| 1 | ; /* |
| 2 | ; * Provide SSE luma and chroma mc functions for HEVC decoding |
| 3 | ; * Copyright (c) 2013 Pierre-Edouard LEPERE |
| 4 | ; * |
| 5 | ; * This file is part of FFmpeg. |
| 6 | ; * |
| 7 | ; * FFmpeg is free software; you can redistribute it and/or |
| 8 | ; * modify it under the terms of the GNU Lesser General Public |
| 9 | ; * License as published by the Free Software Foundation; either |
| 10 | ; * version 2.1 of the License, or (at your option) any later version. |
| 11 | ; * |
| 12 | ; * FFmpeg is distributed in the hope that it will be useful, |
| 13 | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | ; * Lesser General Public License for more details. |
| 16 | ; * |
| 17 | ; * You should have received a copy of the GNU Lesser General Public |
| 18 | ; * License along with FFmpeg; if not, write to the Free Software |
| 19 | ; * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | ; */ |
| 21 | %include "libavutil/x86/x86util.asm" |
| 22 | |
| 23 | SECTION_RODATA |
| 24 | pw_8: times 8 dw (1 << 9) |
| 25 | pw_10: times 8 dw (1 << 11) |
| 26 | pw_12: times 8 dw (1 << 13) |
| 27 | pw_bi_8: times 8 dw (1 << 8) |
| 28 | pw_bi_10: times 8 dw (1 << 10) |
| 29 | pw_bi_12: times 8 dw (1 << 12) |
| 30 | max_pixels_10: times 8 dw ((1 << 10)-1) |
| 31 | max_pixels_12: times 8 dw ((1 << 12)-1) |
| 32 | zero: times 4 dd 0 |
| 33 | one_per_32: times 4 dd 1 |
| 34 | |
| 35 | SECTION .text |
| 36 | %macro EPEL_TABLE 4 |
| 37 | hevc_epel_filters_%4_%1 times %2 d%3 -2, 58 |
| 38 | times %2 d%3 10, -2 |
| 39 | times %2 d%3 -4, 54 |
| 40 | times %2 d%3 16, -2 |
| 41 | times %2 d%3 -6, 46 |
| 42 | times %2 d%3 28, -4 |
| 43 | times %2 d%3 -4, 36 |
| 44 | times %2 d%3 36, -4 |
| 45 | times %2 d%3 -4, 28 |
| 46 | times %2 d%3 46, -6 |
| 47 | times %2 d%3 -2, 16 |
| 48 | times %2 d%3 54, -4 |
| 49 | times %2 d%3 -2, 10 |
| 50 | times %2 d%3 58, -2 |
| 51 | %endmacro |
| 52 | |
| 53 | |
| 54 | |
| 55 | EPEL_TABLE 8, 8, b, sse4 |
| 56 | EPEL_TABLE 10, 4, w, sse4 |
| 57 | EPEL_TABLE 12, 4, w, sse4 |
| 58 | |
| 59 | %macro QPEL_TABLE 4 |
| 60 | hevc_qpel_filters_%4_%1 times %2 d%3 -1, 4 |
| 61 | times %2 d%3 -10, 58 |
| 62 | times %2 d%3 17, -5 |
| 63 | times %2 d%3 1, 0 |
| 64 | times %2 d%3 -1, 4 |
| 65 | times %2 d%3 -11, 40 |
| 66 | times %2 d%3 40,-11 |
| 67 | times %2 d%3 4, -1 |
| 68 | times %2 d%3 0, 1 |
| 69 | times %2 d%3 -5, 17 |
| 70 | times %2 d%3 58,-10 |
| 71 | times %2 d%3 4, -1 |
| 72 | %endmacro |
| 73 | |
| 74 | QPEL_TABLE 8, 8, b, sse4 |
| 75 | QPEL_TABLE 10, 4, w, sse4 |
| 76 | QPEL_TABLE 12, 4, w, sse4 |
| 77 | |
| 78 | %define MAX_PB_SIZE 64 |
| 79 | |
| 80 | %define hevc_qpel_filters_sse4_14 hevc_qpel_filters_sse4_10 |
| 81 | |
| 82 | %if ARCH_X86_64 |
| 83 | |
| 84 | %macro SIMPLE_BILOAD 4 ;width, tab, r1, r2 |
| 85 | %if %1 <= 4 |
| 86 | movq %3, [%2] ; load data from source2 |
| 87 | %elif %1 <= 8 |
| 88 | movdqa %3, [%2] ; load data from source2 |
| 89 | %elif %1 <= 12 |
| 90 | movdqa %3, [%2] ; load data from source2 |
| 91 | movq %4, [%2+16] ; load data from source2 |
| 92 | %else |
| 93 | movdqa %3, [%2] ; load data from source2 |
| 94 | movdqa %4, [%2+16] ; load data from source2 |
| 95 | %endif |
| 96 | %endmacro |
| 97 | |
| 98 | %macro SIMPLE_LOAD 4 ;width, bitd, tab, r1 |
| 99 | %if %1 == 2 || (%2 == 8 && %1 <= 4) |
| 100 | movd %4, [%3] ; load data from source |
| 101 | %elif %1 == 4 || (%2 == 8 && %1 <= 8) |
| 102 | movq %4, [%3] ; load data from source |
| 103 | %else |
| 104 | movdqu %4, [%3] ; load data from source |
| 105 | %endif |
| 106 | %endmacro |
| 107 | |
| 108 | %macro SIMPLE_8LOAD 5 ;width, bitd, tab, r1, r2 |
| 109 | %if %1 == 2 || (%2 == 8 && %1 <= 4) |
| 110 | movq %4, [%3] ; load data from source2 |
| 111 | %elif %1 == 4 || (%2 == 8 && %1 <= 8) |
| 112 | movdqa %4, [%3] ; load data from source2 |
| 113 | %elif %1 <= 12 |
| 114 | movdqa %4, [%3] ; load data from source2 |
| 115 | movq %5, [%3+16] ; load data from source2 |
| 116 | %else |
| 117 | movdqa %4, [%3] ; load data from source2 |
| 118 | movdqa %5, [%3+16] ; load data from source2 |
| 119 | %endif |
| 120 | %endmacro |
| 121 | |
| 122 | %macro EPEL_FILTER 2-4 ; bit depth, filter index |
| 123 | %ifdef PIC |
| 124 | lea rfilterq, [hevc_epel_filters_sse4_%1] |
| 125 | %else |
| 126 | %define rfilterq hevc_epel_filters_sse4_%1 |
| 127 | %endif |
| 128 | sub %2q, 1 |
| 129 | shl %2q, 5 ; multiply by 32 |
| 130 | %if %0 == 2 |
| 131 | movdqa m14, [rfilterq + %2q] ; get 2 first values of filters |
| 132 | movdqa m15, [rfilterq + %2q+16] ; get 2 last values of filters |
| 133 | %else |
| 134 | movdqa %3, [rfilterq + %2q] ; get 2 first values of filters |
| 135 | movdqa %4, [rfilterq + %2q+16] ; get 2 last values of filters |
| 136 | %endif |
| 137 | %endmacro |
| 138 | |
| 139 | %macro EPEL_HV_FILTER 1 |
| 140 | %ifdef PIC |
| 141 | lea rfilterq, [hevc_epel_filters_sse4_%1] |
| 142 | %else |
| 143 | %define rfilterq hevc_epel_filters_sse4_%1 |
| 144 | %endif |
| 145 | sub mxq, 1 |
| 146 | sub myq, 1 |
| 147 | shl mxq, 5 ; multiply by 32 |
| 148 | shl myq, 5 ; multiply by 32 |
| 149 | movdqa m14, [rfilterq + mxq] ; get 2 first values of filters |
| 150 | movdqa m15, [rfilterq + mxq+16] ; get 2 last values of filters |
| 151 | lea r3srcq, [srcstrideq*3] |
| 152 | |
| 153 | %ifdef PIC |
| 154 | lea rfilterq, [hevc_epel_filters_sse4_10] |
| 155 | %else |
| 156 | %define rfilterq hevc_epel_filters_sse4_10 |
| 157 | %endif |
| 158 | movdqa m12, [rfilterq + myq] ; get 2 first values of filters |
| 159 | movdqa m13, [rfilterq + myq+16] ; get 2 last values of filters |
| 160 | %endmacro |
| 161 | |
| 162 | %macro QPEL_FILTER 2 |
| 163 | %ifdef PIC |
| 164 | lea rfilterq, [hevc_qpel_filters_sse4_%1] |
| 165 | %else |
| 166 | %define rfilterq hevc_qpel_filters_sse4_%1 |
| 167 | %endif |
| 168 | lea %2q, [%2q*8-8] |
| 169 | movdqa m12, [rfilterq + %2q*8] ; get 4 first values of filters |
| 170 | movdqa m13, [rfilterq + %2q*8 + 16] ; get 4 first values of filters |
| 171 | movdqa m14, [rfilterq + %2q*8 + 32] ; get 4 first values of filters |
| 172 | movdqa m15, [rfilterq + %2q*8 + 48] ; get 4 first values of filters |
| 173 | %endmacro |
| 174 | |
| 175 | %macro EPEL_LOAD 4 |
| 176 | %ifdef PIC |
| 177 | lea rfilterq, [%2] |
| 178 | %else |
| 179 | %define rfilterq %2 |
| 180 | %endif |
| 181 | %if (%1 == 8 && %4 <= 4) |
| 182 | %define %%load movd |
| 183 | %elif (%1 == 8 && %4 <= 8) || (%1 > 8 && %4 <= 4) |
| 184 | %define %%load movq |
| 185 | %else |
| 186 | %define %%load movdqu |
| 187 | %endif |
| 188 | |
| 189 | %%load m0, [rfilterq ] |
| 190 | %ifnum %3 |
| 191 | %%load m1, [rfilterq+ %3] |
| 192 | %%load m2, [rfilterq+2*%3] |
| 193 | %%load m3, [rfilterq+3*%3] |
| 194 | %else |
| 195 | %%load m1, [rfilterq+ %3q] |
| 196 | %%load m2, [rfilterq+2*%3q] |
| 197 | %%load m3, [rfilterq+r3srcq] |
| 198 | %endif |
| 199 | |
| 200 | %if %1 == 8 |
| 201 | %if %4 > 8 |
| 202 | SBUTTERFLY bw, 0, 1, 10 |
| 203 | SBUTTERFLY bw, 2, 3, 10 |
| 204 | %else |
| 205 | punpcklbw m0, m1 |
| 206 | punpcklbw m2, m3 |
| 207 | %endif |
| 208 | %else |
| 209 | %if %4 > 4 |
| 210 | SBUTTERFLY wd, 0, 1, 10 |
| 211 | SBUTTERFLY wd, 2, 3, 10 |
| 212 | %else |
| 213 | punpcklwd m0, m1 |
| 214 | punpcklwd m2, m3 |
| 215 | %endif |
| 216 | %endif |
| 217 | %endmacro |
| 218 | |
| 219 | |
| 220 | %macro QPEL_H_LOAD 4 |
| 221 | %assign %%stride (%1+7)/8 |
| 222 | %if %1 == 8 |
| 223 | %if %3 <= 4 |
| 224 | %define %%load movd |
| 225 | %elif %3 == 8 |
| 226 | %define %%load movq |
| 227 | %else |
| 228 | %define %%load movdqu |
| 229 | %endif |
| 230 | %else |
| 231 | %if %3 == 2 |
| 232 | %define %%load movd |
| 233 | %elif %3 == 4 |
| 234 | %define %%load movq |
| 235 | %else |
| 236 | %define %%load movdqu |
| 237 | %endif |
| 238 | %endif |
| 239 | %%load m0, [%2-3*%%stride] ;load data from source |
| 240 | %%load m1, [%2-2*%%stride] |
| 241 | %%load m2, [%2-%%stride ] |
| 242 | %%load m3, [%2 ] |
| 243 | %%load m4, [%2+%%stride ] |
| 244 | %%load m5, [%2+2*%%stride] |
| 245 | %%load m6, [%2+3*%%stride] |
| 246 | %%load m7, [%2+4*%%stride] |
| 247 | |
| 248 | %if %1 == 8 |
| 249 | %if %3 > 8 |
| 250 | SBUTTERFLY wd, 0, 1, %4 |
| 251 | SBUTTERFLY wd, 2, 3, %4 |
| 252 | SBUTTERFLY wd, 4, 5, %4 |
| 253 | SBUTTERFLY wd, 6, 7, %4 |
| 254 | %else |
| 255 | punpcklwd m0, m1 |
| 256 | punpcklwd m2, m3 |
| 257 | punpcklwd m4, m5 |
| 258 | punpcklwd m6, m7 |
| 259 | %endif |
| 260 | %else |
| 261 | %if %3 > 4 |
| 262 | SBUTTERFLY dq, 0, 1, %4 |
| 263 | SBUTTERFLY dq, 2, 3, %4 |
| 264 | SBUTTERFLY dq, 4, 5, %4 |
| 265 | SBUTTERFLY dq, 6, 7, %4 |
| 266 | %else |
| 267 | punpckldq m0, m1 |
| 268 | punpckldq m2, m3 |
| 269 | punpckldq m4, m5 |
| 270 | punpckldq m6, m7 |
| 271 | %endif |
| 272 | %endif |
| 273 | %endmacro |
| 274 | |
| 275 | %macro QPEL_V_LOAD 5 |
| 276 | lea %5q, [%2] |
| 277 | sub %5q, r3srcq |
| 278 | movdqu m0, [%5q ] ;load x- 3*srcstride |
| 279 | movdqu m1, [%5q+ %3q ] ;load x- 2*srcstride |
| 280 | movdqu m2, [%5q+ 2*%3q ] ;load x-srcstride |
| 281 | movdqu m3, [%2 ] ;load x |
| 282 | movdqu m4, [%2+ %3q] ;load x+stride |
| 283 | movdqu m5, [%2+ 2*%3q] ;load x+2*stride |
| 284 | movdqu m6, [%2+r3srcq] ;load x+3*stride |
| 285 | movdqu m7, [%2+ 4*%3q] ;load x+4*stride |
| 286 | %if %1 == 8 |
| 287 | %if %4 > 8 |
| 288 | SBUTTERFLY bw, 0, 1, 8 |
| 289 | SBUTTERFLY bw, 2, 3, 8 |
| 290 | SBUTTERFLY bw, 4, 5, 8 |
| 291 | SBUTTERFLY bw, 6, 7, 8 |
| 292 | %else |
| 293 | punpcklbw m0, m1 |
| 294 | punpcklbw m2, m3 |
| 295 | punpcklbw m4, m5 |
| 296 | punpcklbw m6, m7 |
| 297 | %endif |
| 298 | %else |
| 299 | %if %4 > 4 |
| 300 | SBUTTERFLY wd, 0, 1, 8 |
| 301 | SBUTTERFLY wd, 2, 3, 8 |
| 302 | SBUTTERFLY wd, 4, 5, 8 |
| 303 | SBUTTERFLY wd, 6, 7, 8 |
| 304 | %else |
| 305 | punpcklwd m0, m1 |
| 306 | punpcklwd m2, m3 |
| 307 | punpcklwd m4, m5 |
| 308 | punpcklwd m6, m7 |
| 309 | %endif |
| 310 | %endif |
| 311 | %endmacro |
| 312 | |
| 313 | %macro PEL_12STORE2 3 |
| 314 | movd [%1], %2 |
| 315 | %endmacro |
| 316 | %macro PEL_12STORE4 3 |
| 317 | movq [%1], %2 |
| 318 | %endmacro |
| 319 | %macro PEL_12STORE6 3 |
| 320 | movq [%1], %2 |
| 321 | psrldq %2, 8 |
| 322 | movd [%1+8], %2 |
| 323 | %endmacro |
| 324 | %macro PEL_12STORE8 3 |
| 325 | movdqa [%1], %2 |
| 326 | %endmacro |
| 327 | %macro PEL_12STORE12 3 |
| 328 | movdqa [%1], %2 |
| 329 | movq [%1+16], %3 |
| 330 | %endmacro |
| 331 | %macro PEL_12STORE16 3 |
| 332 | PEL_12STORE8 %1, %2, %3 |
| 333 | movdqa [%1+16], %3 |
| 334 | %endmacro |
| 335 | |
| 336 | %macro PEL_10STORE2 3 |
| 337 | movd [%1], %2 |
| 338 | %endmacro |
| 339 | %macro PEL_10STORE4 3 |
| 340 | movq [%1], %2 |
| 341 | %endmacro |
| 342 | %macro PEL_10STORE6 3 |
| 343 | movq [%1], %2 |
| 344 | psrldq %2, 8 |
| 345 | movd [%1+8], %2 |
| 346 | %endmacro |
| 347 | %macro PEL_10STORE8 3 |
| 348 | movdqa [%1], %2 |
| 349 | %endmacro |
| 350 | %macro PEL_10STORE12 3 |
| 351 | movdqa [%1], %2 |
| 352 | movq [%1+16], %3 |
| 353 | %endmacro |
| 354 | %macro PEL_10STORE16 3 |
| 355 | PEL_10STORE8 %1, %2, %3 |
| 356 | movdqa [%1+16], %3 |
| 357 | %endmacro |
| 358 | |
| 359 | %macro PEL_8STORE2 3 |
| 360 | pextrw [%1], %2, 0 |
| 361 | %endmacro |
| 362 | %macro PEL_8STORE4 3 |
| 363 | movd [%1], %2 |
| 364 | %endmacro |
| 365 | %macro PEL_8STORE6 3 |
| 366 | movd [%1], %2 |
| 367 | pextrw [%1+4], %2, 2 |
| 368 | %endmacro |
| 369 | %macro PEL_8STORE8 3 |
| 370 | movq [%1], %2 |
| 371 | %endmacro |
| 372 | %macro PEL_8STORE12 3 |
| 373 | movq [%1], %2 |
| 374 | psrldq %2, 8 |
| 375 | movd [%1+8], %2 |
| 376 | %endmacro |
| 377 | %macro PEL_8STORE16 3 |
| 378 | movdqa [%1], %2 |
| 379 | %endmacro |
| 380 | |
| 381 | %macro LOOP_END 3 |
| 382 | add %1q, 2*MAX_PB_SIZE ; dst += dststride |
| 383 | add %2q, %3q ; src += srcstride |
| 384 | dec heightd ; cmp height |
| 385 | jnz .loop ; height loop |
| 386 | %endmacro |
| 387 | |
| 388 | |
| 389 | %macro MC_PIXEL_COMPUTE 2 ;width, bitdepth |
| 390 | %if %2 == 8 |
| 391 | %if %1 > 8 |
| 392 | punpckhbw m1, m0, m2 |
| 393 | psllw m1, 14-%2 |
| 394 | %endif |
| 395 | punpcklbw m0, m2 |
| 396 | %endif |
| 397 | psllw m0, 14-%2 |
| 398 | %endmacro |
| 399 | |
| 400 | |
| 401 | %macro EPEL_COMPUTE 4 ; bitdepth, width, filter1, filter2 |
| 402 | %if %1 == 8 |
| 403 | pmaddubsw m0, %3 ;x1*c1+x2*c2 |
| 404 | pmaddubsw m2, %4 ;x3*c3+x4*c4 |
| 405 | paddw m0, m2 |
| 406 | %if %2 > 8 |
| 407 | pmaddubsw m1, %3 |
| 408 | pmaddubsw m3, %4 |
| 409 | paddw m1, m3 |
| 410 | %endif |
| 411 | %else |
| 412 | pmaddwd m0, %3 |
| 413 | pmaddwd m2, %4 |
| 414 | paddd m0, m2 |
| 415 | %if %2 > 4 |
| 416 | pmaddwd m1, %3 |
| 417 | pmaddwd m3, %4 |
| 418 | paddd m1, m3 |
| 419 | %endif |
| 420 | %if %1 != 8 |
| 421 | psrad m0, %1-8 |
| 422 | psrad m1, %1-8 |
| 423 | %endif |
| 424 | packssdw m0, m1 |
| 425 | %endif |
| 426 | %endmacro |
| 427 | |
| 428 | %macro QPEL_HV_COMPUTE 4 ; width, bitdepth, filter idx |
| 429 | %ifdef PIC |
| 430 | lea rfilterq, [hevc_qpel_filters_sse4_%2] |
| 431 | %else |
| 432 | %define rfilterq hevc_qpel_filters_sse4_%2 |
| 433 | %endif |
| 434 | |
| 435 | %if %2 == 8 |
| 436 | pmaddubsw m0, [rfilterq + %3q*8 ] ;x1*c1+x2*c2 |
| 437 | pmaddubsw m2, [rfilterq + %3q*8+16] ;x3*c3+x4*c4 |
| 438 | pmaddubsw m4, [rfilterq + %3q*8+32] ;x5*c5+x6*c6 |
| 439 | pmaddubsw m6, [rfilterq + %3q*8+48] ;x7*c7+x8*c8 |
| 440 | paddw m0, m2 |
| 441 | paddw m4, m6 |
| 442 | paddw m0, m4 |
| 443 | %else |
| 444 | pmaddwd m0, [rfilterq + %3q*8 ] |
| 445 | pmaddwd m2, [rfilterq + %3q*8+16] |
| 446 | pmaddwd m4, [rfilterq + %3q*8+32] |
| 447 | pmaddwd m6, [rfilterq + %3q*8+48] |
| 448 | paddd m0, m2 |
| 449 | paddd m4, m6 |
| 450 | paddd m0, m4 |
| 451 | %if %2 != 8 |
| 452 | psrad m0, %2-8 |
| 453 | %endif |
| 454 | %if %1 > 4 |
| 455 | pmaddwd m1, [rfilterq + %3q*8 ] |
| 456 | pmaddwd m3, [rfilterq + %3q*8+16] |
| 457 | pmaddwd m5, [rfilterq + %3q*8+32] |
| 458 | pmaddwd m7, [rfilterq + %3q*8+48] |
| 459 | paddd m1, m3 |
| 460 | paddd m5, m7 |
| 461 | paddd m1, m5 |
| 462 | %if %2 != 8 |
| 463 | psrad m1, %2-8 |
| 464 | %endif |
| 465 | %endif |
| 466 | p%4 m0, m1 |
| 467 | %endif |
| 468 | %endmacro |
| 469 | |
| 470 | %macro QPEL_COMPUTE 2 ; width, bitdepth |
| 471 | %if %2 == 8 |
| 472 | pmaddubsw m0, m12 ;x1*c1+x2*c2 |
| 473 | pmaddubsw m2, m13 ;x3*c3+x4*c4 |
| 474 | pmaddubsw m4, m14 ;x5*c5+x6*c6 |
| 475 | pmaddubsw m6, m15 ;x7*c7+x8*c8 |
| 476 | paddw m0, m2 |
| 477 | paddw m4, m6 |
| 478 | paddw m0, m4 |
| 479 | %if %1 > 8 |
| 480 | pmaddubsw m1, m12 |
| 481 | pmaddubsw m3, m13 |
| 482 | pmaddubsw m5, m14 |
| 483 | pmaddubsw m7, m15 |
| 484 | paddw m1, m3 |
| 485 | paddw m5, m7 |
| 486 | paddw m1, m5 |
| 487 | %endif |
| 488 | %else |
| 489 | pmaddwd m0, m12 |
| 490 | pmaddwd m2, m13 |
| 491 | pmaddwd m4, m14 |
| 492 | pmaddwd m6, m15 |
| 493 | paddd m0, m2 |
| 494 | paddd m4, m6 |
| 495 | paddd m0, m4 |
| 496 | %if %2 != 8 |
| 497 | psrad m0, %2-8 |
| 498 | %endif |
| 499 | %if %1 > 4 |
| 500 | pmaddwd m1, m12 |
| 501 | pmaddwd m3, m13 |
| 502 | pmaddwd m5, m14 |
| 503 | pmaddwd m7, m15 |
| 504 | paddd m1, m3 |
| 505 | paddd m5, m7 |
| 506 | paddd m1, m5 |
| 507 | %if %2 != 8 |
| 508 | psrad m1, %2-8 |
| 509 | %endif |
| 510 | %endif |
| 511 | %endif |
| 512 | %endmacro |
| 513 | |
| 514 | %macro BI_COMPUTE 7 ; width, bitd, src1l, src1h, scr2l, scr2h, pw |
| 515 | paddsw %3, %5 |
| 516 | %if %1 > 8 |
| 517 | paddsw %4, %6 |
| 518 | %endif |
| 519 | UNI_COMPUTE %1, %2, %3, %4, %7 |
| 520 | %endmacro |
| 521 | |
| 522 | %macro UNI_COMPUTE 5 |
| 523 | pmulhrsw %3, %5 |
| 524 | %if %1 > 8 || (%2 > 8 && %1 > 4) |
| 525 | pmulhrsw %4, %5 |
| 526 | %endif |
| 527 | %if %2 == 8 |
| 528 | packuswb %3, %4 |
| 529 | %else |
| 530 | pminsw %3, [max_pixels_%2] |
| 531 | pmaxsw %3, [zero] |
| 532 | %if %1 > 8 |
| 533 | pminsw %4, [max_pixels_%2] |
| 534 | pmaxsw %4, [zero] |
| 535 | %endif |
| 536 | %endif |
| 537 | %endmacro |
| 538 | |
| 539 | INIT_XMM sse4 ; adds ff_ and _sse4 to function name |
| 540 | ; ****************************** |
| 541 | ; void put_hevc_mc_pixels(int16_t *dst, ptrdiff_t dststride, |
| 542 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 543 | ; int height, int mx, int my) |
| 544 | ; ****************************** |
| 545 | |
| 546 | %macro HEVC_PUT_HEVC_PEL_PIXELS 2 |
| 547 | cglobal hevc_put_hevc_pel_pixels%1_%2, 4, 4, 3, dst, src, srcstride,height |
| 548 | pxor m2, m2 |
| 549 | .loop |
| 550 | SIMPLE_LOAD %1, %2, srcq, m0 |
| 551 | MC_PIXEL_COMPUTE %1, %2 |
| 552 | PEL_10STORE%1 dstq, m0, m1 |
| 553 | LOOP_END dst, src, srcstride |
| 554 | RET |
| 555 | |
| 556 | cglobal hevc_put_hevc_uni_pel_pixels%1_%2, 5, 5, 2, dst, dststride, src, srcstride,height |
| 557 | .loop |
| 558 | SIMPLE_LOAD %1, %2, srcq, m0 |
| 559 | PEL_%2STORE%1 dstq, m0, m1 |
| 560 | add dstq, dststrideq ; dst += dststride |
| 561 | add srcq, srcstrideq ; src += srcstride |
| 562 | dec heightd ; cmp height |
| 563 | jnz .loop ; height loop |
| 564 | RET |
| 565 | |
| 566 | cglobal hevc_put_hevc_bi_pel_pixels%1_%2, 6, 6, 6, dst, dststride, src, srcstride, src2, height |
| 567 | pxor m2, m2 |
| 568 | movdqa m5, [pw_bi_%2] |
| 569 | .loop |
| 570 | SIMPLE_LOAD %1, %2, srcq, m0 |
| 571 | SIMPLE_BILOAD %1, src2q, m3, m4 |
| 572 | MC_PIXEL_COMPUTE %1, %2 |
| 573 | BI_COMPUTE %1, %2, m0, m1, m3, m4, m5 |
| 574 | PEL_%2STORE%1 dstq, m0, m1 |
| 575 | add dstq, dststrideq ; dst += dststride |
| 576 | add srcq, srcstrideq ; src += srcstride |
| 577 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 578 | dec heightd ; cmp height |
| 579 | jnz .loop ; height loop |
| 580 | RET |
| 581 | |
| 582 | %endmacro |
| 583 | |
| 584 | |
| 585 | ; ****************************** |
| 586 | ; void put_hevc_epel_hX(int16_t *dst, ptrdiff_t dststride, |
| 587 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 588 | ; int width, int height, int mx, int my, |
| 589 | ; int16_t* mcbuffer) |
| 590 | ; ****************************** |
| 591 | |
| 592 | |
| 593 | %macro HEVC_PUT_HEVC_EPEL 2 |
| 594 | cglobal hevc_put_hevc_epel_h%1_%2, 5, 6, 6, dst, src, srcstride, height, mx, rfilter |
| 595 | %assign %%stride ((%2 + 7)/8) |
| 596 | EPEL_FILTER %2, mx, m4, m5 |
| 597 | .loop |
| 598 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 599 | EPEL_COMPUTE %2, %1, m4, m5 |
| 600 | PEL_10STORE%1 dstq, m0, m1 |
| 601 | LOOP_END dst, src, srcstride |
| 602 | RET |
| 603 | |
| 604 | cglobal hevc_put_hevc_uni_epel_h%1_%2, 6, 7, 7, dst, dststride, src, srcstride, height, mx, rfilter |
| 605 | %assign %%stride ((%2 + 7)/8) |
| 606 | movdqa m6, [pw_%2] |
| 607 | EPEL_FILTER %2, mx, m4, m5 |
| 608 | .loop |
| 609 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 610 | EPEL_COMPUTE %2, %1, m4, m5 |
| 611 | UNI_COMPUTE %1, %2, m0, m1, m6 |
| 612 | PEL_%2STORE%1 dstq, m0, m1 |
| 613 | add dstq, dststrideq ; dst += dststride |
| 614 | add srcq, srcstrideq ; src += srcstride |
| 615 | dec heightd ; cmp height |
| 616 | jnz .loop ; height loop |
| 617 | RET |
| 618 | |
| 619 | cglobal hevc_put_hevc_bi_epel_h%1_%2, 7, 8, 7, dst, dststride, src, srcstride, src2, height, mx, rfilter |
| 620 | movdqa m6, [pw_bi_%2] |
| 621 | EPEL_FILTER %2, mx, m4, m5 |
| 622 | .loop |
| 623 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 624 | EPEL_COMPUTE %2, %1, m4, m5 |
| 625 | SIMPLE_BILOAD %1, src2q, m2, m3 |
| 626 | BI_COMPUTE %1, %2, m0, m1, m2, m3, m6 |
| 627 | PEL_%2STORE%1 dstq, m0, m1 |
| 628 | add dstq, dststrideq ; dst += dststride |
| 629 | add srcq, srcstrideq ; src += srcstride |
| 630 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 631 | dec heightd ; cmp height |
| 632 | jnz .loop ; height loop |
| 633 | RET |
| 634 | |
| 635 | ; ****************************** |
| 636 | ; void put_hevc_epel_v(int16_t *dst, ptrdiff_t dststride, |
| 637 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 638 | ; int width, int height, int mx, int my, |
| 639 | ; int16_t* mcbuffer) |
| 640 | ; ****************************** |
| 641 | |
| 642 | cglobal hevc_put_hevc_epel_v%1_%2, 6, 7, 6, dst, src, srcstride, height, r3src, my, rfilter |
| 643 | lea r3srcq, [srcstrideq*3] |
| 644 | sub srcq, srcstrideq |
| 645 | EPEL_FILTER %2, my, m4, m5 |
| 646 | .loop |
| 647 | EPEL_LOAD %2, srcq, srcstride, %1 |
| 648 | EPEL_COMPUTE %2, %1, m4, m5 |
| 649 | PEL_10STORE%1 dstq, m0, m1 |
| 650 | LOOP_END dst, src, srcstride |
| 651 | RET |
| 652 | |
| 653 | cglobal hevc_put_hevc_uni_epel_v%1_%2, 7, 8, 7, dst, dststride, src, srcstride, height, r3src, my, rfilter |
| 654 | lea r3srcq, [srcstrideq*3] |
| 655 | movdqa m6, [pw_%2] |
| 656 | sub srcq, srcstrideq |
| 657 | EPEL_FILTER %2, my, m4, m5 |
| 658 | .loop |
| 659 | EPEL_LOAD %2, srcq, srcstride, %1 |
| 660 | EPEL_COMPUTE %2, %1, m4, m5 |
| 661 | UNI_COMPUTE %1, %2, m0, m1, m6 |
| 662 | PEL_%2STORE%1 dstq, m0, m1 |
| 663 | add dstq, dststrideq ; dst += dststride |
| 664 | add srcq, srcstrideq ; src += srcstride |
| 665 | dec heightd ; cmp height |
| 666 | jnz .loop ; height loop |
| 667 | RET |
| 668 | |
| 669 | |
| 670 | cglobal hevc_put_hevc_bi_epel_v%1_%2, 8, 9, 7, dst, dststride, src, srcstride, src2, height, r3src, my, rfilter |
| 671 | lea r3srcq, [srcstrideq*3] |
| 672 | movdqa m6, [pw_bi_%2] |
| 673 | sub srcq, srcstrideq |
| 674 | EPEL_FILTER %2, my, m4, m5 |
| 675 | .loop |
| 676 | EPEL_LOAD %2, srcq, srcstride, %1 |
| 677 | EPEL_COMPUTE %2, %1, m4, m5 |
| 678 | SIMPLE_BILOAD %1, src2q, m2, m3 |
| 679 | BI_COMPUTE %1, %2, m0, m1, m2, m3, m6 |
| 680 | PEL_%2STORE%1 dstq, m0, m1 |
| 681 | add dstq, dststrideq ; dst += dststride |
| 682 | add srcq, srcstrideq ; src += srcstride |
| 683 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 684 | dec heightd ; cmp height |
| 685 | jnz .loop ; height loop |
| 686 | RET |
| 687 | %endmacro |
| 688 | |
| 689 | |
| 690 | ; ****************************** |
| 691 | ; void put_hevc_epel_hv(int16_t *dst, ptrdiff_t dststride, |
| 692 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 693 | ; int width, int height, int mx, int my) |
| 694 | ; ****************************** |
| 695 | |
| 696 | %macro HEVC_PUT_HEVC_EPEL_HV 2 |
| 697 | cglobal hevc_put_hevc_epel_hv%1_%2, 6, 8, 12 , dst, src, srcstride, height, mx, my, r3src, rfilter |
| 698 | %assign %%stride ((%2 + 7)/8) |
| 699 | sub srcq, srcstrideq |
| 700 | EPEL_HV_FILTER %2 |
| 701 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 702 | EPEL_COMPUTE %2, %1, m14, m15 |
| 703 | SWAP m4, m0 |
| 704 | add srcq, srcstrideq |
| 705 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 706 | EPEL_COMPUTE %2, %1, m14, m15 |
| 707 | SWAP m5, m0 |
| 708 | add srcq, srcstrideq |
| 709 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 710 | EPEL_COMPUTE %2, %1, m14, m15 |
| 711 | SWAP m6, m0 |
| 712 | add srcq, srcstrideq |
| 713 | .loop |
| 714 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 715 | EPEL_COMPUTE %2, %1, m14, m15 |
| 716 | SWAP m7, m0 |
| 717 | punpcklwd m0, m4, m5 |
| 718 | punpcklwd m2, m6, m7 |
| 719 | %if %1 > 4 |
| 720 | punpckhwd m1, m4, m5 |
| 721 | punpckhwd m3, m6, m7 |
| 722 | %endif |
| 723 | EPEL_COMPUTE 14, %1, m12, m13 |
| 724 | PEL_10STORE%1 dstq, m0, m1 |
| 725 | movdqa m4, m5 |
| 726 | movdqa m5, m6 |
| 727 | movdqa m6, m7 |
| 728 | LOOP_END dst, src, srcstride |
| 729 | RET |
| 730 | |
| 731 | cglobal hevc_put_hevc_uni_epel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, height, mx, my, r3src, rfilter |
| 732 | %assign %%stride ((%2 + 7)/8) |
| 733 | sub srcq, srcstrideq |
| 734 | EPEL_HV_FILTER %2 |
| 735 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 736 | EPEL_COMPUTE %2, %1, m14, m15 |
| 737 | SWAP m4, m0 |
| 738 | add srcq, srcstrideq |
| 739 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 740 | EPEL_COMPUTE %2, %1, m14, m15 |
| 741 | SWAP m5, m0 |
| 742 | add srcq, srcstrideq |
| 743 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 744 | EPEL_COMPUTE %2, %1, m14, m15 |
| 745 | SWAP m6, m0 |
| 746 | add srcq, srcstrideq |
| 747 | .loop |
| 748 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 749 | EPEL_COMPUTE %2, %1, m14, m15 |
| 750 | SWAP m7, m0 |
| 751 | punpcklwd m0, m4, m5 |
| 752 | punpcklwd m2, m6, m7 |
| 753 | %if %1 > 4 |
| 754 | punpckhwd m1, m4, m5 |
| 755 | punpckhwd m3, m6, m7 |
| 756 | %endif |
| 757 | EPEL_COMPUTE 14, %1, m12, m13 |
| 758 | UNI_COMPUTE %1, %2, m0, m1, [pw_%2] |
| 759 | PEL_%2STORE%1 dstq, m0, m1 |
| 760 | movdqa m4, m5 |
| 761 | movdqa m5, m6 |
| 762 | movdqa m6, m7 |
| 763 | add dstq, dststrideq ; dst += dststride |
| 764 | add srcq, srcstrideq ; src += srcstride |
| 765 | dec heightd ; cmp height |
| 766 | jnz .loop ; height loop |
| 767 | RET |
| 768 | |
| 769 | |
| 770 | cglobal hevc_put_hevc_bi_epel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, mx, my, r3src, rfilter |
| 771 | %assign %%stride ((%2 + 7)/8) |
| 772 | sub srcq, srcstrideq |
| 773 | EPEL_HV_FILTER %2 |
| 774 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 775 | EPEL_COMPUTE %2, %1, m14, m15 |
| 776 | SWAP m4, m0 |
| 777 | add srcq, srcstrideq |
| 778 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 779 | EPEL_COMPUTE %2, %1, m14, m15 |
| 780 | SWAP m5, m0 |
| 781 | add srcq, srcstrideq |
| 782 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 783 | EPEL_COMPUTE %2, %1, m14, m15 |
| 784 | SWAP m6, m0 |
| 785 | add srcq, srcstrideq |
| 786 | .loop |
| 787 | EPEL_LOAD %2, srcq-%%stride, %%stride, %1 |
| 788 | EPEL_COMPUTE %2, %1, m14, m15 |
| 789 | SWAP m7, m0 |
| 790 | punpcklwd m0, m4, m5 |
| 791 | punpcklwd m2, m6, m7 |
| 792 | %if %1 > 4 |
| 793 | punpckhwd m1, m4, m5 |
| 794 | punpckhwd m3, m6, m7 |
| 795 | %endif |
| 796 | EPEL_COMPUTE 14, %1, m12, m13 |
| 797 | SIMPLE_BILOAD %1, src2q, m8, m9 |
| 798 | BI_COMPUTE %1, %2, m0, m1, m8, m9, [pw_bi_%2] |
| 799 | PEL_%2STORE%1 dstq, m0, m1 |
| 800 | movdqa m4, m5 |
| 801 | movdqa m5, m6 |
| 802 | movdqa m6, m7 |
| 803 | add dstq, dststrideq ; dst += dststride |
| 804 | add srcq, srcstrideq ; src += srcstride |
| 805 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 806 | dec heightd ; cmp height |
| 807 | jnz .loop ; height loop |
| 808 | RET |
| 809 | %endmacro |
| 810 | |
| 811 | ; ****************************** |
| 812 | ; void put_hevc_qpel_hX_X_X(int16_t *dst, ptrdiff_t dststride, |
| 813 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 814 | ; int width, int height, int mx, int my) |
| 815 | ; ****************************** |
| 816 | |
| 817 | %macro HEVC_PUT_HEVC_QPEL 2 |
| 818 | cglobal hevc_put_hevc_qpel_h%1_%2, 5, 6, 15, dst, src, srcstride, height, mx, rfilter |
| 819 | QPEL_FILTER %2, mx |
| 820 | .loop |
| 821 | QPEL_H_LOAD %2, srcq, %1, 10 |
| 822 | QPEL_COMPUTE %1, %2 |
| 823 | %if %2 > 8 |
| 824 | packssdw m0, m1 |
| 825 | %endif |
| 826 | PEL_10STORE%1 dstq, m0, m1 |
| 827 | LOOP_END dst, src, srcstride |
| 828 | RET |
| 829 | |
| 830 | cglobal hevc_put_hevc_uni_qpel_h%1_%2, 6, 7, 15 , dst, dststride, src, srcstride, height, mx, rfilter |
| 831 | movdqa m9, [pw_%2] |
| 832 | QPEL_FILTER %2, mx |
| 833 | .loop |
| 834 | QPEL_H_LOAD %2, srcq, %1, 10 |
| 835 | QPEL_COMPUTE %1, %2 |
| 836 | %if %2 > 8 |
| 837 | packssdw m0, m1 |
| 838 | %endif |
| 839 | UNI_COMPUTE %1, %2, m0, m1, m9 |
| 840 | PEL_%2STORE%1 dstq, m0, m1 |
| 841 | add dstq, dststrideq ; dst += dststride |
| 842 | add srcq, srcstrideq ; src += srcstride |
| 843 | dec heightd ; cmp height |
| 844 | jnz .loop ; height loop |
| 845 | RET |
| 846 | |
| 847 | cglobal hevc_put_hevc_bi_qpel_h%1_%2, 7, 8, 16 , dst, dststride, src, srcstride, src2, height, mx, rfilter |
| 848 | movdqa m9, [pw_bi_%2] |
| 849 | QPEL_FILTER %2, mx |
| 850 | .loop |
| 851 | QPEL_H_LOAD %2, srcq, %1, 10 |
| 852 | QPEL_COMPUTE %1, %2 |
| 853 | %if %2 > 8 |
| 854 | packssdw m0, m1 |
| 855 | %endif |
| 856 | SIMPLE_BILOAD %1, src2q, m10, m11 |
| 857 | BI_COMPUTE %1, %2, m0, m1, m10, m11, m9 |
| 858 | PEL_%2STORE%1 dstq, m0, m1 |
| 859 | add dstq, dststrideq ; dst += dststride |
| 860 | add srcq, srcstrideq ; src += srcstride |
| 861 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 862 | dec heightd ; cmp height |
| 863 | jnz .loop ; height loop |
| 864 | RET |
| 865 | |
| 866 | |
| 867 | ; ****************************** |
| 868 | ; void put_hevc_qpel_vX_X_X(int16_t *dst, ptrdiff_t dststride, |
| 869 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 870 | ; int width, int height, int mx, int my) |
| 871 | ; ****************************** |
| 872 | |
| 873 | cglobal hevc_put_hevc_qpel_v%1_%2, 6, 8, 15, dst, src, srcstride, height, r3src, my, rfilter |
| 874 | lea r3srcq, [srcstrideq*3] |
| 875 | QPEL_FILTER %2, my |
| 876 | .loop |
| 877 | QPEL_V_LOAD %2, srcq, srcstride, %1, r7 |
| 878 | QPEL_COMPUTE %1, %2 |
| 879 | %if %2 > 8 |
| 880 | packssdw m0, m1 |
| 881 | %endif |
| 882 | PEL_10STORE%1 dstq, m0, m1 |
| 883 | LOOP_END dst, src, srcstride |
| 884 | RET |
| 885 | |
| 886 | cglobal hevc_put_hevc_uni_qpel_v%1_%2, 7, 9, 15, dst, dststride, src, srcstride, height, r3src, my, rfilter |
| 887 | movdqa m9, [pw_%2] |
| 888 | lea r3srcq, [srcstrideq*3] |
| 889 | QPEL_FILTER %2, my |
| 890 | .loop |
| 891 | QPEL_V_LOAD %2, srcq, srcstride, %1, r8 |
| 892 | QPEL_COMPUTE %1, %2 |
| 893 | %if %2 > 8 |
| 894 | packssdw m0, m1 |
| 895 | %endif |
| 896 | UNI_COMPUTE %1, %2, m0, m1, m9 |
| 897 | PEL_%2STORE%1 dstq, m0, m1 |
| 898 | add dstq, dststrideq ; dst += dststride |
| 899 | add srcq, srcstrideq ; src += srcstride |
| 900 | dec heightd ; cmp height |
| 901 | jnz .loop ; height loop |
| 902 | RET |
| 903 | |
| 904 | cglobal hevc_put_hevc_bi_qpel_v%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, r3src, my, rfilter |
| 905 | movdqa m9, [pw_bi_%2] |
| 906 | lea r3srcq, [srcstrideq*3] |
| 907 | QPEL_FILTER %2, my |
| 908 | .loop |
| 909 | SIMPLE_BILOAD %1, src2q, m10, m11 |
| 910 | QPEL_V_LOAD %2, srcq, srcstride, %1, r9 |
| 911 | QPEL_COMPUTE %1, %2 |
| 912 | %if %2 > 8 |
| 913 | packssdw m0, m1 |
| 914 | %endif |
| 915 | BI_COMPUTE %1, %2, m0, m1, m10, m11, m9 |
| 916 | PEL_%2STORE%1 dstq, m0, m1 |
| 917 | add dstq, dststrideq ; dst += dststride |
| 918 | add srcq, srcstrideq ; src += srcstride |
| 919 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 920 | dec heightd ; cmp height |
| 921 | jnz .loop ; height loop |
| 922 | RET |
| 923 | %endmacro |
| 924 | |
| 925 | |
| 926 | ; ****************************** |
| 927 | ; void put_hevc_qpel_hvX_X(int16_t *dst, ptrdiff_t dststride, |
| 928 | ; uint8_t *_src, ptrdiff_t _srcstride, |
| 929 | ; int height, int mx, int my) |
| 930 | ; ****************************** |
| 931 | %macro HEVC_PUT_HEVC_QPEL_HV 2 |
| 932 | cglobal hevc_put_hevc_qpel_hv%1_%2, 6, 8, 12, dst, src, srcstride, height, mx, my, r3src, rfilter |
| 933 | lea mxq, [mxq*8-8] |
| 934 | lea myq, [myq*8-8] |
| 935 | lea r3srcq, [srcstrideq*3] |
| 936 | sub srcq, r3srcq |
| 937 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 938 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 939 | SWAP m8, m0 |
| 940 | add srcq, srcstrideq |
| 941 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 942 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 943 | SWAP m9, m0 |
| 944 | add srcq, srcstrideq |
| 945 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 946 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 947 | SWAP m10, m0 |
| 948 | add srcq, srcstrideq |
| 949 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 950 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 951 | SWAP m11, m0 |
| 952 | add srcq, srcstrideq |
| 953 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 954 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 955 | SWAP m12, m0 |
| 956 | add srcq, srcstrideq |
| 957 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 958 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 959 | SWAP m13, m0 |
| 960 | add srcq, srcstrideq |
| 961 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 962 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 963 | SWAP m14, m0 |
| 964 | add srcq, srcstrideq |
| 965 | .loop |
| 966 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 967 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 968 | SWAP m15, m0 |
| 969 | punpcklwd m0, m8, m9 |
| 970 | punpcklwd m2, m10, m11 |
| 971 | punpcklwd m4, m12, m13 |
| 972 | punpcklwd m6, m14, m15 |
| 973 | %if %1 > 4 |
| 974 | punpckhwd m1, m8, m9 |
| 975 | punpckhwd m3, m10, m11 |
| 976 | punpckhwd m5, m12, m13 |
| 977 | punpckhwd m7, m14, m15 |
| 978 | %endif |
| 979 | QPEL_HV_COMPUTE %1, 14, my, ackssdw |
| 980 | PEL_10STORE%1 dstq, m0, m1 |
| 981 | %if %1 <= 4 |
| 982 | movq m8, m9 |
| 983 | movq m9, m10 |
| 984 | movq m10, m11 |
| 985 | movq m11, m12 |
| 986 | movq m12, m13 |
| 987 | movq m13, m14 |
| 988 | movq m14, m15 |
| 989 | %else |
| 990 | movdqa m8, m9 |
| 991 | movdqa m9, m10 |
| 992 | movdqa m10, m11 |
| 993 | movdqa m11, m12 |
| 994 | movdqa m12, m13 |
| 995 | movdqa m13, m14 |
| 996 | movdqa m14, m15 |
| 997 | %endif |
| 998 | LOOP_END dst, src, srcstride |
| 999 | RET |
| 1000 | |
| 1001 | cglobal hevc_put_hevc_uni_qpel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, height, mx, my, r3src, rfilter |
| 1002 | lea mxq, [mxq*8-8] |
| 1003 | lea myq, [myq*8-8] |
| 1004 | lea r3srcq, [srcstrideq*3] |
| 1005 | sub srcq, r3srcq |
| 1006 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1007 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1008 | SWAP m8, m0 |
| 1009 | add srcq, srcstrideq |
| 1010 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1011 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1012 | SWAP m9, m0 |
| 1013 | add srcq, srcstrideq |
| 1014 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1015 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1016 | SWAP m10, m0 |
| 1017 | add srcq, srcstrideq |
| 1018 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1019 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1020 | SWAP m11, m0 |
| 1021 | add srcq, srcstrideq |
| 1022 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1023 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1024 | SWAP m12, m0 |
| 1025 | add srcq, srcstrideq |
| 1026 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1027 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1028 | SWAP m13, m0 |
| 1029 | add srcq, srcstrideq |
| 1030 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1031 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1032 | SWAP m14, m0 |
| 1033 | add srcq, srcstrideq |
| 1034 | .loop |
| 1035 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1036 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1037 | SWAP m15, m0 |
| 1038 | punpcklwd m0, m8, m9 |
| 1039 | punpcklwd m2, m10, m11 |
| 1040 | punpcklwd m4, m12, m13 |
| 1041 | punpcklwd m6, m14, m15 |
| 1042 | %if %1 > 4 |
| 1043 | punpckhwd m1, m8, m9 |
| 1044 | punpckhwd m3, m10, m11 |
| 1045 | punpckhwd m5, m12, m13 |
| 1046 | punpckhwd m7, m14, m15 |
| 1047 | %endif |
| 1048 | QPEL_HV_COMPUTE %1, 14, my, ackusdw |
| 1049 | UNI_COMPUTE %1, %2, m0, m1, [pw_%2] |
| 1050 | PEL_%2STORE%1 dstq, m0, m1 |
| 1051 | |
| 1052 | %if %1 <= 4 |
| 1053 | movq m8, m9 |
| 1054 | movq m9, m10 |
| 1055 | movq m10, m11 |
| 1056 | movq m11, m12 |
| 1057 | movq m12, m13 |
| 1058 | movq m13, m14 |
| 1059 | movq m14, m15 |
| 1060 | %else |
| 1061 | movdqa m8, m9 |
| 1062 | movdqa m9, m10 |
| 1063 | movdqa m10, m11 |
| 1064 | movdqa m11, m12 |
| 1065 | movdqa m12, m13 |
| 1066 | movdqa m13, m14 |
| 1067 | movdqa m14, m15 |
| 1068 | %endif |
| 1069 | add dstq, dststrideq ; dst += dststride |
| 1070 | add srcq, srcstrideq ; src += srcstride |
| 1071 | dec heightd ; cmp height |
| 1072 | jnz .loop ; height loop |
| 1073 | RET |
| 1074 | |
| 1075 | cglobal hevc_put_hevc_bi_qpel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, mx, my, r3src, rfilter |
| 1076 | lea mxq, [mxq*8-8] |
| 1077 | lea myq, [myq*8-8] |
| 1078 | lea r3srcq, [srcstrideq*3] |
| 1079 | sub srcq, r3srcq |
| 1080 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1081 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1082 | SWAP m8, m0 |
| 1083 | add srcq, srcstrideq |
| 1084 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1085 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1086 | SWAP m9, m0 |
| 1087 | add srcq, srcstrideq |
| 1088 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1089 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1090 | SWAP m10, m0 |
| 1091 | add srcq, srcstrideq |
| 1092 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1093 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1094 | SWAP m11, m0 |
| 1095 | add srcq, srcstrideq |
| 1096 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1097 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1098 | SWAP m12, m0 |
| 1099 | add srcq, srcstrideq |
| 1100 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1101 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1102 | SWAP m13, m0 |
| 1103 | add srcq, srcstrideq |
| 1104 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1105 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1106 | SWAP m14, m0 |
| 1107 | add srcq, srcstrideq |
| 1108 | .loop |
| 1109 | QPEL_H_LOAD %2, srcq, %1, 15 |
| 1110 | QPEL_HV_COMPUTE %1, %2, mx, ackssdw |
| 1111 | SWAP m15, m0 |
| 1112 | punpcklwd m0, m8, m9 |
| 1113 | punpcklwd m2, m10, m11 |
| 1114 | punpcklwd m4, m12, m13 |
| 1115 | punpcklwd m6, m14, m15 |
| 1116 | %if %1 > 4 |
| 1117 | punpckhwd m1, m8, m9 |
| 1118 | punpckhwd m3, m10, m11 |
| 1119 | punpckhwd m5, m12, m13 |
| 1120 | punpckhwd m7, m14, m15 |
| 1121 | %endif |
| 1122 | QPEL_HV_COMPUTE %1, 14, my, ackssdw |
| 1123 | SIMPLE_BILOAD %1, src2q, m8, m9 ;m9 not used in this case |
| 1124 | BI_COMPUTE %1, %2, m0, m1, m8, m9, [pw_bi_%2] |
| 1125 | PEL_%2STORE%1 dstq, m0, m1 |
| 1126 | |
| 1127 | %if %1 <= 4 |
| 1128 | movq m8, m9 |
| 1129 | movq m9, m10 |
| 1130 | movq m10, m11 |
| 1131 | movq m11, m12 |
| 1132 | movq m12, m13 |
| 1133 | movq m13, m14 |
| 1134 | movq m14, m15 |
| 1135 | %else |
| 1136 | movdqa m8, m9 |
| 1137 | movdqa m9, m10 |
| 1138 | movdqa m10, m11 |
| 1139 | movdqa m11, m12 |
| 1140 | movdqa m12, m13 |
| 1141 | movdqa m13, m14 |
| 1142 | movdqa m14, m15 |
| 1143 | %endif |
| 1144 | add dstq, dststrideq ; dst += dststride |
| 1145 | add srcq, srcstrideq ; src += srcstride |
| 1146 | add src2q, 2*MAX_PB_SIZE ; src += srcstride |
| 1147 | dec heightd ; cmp height |
| 1148 | jnz .loop ; height loop |
| 1149 | RET |
| 1150 | %endmacro |
| 1151 | |
| 1152 | %macro WEIGHTING_FUNCS 2 |
| 1153 | %if WIN64 || ARCH_X86_32 |
| 1154 | cglobal hevc_put_hevc_uni_w%1_%2, 4, 5, 7, dst, dststride, src, srcstride, height, denom, wx, ox |
| 1155 | mov r4d, denomm |
| 1156 | %define SHIFT r4d |
| 1157 | %else |
| 1158 | cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, height, denom, wx, ox |
| 1159 | %define SHIFT denomd |
| 1160 | %endif |
| 1161 | lea SHIFT, [SHIFT+14-%2] ; shift = 14 - bitd + denom |
| 1162 | %if %1 <= 4 |
| 1163 | pxor m1, m1 |
| 1164 | %endif |
| 1165 | movd m2, wxm ; WX |
| 1166 | movd m4, SHIFT ; shift |
| 1167 | %if %1 <= 4 |
| 1168 | punpcklwd m2, m1 |
| 1169 | %else |
| 1170 | punpcklwd m2, m2 |
| 1171 | %endif |
| 1172 | dec SHIFT |
| 1173 | movdqu m5, [one_per_32] |
| 1174 | movd m6, SHIFT |
| 1175 | pshufd m2, m2, 0 |
| 1176 | mov SHIFT, oxm |
| 1177 | pslld m5, m6 |
| 1178 | %if %2 != 8 |
| 1179 | shl SHIFT, %2-8 ; ox << (bitd - 8) |
| 1180 | %endif |
| 1181 | movd m3, SHIFT ; OX |
| 1182 | pshufd m3, m3, 0 |
| 1183 | %if WIN64 || ARCH_X86_32 |
| 1184 | mov SHIFT, heightm |
| 1185 | %endif |
| 1186 | .loop |
| 1187 | SIMPLE_LOAD %1, 10, srcq, m0 |
| 1188 | %if %1 <= 4 |
| 1189 | punpcklwd m0, m1 |
| 1190 | pmaddwd m0, m2 |
| 1191 | paddd m0, m5 |
| 1192 | psrad m0, m4 |
| 1193 | paddd m0, m3 |
| 1194 | %else |
| 1195 | pmulhw m6, m0, m2 |
| 1196 | pmullw m0, m2 |
| 1197 | punpckhwd m1, m0, m6 |
| 1198 | punpcklwd m0, m6 |
| 1199 | paddd m0, m5 |
| 1200 | paddd m1, m5 |
| 1201 | psrad m0, m4 |
| 1202 | psrad m1, m4 |
| 1203 | paddd m0, m3 |
| 1204 | paddd m1, m3 |
| 1205 | %endif |
| 1206 | packusdw m0, m1 |
| 1207 | %if %2 == 8 |
| 1208 | packuswb m0, m0 |
| 1209 | %else |
| 1210 | pminsw m0, [max_pixels_%2] |
| 1211 | %endif |
| 1212 | PEL_%2STORE%1 dstq, m0, m1 |
| 1213 | add dstq, dststrideq ; dst += dststride |
| 1214 | add srcq, 2*MAX_PB_SIZE ; src += srcstride |
| 1215 | dec heightd ; cmp height |
| 1216 | jnz .loop ; height loop |
| 1217 | RET |
| 1218 | |
| 1219 | cglobal hevc_put_hevc_bi_w%1_%2, 5, 7, 10, dst, dststride, src, srcstride, src2, height, denom, wx0, wx1, ox0, ox1 |
| 1220 | mov r6d, denomm |
| 1221 | %if %1 <= 4 |
| 1222 | pxor m1, m1 |
| 1223 | %endif |
| 1224 | movd m2, wx0m ; WX0 |
| 1225 | lea r6d, [r6d+14-%2] ; shift = 14 - bitd + denom |
| 1226 | movd m3, wx1m ; WX1 |
| 1227 | movd m0, r6d ; shift |
| 1228 | %if %1 <= 4 |
| 1229 | punpcklwd m2, m1 |
| 1230 | punpcklwd m3, m1 |
| 1231 | %else |
| 1232 | punpcklwd m2, m2 |
| 1233 | punpcklwd m3, m3 |
| 1234 | %endif |
| 1235 | inc r6d |
| 1236 | movd m5, r6d ; shift+1 |
| 1237 | pshufd m2, m2, 0 |
| 1238 | mov r6d, ox0m |
| 1239 | pshufd m3, m3, 0 |
| 1240 | add r6d, ox1m |
| 1241 | %if %2 != 8 |
| 1242 | shl r6d, %2-8 ; ox << (bitd - 8) |
| 1243 | %endif |
| 1244 | inc r6d |
| 1245 | movd m4, r6d ; offset |
| 1246 | pshufd m4, m4, 0 |
| 1247 | mov r6d, heightm |
| 1248 | pslld m4, m0 |
| 1249 | |
| 1250 | .loop |
| 1251 | SIMPLE_LOAD %1, 10, srcq, m0 |
| 1252 | SIMPLE_LOAD %1, 10, src2q, m8 |
| 1253 | %if %1 <= 4 |
| 1254 | punpcklwd m0, m1 |
| 1255 | punpcklwd m8, m1 |
| 1256 | pmaddwd m0, m3 |
| 1257 | pmaddwd m8, m2 |
| 1258 | paddd m0, m4 |
| 1259 | paddd m0, m8 |
| 1260 | psrad m0, m5 |
| 1261 | %else |
| 1262 | pmulhw m6, m0, m3 |
| 1263 | pmullw m0, m3 |
| 1264 | pmulhw m7, m8, m2 |
| 1265 | pmullw m8, m2 |
| 1266 | punpckhwd m1, m0, m6 |
| 1267 | punpcklwd m0, m6 |
| 1268 | punpckhwd m9, m8, m7 |
| 1269 | punpcklwd m8, m7 |
| 1270 | paddd m0, m8 |
| 1271 | paddd m1, m9 |
| 1272 | paddd m0, m4 |
| 1273 | paddd m1, m4 |
| 1274 | psrad m0, m5 |
| 1275 | psrad m1, m5 |
| 1276 | %endif |
| 1277 | packusdw m0, m1 |
| 1278 | %if %2 == 8 |
| 1279 | packuswb m0, m0 |
| 1280 | %else |
| 1281 | pminsw m0, [max_pixels_%2] |
| 1282 | %endif |
| 1283 | PEL_%2STORE%1 dstq, m0, m1 |
| 1284 | add dstq, dststrideq ; dst += dststride |
| 1285 | add srcq, 2*MAX_PB_SIZE ; src += srcstride |
| 1286 | add src2q, 2*MAX_PB_SIZE ; src2 += srcstride |
| 1287 | dec r6d ; cmp height |
| 1288 | jnz .loop ; height loop |
| 1289 | RET |
| 1290 | %endmacro |
| 1291 | |
| 1292 | WEIGHTING_FUNCS 2, 8 |
| 1293 | WEIGHTING_FUNCS 4, 8 |
| 1294 | WEIGHTING_FUNCS 6, 8 |
| 1295 | WEIGHTING_FUNCS 8, 8 |
| 1296 | |
| 1297 | WEIGHTING_FUNCS 2, 10 |
| 1298 | WEIGHTING_FUNCS 4, 10 |
| 1299 | WEIGHTING_FUNCS 6, 10 |
| 1300 | WEIGHTING_FUNCS 8, 10 |
| 1301 | |
| 1302 | WEIGHTING_FUNCS 2, 12 |
| 1303 | WEIGHTING_FUNCS 4, 12 |
| 1304 | WEIGHTING_FUNCS 6, 12 |
| 1305 | WEIGHTING_FUNCS 8, 12 |
| 1306 | |
| 1307 | HEVC_PUT_HEVC_PEL_PIXELS 2, 8 |
| 1308 | HEVC_PUT_HEVC_PEL_PIXELS 4, 8 |
| 1309 | HEVC_PUT_HEVC_PEL_PIXELS 6, 8 |
| 1310 | HEVC_PUT_HEVC_PEL_PIXELS 8, 8 |
| 1311 | HEVC_PUT_HEVC_PEL_PIXELS 12, 8 |
| 1312 | HEVC_PUT_HEVC_PEL_PIXELS 16, 8 |
| 1313 | |
| 1314 | HEVC_PUT_HEVC_PEL_PIXELS 2, 10 |
| 1315 | HEVC_PUT_HEVC_PEL_PIXELS 4, 10 |
| 1316 | HEVC_PUT_HEVC_PEL_PIXELS 6, 10 |
| 1317 | HEVC_PUT_HEVC_PEL_PIXELS 8, 10 |
| 1318 | |
| 1319 | HEVC_PUT_HEVC_PEL_PIXELS 2, 12 |
| 1320 | HEVC_PUT_HEVC_PEL_PIXELS 4, 12 |
| 1321 | HEVC_PUT_HEVC_PEL_PIXELS 6, 12 |
| 1322 | HEVC_PUT_HEVC_PEL_PIXELS 8, 12 |
| 1323 | |
| 1324 | HEVC_PUT_HEVC_EPEL 2, 8 |
| 1325 | HEVC_PUT_HEVC_EPEL 4, 8 |
| 1326 | HEVC_PUT_HEVC_EPEL 6, 8 |
| 1327 | HEVC_PUT_HEVC_EPEL 8, 8 |
| 1328 | HEVC_PUT_HEVC_EPEL 12, 8 |
| 1329 | HEVC_PUT_HEVC_EPEL 16, 8 |
| 1330 | |
| 1331 | |
| 1332 | HEVC_PUT_HEVC_EPEL 2, 10 |
| 1333 | HEVC_PUT_HEVC_EPEL 4, 10 |
| 1334 | HEVC_PUT_HEVC_EPEL 6, 10 |
| 1335 | HEVC_PUT_HEVC_EPEL 8, 10 |
| 1336 | |
| 1337 | HEVC_PUT_HEVC_EPEL 2, 12 |
| 1338 | HEVC_PUT_HEVC_EPEL 4, 12 |
| 1339 | HEVC_PUT_HEVC_EPEL 6, 12 |
| 1340 | HEVC_PUT_HEVC_EPEL 8, 12 |
| 1341 | |
| 1342 | HEVC_PUT_HEVC_EPEL_HV 2, 8 |
| 1343 | HEVC_PUT_HEVC_EPEL_HV 4, 8 |
| 1344 | HEVC_PUT_HEVC_EPEL_HV 6, 8 |
| 1345 | HEVC_PUT_HEVC_EPEL_HV 8, 8 |
| 1346 | |
| 1347 | HEVC_PUT_HEVC_EPEL_HV 2, 10 |
| 1348 | HEVC_PUT_HEVC_EPEL_HV 4, 10 |
| 1349 | HEVC_PUT_HEVC_EPEL_HV 6, 10 |
| 1350 | HEVC_PUT_HEVC_EPEL_HV 8, 10 |
| 1351 | |
| 1352 | HEVC_PUT_HEVC_EPEL_HV 2, 12 |
| 1353 | HEVC_PUT_HEVC_EPEL_HV 4, 12 |
| 1354 | HEVC_PUT_HEVC_EPEL_HV 6, 12 |
| 1355 | HEVC_PUT_HEVC_EPEL_HV 8, 12 |
| 1356 | |
| 1357 | HEVC_PUT_HEVC_QPEL 4, 8 |
| 1358 | HEVC_PUT_HEVC_QPEL 8, 8 |
| 1359 | HEVC_PUT_HEVC_QPEL 12, 8 |
| 1360 | HEVC_PUT_HEVC_QPEL 16, 8 |
| 1361 | |
| 1362 | HEVC_PUT_HEVC_QPEL 4, 10 |
| 1363 | HEVC_PUT_HEVC_QPEL 8, 10 |
| 1364 | |
| 1365 | HEVC_PUT_HEVC_QPEL 4, 12 |
| 1366 | HEVC_PUT_HEVC_QPEL 8, 12 |
| 1367 | |
| 1368 | HEVC_PUT_HEVC_QPEL_HV 2, 8 |
| 1369 | HEVC_PUT_HEVC_QPEL_HV 4, 8 |
| 1370 | HEVC_PUT_HEVC_QPEL_HV 6, 8 |
| 1371 | HEVC_PUT_HEVC_QPEL_HV 8, 8 |
| 1372 | |
| 1373 | HEVC_PUT_HEVC_QPEL_HV 2, 10 |
| 1374 | HEVC_PUT_HEVC_QPEL_HV 4, 10 |
| 1375 | HEVC_PUT_HEVC_QPEL_HV 6, 10 |
| 1376 | HEVC_PUT_HEVC_QPEL_HV 8, 10 |
| 1377 | |
| 1378 | HEVC_PUT_HEVC_QPEL_HV 2, 12 |
| 1379 | HEVC_PUT_HEVC_QPEL_HV 4, 12 |
| 1380 | HEVC_PUT_HEVC_QPEL_HV 6, 12 |
| 1381 | HEVC_PUT_HEVC_QPEL_HV 8, 12 |
| 1382 | |
| 1383 | %endif ; ARCH_X86_64 |