ffmpeg / libavcodec / x86 / vp8dsp-init.c @ f2a30bd8
History | View | Annotate | Download (14.2 KB)
| 1 |
/*
|
|---|---|
| 2 |
* VP8 DSP functions x86-optimized
|
| 3 |
* Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
|
| 4 |
* Copyright (c) 2010 Jason Garrett-Glaser <darkshikari@gmail.com>
|
| 5 |
*
|
| 6 |
* This file is part of FFmpeg.
|
| 7 |
*
|
| 8 |
* FFmpeg is free software; you can redistribute it and/or
|
| 9 |
* modify it under the terms of the GNU Lesser General Public
|
| 10 |
* License as published by the Free Software Foundation; either
|
| 11 |
* version 2.1 of the License, or (at your option) any later version.
|
| 12 |
*
|
| 13 |
* FFmpeg is distributed in the hope that it will be useful,
|
| 14 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 15 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 16 |
* Lesser General Public License for more details.
|
| 17 |
*
|
| 18 |
* You should have received a copy of the GNU Lesser General Public
|
| 19 |
* License along with FFmpeg; if not, write to the Free Software
|
| 20 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
| 21 |
*/
|
| 22 |
|
| 23 |
#include "libavutil/x86_cpu.h" |
| 24 |
#include "libavcodec/vp8dsp.h" |
| 25 |
|
| 26 |
#if HAVE_YASM
|
| 27 |
|
| 28 |
/*
|
| 29 |
* MC functions
|
| 30 |
*/
|
| 31 |
extern void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, int dststride, |
| 32 |
uint8_t *src, int srcstride,
|
| 33 |
int height, int mx, int my); |
| 34 |
extern void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, int dststride, |
| 35 |
uint8_t *src, int srcstride,
|
| 36 |
int height, int mx, int my); |
| 37 |
extern void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, int dststride, |
| 38 |
uint8_t *src, int srcstride,
|
| 39 |
int height, int mx, int my); |
| 40 |
extern void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, int dststride, |
| 41 |
uint8_t *src, int srcstride,
|
| 42 |
int height, int mx, int my); |
| 43 |
|
| 44 |
extern void ff_put_vp8_epel8_h4_sse2 (uint8_t *dst, int dststride, |
| 45 |
uint8_t *src, int srcstride,
|
| 46 |
int height, int mx, int my); |
| 47 |
extern void ff_put_vp8_epel8_h6_sse2 (uint8_t *dst, int dststride, |
| 48 |
uint8_t *src, int srcstride,
|
| 49 |
int height, int mx, int my); |
| 50 |
extern void ff_put_vp8_epel8_v4_sse2 (uint8_t *dst, int dststride, |
| 51 |
uint8_t *src, int srcstride,
|
| 52 |
int height, int mx, int my); |
| 53 |
extern void ff_put_vp8_epel8_v6_sse2 (uint8_t *dst, int dststride, |
| 54 |
uint8_t *src, int srcstride,
|
| 55 |
int height, int mx, int my); |
| 56 |
|
| 57 |
extern void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, int dststride, |
| 58 |
uint8_t *src, int srcstride,
|
| 59 |
int height, int mx, int my); |
| 60 |
extern void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, int dststride, |
| 61 |
uint8_t *src, int srcstride,
|
| 62 |
int height, int mx, int my); |
| 63 |
extern void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, int dststride, |
| 64 |
uint8_t *src, int srcstride,
|
| 65 |
int height, int mx, int my); |
| 66 |
extern void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, int dststride, |
| 67 |
uint8_t *src, int srcstride,
|
| 68 |
int height, int mx, int my); |
| 69 |
extern void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, int dststride, |
| 70 |
uint8_t *src, int srcstride,
|
| 71 |
int height, int mx, int my); |
| 72 |
extern void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, int dststride, |
| 73 |
uint8_t *src, int srcstride,
|
| 74 |
int height, int mx, int my); |
| 75 |
extern void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, int dststride, |
| 76 |
uint8_t *src, int srcstride,
|
| 77 |
int height, int mx, int my); |
| 78 |
extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, int dststride, |
| 79 |
uint8_t *src, int srcstride,
|
| 80 |
int height, int mx, int my); |
| 81 |
|
| 82 |
extern void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, int dststride, |
| 83 |
uint8_t *src, int srcstride,
|
| 84 |
int height, int mx, int my); |
| 85 |
extern void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, int dststride, |
| 86 |
uint8_t *src, int srcstride,
|
| 87 |
int height, int mx, int my); |
| 88 |
extern void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, int dststride, |
| 89 |
uint8_t *src, int srcstride,
|
| 90 |
int height, int mx, int my); |
| 91 |
extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, int dststride, |
| 92 |
uint8_t *src, int srcstride,
|
| 93 |
int height, int mx, int my); |
| 94 |
|
| 95 |
extern void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, int dststride, |
| 96 |
uint8_t *src, int srcstride,
|
| 97 |
int height, int mx, int my); |
| 98 |
extern void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, int dststride, |
| 99 |
uint8_t *src, int srcstride,
|
| 100 |
int height, int mx, int my); |
| 101 |
extern void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, int dststride, |
| 102 |
uint8_t *src, int srcstride,
|
| 103 |
int height, int mx, int my); |
| 104 |
extern void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, int dststride, |
| 105 |
uint8_t *src, int srcstride,
|
| 106 |
int height, int mx, int my); |
| 107 |
|
| 108 |
|
| 109 |
extern void ff_put_vp8_pixels8_mmx (uint8_t *dst, int dststride, |
| 110 |
uint8_t *src, int srcstride,
|
| 111 |
int height, int mx, int my); |
| 112 |
extern void ff_put_vp8_pixels16_mmx(uint8_t *dst, int dststride, |
| 113 |
uint8_t *src, int srcstride,
|
| 114 |
int height, int mx, int my); |
| 115 |
extern void ff_put_vp8_pixels16_sse(uint8_t *dst, int dststride, |
| 116 |
uint8_t *src, int srcstride,
|
| 117 |
int height, int mx, int my); |
| 118 |
|
| 119 |
#define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
|
| 120 |
static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \ |
| 121 |
uint8_t *dst, int dststride, uint8_t *src, \
|
| 122 |
int srcstride, int height, int mx, int my) \ |
| 123 |
{ \
|
| 124 |
ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 125 |
dst, dststride, src, srcstride, height, mx, my); \ |
| 126 |
ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 127 |
dst + 8, dststride, src + 8, srcstride, height, mx, my); \ |
| 128 |
} |
| 129 |
#define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \
|
| 130 |
static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 131 |
uint8_t *dst, int dststride, uint8_t *src, \
|
| 132 |
int srcstride, int height, int mx, int my) \ |
| 133 |
{ \
|
| 134 |
ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
| 135 |
dst, dststride, src, srcstride, height, mx, my); \ |
| 136 |
ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
| 137 |
dst + 4, dststride, src + 4, srcstride, height, mx, my); \ |
| 138 |
} |
| 139 |
|
| 140 |
TAP_W8 (mmxext, epel, h4) |
| 141 |
TAP_W8 (mmxext, epel, h6) |
| 142 |
TAP_W16(mmxext, epel, h6) |
| 143 |
TAP_W8 (mmxext, epel, v4) |
| 144 |
TAP_W8 (mmxext, epel, v6) |
| 145 |
TAP_W16(mmxext, epel, v6) |
| 146 |
TAP_W8 (mmxext, bilinear, h) |
| 147 |
TAP_W16(mmxext, bilinear, h) |
| 148 |
TAP_W8 (mmxext, bilinear, v) |
| 149 |
TAP_W16(mmxext, bilinear, v) |
| 150 |
|
| 151 |
TAP_W16(sse2, epel, h6) |
| 152 |
TAP_W16(sse2, epel, v6) |
| 153 |
TAP_W16(sse2, bilinear, h) |
| 154 |
TAP_W16(sse2, bilinear, v) |
| 155 |
|
| 156 |
TAP_W16(ssse3, epel, h6) |
| 157 |
TAP_W16(ssse3, epel, v6) |
| 158 |
TAP_W16(ssse3, bilinear, h) |
| 159 |
TAP_W16(ssse3, bilinear, v) |
| 160 |
|
| 161 |
#define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \
|
| 162 |
static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \ |
| 163 |
uint8_t *dst, int dststride, uint8_t *src, \
|
| 164 |
int srcstride, int height, int mx, int my) \ |
| 165 |
{ \
|
| 166 |
DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
|
| 167 |
uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \ |
| 168 |
src -= srcstride * (TAPNUMY / 2 - 1); \ |
| 169 |
ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \ |
| 170 |
tmp, SIZE, src, srcstride, height + TAPNUMY - 1, mx, my); \
|
| 171 |
ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \ |
| 172 |
dst, dststride, tmpptr, SIZE, height, mx, my); \ |
| 173 |
} |
| 174 |
|
| 175 |
#define HVTAPMMX(x, y) \
|
| 176 |
HVTAP(mmxext, 8, x, y, 4, 8) \ |
| 177 |
HVTAP(mmxext, 8, x, y, 8, 16) |
| 178 |
|
| 179 |
HVTAPMMX(4, 4) |
| 180 |
HVTAPMMX(4, 6) |
| 181 |
HVTAPMMX(6, 4) |
| 182 |
HVTAPMMX(6, 6) |
| 183 |
HVTAP(mmxext, 8, 6, 6, 16, 16) |
| 184 |
|
| 185 |
#define HVTAPSSE2(x, y, w) \
|
| 186 |
HVTAP(sse2, 16, x, y, w, 16) \ |
| 187 |
HVTAP(ssse3, 16, x, y, w, 16) |
| 188 |
|
| 189 |
HVTAPSSE2(4, 4, 8) |
| 190 |
HVTAPSSE2(4, 6, 8) |
| 191 |
HVTAPSSE2(6, 4, 8) |
| 192 |
HVTAPSSE2(6, 6, 8) |
| 193 |
HVTAPSSE2(6, 6, 16) |
| 194 |
|
| 195 |
HVTAP(ssse3, 16, 4, 4, 4, 8) |
| 196 |
HVTAP(ssse3, 16, 4, 6, 4, 8) |
| 197 |
HVTAP(ssse3, 16, 6, 4, 4, 8) |
| 198 |
HVTAP(ssse3, 16, 6, 6, 4, 8) |
| 199 |
|
| 200 |
#define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \
|
| 201 |
static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \ |
| 202 |
uint8_t *dst, int dststride, uint8_t *src, \
|
| 203 |
int srcstride, int height, int mx, int my) \ |
| 204 |
{ \
|
| 205 |
DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \
|
| 206 |
ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \ |
| 207 |
tmp, SIZE, src, srcstride, height + 1, mx, my); \
|
| 208 |
ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \ |
| 209 |
dst, dststride, tmp, SIZE, height, mx, my); \ |
| 210 |
} |
| 211 |
|
| 212 |
HVBILIN(mmxext, 8, 4, 8) |
| 213 |
HVBILIN(mmxext, 8, 8, 16) |
| 214 |
HVBILIN(mmxext, 8, 16, 16) |
| 215 |
HVBILIN(sse2, 8, 8, 16) |
| 216 |
HVBILIN(sse2, 8, 16, 16) |
| 217 |
HVBILIN(ssse3, 8, 4, 8) |
| 218 |
HVBILIN(ssse3, 8, 8, 16) |
| 219 |
HVBILIN(ssse3, 8, 16, 16) |
| 220 |
|
| 221 |
extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, DCTELEM block[16], int stride); |
| 222 |
extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, DCTELEM block[16], int stride); |
| 223 |
extern void ff_vp8_luma_dc_wht_mmxext(DCTELEM block[4][4][16], DCTELEM dc[16]); |
| 224 |
extern void ff_vp8_idct_add_mmx(uint8_t *dst, DCTELEM block[16], int stride); |
| 225 |
|
| 226 |
extern void ff_vp8_v_loop_filter_simple_mmx (uint8_t *dst, int stride, int flim); |
| 227 |
extern void ff_vp8_v_loop_filter_simple_mmxext(uint8_t *dst, int stride, int flim); |
| 228 |
extern void ff_vp8_v_loop_filter_simple_sse2 (uint8_t *dst, int stride, int flim); |
| 229 |
extern void ff_vp8_h_loop_filter_simple_mmx (uint8_t *dst, int stride, int flim); |
| 230 |
extern void ff_vp8_h_loop_filter_simple_mmxext(uint8_t *dst, int stride, int flim); |
| 231 |
extern void ff_vp8_h_loop_filter_simple_sse2 (uint8_t *dst, int stride, int flim); |
| 232 |
#endif
|
| 233 |
|
| 234 |
#define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \
|
| 235 |
c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \ |
| 236 |
c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \ |
| 237 |
c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT |
| 238 |
|
| 239 |
#define VP8_MC_FUNC(IDX, SIZE, OPT) \
|
| 240 |
c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \ |
| 241 |
c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \ |
| 242 |
c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \ |
| 243 |
c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \ |
| 244 |
c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \ |
| 245 |
VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) |
| 246 |
|
| 247 |
#define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \
|
| 248 |
c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
| 249 |
c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
| 250 |
c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
| 251 |
c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 252 |
c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 253 |
c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
| 254 |
c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 255 |
c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT |
| 256 |
|
| 257 |
|
| 258 |
av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
|
| 259 |
{
|
| 260 |
mm_flags = mm_support(); |
| 261 |
|
| 262 |
#if HAVE_YASM
|
| 263 |
if (mm_flags & FF_MM_MMX) {
|
| 264 |
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; |
| 265 |
c->vp8_idct_add = ff_vp8_idct_add_mmx; |
| 266 |
c->put_vp8_epel_pixels_tab[0][0][0] = |
| 267 |
c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx; |
| 268 |
c->put_vp8_epel_pixels_tab[1][0][0] = |
| 269 |
c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx; |
| 270 |
|
| 271 |
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx; |
| 272 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx; |
| 273 |
} |
| 274 |
|
| 275 |
/* note that 4-tap width=16 functions are missing because w=16
|
| 276 |
* is only used for luma, and luma is always a copy or sixtap. */
|
| 277 |
if (mm_flags & FF_MM_MMX2) {
|
| 278 |
c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmxext; |
| 279 |
VP8_LUMA_MC_FUNC(0, 16, mmxext); |
| 280 |
VP8_MC_FUNC(1, 8, mmxext); |
| 281 |
VP8_MC_FUNC(2, 4, mmxext); |
| 282 |
VP8_BILINEAR_MC_FUNC(0, 16, mmxext); |
| 283 |
VP8_BILINEAR_MC_FUNC(1, 8, mmxext); |
| 284 |
VP8_BILINEAR_MC_FUNC(2, 4, mmxext); |
| 285 |
|
| 286 |
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmxext; |
| 287 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmxext; |
| 288 |
} |
| 289 |
|
| 290 |
if (mm_flags & FF_MM_SSE) {
|
| 291 |
c->put_vp8_epel_pixels_tab[0][0][0] = |
| 292 |
c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; |
| 293 |
} |
| 294 |
|
| 295 |
if (mm_flags & FF_MM_SSE2) {
|
| 296 |
VP8_LUMA_MC_FUNC(0, 16, sse2); |
| 297 |
VP8_MC_FUNC(1, 8, sse2); |
| 298 |
VP8_BILINEAR_MC_FUNC(0, 16, sse2); |
| 299 |
VP8_BILINEAR_MC_FUNC(1, 8, sse2); |
| 300 |
|
| 301 |
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; |
| 302 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2; |
| 303 |
} |
| 304 |
|
| 305 |
if (mm_flags & FF_MM_SSSE3) {
|
| 306 |
VP8_LUMA_MC_FUNC(0, 16, ssse3); |
| 307 |
VP8_MC_FUNC(1, 8, ssse3); |
| 308 |
VP8_MC_FUNC(2, 4, ssse3); |
| 309 |
VP8_BILINEAR_MC_FUNC(0, 16, ssse3); |
| 310 |
VP8_BILINEAR_MC_FUNC(1, 8, ssse3); |
| 311 |
VP8_BILINEAR_MC_FUNC(2, 4, ssse3); |
| 312 |
} |
| 313 |
|
| 314 |
if (mm_flags & FF_MM_SSE4) {
|
| 315 |
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4; |
| 316 |
} |
| 317 |
#endif
|
| 318 |
} |