| 1 | /* |
| 2 | * VP8 DSP functions x86-optimized |
| 3 | * Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com> |
| 4 | * Copyright (c) 2010 Fiona Glaser <fiona@x264.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/attributes.h" |
| 24 | #include "libavutil/cpu.h" |
| 25 | #include "libavutil/mem.h" |
| 26 | #include "libavutil/x86/asm.h" |
| 27 | #include "libavutil/x86/cpu.h" |
| 28 | #include "libavcodec/vp8dsp.h" |
| 29 | |
| 30 | #if HAVE_YASM |
| 31 | |
| 32 | /* |
| 33 | * MC functions |
| 34 | */ |
| 35 | void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 36 | uint8_t *src, ptrdiff_t srcstride, |
| 37 | int height, int mx, int my); |
| 38 | void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 39 | uint8_t *src, ptrdiff_t srcstride, |
| 40 | int height, int mx, int my); |
| 41 | void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 42 | uint8_t *src, ptrdiff_t srcstride, |
| 43 | int height, int mx, int my); |
| 44 | void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 45 | uint8_t *src, ptrdiff_t srcstride, |
| 46 | int height, int mx, int my); |
| 47 | |
| 48 | void ff_put_vp8_epel8_h4_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 49 | uint8_t *src, ptrdiff_t srcstride, |
| 50 | int height, int mx, int my); |
| 51 | void ff_put_vp8_epel8_h6_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 52 | uint8_t *src, ptrdiff_t srcstride, |
| 53 | int height, int mx, int my); |
| 54 | void ff_put_vp8_epel8_v4_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 55 | uint8_t *src, ptrdiff_t srcstride, |
| 56 | int height, int mx, int my); |
| 57 | void ff_put_vp8_epel8_v6_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 58 | uint8_t *src, ptrdiff_t srcstride, |
| 59 | int height, int mx, int my); |
| 60 | |
| 61 | void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 62 | uint8_t *src, ptrdiff_t srcstride, |
| 63 | int height, int mx, int my); |
| 64 | void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 65 | uint8_t *src, ptrdiff_t srcstride, |
| 66 | int height, int mx, int my); |
| 67 | void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 68 | uint8_t *src, ptrdiff_t srcstride, |
| 69 | int height, int mx, int my); |
| 70 | void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 71 | uint8_t *src, ptrdiff_t srcstride, |
| 72 | int height, int mx, int my); |
| 73 | void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 74 | uint8_t *src, ptrdiff_t srcstride, |
| 75 | int height, int mx, int my); |
| 76 | void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 77 | uint8_t *src, ptrdiff_t srcstride, |
| 78 | int height, int mx, int my); |
| 79 | void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 80 | uint8_t *src, ptrdiff_t srcstride, |
| 81 | int height, int mx, int my); |
| 82 | void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 83 | uint8_t *src, ptrdiff_t srcstride, |
| 84 | int height, int mx, int my); |
| 85 | |
| 86 | void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 87 | uint8_t *src, ptrdiff_t srcstride, |
| 88 | int height, int mx, int my); |
| 89 | void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 90 | uint8_t *src, ptrdiff_t srcstride, |
| 91 | int height, int mx, int my); |
| 92 | void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 93 | uint8_t *src, ptrdiff_t srcstride, |
| 94 | int height, int mx, int my); |
| 95 | void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 96 | uint8_t *src, ptrdiff_t srcstride, |
| 97 | int height, int mx, int my); |
| 98 | |
| 99 | void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, ptrdiff_t dststride, |
| 100 | uint8_t *src, ptrdiff_t srcstride, |
| 101 | int height, int mx, int my); |
| 102 | void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, ptrdiff_t dststride, |
| 103 | uint8_t *src, ptrdiff_t srcstride, |
| 104 | int height, int mx, int my); |
| 105 | void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 106 | uint8_t *src, ptrdiff_t srcstride, |
| 107 | int height, int mx, int my); |
| 108 | void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
| 109 | uint8_t *src, ptrdiff_t srcstride, |
| 110 | int height, int mx, int my); |
| 111 | |
| 112 | |
| 113 | void ff_put_vp8_pixels8_mmx (uint8_t *dst, ptrdiff_t dststride, |
| 114 | uint8_t *src, ptrdiff_t srcstride, |
| 115 | int height, int mx, int my); |
| 116 | void ff_put_vp8_pixels16_mmx(uint8_t *dst, ptrdiff_t dststride, |
| 117 | uint8_t *src, ptrdiff_t srcstride, |
| 118 | int height, int mx, int my); |
| 119 | void ff_put_vp8_pixels16_sse(uint8_t *dst, ptrdiff_t dststride, |
| 120 | uint8_t *src, ptrdiff_t srcstride, |
| 121 | int height, int mx, int my); |
| 122 | |
| 123 | #define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \ |
| 124 | static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \ |
| 125 | uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
| 126 | ptrdiff_t srcstride, int height, int mx, int my) \ |
| 127 | { \ |
| 128 | ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 129 | dst, dststride, src, srcstride, height, mx, my); \ |
| 130 | ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 131 | dst + 8, dststride, src + 8, srcstride, height, mx, my); \ |
| 132 | } |
| 133 | #define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \ |
| 134 | static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
| 135 | uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
| 136 | ptrdiff_t srcstride, int height, int mx, int my) \ |
| 137 | { \ |
| 138 | ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
| 139 | dst, dststride, src, srcstride, height, mx, my); \ |
| 140 | ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
| 141 | dst + 4, dststride, src + 4, srcstride, height, mx, my); \ |
| 142 | } |
| 143 | |
| 144 | #if ARCH_X86_32 |
| 145 | TAP_W8 (mmxext, epel, h4) |
| 146 | TAP_W8 (mmxext, epel, h6) |
| 147 | TAP_W16(mmxext, epel, h6) |
| 148 | TAP_W8 (mmxext, epel, v4) |
| 149 | TAP_W8 (mmxext, epel, v6) |
| 150 | TAP_W16(mmxext, epel, v6) |
| 151 | TAP_W8 (mmxext, bilinear, h) |
| 152 | TAP_W16(mmxext, bilinear, h) |
| 153 | TAP_W8 (mmxext, bilinear, v) |
| 154 | TAP_W16(mmxext, bilinear, v) |
| 155 | #endif |
| 156 | |
| 157 | TAP_W16(sse2, epel, h6) |
| 158 | TAP_W16(sse2, epel, v6) |
| 159 | TAP_W16(sse2, bilinear, h) |
| 160 | TAP_W16(sse2, bilinear, v) |
| 161 | |
| 162 | TAP_W16(ssse3, epel, h6) |
| 163 | TAP_W16(ssse3, epel, v6) |
| 164 | TAP_W16(ssse3, bilinear, h) |
| 165 | TAP_W16(ssse3, bilinear, v) |
| 166 | |
| 167 | #define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \ |
| 168 | static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \ |
| 169 | uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
| 170 | ptrdiff_t srcstride, int height, int mx, int my) \ |
| 171 | { \ |
| 172 | DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \ |
| 173 | uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \ |
| 174 | src -= srcstride * (TAPNUMY / 2 - 1); \ |
| 175 | ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \ |
| 176 | tmp, SIZE, src, srcstride, height + TAPNUMY - 1, mx, my); \ |
| 177 | ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \ |
| 178 | dst, dststride, tmpptr, SIZE, height, mx, my); \ |
| 179 | } |
| 180 | |
| 181 | #if ARCH_X86_32 |
| 182 | #define HVTAPMMX(x, y) \ |
| 183 | HVTAP(mmxext, 8, x, y, 4, 8) \ |
| 184 | HVTAP(mmxext, 8, x, y, 8, 16) |
| 185 | |
| 186 | HVTAP(mmxext, 8, 6, 6, 16, 16) |
| 187 | #else |
| 188 | #define HVTAPMMX(x, y) \ |
| 189 | HVTAP(mmxext, 8, x, y, 4, 8) |
| 190 | #endif |
| 191 | |
| 192 | HVTAPMMX(4, 4) |
| 193 | HVTAPMMX(4, 6) |
| 194 | HVTAPMMX(6, 4) |
| 195 | HVTAPMMX(6, 6) |
| 196 | |
| 197 | #define HVTAPSSE2(x, y, w) \ |
| 198 | HVTAP(sse2, 16, x, y, w, 16) \ |
| 199 | HVTAP(ssse3, 16, x, y, w, 16) |
| 200 | |
| 201 | HVTAPSSE2(4, 4, 8) |
| 202 | HVTAPSSE2(4, 6, 8) |
| 203 | HVTAPSSE2(6, 4, 8) |
| 204 | HVTAPSSE2(6, 6, 8) |
| 205 | HVTAPSSE2(6, 6, 16) |
| 206 | |
| 207 | HVTAP(ssse3, 16, 4, 4, 4, 8) |
| 208 | HVTAP(ssse3, 16, 4, 6, 4, 8) |
| 209 | HVTAP(ssse3, 16, 6, 4, 4, 8) |
| 210 | HVTAP(ssse3, 16, 6, 6, 4, 8) |
| 211 | |
| 212 | #define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \ |
| 213 | static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \ |
| 214 | uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
| 215 | ptrdiff_t srcstride, int height, int mx, int my) \ |
| 216 | { \ |
| 217 | DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \ |
| 218 | ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \ |
| 219 | tmp, SIZE, src, srcstride, height + 1, mx, my); \ |
| 220 | ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \ |
| 221 | dst, dststride, tmp, SIZE, height, mx, my); \ |
| 222 | } |
| 223 | |
| 224 | HVBILIN(mmxext, 8, 4, 8) |
| 225 | #if ARCH_X86_32 |
| 226 | HVBILIN(mmxext, 8, 8, 16) |
| 227 | HVBILIN(mmxext, 8, 16, 16) |
| 228 | #endif |
| 229 | HVBILIN(sse2, 8, 8, 16) |
| 230 | HVBILIN(sse2, 8, 16, 16) |
| 231 | HVBILIN(ssse3, 8, 4, 8) |
| 232 | HVBILIN(ssse3, 8, 8, 16) |
| 233 | HVBILIN(ssse3, 8, 16, 16) |
| 234 | |
| 235 | void ff_vp8_idct_dc_add_mmx(uint8_t *dst, int16_t block[16], |
| 236 | ptrdiff_t stride); |
| 237 | void ff_vp8_idct_dc_add_sse4(uint8_t *dst, int16_t block[16], |
| 238 | ptrdiff_t stride); |
| 239 | void ff_vp8_idct_dc_add4y_mmx(uint8_t *dst, int16_t block[4][16], |
| 240 | ptrdiff_t stride); |
| 241 | void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, int16_t block[4][16], |
| 242 | ptrdiff_t stride); |
| 243 | void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, int16_t block[2][16], |
| 244 | ptrdiff_t stride); |
| 245 | void ff_vp8_luma_dc_wht_mmx(int16_t block[4][4][16], int16_t dc[16]); |
| 246 | void ff_vp8_luma_dc_wht_sse(int16_t block[4][4][16], int16_t dc[16]); |
| 247 | void ff_vp8_idct_add_mmx(uint8_t *dst, int16_t block[16], ptrdiff_t stride); |
| 248 | void ff_vp8_idct_add_sse(uint8_t *dst, int16_t block[16], ptrdiff_t stride); |
| 249 | |
| 250 | #define DECLARE_LOOP_FILTER(NAME) \ |
| 251 | void ff_vp8_v_loop_filter_simple_ ## NAME(uint8_t *dst, \ |
| 252 | ptrdiff_t stride, \ |
| 253 | int flim); \ |
| 254 | void ff_vp8_h_loop_filter_simple_ ## NAME(uint8_t *dst, \ |
| 255 | ptrdiff_t stride, \ |
| 256 | int flim); \ |
| 257 | void ff_vp8_v_loop_filter16y_inner_ ## NAME (uint8_t *dst, \ |
| 258 | ptrdiff_t stride, \ |
| 259 | int e, int i, int hvt); \ |
| 260 | void ff_vp8_h_loop_filter16y_inner_ ## NAME (uint8_t *dst, \ |
| 261 | ptrdiff_t stride, \ |
| 262 | int e, int i, int hvt); \ |
| 263 | void ff_vp8_v_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \ |
| 264 | uint8_t *dstV, \ |
| 265 | ptrdiff_t s, \ |
| 266 | int e, int i, int hvt); \ |
| 267 | void ff_vp8_h_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \ |
| 268 | uint8_t *dstV, \ |
| 269 | ptrdiff_t s, \ |
| 270 | int e, int i, int hvt); \ |
| 271 | void ff_vp8_v_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \ |
| 272 | ptrdiff_t stride, \ |
| 273 | int e, int i, int hvt); \ |
| 274 | void ff_vp8_h_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \ |
| 275 | ptrdiff_t stride, \ |
| 276 | int e, int i, int hvt); \ |
| 277 | void ff_vp8_v_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \ |
| 278 | uint8_t *dstV, \ |
| 279 | ptrdiff_t s, \ |
| 280 | int e, int i, int hvt); \ |
| 281 | void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \ |
| 282 | uint8_t *dstV, \ |
| 283 | ptrdiff_t s, \ |
| 284 | int e, int i, int hvt); |
| 285 | |
| 286 | DECLARE_LOOP_FILTER(mmx) |
| 287 | DECLARE_LOOP_FILTER(mmxext) |
| 288 | DECLARE_LOOP_FILTER(sse2) |
| 289 | DECLARE_LOOP_FILTER(ssse3) |
| 290 | DECLARE_LOOP_FILTER(sse4) |
| 291 | |
| 292 | #endif /* HAVE_YASM */ |
| 293 | |
| 294 | #define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \ |
| 295 | c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \ |
| 296 | c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \ |
| 297 | c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT |
| 298 | |
| 299 | #define VP8_MC_FUNC(IDX, SIZE, OPT) \ |
| 300 | c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \ |
| 301 | c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \ |
| 302 | c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \ |
| 303 | c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \ |
| 304 | c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \ |
| 305 | VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) |
| 306 | |
| 307 | #define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \ |
| 308 | c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
| 309 | c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
| 310 | c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
| 311 | c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 312 | c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 313 | c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
| 314 | c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
| 315 | c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT |
| 316 | |
| 317 | |
| 318 | av_cold void ff_vp78dsp_init_x86(VP8DSPContext *c) |
| 319 | { |
| 320 | #if HAVE_YASM |
| 321 | int cpu_flags = av_get_cpu_flags(); |
| 322 | |
| 323 | if (EXTERNAL_MMX(cpu_flags)) { |
| 324 | #if ARCH_X86_32 |
| 325 | c->put_vp8_epel_pixels_tab[0][0][0] = |
| 326 | c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx; |
| 327 | #endif |
| 328 | c->put_vp8_epel_pixels_tab[1][0][0] = |
| 329 | c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx; |
| 330 | } |
| 331 | |
| 332 | /* note that 4-tap width=16 functions are missing because w=16 |
| 333 | * is only used for luma, and luma is always a copy or sixtap. */ |
| 334 | if (EXTERNAL_MMXEXT(cpu_flags)) { |
| 335 | VP8_MC_FUNC(2, 4, mmxext); |
| 336 | VP8_BILINEAR_MC_FUNC(2, 4, mmxext); |
| 337 | #if ARCH_X86_32 |
| 338 | VP8_LUMA_MC_FUNC(0, 16, mmxext); |
| 339 | VP8_MC_FUNC(1, 8, mmxext); |
| 340 | VP8_BILINEAR_MC_FUNC(0, 16, mmxext); |
| 341 | VP8_BILINEAR_MC_FUNC(1, 8, mmxext); |
| 342 | #endif |
| 343 | } |
| 344 | |
| 345 | if (EXTERNAL_SSE(cpu_flags)) { |
| 346 | c->put_vp8_epel_pixels_tab[0][0][0] = |
| 347 | c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; |
| 348 | } |
| 349 | |
| 350 | if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { |
| 351 | VP8_LUMA_MC_FUNC(0, 16, sse2); |
| 352 | VP8_MC_FUNC(1, 8, sse2); |
| 353 | VP8_BILINEAR_MC_FUNC(0, 16, sse2); |
| 354 | VP8_BILINEAR_MC_FUNC(1, 8, sse2); |
| 355 | } |
| 356 | |
| 357 | if (EXTERNAL_SSSE3(cpu_flags)) { |
| 358 | VP8_LUMA_MC_FUNC(0, 16, ssse3); |
| 359 | VP8_MC_FUNC(1, 8, ssse3); |
| 360 | VP8_MC_FUNC(2, 4, ssse3); |
| 361 | VP8_BILINEAR_MC_FUNC(0, 16, ssse3); |
| 362 | VP8_BILINEAR_MC_FUNC(1, 8, ssse3); |
| 363 | VP8_BILINEAR_MC_FUNC(2, 4, ssse3); |
| 364 | } |
| 365 | #endif /* HAVE_YASM */ |
| 366 | } |
| 367 | |
| 368 | av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c) |
| 369 | { |
| 370 | #if HAVE_YASM |
| 371 | int cpu_flags = av_get_cpu_flags(); |
| 372 | |
| 373 | if (EXTERNAL_MMX(cpu_flags)) { |
| 374 | c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; |
| 375 | c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx; |
| 376 | #if ARCH_X86_32 |
| 377 | c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx; |
| 378 | c->vp8_idct_add = ff_vp8_idct_add_mmx; |
| 379 | c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx; |
| 380 | |
| 381 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx; |
| 382 | c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx; |
| 383 | |
| 384 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx; |
| 385 | c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx; |
| 386 | c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx; |
| 387 | c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx; |
| 388 | |
| 389 | c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmx; |
| 390 | c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmx; |
| 391 | c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmx; |
| 392 | c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmx; |
| 393 | #endif |
| 394 | } |
| 395 | |
| 396 | /* note that 4-tap width=16 functions are missing because w=16 |
| 397 | * is only used for luma, and luma is always a copy or sixtap. */ |
| 398 | if (EXTERNAL_MMXEXT(cpu_flags)) { |
| 399 | #if ARCH_X86_32 |
| 400 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmxext; |
| 401 | c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmxext; |
| 402 | |
| 403 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmxext; |
| 404 | c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmxext; |
| 405 | c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmxext; |
| 406 | c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmxext; |
| 407 | |
| 408 | c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmxext; |
| 409 | c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmxext; |
| 410 | c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmxext; |
| 411 | c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmxext; |
| 412 | #endif |
| 413 | } |
| 414 | |
| 415 | if (EXTERNAL_SSE(cpu_flags)) { |
| 416 | c->vp8_idct_add = ff_vp8_idct_add_sse; |
| 417 | c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; |
| 418 | } |
| 419 | |
| 420 | if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { |
| 421 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; |
| 422 | |
| 423 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; |
| 424 | c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_sse2; |
| 425 | |
| 426 | c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_sse2; |
| 427 | c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_sse2; |
| 428 | } |
| 429 | |
| 430 | if (EXTERNAL_SSE2(cpu_flags)) { |
| 431 | c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_sse2; |
| 432 | |
| 433 | c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2; |
| 434 | |
| 435 | c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2; |
| 436 | c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2; |
| 437 | |
| 438 | c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse2; |
| 439 | c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse2; |
| 440 | } |
| 441 | |
| 442 | if (EXTERNAL_SSSE3(cpu_flags)) { |
| 443 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3; |
| 444 | c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3; |
| 445 | |
| 446 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_ssse3; |
| 447 | c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_ssse3; |
| 448 | c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_ssse3; |
| 449 | c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_ssse3; |
| 450 | |
| 451 | c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_ssse3; |
| 452 | c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_ssse3; |
| 453 | c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_ssse3; |
| 454 | c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_ssse3; |
| 455 | } |
| 456 | |
| 457 | if (EXTERNAL_SSE4(cpu_flags)) { |
| 458 | c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4; |
| 459 | |
| 460 | c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse4; |
| 461 | c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse4; |
| 462 | c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse4; |
| 463 | } |
| 464 | #endif /* HAVE_YASM */ |
| 465 | } |