mirror of https://github.com/FFmpeg/FFmpeg.git
* commit 'fab9df63a3156ffe1f9490aafaea41e03ef60ddf': dsputil: Split off global motion compensation bits into a separate context Conflicts: libavcodec/dsputil.c libavcodec/dsputil.h libavcodec/ppc/dsputil_altivec.h libavcodec/x86/dsputil_init.c libavcodec/x86/dsputil_mmx.c libavcodec/x86/dsputil_x86.h Merged-by: Michael Niedermayer <michaelni@gmx.at>pull/76/merge
Michael Niedermayer
11 years ago
commit
d7463c6813
17 changed files with 382 additions and 283 deletions
@ -0,0 +1,119 @@ |
||||
/*
|
||||
* This file is part of FFmpeg. |
||||
* |
||||
* FFmpeg is free software; you can redistribute it and/or |
||||
* 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. |
||||
* |
||||
* FFmpeg 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 |
||||
* Lesser General Public License for more details. |
||||
* |
||||
* You should have received a copy of the GNU Lesser General Public |
||||
* License along with FFmpeg; if not, write to the Free Software |
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
||||
*/ |
||||
|
||||
#include "config.h" |
||||
#include "libavutil/attributes.h" |
||||
#include "libavutil/common.h" |
||||
#include "mpegvideodsp.h" |
||||
|
||||
static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, |
||||
int x16, int y16, int rounder) |
||||
{ |
||||
const int A = (16 - x16) * (16 - y16); |
||||
const int B = (x16) * (16 - y16); |
||||
const int C = (16 - x16) * (y16); |
||||
const int D = (x16) * (y16); |
||||
int i; |
||||
|
||||
for (i = 0; i < h; i++) { |
||||
dst[0] = (A * src[0] + B * src[1] + C * src[stride + 0] + D * src[stride + 1] + rounder) >> 8; |
||||
dst[1] = (A * src[1] + B * src[2] + C * src[stride + 1] + D * src[stride + 2] + rounder) >> 8; |
||||
dst[2] = (A * src[2] + B * src[3] + C * src[stride + 2] + D * src[stride + 3] + rounder) >> 8; |
||||
dst[3] = (A * src[3] + B * src[4] + C * src[stride + 3] + D * src[stride + 4] + rounder) >> 8; |
||||
dst[4] = (A * src[4] + B * src[5] + C * src[stride + 4] + D * src[stride + 5] + rounder) >> 8; |
||||
dst[5] = (A * src[5] + B * src[6] + C * src[stride + 5] + D * src[stride + 6] + rounder) >> 8; |
||||
dst[6] = (A * src[6] + B * src[7] + C * src[stride + 6] + D * src[stride + 7] + rounder) >> 8; |
||||
dst[7] = (A * src[7] + B * src[8] + C * src[stride + 7] + D * src[stride + 8] + rounder) >> 8; |
||||
dst += stride; |
||||
src += stride; |
||||
} |
||||
} |
||||
|
||||
void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, |
||||
int dxx, int dxy, int dyx, int dyy, int shift, int r, |
||||
int width, int height) |
||||
{ |
||||
int y, vx, vy; |
||||
const int s = 1 << shift; |
||||
|
||||
width--; |
||||
height--; |
||||
|
||||
for (y = 0; y < h; y++) { |
||||
int x; |
||||
|
||||
vx = ox; |
||||
vy = oy; |
||||
for (x = 0; x < 8; x++) { // FIXME: optimize
|
||||
int index; |
||||
int src_x = vx >> 16; |
||||
int src_y = vy >> 16; |
||||
int frac_x = src_x & (s - 1); |
||||
int frac_y = src_y & (s - 1); |
||||
|
||||
src_x >>= shift; |
||||
src_y >>= shift; |
||||
|
||||
if ((unsigned) src_x < width) { |
||||
if ((unsigned) src_y < height) { |
||||
index = src_x + src_y * stride; |
||||
dst[y * stride + x] = |
||||
((src[index] * (s - frac_x) + |
||||
src[index + 1] * frac_x) * (s - frac_y) + |
||||
(src[index + stride] * (s - frac_x) + |
||||
src[index + stride + 1] * frac_x) * frac_y + |
||||
r) >> (shift * 2); |
||||
} else { |
||||
index = src_x + av_clip(src_y, 0, height) * stride; |
||||
dst[y * stride + x] = |
||||
((src[index] * (s - frac_x) + |
||||
src[index + 1] * frac_x) * s + |
||||
r) >> (shift * 2); |
||||
} |
||||
} else { |
||||
if ((unsigned) src_y < height) { |
||||
index = av_clip(src_x, 0, width) + src_y * stride; |
||||
dst[y * stride + x] = |
||||
((src[index] * (s - frac_y) + |
||||
src[index + stride] * frac_y) * s + |
||||
r) >> (shift * 2); |
||||
} else { |
||||
index = av_clip(src_x, 0, width) + |
||||
av_clip(src_y, 0, height) * stride; |
||||
dst[y * stride + x] = src[index]; |
||||
} |
||||
} |
||||
|
||||
vx += dxx; |
||||
vy += dyx; |
||||
} |
||||
ox += dxy; |
||||
oy += dyy; |
||||
} |
||||
} |
||||
|
||||
av_cold void ff_mpegvideodsp_init(MpegVideoDSPContext *c) |
||||
{ |
||||
c->gmc1 = gmc1_c; |
||||
c->gmc = ff_gmc_c; |
||||
|
||||
if (ARCH_PPC) |
||||
ff_mpegvideodsp_init_ppc(c); |
||||
if (ARCH_X86) |
||||
ff_mpegvideodsp_init_x86(c); |
||||
} |
||||
@ -0,0 +1,47 @@ |
||||
/*
|
||||
* This file is part of FFmpeg. |
||||
* |
||||
* FFmpeg is free software; you can redistribute it and/or |
||||
* 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. |
||||
* |
||||
* FFmpeg 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 |
||||
* Lesser General Public License for more details. |
||||
* |
||||
* You should have received a copy of the GNU Lesser General Public |
||||
* License along with FFmpeg; if not, write to the Free Software |
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
||||
*/ |
||||
|
||||
#ifndef AVCODEC_MPEGVIDEODSP_H |
||||
#define AVCODEC_MPEGVIDEODSP_H |
||||
|
||||
#include <stdint.h> |
||||
|
||||
void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, |
||||
int dxx, int dxy, int dyx, int dyy, int shift, int r, |
||||
int width, int height); |
||||
|
||||
typedef struct MpegVideoDSPContext { |
||||
/**
|
||||
* translational global motion compensation. |
||||
*/ |
||||
void (*gmc1)(uint8_t *dst /* align 8 */, uint8_t *src /* align 1 */, |
||||
int srcStride, int h, int x16, int y16, int rounder); |
||||
/**
|
||||
* global motion compensation. |
||||
*/ |
||||
void (*gmc)(uint8_t *dst /* align 8 */, uint8_t *src /* align 1 */, |
||||
int stride, int h, int ox, int oy, |
||||
int dxx, int dxy, int dyx, int dyy, |
||||
int shift, int r, int width, int height); |
||||
} MpegVideoDSPContext; |
||||
|
||||
void ff_mpegvideodsp_init(MpegVideoDSPContext *c); |
||||
void ff_mpegvideodsp_init_ppc(MpegVideoDSPContext *c); |
||||
void ff_mpegvideodsp_init_x86(MpegVideoDSPContext *c); |
||||
|
||||
#endif /* AVCODEC_MPEGVIDEODSP_H */ |
||||
@ -0,0 +1,161 @@ |
||||
/*
|
||||
* This file is part of FFmpeg. |
||||
* |
||||
* FFmpeg is free software; you can redistribute it and/or |
||||
* 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. |
||||
* |
||||
* FFmpeg 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 |
||||
* Lesser General Public License for more details. |
||||
* |
||||
* You should have received a copy of the GNU Lesser General Public |
||||
* License along with FFmpeg; if not, write to the Free Software |
||||
* 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/x86/asm.h" |
||||
#include "libavutil/x86/cpu.h" |
||||
#include "libavcodec/mpegvideodsp.h" |
||||
#include "libavcodec/videodsp.h" |
||||
|
||||
#if HAVE_INLINE_ASM |
||||
|
||||
static void gmc_mmx(uint8_t *dst, uint8_t *src, |
||||
int stride, int h, int ox, int oy, |
||||
int dxx, int dxy, int dyx, int dyy, |
||||
int shift, int r, int width, int height) |
||||
{ |
||||
const int w = 8; |
||||
const int ix = ox >> (16 + shift); |
||||
const int iy = oy >> (16 + shift); |
||||
const int oxs = ox >> 4; |
||||
const int oys = oy >> 4; |
||||
const int dxxs = dxx >> 4; |
||||
const int dxys = dxy >> 4; |
||||
const int dyxs = dyx >> 4; |
||||
const int dyys = dyy >> 4; |
||||
const uint16_t r4[4] = { r, r, r, r }; |
||||
const uint16_t dxy4[4] = { dxys, dxys, dxys, dxys }; |
||||
const uint16_t dyy4[4] = { dyys, dyys, dyys, dyys }; |
||||
const uint64_t shift2 = 2 * shift; |
||||
#define MAX_STRIDE 4096U |
||||
#define MAX_H 8U |
||||
uint8_t edge_buf[(MAX_H + 1) * MAX_STRIDE]; |
||||
int x, y; |
||||
|
||||
const int dxw = (dxx - (1 << (16 + shift))) * (w - 1); |
||||
const int dyh = (dyy - (1 << (16 + shift))) * (h - 1); |
||||
const int dxh = dxy * (h - 1); |
||||
const int dyw = dyx * (w - 1); |
||||
int need_emu = (unsigned) ix >= width - w || |
||||
(unsigned) iy >= height - h; |
||||
|
||||
if ( // non-constant fullpel offset (3% of blocks)
|
||||
((ox ^ (ox + dxw)) | (ox ^ (ox + dxh)) | (ox ^ (ox + dxw + dxh)) | |
||||
(oy ^ (oy + dyw)) | (oy ^ (oy + dyh)) | (oy ^ (oy + dyw + dyh))) >> (16 + shift) || |
||||
// uses more than 16 bits of subpel mv (only at huge resolution)
|
||||
(dxx | dxy | dyx | dyy) & 15 || |
||||
(need_emu && (h > MAX_H || stride > MAX_STRIDE))) { |
||||
// FIXME could still use mmx for some of the rows
|
||||
ff_gmc_c(dst, src, stride, h, ox, oy, dxx, dxy, dyx, dyy, |
||||
shift, r, width, height); |
||||
return; |
||||
} |
||||
|
||||
src += ix + iy * stride; |
||||
if (need_emu) { |
||||
ff_emulated_edge_mc_8(edge_buf, src, stride, stride, w + 1, h + 1, ix, iy, width, height); |
||||
src = edge_buf; |
||||
} |
||||
|
||||
__asm__ volatile ( |
||||
"movd %0, %%mm6 \n\t" |
||||
"pxor %%mm7, %%mm7 \n\t" |
||||
"punpcklwd %%mm6, %%mm6 \n\t" |
||||
"punpcklwd %%mm6, %%mm6 \n\t" |
||||
:: "r" (1 << shift)); |
||||
|
||||
for (x = 0; x < w; x += 4) { |
||||
uint16_t dx4[4] = { oxs - dxys + dxxs * (x + 0), |
||||
oxs - dxys + dxxs * (x + 1), |
||||
oxs - dxys + dxxs * (x + 2), |
||||
oxs - dxys + dxxs * (x + 3) }; |
||||
uint16_t dy4[4] = { oys - dyys + dyxs * (x + 0), |
||||
oys - dyys + dyxs * (x + 1), |
||||
oys - dyys + dyxs * (x + 2), |
||||
oys - dyys + dyxs * (x + 3) }; |
||||
|
||||
for (y = 0; y < h; y++) { |
||||
__asm__ volatile ( |
||||
"movq %0, %%mm4 \n\t" |
||||
"movq %1, %%mm5 \n\t" |
||||
"paddw %2, %%mm4 \n\t" |
||||
"paddw %3, %%mm5 \n\t" |
||||
"movq %%mm4, %0 \n\t" |
||||
"movq %%mm5, %1 \n\t" |
||||
"psrlw $12, %%mm4 \n\t" |
||||
"psrlw $12, %%mm5 \n\t" |
||||
: "+m" (*dx4), "+m" (*dy4) |
||||
: "m" (*dxy4), "m" (*dyy4)); |
||||
|
||||
__asm__ volatile ( |
||||
"movq %%mm6, %%mm2 \n\t" |
||||
"movq %%mm6, %%mm1 \n\t" |
||||
"psubw %%mm4, %%mm2 \n\t" |
||||
"psubw %%mm5, %%mm1 \n\t" |
||||
"movq %%mm2, %%mm0 \n\t" |
||||
"movq %%mm4, %%mm3 \n\t" |
||||
"pmullw %%mm1, %%mm0 \n\t" // (s - dx) * (s - dy)
|
||||
"pmullw %%mm5, %%mm3 \n\t" // dx * dy
|
||||
"pmullw %%mm5, %%mm2 \n\t" // (s - dx) * dy
|
||||
"pmullw %%mm4, %%mm1 \n\t" // dx * (s - dy)
|
||||
|
||||
"movd %4, %%mm5 \n\t" |
||||
"movd %3, %%mm4 \n\t" |
||||
"punpcklbw %%mm7, %%mm5 \n\t" |
||||
"punpcklbw %%mm7, %%mm4 \n\t" |
||||
"pmullw %%mm5, %%mm3 \n\t" // src[1, 1] * dx * dy
|
||||
"pmullw %%mm4, %%mm2 \n\t" // src[0, 1] * (s - dx) * dy
|
||||
|
||||
"movd %2, %%mm5 \n\t" |
||||
"movd %1, %%mm4 \n\t" |
||||
"punpcklbw %%mm7, %%mm5 \n\t" |
||||
"punpcklbw %%mm7, %%mm4 \n\t" |
||||
"pmullw %%mm5, %%mm1 \n\t" // src[1, 0] * dx * (s - dy)
|
||||
"pmullw %%mm4, %%mm0 \n\t" // src[0, 0] * (s - dx) * (s - dy)
|
||||
"paddw %5, %%mm1 \n\t" |
||||
"paddw %%mm3, %%mm2 \n\t" |
||||
"paddw %%mm1, %%mm0 \n\t" |
||||
"paddw %%mm2, %%mm0 \n\t" |
||||
|
||||
"psrlw %6, %%mm0 \n\t" |
||||
"packuswb %%mm0, %%mm0 \n\t" |
||||
"movd %%mm0, %0 \n\t" |
||||
|
||||
: "=m" (dst[x + y * stride]) |
||||
: "m" (src[0]), "m" (src[1]), |
||||
"m" (src[stride]), "m" (src[stride + 1]), |
||||
"m" (*r4), "m" (shift2)); |
||||
src += stride; |
||||
} |
||||
src += 4 - h * stride; |
||||
} |
||||
} |
||||
|
||||
#endif /* HAVE_INLINE_ASM */ |
||||
|
||||
av_cold void ff_mpegvideodsp_init_x86(MpegVideoDSPContext *c) |
||||
{ |
||||
#if HAVE_INLINE_ASM |
||||
int cpu_flags = av_get_cpu_flags(); |
||||
|
||||
if (INLINE_MMX(cpu_flags)) |
||||
c->gmc = gmc_mmx; |
||||
#endif /* HAVE_INLINE_ASM */ |
||||
} |
||||
Loading…
Reference in new issue