f46bb608 |
/*
* Copyright (c) 2002 Brian Foley
* Copyright (c) 2002 Dieter Shirley
* Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
* |
2d5e9451 |
* This file is part of FFmpeg. |
f46bb608 |
* |
2d5e9451 |
* FFmpeg is free software; you can redistribute it and/or |
f46bb608 |
* 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.
* |
2d5e9451 |
* FFmpeg is distributed in the hope that it will be useful, |
f46bb608 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public |
2d5e9451 |
* License along with FFmpeg; if not, write to the Free Software |
f46bb608 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "config.h"
#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/ppc/cpu.h"
#include "libavutil/ppc/util_altivec.h" |
ea7ee4b4 |
|
f46bb608 |
#include "libavcodec/avcodec.h"
#include "libavcodec/pixblockdsp.h"
#if HAVE_ALTIVEC
|
31dea051 |
#if HAVE_VSX
static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels, |
e07fa300 |
ptrdiff_t stride) |
31dea051 |
{
int i;
vector unsigned char perm =
(vector unsigned char) {0x00,0x10, 0x01,0x11,0x02,0x12,0x03,0x13,\
0x04,0x14,0x05,0x15,0x06,0x16,0x07,0x17};
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
for (i = 0; i < 8; i++) {
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
vector unsigned char bytes = vec_vsx_ld(0, pixels);
// Convert the bytes into shorts.
//vector signed short shorts = (vector signed short) vec_perm(zero, bytes, perm);
vector signed short shorts = (vector signed short) vec_perm(bytes, zero, perm);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_vsx_st(shorts, i * 16, (vector signed short *) block);
|
e07fa300 |
pixels += stride; |
31dea051 |
}
}
#else |
f46bb608 |
static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels, |
de452e50 |
ptrdiff_t stride) |
f46bb608 |
{
int i; |
7014b659 |
const vec_u8 zero = (const vec_u8)vec_splat_u8(0); |
f46bb608 |
for (i = 0; i < 8; i++) { |
3932ccc4 |
vec_u8 perm = vec_lvsl(0, pixels); |
f46bb608 |
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */ |
7014b659 |
vec_u8 pixl = vec_ld(0, pixels);
vec_u8 pixr = vec_ld(7, pixels);
vec_u8 bytes = vec_perm(pixl, pixr, perm); |
f46bb608 |
// Convert the bytes into shorts. |
7014b659 |
vec_s16 shorts = (vec_s16)vec_mergeh(zero, bytes); |
f46bb608 |
// Save the data to the block, we assume the block is 16-byte aligned. |
7014b659 |
vec_st(shorts, i * 16, (vec_s16 *)block); |
f46bb608 |
|
de452e50 |
pixels += stride; |
f46bb608 |
}
}
|
31dea051 |
#endif /* HAVE_VSX */
#if HAVE_VSX
static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1, |
e07fa300 |
const uint8_t *s2, ptrdiff_t stride) |
31dea051 |
{
int i;
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
vector signed short shorts1, shorts2;
for (i = 0; i < 4; i++) {
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
vector unsigned char bytes = vec_vsx_ld(0, s1);
// Convert the bytes into shorts.
shorts1 = (vector signed short) vec_mergeh(bytes, zero);
// Do the same for the second block of pixels.
bytes =vec_vsx_ld(0, s2);
// Convert the bytes into shorts.
shorts2 = (vector signed short) vec_mergeh(bytes, zero);
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_vsx_st(shorts1, 0, (vector signed short *) block);
s1 += stride;
s2 += stride;
block += 8;
/* The code below is a copy of the code above...
* This is a manual unroll. */
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
bytes = vec_vsx_ld(0, s1);
// Convert the bytes into shorts.
shorts1 = (vector signed short) vec_mergeh(bytes, zero);
// Do the same for the second block of pixels.
bytes = vec_vsx_ld(0, s2);
// Convert the bytes into shorts.
shorts2 = (vector signed short) vec_mergeh(bytes, zero);
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_vsx_st(shorts1, 0, (vector signed short *) block);
s1 += stride;
s2 += stride;
block += 8;
}
}
#else |
f46bb608 |
static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1, |
de452e50 |
const uint8_t *s2, ptrdiff_t stride) |
f46bb608 |
{
int i; |
3932ccc4 |
vec_u8 perm; |
7014b659 |
const vec_u8 zero = (const vec_u8)vec_splat_u8(0);
vec_s16 shorts1, shorts2; |
f46bb608 |
for (i = 0; i < 4; i++) {
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */ |
3932ccc4 |
perm = vec_lvsl(0, s1); |
7014b659 |
vec_u8 pixl = vec_ld(0, s1);
vec_u8 pixr = vec_ld(15, s1); |
3932ccc4 |
vec_u8 bytes = vec_perm(pixl, pixr, perm); |
f46bb608 |
// Convert the bytes into shorts. |
7014b659 |
shorts1 = (vec_s16)vec_mergeh(zero, bytes); |
f46bb608 |
// Do the same for the second block of pixels. |
3932ccc4 |
perm = vec_lvsl(0, s2); |
f46bb608 |
pixl = vec_ld(0, s2);
pixr = vec_ld(15, s2); |
3932ccc4 |
bytes = vec_perm(pixl, pixr, perm); |
f46bb608 |
// Convert the bytes into shorts. |
7014b659 |
shorts2 = (vec_s16)vec_mergeh(zero, bytes); |
f46bb608 |
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned. |
7014b659 |
vec_st(shorts1, 0, (vec_s16 *)block); |
f46bb608 |
s1 += stride;
s2 += stride;
block += 8;
/* The code below is a copy of the code above...
* This is a manual unroll. */
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */ |
3932ccc4 |
perm = vec_lvsl(0, s1); |
f46bb608 |
pixl = vec_ld(0, s1);
pixr = vec_ld(15, s1); |
3932ccc4 |
bytes = vec_perm(pixl, pixr, perm); |
f46bb608 |
// Convert the bytes into shorts. |
7014b659 |
shorts1 = (vec_s16)vec_mergeh(zero, bytes); |
f46bb608 |
// Do the same for the second block of pixels. |
3932ccc4 |
perm = vec_lvsl(0, s2); |
f46bb608 |
pixl = vec_ld(0, s2);
pixr = vec_ld(15, s2); |
3932ccc4 |
bytes = vec_perm(pixl, pixr, perm); |
f46bb608 |
// Convert the bytes into shorts. |
7014b659 |
shorts2 = (vec_s16)vec_mergeh(zero, bytes); |
f46bb608 |
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned. |
7014b659 |
vec_st(shorts1, 0, (vec_s16 *)block); |
f46bb608 |
s1 += stride;
s2 += stride;
block += 8;
}
}
|
31dea051 |
#endif /* HAVE_VSX */
|
f46bb608 |
#endif /* HAVE_ALTIVEC */
|
d0bf20a4 |
#if HAVE_VSX
static void get_pixels_vsx(int16_t *restrict block, const uint8_t *pixels, |
de452e50 |
ptrdiff_t stride) |
d0bf20a4 |
{
int i;
for (i = 0; i < 8; i++) {
vec_s16 shorts = vsx_ld_u8_s16(0, pixels);
vec_vsx_st(shorts, i * 16, block);
|
de452e50 |
pixels += stride; |
d0bf20a4 |
}
}
static void diff_pixels_vsx(int16_t *restrict block, const uint8_t *s1, |
de452e50 |
const uint8_t *s2, ptrdiff_t stride) |
d0bf20a4 |
{
int i;
vec_s16 shorts1, shorts2;
for (i = 0; i < 8; i++) {
shorts1 = vsx_ld_u8_s16(0, s1);
shorts2 = vsx_ld_u8_s16(0, s2);
shorts1 = vec_sub(shorts1, shorts2);
vec_vsx_st(shorts1, 0, block);
s1 += stride;
s2 += stride;
block += 8;
}
}
#endif /* HAVE_VSX */
|
f46bb608 |
av_cold void ff_pixblockdsp_init_ppc(PixblockDSPContext *c,
AVCodecContext *avctx,
unsigned high_bit_depth)
{
#if HAVE_ALTIVEC
if (!PPC_ALTIVEC(av_get_cpu_flags()))
return;
c->diff_pixels = diff_pixels_altivec;
if (!high_bit_depth) {
c->get_pixels = get_pixels_altivec;
}
#endif /* HAVE_ALTIVEC */ |
d0bf20a4 |
#if HAVE_VSX
if (!PPC_VSX(av_get_cpu_flags()))
return;
c->diff_pixels = diff_pixels_vsx;
if (!high_bit_depth)
c->get_pixels = get_pixels_vsx;
#endif /* HAVE_VSX */ |
f46bb608 |
} |