| 1 | /* |
| 2 | * Copyright (c) 2012 |
| 3 | * MIPS Technologies, Inc., California. |
| 4 | * |
| 5 | * Redistribution and use in source and binary forms, with or without |
| 6 | * modification, are permitted provided that the following conditions |
| 7 | * are met: |
| 8 | * 1. Redistributions of source code must retain the above copyright |
| 9 | * notice, this list of conditions and the following disclaimer. |
| 10 | * 2. Redistributions in binary form must reproduce the above copyright |
| 11 | * notice, this list of conditions and the following disclaimer in the |
| 12 | * documentation and/or other materials provided with the distribution. |
| 13 | * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its |
| 14 | * contributors may be used to endorse or promote products derived from |
| 15 | * this software without specific prior written permission. |
| 16 | * |
| 17 | * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND |
| 18 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 19 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 20 | * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE |
| 21 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 22 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 23 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 24 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 25 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 26 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 27 | * SUCH DAMAGE. |
| 28 | * |
| 29 | * Authors: Darko Laus (darko@mips.com) |
| 30 | * Djordje Pesut (djordje@mips.com) |
| 31 | * Mirjana Vulin (mvulin@mips.com) |
| 32 | * |
| 33 | * This file is part of FFmpeg. |
| 34 | * |
| 35 | * FFmpeg is free software; you can redistribute it and/or |
| 36 | * modify it under the terms of the GNU Lesser General Public |
| 37 | * License as published by the Free Software Foundation; either |
| 38 | * version 2.1 of the License, or (at your option) any later version. |
| 39 | * |
| 40 | * FFmpeg is distributed in the hope that it will be useful, |
| 41 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 42 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 43 | * Lesser General Public License for more details. |
| 44 | * |
| 45 | * You should have received a copy of the GNU Lesser General Public |
| 46 | * License along with FFmpeg; if not, write to the Free Software |
| 47 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 48 | */ |
| 49 | |
| 50 | /** |
| 51 | * @file |
| 52 | * Reference: libavcodec/aacpsdsp.c |
| 53 | */ |
| 54 | |
| 55 | #include "config.h" |
| 56 | #include "libavcodec/aacpsdsp.h" |
| 57 | |
| 58 | #if HAVE_INLINE_ASM |
| 59 | static void ps_hybrid_analysis_ileave_mips(float (*out)[32][2], float L[2][38][64], |
| 60 | int i, int len) |
| 61 | { |
| 62 | int temp0, temp1, temp2, temp3; |
| 63 | int temp4, temp5, temp6, temp7; |
| 64 | float *out1=&out[i][0][0]; |
| 65 | float *L1=&L[0][0][i]; |
| 66 | float *j=out1+ len*2; |
| 67 | |
| 68 | for (; i < 64; i++) { |
| 69 | |
| 70 | /* loop unrolled 8 times */ |
| 71 | __asm__ volatile ( |
| 72 | "1: \n\t" |
| 73 | "lw %[temp0], 0(%[L1]) \n\t" |
| 74 | "lw %[temp1], 9728(%[L1]) \n\t" |
| 75 | "lw %[temp2], 256(%[L1]) \n\t" |
| 76 | "lw %[temp3], 9984(%[L1]) \n\t" |
| 77 | "lw %[temp4], 512(%[L1]) \n\t" |
| 78 | "lw %[temp5], 10240(%[L1]) \n\t" |
| 79 | "lw %[temp6], 768(%[L1]) \n\t" |
| 80 | "lw %[temp7], 10496(%[L1]) \n\t" |
| 81 | "sw %[temp0], 0(%[out1]) \n\t" |
| 82 | "sw %[temp1], 4(%[out1]) \n\t" |
| 83 | "sw %[temp2], 8(%[out1]) \n\t" |
| 84 | "sw %[temp3], 12(%[out1]) \n\t" |
| 85 | "sw %[temp4], 16(%[out1]) \n\t" |
| 86 | "sw %[temp5], 20(%[out1]) \n\t" |
| 87 | "sw %[temp6], 24(%[out1]) \n\t" |
| 88 | "sw %[temp7], 28(%[out1]) \n\t" |
| 89 | "addiu %[out1], %[out1], 32 \n\t" |
| 90 | "addiu %[L1], %[L1], 1024 \n\t" |
| 91 | "bne %[out1], %[j], 1b \n\t" |
| 92 | |
| 93 | : [out1]"+r"(out1), [L1]"+r"(L1), [j]"+r"(j), |
| 94 | [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 95 | [temp2]"=&r"(temp2), [temp3]"=&r"(temp3), |
| 96 | [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 97 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7) |
| 98 | : [len]"r"(len) |
| 99 | : "memory" |
| 100 | ); |
| 101 | out1-=(len<<1)-64; |
| 102 | L1-=(len<<6)-1; |
| 103 | j+=len*2; |
| 104 | } |
| 105 | } |
| 106 | |
| 107 | static void ps_hybrid_synthesis_deint_mips(float out[2][38][64], |
| 108 | float (*in)[32][2], |
| 109 | int i, int len) |
| 110 | { |
| 111 | int n; |
| 112 | int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; |
| 113 | float *out1 = (float*)out + i; |
| 114 | float *out2 = (float*)out + 2432 + i; |
| 115 | float *in1 = (float*)in + 64 * i; |
| 116 | float *in2 = (float*)in + 64 * i + 1; |
| 117 | |
| 118 | for (; i < 64; i++) { |
| 119 | for (n = 0; n < 7; n++) { |
| 120 | |
| 121 | /* loop unrolled 8 times */ |
| 122 | __asm__ volatile ( |
| 123 | "lw %[temp0], 0(%[in1]) \n\t" |
| 124 | "lw %[temp1], 0(%[in2]) \n\t" |
| 125 | "lw %[temp2], 8(%[in1]) \n\t" |
| 126 | "lw %[temp3], 8(%[in2]) \n\t" |
| 127 | "lw %[temp4], 16(%[in1]) \n\t" |
| 128 | "lw %[temp5], 16(%[in2]) \n\t" |
| 129 | "lw %[temp6], 24(%[in1]) \n\t" |
| 130 | "lw %[temp7], 24(%[in2]) \n\t" |
| 131 | "addiu %[out1], %[out1], 1024 \n\t" |
| 132 | "addiu %[out2], %[out2], 1024 \n\t" |
| 133 | "addiu %[in1], %[in1], 32 \n\t" |
| 134 | "addiu %[in2], %[in2], 32 \n\t" |
| 135 | "sw %[temp0], -1024(%[out1]) \n\t" |
| 136 | "sw %[temp1], -1024(%[out2]) \n\t" |
| 137 | "sw %[temp2], -768(%[out1]) \n\t" |
| 138 | "sw %[temp3], -768(%[out2]) \n\t" |
| 139 | "sw %[temp4], -512(%[out1]) \n\t" |
| 140 | "sw %[temp5], -512(%[out2]) \n\t" |
| 141 | "sw %[temp6], -256(%[out1]) \n\t" |
| 142 | "sw %[temp7], -256(%[out2]) \n\t" |
| 143 | |
| 144 | : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 145 | [temp2]"=&r"(temp2), [temp3]"=&r"(temp3), |
| 146 | [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 147 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7), |
| 148 | [out1]"+r"(out1), [out2]"+r"(out2), |
| 149 | [in1]"+r"(in1), [in2]"+r"(in2) |
| 150 | : |
| 151 | : "memory" |
| 152 | ); |
| 153 | } |
| 154 | /* loop unrolled 8 times */ |
| 155 | __asm__ volatile ( |
| 156 | "lw %[temp0], 0(%[in1]) \n\t" |
| 157 | "lw %[temp1], 0(%[in2]) \n\t" |
| 158 | "lw %[temp2], 8(%[in1]) \n\t" |
| 159 | "lw %[temp3], 8(%[in2]) \n\t" |
| 160 | "lw %[temp4], 16(%[in1]) \n\t" |
| 161 | "lw %[temp5], 16(%[in2]) \n\t" |
| 162 | "lw %[temp6], 24(%[in1]) \n\t" |
| 163 | "lw %[temp7], 24(%[in2]) \n\t" |
| 164 | "addiu %[out1], %[out1], -7164 \n\t" |
| 165 | "addiu %[out2], %[out2], -7164 \n\t" |
| 166 | "addiu %[in1], %[in1], 32 \n\t" |
| 167 | "addiu %[in2], %[in2], 32 \n\t" |
| 168 | "sw %[temp0], 7164(%[out1]) \n\t" |
| 169 | "sw %[temp1], 7164(%[out2]) \n\t" |
| 170 | "sw %[temp2], 7420(%[out1]) \n\t" |
| 171 | "sw %[temp3], 7420(%[out2]) \n\t" |
| 172 | "sw %[temp4], 7676(%[out1]) \n\t" |
| 173 | "sw %[temp5], 7676(%[out2]) \n\t" |
| 174 | "sw %[temp6], 7932(%[out1]) \n\t" |
| 175 | "sw %[temp7], 7932(%[out2]) \n\t" |
| 176 | |
| 177 | : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 178 | [temp2]"=&r"(temp2), [temp3]"=&r"(temp3), |
| 179 | [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 180 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7), |
| 181 | [out1]"+r"(out1), [out2]"+r"(out2), |
| 182 | [in1]"+r"(in1), [in2]"+r"(in2) |
| 183 | : |
| 184 | : "memory" |
| 185 | ); |
| 186 | } |
| 187 | } |
| 188 | |
| 189 | #if HAVE_MIPSFPU |
| 190 | static void ps_add_squares_mips(float *dst, const float (*src)[2], int n) |
| 191 | { |
| 192 | int i; |
| 193 | float temp0, temp1, temp2, temp3, temp4, temp5; |
| 194 | float temp6, temp7, temp8, temp9, temp10, temp11; |
| 195 | float *src0 = (float*)&src[0][0]; |
| 196 | float *dst0 = &dst[0]; |
| 197 | |
| 198 | for (i = 0; i < 8; i++) { |
| 199 | /* loop unrolled 4 times */ |
| 200 | __asm__ volatile ( |
| 201 | "lwc1 %[temp0], 0(%[src0]) \n\t" |
| 202 | "lwc1 %[temp1], 4(%[src0]) \n\t" |
| 203 | "lwc1 %[temp2], 8(%[src0]) \n\t" |
| 204 | "lwc1 %[temp3], 12(%[src0]) \n\t" |
| 205 | "lwc1 %[temp4], 16(%[src0]) \n\t" |
| 206 | "lwc1 %[temp5], 20(%[src0]) \n\t" |
| 207 | "lwc1 %[temp6], 24(%[src0]) \n\t" |
| 208 | "lwc1 %[temp7], 28(%[src0]) \n\t" |
| 209 | "lwc1 %[temp8], 0(%[dst0]) \n\t" |
| 210 | "lwc1 %[temp9], 4(%[dst0]) \n\t" |
| 211 | "lwc1 %[temp10], 8(%[dst0]) \n\t" |
| 212 | "lwc1 %[temp11], 12(%[dst0]) \n\t" |
| 213 | "mul.s %[temp1], %[temp1], %[temp1] \n\t" |
| 214 | "mul.s %[temp3], %[temp3], %[temp3] \n\t" |
| 215 | "mul.s %[temp5], %[temp5], %[temp5] \n\t" |
| 216 | "mul.s %[temp7], %[temp7], %[temp7] \n\t" |
| 217 | "madd.s %[temp0], %[temp1], %[temp0], %[temp0] \n\t" |
| 218 | "madd.s %[temp2], %[temp3], %[temp2], %[temp2] \n\t" |
| 219 | "madd.s %[temp4], %[temp5], %[temp4], %[temp4] \n\t" |
| 220 | "madd.s %[temp6], %[temp7], %[temp6], %[temp6] \n\t" |
| 221 | "add.s %[temp0], %[temp8], %[temp0] \n\t" |
| 222 | "add.s %[temp2], %[temp9], %[temp2] \n\t" |
| 223 | "add.s %[temp4], %[temp10], %[temp4] \n\t" |
| 224 | "add.s %[temp6], %[temp11], %[temp6] \n\t" |
| 225 | "swc1 %[temp0], 0(%[dst0]) \n\t" |
| 226 | "swc1 %[temp2], 4(%[dst0]) \n\t" |
| 227 | "swc1 %[temp4], 8(%[dst0]) \n\t" |
| 228 | "swc1 %[temp6], 12(%[dst0]) \n\t" |
| 229 | "addiu %[dst0], %[dst0], 16 \n\t" |
| 230 | "addiu %[src0], %[src0], 32 \n\t" |
| 231 | |
| 232 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 233 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 234 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8), |
| 235 | [temp9]"=&f"(temp9), [dst0]"+r"(dst0), [src0]"+r"(src0), |
| 236 | [temp10]"=&f"(temp10), [temp11]"=&f"(temp11) |
| 237 | : |
| 238 | : "memory" |
| 239 | ); |
| 240 | } |
| 241 | } |
| 242 | |
| 243 | static void ps_mul_pair_single_mips(float (*dst)[2], float (*src0)[2], float *src1, |
| 244 | int n) |
| 245 | { |
| 246 | float temp0, temp1, temp2; |
| 247 | float *p_d, *p_s0, *p_s1, *end; |
| 248 | p_d = &dst[0][0]; |
| 249 | p_s0 = &src0[0][0]; |
| 250 | p_s1 = &src1[0]; |
| 251 | end = p_s1 + n; |
| 252 | |
| 253 | __asm__ volatile( |
| 254 | ".set push \n\t" |
| 255 | ".set noreorder \n\t" |
| 256 | "1: \n\t" |
| 257 | "lwc1 %[temp2], 0(%[p_s1]) \n\t" |
| 258 | "lwc1 %[temp0], 0(%[p_s0]) \n\t" |
| 259 | "lwc1 %[temp1], 4(%[p_s0]) \n\t" |
| 260 | "addiu %[p_d], %[p_d], 8 \n\t" |
| 261 | "mul.s %[temp0], %[temp0], %[temp2] \n\t" |
| 262 | "mul.s %[temp1], %[temp1], %[temp2] \n\t" |
| 263 | "addiu %[p_s0], %[p_s0], 8 \n\t" |
| 264 | "swc1 %[temp0], -8(%[p_d]) \n\t" |
| 265 | "swc1 %[temp1], -4(%[p_d]) \n\t" |
| 266 | "bne %[p_s1], %[end], 1b \n\t" |
| 267 | " addiu %[p_s1], %[p_s1], 4 \n\t" |
| 268 | ".set pop \n\t" |
| 269 | |
| 270 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), |
| 271 | [temp2]"=&f"(temp2), [p_d]"+r"(p_d), |
| 272 | [p_s0]"+r"(p_s0), [p_s1]"+r"(p_s1) |
| 273 | : [end]"r"(end) |
| 274 | : "memory" |
| 275 | ); |
| 276 | } |
| 277 | |
| 278 | static void ps_decorrelate_mips(float (*out)[2], float (*delay)[2], |
| 279 | float (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2], |
| 280 | const float phi_fract[2], float (*Q_fract)[2], |
| 281 | const float *transient_gain, |
| 282 | float g_decay_slope, |
| 283 | int len) |
| 284 | { |
| 285 | float *p_delay = &delay[0][0]; |
| 286 | float *p_out = &out[0][0]; |
| 287 | float *p_ap_delay = &ap_delay[0][0][0]; |
| 288 | float *p_t_gain = (float*)transient_gain; |
| 289 | float *p_Q_fract = &Q_fract[0][0]; |
| 290 | float ag0, ag1, ag2; |
| 291 | float phi_fract0 = phi_fract[0]; |
| 292 | float phi_fract1 = phi_fract[1]; |
| 293 | float temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9; |
| 294 | |
| 295 | len = (int)((int*)p_delay + (len << 1)); |
| 296 | |
| 297 | /* merged 2 loops */ |
| 298 | __asm__ volatile( |
| 299 | ".set push \n\t" |
| 300 | ".set noreorder \n\t" |
| 301 | "li.s %[ag0], 0.65143905753106 \n\t" |
| 302 | "li.s %[ag1], 0.56471812200776 \n\t" |
| 303 | "li.s %[ag2], 0.48954165955695 \n\t" |
| 304 | "mul.s %[ag0], %[ag0], %[g_decay_slope] \n\t" |
| 305 | "mul.s %[ag1], %[ag1], %[g_decay_slope] \n\t" |
| 306 | "mul.s %[ag2], %[ag2], %[g_decay_slope] \n\t" |
| 307 | "1: \n\t" |
| 308 | "lwc1 %[temp0], 0(%[p_delay]) \n\t" |
| 309 | "lwc1 %[temp1], 4(%[p_delay]) \n\t" |
| 310 | "lwc1 %[temp4], 16(%[p_ap_delay]) \n\t" |
| 311 | "lwc1 %[temp5], 20(%[p_ap_delay]) \n\t" |
| 312 | "mul.s %[temp3], %[temp0], %[phi_fract1] \n\t" |
| 313 | "lwc1 %[temp6], 0(%[p_Q_fract]) \n\t" |
| 314 | "mul.s %[temp2], %[temp1], %[phi_fract1] \n\t" |
| 315 | "lwc1 %[temp7], 4(%[p_Q_fract]) \n\t" |
| 316 | "madd.s %[temp3], %[temp3], %[temp1], %[phi_fract0] \n\t" |
| 317 | "msub.s %[temp2], %[temp2], %[temp0], %[phi_fract0] \n\t" |
| 318 | "mul.s %[temp8], %[temp5], %[temp7] \n\t" |
| 319 | "mul.s %[temp9], %[temp4], %[temp7] \n\t" |
| 320 | "lwc1 %[temp7], 12(%[p_Q_fract]) \n\t" |
| 321 | "mul.s %[temp0], %[ag0], %[temp2] \n\t" |
| 322 | "mul.s %[temp1], %[ag0], %[temp3] \n\t" |
| 323 | "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t" |
| 324 | "lwc1 %[temp4], 304(%[p_ap_delay]) \n\t" |
| 325 | "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t" |
| 326 | "lwc1 %[temp5], 308(%[p_ap_delay]) \n\t" |
| 327 | "sub.s %[temp0], %[temp8], %[temp0] \n\t" |
| 328 | "sub.s %[temp1], %[temp9], %[temp1] \n\t" |
| 329 | "madd.s %[temp2], %[temp2], %[ag0], %[temp0] \n\t" |
| 330 | "lwc1 %[temp6], 8(%[p_Q_fract]) \n\t" |
| 331 | "madd.s %[temp3], %[temp3], %[ag0], %[temp1] \n\t" |
| 332 | "mul.s %[temp8], %[temp5], %[temp7] \n\t" |
| 333 | "mul.s %[temp9], %[temp4], %[temp7] \n\t" |
| 334 | "lwc1 %[temp7], 20(%[p_Q_fract]) \n\t" |
| 335 | "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t" |
| 336 | "swc1 %[temp2], 40(%[p_ap_delay]) \n\t" |
| 337 | "mul.s %[temp2], %[ag1], %[temp0] \n\t" |
| 338 | "swc1 %[temp3], 44(%[p_ap_delay]) \n\t" |
| 339 | "mul.s %[temp3], %[ag1], %[temp1] \n\t" |
| 340 | "lwc1 %[temp4], 592(%[p_ap_delay]) \n\t" |
| 341 | "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t" |
| 342 | "lwc1 %[temp5], 596(%[p_ap_delay]) \n\t" |
| 343 | "sub.s %[temp2], %[temp8], %[temp2] \n\t" |
| 344 | "sub.s %[temp3], %[temp9], %[temp3] \n\t" |
| 345 | "lwc1 %[temp6], 16(%[p_Q_fract]) \n\t" |
| 346 | "madd.s %[temp0], %[temp0], %[ag1], %[temp2] \n\t" |
| 347 | "madd.s %[temp1], %[temp1], %[ag1], %[temp3] \n\t" |
| 348 | "mul.s %[temp8], %[temp5], %[temp7] \n\t" |
| 349 | "mul.s %[temp9], %[temp4], %[temp7] \n\t" |
| 350 | "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t" |
| 351 | "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t" |
| 352 | "swc1 %[temp0], 336(%[p_ap_delay]) \n\t" |
| 353 | "mul.s %[temp0], %[ag2], %[temp2] \n\t" |
| 354 | "swc1 %[temp1], 340(%[p_ap_delay]) \n\t" |
| 355 | "mul.s %[temp1], %[ag2], %[temp3] \n\t" |
| 356 | "lwc1 %[temp4], 0(%[p_t_gain]) \n\t" |
| 357 | "sub.s %[temp0], %[temp8], %[temp0] \n\t" |
| 358 | "addiu %[p_ap_delay], %[p_ap_delay], 8 \n\t" |
| 359 | "sub.s %[temp1], %[temp9], %[temp1] \n\t" |
| 360 | "addiu %[p_t_gain], %[p_t_gain], 4 \n\t" |
| 361 | "madd.s %[temp2], %[temp2], %[ag2], %[temp0] \n\t" |
| 362 | "addiu %[p_delay], %[p_delay], 8 \n\t" |
| 363 | "madd.s %[temp3], %[temp3], %[ag2], %[temp1] \n\t" |
| 364 | "addiu %[p_out], %[p_out], 8 \n\t" |
| 365 | "mul.s %[temp5], %[temp4], %[temp0] \n\t" |
| 366 | "mul.s %[temp6], %[temp4], %[temp1] \n\t" |
| 367 | "swc1 %[temp2], 624(%[p_ap_delay]) \n\t" |
| 368 | "swc1 %[temp3], 628(%[p_ap_delay]) \n\t" |
| 369 | "swc1 %[temp5], -8(%[p_out]) \n\t" |
| 370 | "swc1 %[temp6], -4(%[p_out]) \n\t" |
| 371 | "bne %[p_delay], %[len], 1b \n\t" |
| 372 | " swc1 %[temp6], -4(%[p_out]) \n\t" |
| 373 | ".set pop \n\t" |
| 374 | |
| 375 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 376 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 377 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8), |
| 378 | [temp9]"=&f"(temp9), [p_delay]"+r"(p_delay), [p_ap_delay]"+r"(p_ap_delay), |
| 379 | [p_Q_fract]"+r"(p_Q_fract), [p_t_gain]"+r"(p_t_gain), [p_out]"+r"(p_out), |
| 380 | [ag0]"=&f"(ag0), [ag1]"=&f"(ag1), [ag2]"=&f"(ag2) |
| 381 | : [phi_fract0]"f"(phi_fract0), [phi_fract1]"f"(phi_fract1), |
| 382 | [len]"r"(len), [g_decay_slope]"f"(g_decay_slope) |
| 383 | : "memory" |
| 384 | ); |
| 385 | } |
| 386 | |
| 387 | static void ps_stereo_interpolate_mips(float (*l)[2], float (*r)[2], |
| 388 | float h[2][4], float h_step[2][4], |
| 389 | int len) |
| 390 | { |
| 391 | float h0 = h[0][0]; |
| 392 | float h1 = h[0][1]; |
| 393 | float h2 = h[0][2]; |
| 394 | float h3 = h[0][3]; |
| 395 | float hs0 = h_step[0][0]; |
| 396 | float hs1 = h_step[0][1]; |
| 397 | float hs2 = h_step[0][2]; |
| 398 | float hs3 = h_step[0][3]; |
| 399 | float temp0, temp1, temp2, temp3; |
| 400 | float l_re, l_im, r_re, r_im; |
| 401 | |
| 402 | len = (int)((int*)l + (len << 1)); |
| 403 | |
| 404 | __asm__ volatile( |
| 405 | ".set push \n\t" |
| 406 | ".set noreorder \n\t" |
| 407 | "1: \n\t" |
| 408 | "add.s %[h0], %[h0], %[hs0] \n\t" |
| 409 | "lwc1 %[l_re], 0(%[l]) \n\t" |
| 410 | "add.s %[h1], %[h1], %[hs1] \n\t" |
| 411 | "lwc1 %[r_re], 0(%[r]) \n\t" |
| 412 | "add.s %[h2], %[h2], %[hs2] \n\t" |
| 413 | "lwc1 %[l_im], 4(%[l]) \n\t" |
| 414 | "add.s %[h3], %[h3], %[hs3] \n\t" |
| 415 | "lwc1 %[r_im], 4(%[r]) \n\t" |
| 416 | "mul.s %[temp0], %[h0], %[l_re] \n\t" |
| 417 | "addiu %[l], %[l], 8 \n\t" |
| 418 | "mul.s %[temp2], %[h1], %[l_re] \n\t" |
| 419 | "addiu %[r], %[r], 8 \n\t" |
| 420 | "madd.s %[temp0], %[temp0], %[h2], %[r_re] \n\t" |
| 421 | "madd.s %[temp2], %[temp2], %[h3], %[r_re] \n\t" |
| 422 | "mul.s %[temp1], %[h0], %[l_im] \n\t" |
| 423 | "mul.s %[temp3], %[h1], %[l_im] \n\t" |
| 424 | "madd.s %[temp1], %[temp1], %[h2], %[r_im] \n\t" |
| 425 | "madd.s %[temp3], %[temp3], %[h3], %[r_im] \n\t" |
| 426 | "swc1 %[temp0], -8(%[l]) \n\t" |
| 427 | "swc1 %[temp2], -8(%[r]) \n\t" |
| 428 | "swc1 %[temp1], -4(%[l]) \n\t" |
| 429 | "bne %[l], %[len], 1b \n\t" |
| 430 | " swc1 %[temp3], -4(%[r]) \n\t" |
| 431 | ".set pop \n\t" |
| 432 | |
| 433 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), |
| 434 | [temp2]"=&f"(temp2), [temp3]"=&f"(temp3), |
| 435 | [h0]"+f"(h0), [h1]"+f"(h1), [h2]"+f"(h2), |
| 436 | [h3]"+f"(h3), [l]"+r"(l), [r]"+r"(r), |
| 437 | [l_re]"=&f"(l_re), [l_im]"=&f"(l_im), |
| 438 | [r_re]"=&f"(r_re), [r_im]"=&f"(r_im) |
| 439 | : [hs0]"f"(hs0), [hs1]"f"(hs1), [hs2]"f"(hs2), |
| 440 | [hs3]"f"(hs3), [len]"r"(len) |
| 441 | : "memory" |
| 442 | ); |
| 443 | } |
| 444 | #endif /* HAVE_MIPSFPU */ |
| 445 | #endif /* HAVE_INLINE_ASM */ |
| 446 | |
| 447 | void ff_psdsp_init_mips(PSDSPContext *s) |
| 448 | { |
| 449 | #if HAVE_INLINE_ASM |
| 450 | s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_mips; |
| 451 | s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_mips; |
| 452 | #if HAVE_MIPSFPU |
| 453 | s->add_squares = ps_add_squares_mips; |
| 454 | s->mul_pair_single = ps_mul_pair_single_mips; |
| 455 | s->decorrelate = ps_decorrelate_mips; |
| 456 | s->stereo_interpolate[0] = ps_stereo_interpolate_mips; |
| 457 | #endif /* HAVE_MIPSFPU */ |
| 458 | #endif /* HAVE_INLINE_ASM */ |
| 459 | } |