| @@ -0,0 +1,533 @@ |
| 1 | +/* |
| 2 | + * Copyright (c) 2014 RISC OS Open Ltd |
| 3 | + * Author: Ben Avison <bavison@riscosopen.org> |
| 4 | + * |
| 5 | + * This file is part of Libav. |
| 6 | + * |
| 7 | + * Libav 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 | + * Libav 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 Libav; if not, write to the Free Software |
| 19 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | + */ |
| 21 | + |
| 22 | +#include "libavutil/arm/asm.S" |
| 23 | + |
| 24 | +.macro loadregoffsh2 group, index, base, offgroup, offindex |
| 25 | + .altmacro |
| 26 | + loadregoffsh2_ \group, %(\index), \base, \offgroup, %(\offindex) |
| 27 | + .noaltmacro |
| 28 | +.endm |
| 29 | + |
| 30 | +.macro loadregoffsh2_ group, index, base, offgroup, offindex |
| 31 | + ldr \group\index, [\base, \offgroup\offindex, lsl #2] |
| 32 | +.endm |
| 33 | + |
| 34 | +.macro eorlslreg check, data, group, index |
| 35 | + .altmacro |
| 36 | + eorlslreg_ \check, \data, \group, %(\index) |
| 37 | + .noaltmacro |
| 38 | +.endm |
| 39 | + |
| 40 | +.macro eorlslreg_ check, data, group, index |
| 41 | + eor \check, \check, \data, lsl \group\index |
| 42 | +.endm |
| 43 | + |
| 44 | +.macro decr_modulo var, by, modulus |
| 45 | + .set \var, \var - \by |
| 46 | + .if \var == 0 |
| 47 | + .set \var, \modulus |
| 48 | + .endif |
| 49 | +.endm |
| 50 | + |
| 51 | + .macro load_group1 size, channels, r0, r1, r2, r3, pointer_dead=0 |
| 52 | + .if \size == 2 |
| 53 | + ldrd \r0, \r1, [IN], #(\size + 8 - \channels) * 4 |
| 54 | + .else // size == 4 |
| 55 | + .if IDX1 > 4 || \channels==8 |
| 56 | + ldm IN!, {\r0, \r1, \r2, \r3} |
| 57 | + .else |
| 58 | + ldm IN, {\r0, \r1, \r2, \r3} |
| 59 | + .if !\pointer_dead |
| 60 | + add IN, IN, #(4 + 8 - \channels) * 4 |
| 61 | + .endif |
| 62 | + .endif |
| 63 | + .endif |
| 64 | + decr_modulo IDX1, \size, \channels |
| 65 | + .endm |
| 66 | + |
| 67 | + .macro load_group2 size, channels, r0, r1, r2, r3, pointer_dead=0 |
| 68 | + .if \size == 2 |
| 69 | + .if IDX1 > 2 |
| 70 | + ldm IN!, {\r2, \r3} |
| 71 | + .else |
| 72 | +//A .ifc \r2, ip |
| 73 | +//A .if \pointer_dead |
| 74 | +//A ldm IN, {\r2, \r3} |
| 75 | +//A .else |
| 76 | +//A ldr \r2, [IN], #4 |
| 77 | +//A ldr \r3, [IN], #(\size - 1 + 8 - \channels) * 4 |
| 78 | +//A .endif |
| 79 | +//A .else |
| 80 | + ldrd \r2, \r3, [IN], #(\size + 8 - \channels) * 4 |
| 81 | +//A .endif |
| 82 | + .endif |
| 83 | + .endif |
| 84 | + decr_modulo IDX1, \size, \channels |
| 85 | + .endm |
| 86 | + |
| 87 | +.macro implement_pack inorder, channels, shift |
| 88 | +.if \inorder |
| 89 | +.ifc \shift, mixed |
| 90 | + |
| 91 | +CHECK .req a1 |
| 92 | +COUNT .req a2 |
| 93 | +IN .req a3 |
| 94 | +OUT .req a4 |
| 95 | +DAT0 .req v1 |
| 96 | +DAT1 .req v2 |
| 97 | +DAT2 .req v3 |
| 98 | +DAT3 .req v4 |
| 99 | +SHIFT0 .req v5 |
| 100 | +SHIFT1 .req v6 |
| 101 | +SHIFT2 .req sl |
| 102 | +SHIFT3 .req fp |
| 103 | +SHIFT4 .req ip |
| 104 | +SHIFT5 .req lr |
| 105 | + |
| 106 | + .macro output4words |
| 107 | + .set SIZE_GROUP1, IDX1 |
| 108 | + .if SIZE_GROUP1 > 4 |
| 109 | + .set SIZE_GROUP1, 4 |
| 110 | + .endif |
| 111 | + .set SIZE_GROUP2, 4 - SIZE_GROUP1 |
| 112 | + load_group1 SIZE_GROUP1, \channels, DAT0, DAT1, DAT2, DAT3 |
| 113 | + load_group2 SIZE_GROUP2, \channels, DAT0, DAT1, DAT2, DAT3 |
| 114 | + .if \channels == 2 |
| 115 | + lsl DAT0, SHIFT0 |
| 116 | + lsl DAT1, SHIFT1 |
| 117 | + lsl DAT2, SHIFT0 |
| 118 | + lsl DAT3, SHIFT1 |
| 119 | + .elseif \channels == 6 |
| 120 | + .if IDX2 == 6 |
| 121 | + lsl DAT0, SHIFT0 |
| 122 | + lsl DAT1, SHIFT1 |
| 123 | + lsl DAT2, SHIFT2 |
| 124 | + lsl DAT3, SHIFT3 |
| 125 | + .elseif IDX2 == 2 |
| 126 | + lsl DAT0, SHIFT4 |
| 127 | + lsl DAT1, SHIFT5 |
| 128 | + lsl DAT2, SHIFT0 |
| 129 | + lsl DAT3, SHIFT1 |
| 130 | + .else // IDX2 == 4 |
| 131 | + lsl DAT0, SHIFT2 |
| 132 | + lsl DAT1, SHIFT3 |
| 133 | + lsl DAT2, SHIFT4 |
| 134 | + lsl DAT3, SHIFT5 |
| 135 | + .endif |
| 136 | + .elseif \channels == 8 |
| 137 | + .if IDX2 == 8 |
| 138 | + uxtb SHIFT0, SHIFT4, ror #0 |
| 139 | + uxtb SHIFT1, SHIFT4, ror #8 |
| 140 | + uxtb SHIFT2, SHIFT4, ror #16 |
| 141 | + uxtb SHIFT3, SHIFT4, ror #24 |
| 142 | + .else |
| 143 | + uxtb SHIFT0, SHIFT5, ror #0 |
| 144 | + uxtb SHIFT1, SHIFT5, ror #8 |
| 145 | + uxtb SHIFT2, SHIFT5, ror #16 |
| 146 | + uxtb SHIFT3, SHIFT5, ror #24 |
| 147 | + .endif |
| 148 | + lsl DAT0, SHIFT0 |
| 149 | + lsl DAT1, SHIFT1 |
| 150 | + lsl DAT2, SHIFT2 |
| 151 | + lsl DAT3, SHIFT3 |
| 152 | + .endif |
| 153 | + eor CHECK, CHECK, DAT0, lsr #8 - (\channels - IDX2) |
| 154 | + eor CHECK, CHECK, DAT1, lsr #7 - (\channels - IDX2) |
| 155 | + decr_modulo IDX2, 2, \channels |
| 156 | + eor CHECK, CHECK, DAT2, lsr #8 - (\channels - IDX2) |
| 157 | + eor CHECK, CHECK, DAT3, lsr #7 - (\channels - IDX2) |
| 158 | + decr_modulo IDX2, 2, \channels |
| 159 | + stm OUT!, {DAT0 - DAT3} |
| 160 | + .endm |
| 161 | + |
| 162 | + .set WORDS_PER_LOOP, \channels // calculate LCM (channels, 4) |
| 163 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 164 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 165 | + .endif |
| 166 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 167 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 168 | + .endif |
| 169 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP * 4 |
| 170 | + .set SAMPLES_PER_LOOP, WORDS_PER_LOOP / \channels |
| 171 | + |
| 172 | +function ff_mlp_pack_output_inorder_\channels\()ch_mixedshift_armv6, export=1 |
| 173 | + .if SAMPLES_PER_LOOP > 1 |
| 174 | + tst COUNT, #SAMPLES_PER_LOOP - 1 // always seems to be in practice |
| 175 | + it ne |
| 176 | + bne X(ff_mlp_pack_output) // but just in case, branch to C implementation if not |
| 177 | + .endif |
| 178 | + teq COUNT, #0 |
| 179 | + it eq |
| 180 | + bxeq lr |
| 181 | + push {v1-v6,sl,fp,lr} |
| 182 | + ldr SHIFT0, [sp, #(9+1)*4] // get output_shift from stack |
| 183 | + ldr SHIFT1, =0x08080808 |
| 184 | + ldr SHIFT4, [SHIFT0] |
| 185 | + .if \channels == 2 |
| 186 | + uadd8 SHIFT4, SHIFT4, SHIFT1 // increase all shifts by 8 |
| 187 | + uxtb SHIFT0, SHIFT4, ror #0 |
| 188 | + uxtb SHIFT1, SHIFT4, ror #8 |
| 189 | + .else |
| 190 | + ldr SHIFT5, [SHIFT0, #4] |
| 191 | + uadd8 SHIFT4, SHIFT4, SHIFT1 // increase all shifts by 8 |
| 192 | + uadd8 SHIFT5, SHIFT5, SHIFT1 |
| 193 | + .if \channels == 6 |
| 194 | + uxtb SHIFT0, SHIFT4, ror #0 |
| 195 | + uxtb SHIFT1, SHIFT4, ror #8 |
| 196 | + uxtb SHIFT2, SHIFT4, ror #16 |
| 197 | + uxtb SHIFT3, SHIFT4, ror #24 |
| 198 | + uxtb SHIFT4, SHIFT5, ror #0 |
| 199 | + uxtb SHIFT5, SHIFT5, ror #8 |
| 200 | + .endif |
| 201 | + .endif |
| 202 | + .set IDX1, \channels |
| 203 | + .set IDX2, \channels |
| 204 | +0: |
| 205 | + .rept WORDS_PER_LOOP / 4 |
| 206 | + output4words |
| 207 | + .endr |
| 208 | + subs COUNT, COUNT, #SAMPLES_PER_LOOP |
| 209 | + bne 0b |
| 210 | + pop {v1-v6,sl,fp,pc} |
| 211 | + .ltorg |
| 212 | +endfunc |
| 213 | + .purgem output4words |
| 214 | + |
| 215 | + .unreq CHECK |
| 216 | + .unreq COUNT |
| 217 | + .unreq IN |
| 218 | + .unreq OUT |
| 219 | + .unreq DAT0 |
| 220 | + .unreq DAT1 |
| 221 | + .unreq DAT2 |
| 222 | + .unreq DAT3 |
| 223 | + .unreq SHIFT0 |
| 224 | + .unreq SHIFT1 |
| 225 | + .unreq SHIFT2 |
| 226 | + .unreq SHIFT3 |
| 227 | + .unreq SHIFT4 |
| 228 | + .unreq SHIFT5 |
| 229 | + |
| 230 | +.else // not mixed |
| 231 | + |
| 232 | +CHECK .req a1 |
| 233 | +COUNT .req a2 |
| 234 | +IN .req a3 |
| 235 | +OUT .req a4 |
| 236 | +DAT0 .req v1 |
| 237 | +DAT1 .req v2 |
| 238 | +DAT2 .req v3 |
| 239 | +DAT3 .req v4 |
| 240 | +DAT4 .req v5 |
| 241 | +DAT5 .req v6 |
| 242 | +DAT6 .req sl // use these rather than the otherwise unused |
| 243 | +DAT7 .req fp // ip and lr so that we can load them usinf LDRD |
| 244 | + |
| 245 | + .macro output4words tail, head, r0, r1, r2, r3, r4, r5, r6, r7, pointer_dead=0 |
| 246 | + .if \head |
| 247 | + .set SIZE_GROUP1, IDX1 |
| 248 | + .if SIZE_GROUP1 > 4 |
| 249 | + .set SIZE_GROUP1, 4 |
| 250 | + .endif |
| 251 | + .set SIZE_GROUP2, 4 - SIZE_GROUP1 |
| 252 | + load_group1 SIZE_GROUP1, \channels, \r0, \r1, \r2, \r3, \pointer_dead |
| 253 | + .endif |
| 254 | + .if \tail |
| 255 | + eor CHECK, CHECK, \r4, lsr #8 - (\channels - IDX2) |
| 256 | + eor CHECK, CHECK, \r5, lsr #7 - (\channels - IDX2) |
| 257 | + decr_modulo IDX2, 2, \channels |
| 258 | + .endif |
| 259 | + .if \head |
| 260 | + load_group2 SIZE_GROUP2, \channels, \r0, \r1, \r2, \r3, \pointer_dead |
| 261 | + .endif |
| 262 | + .if \tail |
| 263 | + eor CHECK, CHECK, \r6, lsr #8 - (\channels - IDX2) |
| 264 | + eor CHECK, CHECK, \r7, lsr #7 - (\channels - IDX2) |
| 265 | + decr_modulo IDX2, 2, \channels |
| 266 | + stm OUT!, {\r4, \r5, \r6, \r7} |
| 267 | + .endif |
| 268 | + .if \head |
| 269 | + lsl \r0, #8 + \shift |
| 270 | + lsl \r1, #8 + \shift |
| 271 | + lsl \r2, #8 + \shift |
| 272 | + lsl \r3, #8 + \shift |
| 273 | + .endif |
| 274 | + .endm |
| 275 | + |
| 276 | + .set WORDS_PER_LOOP, \channels // calculate LCM (channels, 8) |
| 277 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 278 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 279 | + .endif |
| 280 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 281 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 282 | + .endif |
| 283 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 284 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 285 | + .endif |
| 286 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP * 8 |
| 287 | + .set SAMPLES_PER_LOOP, WORDS_PER_LOOP / \channels |
| 288 | + |
| 289 | +function ff_mlp_pack_output_inorder_\channels\()ch_\shift\()shift_armv6, export=1 |
| 290 | + .if SAMPLES_PER_LOOP > 1 |
| 291 | + tst COUNT, #SAMPLES_PER_LOOP - 1 // always seems to be in practice |
| 292 | + it ne |
| 293 | + bne X(ff_mlp_pack_output) // but just in case, branch to C implementation if not |
| 294 | + .endif |
| 295 | + subs COUNT, COUNT, #SAMPLES_PER_LOOP |
| 296 | + it lo |
| 297 | + bxlo lr |
| 298 | + push {v1-v6,sl,fp,lr} |
| 299 | + .set IDX1, \channels |
| 300 | + .set IDX2, \channels |
| 301 | + output4words 0, 1, DAT0, DAT1, DAT2, DAT3, DAT4, DAT5, DAT6, DAT7 |
| 302 | +0: beq 1f |
| 303 | + .rept WORDS_PER_LOOP / 8 |
| 304 | + output4words 1, 1, DAT4, DAT5, DAT6, DAT7, DAT0, DAT1, DAT2, DAT3 |
| 305 | + output4words 1, 1, DAT0, DAT1, DAT2, DAT3, DAT4, DAT5, DAT6, DAT7 |
| 306 | + .endr |
| 307 | + subs COUNT, COUNT, #SAMPLES_PER_LOOP |
| 308 | + bne 0b |
| 309 | +1: |
| 310 | + .rept WORDS_PER_LOOP / 8 - 1 |
| 311 | + output4words 1, 1, DAT4, DAT5, DAT6, DAT7, DAT0, DAT1, DAT2, DAT3 |
| 312 | + output4words 1, 1, DAT0, DAT1, DAT2, DAT3, DAT4, DAT5, DAT6, DAT7 |
| 313 | + .endr |
| 314 | + output4words 1, 1, DAT4, DAT5, DAT6, DAT7, DAT0, DAT1, DAT2, DAT3, pointer_dead=1 |
| 315 | + output4words 1, 0, DAT0, DAT1, DAT2, DAT3, DAT4, DAT5, DAT6, DAT7 |
| 316 | + pop {v1-v6,sl,fp,pc} |
| 317 | +endfunc |
| 318 | + .purgem output4words |
| 319 | + |
| 320 | + .unreq CHECK |
| 321 | + .unreq COUNT |
| 322 | + .unreq IN |
| 323 | + .unreq OUT |
| 324 | + .unreq DAT0 |
| 325 | + .unreq DAT1 |
| 326 | + .unreq DAT2 |
| 327 | + .unreq DAT3 |
| 328 | + .unreq DAT4 |
| 329 | + .unreq DAT5 |
| 330 | + .unreq DAT6 |
| 331 | + .unreq DAT7 |
| 332 | + |
| 333 | +.endif // mixed |
| 334 | +.else // not inorder |
| 335 | +.ifc \shift, mixed |
| 336 | + |
| 337 | +// This case not currently handled |
| 338 | + |
| 339 | +.else // not mixed |
| 340 | + |
| 341 | +#if !CONFIG_THUMB |
| 342 | + |
| 343 | +CHECK .req a1 |
| 344 | +COUNT .req a2 |
| 345 | +IN .req a3 |
| 346 | +OUT .req a4 |
| 347 | +DAT0 .req v1 |
| 348 | +DAT1 .req v2 |
| 349 | +DAT2 .req v3 |
| 350 | +DAT3 .req v4 |
| 351 | +CHAN0 .req v5 |
| 352 | +CHAN1 .req v6 |
| 353 | +CHAN2 .req sl |
| 354 | +CHAN3 .req fp |
| 355 | +CHAN4 .req ip |
| 356 | +CHAN5 .req lr |
| 357 | + |
| 358 | + .macro output4words |
| 359 | + .if \channels == 8 |
| 360 | + .if IDX1 == 8 |
| 361 | + uxtb CHAN0, CHAN4, ror #0 |
| 362 | + uxtb CHAN1, CHAN4, ror #8 |
| 363 | + uxtb CHAN2, CHAN4, ror #16 |
| 364 | + uxtb CHAN3, CHAN4, ror #24 |
| 365 | + .else |
| 366 | + uxtb CHAN0, CHAN5, ror #0 |
| 367 | + uxtb CHAN1, CHAN5, ror #8 |
| 368 | + uxtb CHAN2, CHAN5, ror #16 |
| 369 | + uxtb CHAN3, CHAN5, ror #24 |
| 370 | + .endif |
| 371 | + ldr DAT0, [IN, CHAN0, lsl #2] |
| 372 | + ldr DAT1, [IN, CHAN1, lsl #2] |
| 373 | + ldr DAT2, [IN, CHAN2, lsl #2] |
| 374 | + ldr DAT3, [IN, CHAN3, lsl #2] |
| 375 | + .if IDX1 == 4 |
| 376 | + add IN, IN, #8*4 |
| 377 | + .endif |
| 378 | + decr_modulo IDX1, 4, \channels |
| 379 | + .else |
| 380 | + .set SIZE_GROUP1, IDX1 |
| 381 | + .if SIZE_GROUP1 > 4 |
| 382 | + .set SIZE_GROUP1, 4 |
| 383 | + .endif |
| 384 | + .set SIZE_GROUP2, 4 - SIZE_GROUP1 |
| 385 | + .if SIZE_GROUP1 == 2 |
| 386 | + loadregoffsh2 DAT, 0, IN, CHAN, 0 + (\channels - IDX1) |
| 387 | + loadregoffsh2 DAT, 1, IN, CHAN, 1 + (\channels - IDX1) |
| 388 | + add IN, IN, #8*4 |
| 389 | + .else // SIZE_GROUP1 == 4 |
| 390 | + loadregoffsh2 DAT, 0, IN, CHAN, 0 + (\channels - IDX1) |
| 391 | + loadregoffsh2 DAT, 1, IN, CHAN, 1 + (\channels - IDX1) |
| 392 | + loadregoffsh2 DAT, 2, IN, CHAN, 2 + (\channels - IDX1) |
| 393 | + loadregoffsh2 DAT, 3, IN, CHAN, 3 + (\channels - IDX1) |
| 394 | + .if IDX1 == 4 |
| 395 | + add IN, IN, #8*4 |
| 396 | + .endif |
| 397 | + .endif |
| 398 | + decr_modulo IDX1, SIZE_GROUP1, \channels |
| 399 | + .if SIZE_GROUP2 == 2 |
| 400 | + loadregoffsh2 DAT, 2, IN, CHAN, 0 + (\channels - IDX1) |
| 401 | + loadregoffsh2 DAT, 3, IN, CHAN, 1 + (\channels - IDX1) |
| 402 | + .if IDX1 == 2 |
| 403 | + add IN, IN, #8*4 |
| 404 | + .endif |
| 405 | + .endif |
| 406 | + decr_modulo IDX1, SIZE_GROUP2, \channels |
| 407 | + .endif |
| 408 | + .if \channels == 8 // in this case we can corrupt CHAN0-3 |
| 409 | + rsb CHAN0, CHAN0, #8 |
| 410 | + rsb CHAN1, CHAN1, #8 |
| 411 | + rsb CHAN2, CHAN2, #8 |
| 412 | + rsb CHAN3, CHAN3, #8 |
| 413 | + lsl DAT0, #8 + \shift |
| 414 | + lsl DAT1, #8 + \shift |
| 415 | + lsl DAT2, #8 + \shift |
| 416 | + lsl DAT3, #8 + \shift |
| 417 | + eor CHECK, CHECK, DAT0, lsr CHAN0 |
| 418 | + eor CHECK, CHECK, DAT1, lsr CHAN1 |
| 419 | + eor CHECK, CHECK, DAT2, lsr CHAN2 |
| 420 | + eor CHECK, CHECK, DAT3, lsr CHAN3 |
| 421 | + .else |
| 422 | + .if \shift != 0 |
| 423 | + lsl DAT0, #\shift |
| 424 | + lsl DAT1, #\shift |
| 425 | + lsl DAT2, #\shift |
| 426 | + lsl DAT3, #\shift |
| 427 | + .endif |
| 428 | + bic DAT0, DAT0, #0xff000000 |
| 429 | + bic DAT1, DAT1, #0xff000000 |
| 430 | + bic DAT2, DAT2, #0xff000000 |
| 431 | + bic DAT3, DAT3, #0xff000000 |
| 432 | + eorlslreg CHECK, DAT0, CHAN, 0 + (\channels - IDX2) |
| 433 | + eorlslreg CHECK, DAT1, CHAN, 1 + (\channels - IDX2) |
| 434 | + decr_modulo IDX2, 2, \channels |
| 435 | + eorlslreg CHECK, DAT2, CHAN, 0 + (\channels - IDX2) |
| 436 | + eorlslreg CHECK, DAT3, CHAN, 1 + (\channels - IDX2) |
| 437 | + decr_modulo IDX2, 2, \channels |
| 438 | + lsl DAT0, #8 |
| 439 | + lsl DAT1, #8 |
| 440 | + lsl DAT2, #8 |
| 441 | + lsl DAT3, #8 |
| 442 | + .endif |
| 443 | + stm OUT!, {DAT0 - DAT3} |
| 444 | + .endm |
| 445 | + |
| 446 | + .set WORDS_PER_LOOP, \channels // calculate LCM (channels, 4) |
| 447 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 448 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 449 | + .endif |
| 450 | + .if (WORDS_PER_LOOP % 2) == 0 |
| 451 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP / 2 |
| 452 | + .endif |
| 453 | + .set WORDS_PER_LOOP, WORDS_PER_LOOP * 4 |
| 454 | + .set SAMPLES_PER_LOOP, WORDS_PER_LOOP / \channels |
| 455 | + |
| 456 | +function ff_mlp_pack_output_outoforder_\channels\()ch_\shift\()shift_armv6, export=1 |
| 457 | + .if SAMPLES_PER_LOOP > 1 |
| 458 | + tst COUNT, #SAMPLES_PER_LOOP - 1 // always seems to be in practice |
| 459 | + it ne |
| 460 | + bne X(ff_mlp_pack_output) // but just in case, branch to C implementation if not |
| 461 | + .endif |
| 462 | + teq COUNT, #0 |
| 463 | + it eq |
| 464 | + bxeq lr |
| 465 | + push {v1-v6,sl,fp,lr} |
| 466 | + ldr CHAN0, [sp, #(9+0)*4] // get ch_assign from stack |
| 467 | + ldr CHAN4, [CHAN0] |
| 468 | + .if \channels == 2 |
| 469 | + uxtb CHAN0, CHAN4, ror #0 |
| 470 | + uxtb CHAN1, CHAN4, ror #8 |
| 471 | + .else |
| 472 | + ldr CHAN5, [CHAN0, #4] |
| 473 | + .if \channels == 6 |
| 474 | + uxtb CHAN0, CHAN4, ror #0 |
| 475 | + uxtb CHAN1, CHAN4, ror #8 |
| 476 | + uxtb CHAN2, CHAN4, ror #16 |
| 477 | + uxtb CHAN3, CHAN4, ror #24 |
| 478 | + uxtb CHAN4, CHAN5, ror #0 |
| 479 | + uxtb CHAN5, CHAN5, ror #8 |
| 480 | + .endif |
| 481 | + .endif |
| 482 | + .set IDX1, \channels |
| 483 | + .set IDX2, \channels |
| 484 | +0: |
| 485 | + .rept WORDS_PER_LOOP / 4 |
| 486 | + output4words |
| 487 | + .endr |
| 488 | + subs COUNT, COUNT, #SAMPLES_PER_LOOP |
| 489 | + bne 0b |
| 490 | + pop {v1-v6,sl,fp,pc} |
| 491 | + .ltorg |
| 492 | +endfunc |
| 493 | + .purgem output4words |
| 494 | + |
| 495 | + .unreq CHECK |
| 496 | + .unreq COUNT |
| 497 | + .unreq IN |
| 498 | + .unreq OUT |
| 499 | + .unreq DAT0 |
| 500 | + .unreq DAT1 |
| 501 | + .unreq DAT2 |
| 502 | + .unreq DAT3 |
| 503 | + .unreq CHAN0 |
| 504 | + .unreq CHAN1 |
| 505 | + .unreq CHAN2 |
| 506 | + .unreq CHAN3 |
| 507 | + .unreq CHAN4 |
| 508 | + .unreq CHAN5 |
| 509 | + |
| 510 | +#endif // !CONFIG_THUMB |
| 511 | + |
| 512 | +.endif // mixed |
| 513 | +.endif // inorder |
| 514 | +.endm // implement_pack |
| 515 | + |
| 516 | +.macro pack_channels inorder, channels |
| 517 | + implement_pack \inorder, \channels, 0 |
| 518 | + implement_pack \inorder, \channels, 1 |
| 519 | + implement_pack \inorder, \channels, 2 |
| 520 | + implement_pack \inorder, \channels, 3 |
| 521 | + implement_pack \inorder, \channels, 4 |
| 522 | + implement_pack \inorder, \channels, 5 |
| 523 | + implement_pack \inorder, \channels, mixed |
| 524 | +.endm |
| 525 | + |
| 526 | +.macro pack_order inorder |
| 527 | + pack_channels \inorder, 2 |
| 528 | + pack_channels \inorder, 6 |
| 529 | + pack_channels \inorder, 8 |
| 530 | +.endm |
| 531 | + |
| 532 | + pack_order 0 |
| 533 | + pack_order 1 |