterPos[i];
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector unsigned char src_v0 = vec_ld(srcPos, src);
|
|
|
|
vector unsigned char src_v1 = vec_ld(srcPos + 16, src);
|
|
|
|
vector unsigned char src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short src_vA = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
|
|
|
|
vector signed short src_vB = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short filter_v0 = vec_ld(i << 5, filter);
|
|
|
|
vector signed short filter_v1 = vec_ld((i << 5) + 16, filter);
|
|
|
|
// the 5 above are 4 (filterSize == 16) + 1 (sizeof(short) == 2)
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int val_acc = vec_msums(src_vA, filter_v0, (vector signed int)vzero);
|
|
|
|
vector signed int val_v = vec_msums(src_vB, filter_v1, val_acc);
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int val_s = vec_sums(val_v, vzero);
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vec_st(val_s, 0, tempo);
|
2009-02-14 16:39:15 +00:00
|
|
|
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
|
|
|
break;
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
default:
|
2004-04-06 00:25:47 +00:00
|
|
|
{
|
2007-04-28 11:44:49 +00:00
|
|
|
for (i=0; i<dstW; i++) {
|
|
|
|
register int j;
|
|
|
|
register int srcPos = filterPos[i];
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2006-11-01 19:44:35 +00:00
|
|
|
vector signed int val_s, val_v = (vector signed int)vzero;
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short filter_v0R = vec_ld(i * 2 * filterSize, filter);
|
2004-04-06 00:25:47 +00:00
|
|
|
vector unsigned char permF = vec_lvsl((i * 2 * filterSize), filter);
|
|
|
|
|
|
|
|
vector unsigned char src_v0 = vec_ld(srcPos, src);
|
|
|
|
vector unsigned char permS = vec_lvsl(srcPos, src);
|
|
|
|
|
|
|
|
for (j = 0 ; j < filterSize - 15; j += 16) {
|
2007-04-28 11:44:49 +00:00
|
|
|
vector unsigned char src_v1 = vec_ld(srcPos + j + 16, src);
|
|
|
|
vector unsigned char src_vF = vec_perm(src_v0, src_v1, permS);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short src_vA = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
|
|
|
|
vector signed short src_vB = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed short filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
|
|
|
|
vector signed short filter_v2R = vec_ld((i * 2 * filterSize) + (j * 2) + 32, filter);
|
|
|
|
vector signed short filter_v0 = vec_perm(filter_v0R, filter_v1R, permF);
|
|
|
|
vector signed short filter_v1 = vec_perm(filter_v1R, filter_v2R, permF);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
vector signed int val_acc = vec_msums(src_vA, filter_v0, val_v);
|
|
|
|
val_v = vec_msums(src_vB, filter_v1, val_acc);
|
2004-04-06 00:25:47 +00:00
|
|
|
|
2007-04-28 11:44:49 +00:00
|
|
|
filter_v0R = filter_v2R;
|
|
|
|
src_v0 = src_v1;
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
|
|
|
|
2008-01-17 11:07:27 +00:00
|
|
|
if (j < filterSize-7) {
|
2007-04-28 11:44:49 +00:00
|
|
|
// loading src_v0 is useless, it's already done above
|
|
|
|
//vector unsigned char src_v0 = vec_ld(srcPos + j, src);
|
|
|
|
vector unsigned char src_v1, src_vF;
|
|
|
|
vector signed short src_v, filter_v1R, filter_v;
|
|
|
|
if ((((int)src + srcPos)% 16) > 8) {
|
|
|
|
src_v1 = vec_ld(srcPos + j + 16, src);
|
|
|
|
}
|
|
|
|
src_vF = vec_perm(src_v0, src_v1, permS);
|
|
|
|
|
|
|
|
src_v = // vec_unpackh sign-extends...
|
|
|
|
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
|
|
|
|
// loading filter_v0R is useless, it's already done above
|
|
|
|
//vector signed short filter_v0R = vec_ld((i * 2 * filterSize) + j, filter);
|
|
|
|
filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
|
|
|
|
filter_v = vec_perm(filter_v0R, filter_v1R, permF);
|
|
|
|
|
|
|
|
val_v = vec_msums(src_v, filter_v, val_v);
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
|
|
|
|
2006-11-01 19:44:35 +00:00
|
|
|
val_s = vec_sums(val_v, vzero);
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2004-04-06 00:25:47 +00:00
|
|
|
vec_st(val_s, 0, tempo);
|
2009-02-14 16:39:15 +00:00
|
|
|
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2007-04-26 23:07:11 +00:00
|
|
|
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
2007-04-28 11:44:49 +00:00
|
|
|
}
|
2004-04-06 00:25:47 +00:00
|
|
|
}
|
2011-06-03 02:28:56 +00:00
|
|
|
|
2011-06-03 03:04:04 +00:00
|
|
|
void ff_sws_init_swScale_altivec(SwsContext *c)
|
2011-06-03 02:28:56 +00:00
|
|
|
{
|
2011-06-03 03:04:04 +00:00
|
|
|
if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
|
|
|
|
return;
|
|
|
|
|
2011-06-03 02:28:56 +00:00
|
|
|
c->yuv2yuvX = yuv2yuvX_altivec_real;
|
|
|
|
|
|
|
|
/* The following list of supported dstFormat values should
|
|
|
|
* match what's found in the body of ff_yuv2packedX_altivec() */
|
|
|
|
if (!(c->flags & SWS_BITEXACT) && !c->alpPixBuf &&
|
|
|
|
(c->dstFormat==PIX_FMT_ABGR || c->dstFormat==PIX_FMT_BGRA ||
|
|
|
|
c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 ||
|
|
|
|
c->dstFormat==PIX_FMT_RGBA || c->dstFormat==PIX_FMT_ARGB)) {
|
|
|
|
c->yuv2packedX = ff_yuv2packedX_altivec;
|
|
|
|
}
|
|
|
|
}
|