| 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 | * AAC Spectral Band Replication decoding functions optimized for MIPS |
| 34 | * |
| 35 | * This file is part of FFmpeg. |
| 36 | * |
| 37 | * FFmpeg is free software; you can redistribute it and/or |
| 38 | * modify it under the terms of the GNU Lesser General Public |
| 39 | * License as published by the Free Software Foundation; either |
| 40 | * version 2.1 of the License, or (at your option) any later version. |
| 41 | * |
| 42 | * FFmpeg is distributed in the hope that it will be useful, |
| 43 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 44 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 45 | * Lesser General Public License for more details. |
| 46 | * |
| 47 | * You should have received a copy of the GNU Lesser General Public |
| 48 | * License along with FFmpeg; if not, write to the Free Software |
| 49 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 50 | */ |
| 51 | |
| 52 | /** |
| 53 | * @file |
| 54 | * Reference: libavcodec/sbrdsp.c |
| 55 | */ |
| 56 | |
| 57 | #include "config.h" |
| 58 | #include "libavcodec/sbrdsp.h" |
| 59 | |
| 60 | #if HAVE_INLINE_ASM |
| 61 | static void sbr_neg_odd_64_mips(float *x) |
| 62 | { |
| 63 | int Temp1, Temp2, Temp3, Temp4, Temp5; |
| 64 | float *x1 = &x[1]; |
| 65 | float *x_end = x1 + 64; |
| 66 | |
| 67 | /* loop unrolled 4 times */ |
| 68 | __asm__ volatile ( |
| 69 | "lui %[Temp5], 0x8000 \n\t" |
| 70 | "1: \n\t" |
| 71 | "lw %[Temp1], 0(%[x1]) \n\t" |
| 72 | "lw %[Temp2], 8(%[x1]) \n\t" |
| 73 | "lw %[Temp3], 16(%[x1]) \n\t" |
| 74 | "lw %[Temp4], 24(%[x1]) \n\t" |
| 75 | "xor %[Temp1], %[Temp1], %[Temp5] \n\t" |
| 76 | "xor %[Temp2], %[Temp2], %[Temp5] \n\t" |
| 77 | "xor %[Temp3], %[Temp3], %[Temp5] \n\t" |
| 78 | "xor %[Temp4], %[Temp4], %[Temp5] \n\t" |
| 79 | "sw %[Temp1], 0(%[x1]) \n\t" |
| 80 | "sw %[Temp2], 8(%[x1]) \n\t" |
| 81 | "sw %[Temp3], 16(%[x1]) \n\t" |
| 82 | "sw %[Temp4], 24(%[x1]) \n\t" |
| 83 | "addiu %[x1], %[x1], 32 \n\t" |
| 84 | "bne %[x1], %[x_end], 1b \n\t" |
| 85 | |
| 86 | : [Temp1]"=&r"(Temp1), [Temp2]"=&r"(Temp2), |
| 87 | [Temp3]"=&r"(Temp3), [Temp4]"=&r"(Temp4), |
| 88 | [Temp5]"=&r"(Temp5), [x1]"+r"(x1) |
| 89 | : [x_end]"r"(x_end) |
| 90 | : "memory" |
| 91 | ); |
| 92 | } |
| 93 | |
| 94 | static void sbr_qmf_pre_shuffle_mips(float *z) |
| 95 | { |
| 96 | int Temp1, Temp2, Temp3, Temp4, Temp5, Temp6; |
| 97 | float *z1 = &z[66]; |
| 98 | float *z2 = &z[59]; |
| 99 | float *z3 = &z[2]; |
| 100 | float *z4 = z1 + 60; |
| 101 | |
| 102 | /* loop unrolled 5 times */ |
| 103 | __asm__ volatile ( |
| 104 | "lui %[Temp6], 0x8000 \n\t" |
| 105 | "1: \n\t" |
| 106 | "lw %[Temp1], 0(%[z2]) \n\t" |
| 107 | "lw %[Temp2], 4(%[z2]) \n\t" |
| 108 | "lw %[Temp3], 8(%[z2]) \n\t" |
| 109 | "lw %[Temp4], 12(%[z2]) \n\t" |
| 110 | "lw %[Temp5], 16(%[z2]) \n\t" |
| 111 | "xor %[Temp1], %[Temp1], %[Temp6] \n\t" |
| 112 | "xor %[Temp2], %[Temp2], %[Temp6] \n\t" |
| 113 | "xor %[Temp3], %[Temp3], %[Temp6] \n\t" |
| 114 | "xor %[Temp4], %[Temp4], %[Temp6] \n\t" |
| 115 | "xor %[Temp5], %[Temp5], %[Temp6] \n\t" |
| 116 | "addiu %[z2], %[z2], -20 \n\t" |
| 117 | "sw %[Temp1], 32(%[z1]) \n\t" |
| 118 | "sw %[Temp2], 24(%[z1]) \n\t" |
| 119 | "sw %[Temp3], 16(%[z1]) \n\t" |
| 120 | "sw %[Temp4], 8(%[z1]) \n\t" |
| 121 | "sw %[Temp5], 0(%[z1]) \n\t" |
| 122 | "lw %[Temp1], 0(%[z3]) \n\t" |
| 123 | "lw %[Temp2], 4(%[z3]) \n\t" |
| 124 | "lw %[Temp3], 8(%[z3]) \n\t" |
| 125 | "lw %[Temp4], 12(%[z3]) \n\t" |
| 126 | "lw %[Temp5], 16(%[z3]) \n\t" |
| 127 | "sw %[Temp1], 4(%[z1]) \n\t" |
| 128 | "sw %[Temp2], 12(%[z1]) \n\t" |
| 129 | "sw %[Temp3], 20(%[z1]) \n\t" |
| 130 | "sw %[Temp4], 28(%[z1]) \n\t" |
| 131 | "sw %[Temp5], 36(%[z1]) \n\t" |
| 132 | "addiu %[z3], %[z3], 20 \n\t" |
| 133 | "addiu %[z1], %[z1], 40 \n\t" |
| 134 | "bne %[z1], %[z4], 1b \n\t" |
| 135 | "lw %[Temp1], 132(%[z]) \n\t" |
| 136 | "lw %[Temp2], 128(%[z]) \n\t" |
| 137 | "lw %[Temp3], 0(%[z]) \n\t" |
| 138 | "lw %[Temp4], 4(%[z]) \n\t" |
| 139 | "xor %[Temp1], %[Temp1], %[Temp6] \n\t" |
| 140 | "sw %[Temp1], 504(%[z]) \n\t" |
| 141 | "sw %[Temp2], 508(%[z]) \n\t" |
| 142 | "sw %[Temp3], 256(%[z]) \n\t" |
| 143 | "sw %[Temp4], 260(%[z]) \n\t" |
| 144 | |
| 145 | : [Temp1]"=&r"(Temp1), [Temp2]"=&r"(Temp2), |
| 146 | [Temp3]"=&r"(Temp3), [Temp4]"=&r"(Temp4), |
| 147 | [Temp5]"=&r"(Temp5), [Temp6]"=&r"(Temp6), |
| 148 | [z1]"+r"(z1), [z2]"+r"(z2), [z3]"+r"(z3) |
| 149 | : [z4]"r"(z4), [z]"r"(z) |
| 150 | : "memory" |
| 151 | ); |
| 152 | } |
| 153 | |
| 154 | static void sbr_qmf_post_shuffle_mips(float W[32][2], const float *z) |
| 155 | { |
| 156 | int Temp1, Temp2, Temp3, Temp4, Temp5; |
| 157 | float *W_ptr = (float *)W; |
| 158 | float *z1 = (float *)z; |
| 159 | float *z2 = (float *)&z[60]; |
| 160 | float *z_end = z1 + 32; |
| 161 | |
| 162 | /* loop unrolled 4 times */ |
| 163 | __asm__ volatile ( |
| 164 | "lui %[Temp5], 0x8000 \n\t" |
| 165 | "1: \n\t" |
| 166 | "lw %[Temp1], 0(%[z2]) \n\t" |
| 167 | "lw %[Temp2], 4(%[z2]) \n\t" |
| 168 | "lw %[Temp3], 8(%[z2]) \n\t" |
| 169 | "lw %[Temp4], 12(%[z2]) \n\t" |
| 170 | "xor %[Temp1], %[Temp1], %[Temp5] \n\t" |
| 171 | "xor %[Temp2], %[Temp2], %[Temp5] \n\t" |
| 172 | "xor %[Temp3], %[Temp3], %[Temp5] \n\t" |
| 173 | "xor %[Temp4], %[Temp4], %[Temp5] \n\t" |
| 174 | "addiu %[z2], %[z2], -16 \n\t" |
| 175 | "sw %[Temp1], 24(%[W_ptr]) \n\t" |
| 176 | "sw %[Temp2], 16(%[W_ptr]) \n\t" |
| 177 | "sw %[Temp3], 8(%[W_ptr]) \n\t" |
| 178 | "sw %[Temp4], 0(%[W_ptr]) \n\t" |
| 179 | "lw %[Temp1], 0(%[z1]) \n\t" |
| 180 | "lw %[Temp2], 4(%[z1]) \n\t" |
| 181 | "lw %[Temp3], 8(%[z1]) \n\t" |
| 182 | "lw %[Temp4], 12(%[z1]) \n\t" |
| 183 | "sw %[Temp1], 4(%[W_ptr]) \n\t" |
| 184 | "sw %[Temp2], 12(%[W_ptr]) \n\t" |
| 185 | "sw %[Temp3], 20(%[W_ptr]) \n\t" |
| 186 | "sw %[Temp4], 28(%[W_ptr]) \n\t" |
| 187 | "addiu %[z1], %[z1], 16 \n\t" |
| 188 | "addiu %[W_ptr], %[W_ptr], 32 \n\t" |
| 189 | "bne %[z1], %[z_end], 1b \n\t" |
| 190 | |
| 191 | : [Temp1]"=&r"(Temp1), [Temp2]"=&r"(Temp2), |
| 192 | [Temp3]"=&r"(Temp3), [Temp4]"=&r"(Temp4), |
| 193 | [Temp5]"=&r"(Temp5), [z1]"+r"(z1), |
| 194 | [z2]"+r"(z2), [W_ptr]"+r"(W_ptr) |
| 195 | : [z_end]"r"(z_end) |
| 196 | : "memory" |
| 197 | ); |
| 198 | } |
| 199 | |
| 200 | #if HAVE_MIPSFPU |
| 201 | static void sbr_sum64x5_mips(float *z) |
| 202 | { |
| 203 | int k; |
| 204 | float *z1; |
| 205 | float f1, f2, f3, f4, f5, f6, f7, f8; |
| 206 | for (k = 0; k < 64; k += 8) { |
| 207 | |
| 208 | z1 = &z[k]; |
| 209 | |
| 210 | /* loop unrolled 8 times */ |
| 211 | __asm__ volatile ( |
| 212 | "lwc1 $f0, 0(%[z1]) \n\t" |
| 213 | "lwc1 $f1, 256(%[z1]) \n\t" |
| 214 | "lwc1 $f2, 4(%[z1]) \n\t" |
| 215 | "lwc1 $f3, 260(%[z1]) \n\t" |
| 216 | "lwc1 $f4, 8(%[z1]) \n\t" |
| 217 | "add.s %[f1], $f0, $f1 \n\t" |
| 218 | "lwc1 $f5, 264(%[z1]) \n\t" |
| 219 | "add.s %[f2], $f2, $f3 \n\t" |
| 220 | "lwc1 $f6, 12(%[z1]) \n\t" |
| 221 | "lwc1 $f7, 268(%[z1]) \n\t" |
| 222 | "add.s %[f3], $f4, $f5 \n\t" |
| 223 | "lwc1 $f8, 16(%[z1]) \n\t" |
| 224 | "lwc1 $f9, 272(%[z1]) \n\t" |
| 225 | "add.s %[f4], $f6, $f7 \n\t" |
| 226 | "lwc1 $f10, 20(%[z1]) \n\t" |
| 227 | "lwc1 $f11, 276(%[z1]) \n\t" |
| 228 | "add.s %[f5], $f8, $f9 \n\t" |
| 229 | "lwc1 $f12, 24(%[z1]) \n\t" |
| 230 | "lwc1 $f13, 280(%[z1]) \n\t" |
| 231 | "add.s %[f6], $f10, $f11 \n\t" |
| 232 | "lwc1 $f14, 28(%[z1]) \n\t" |
| 233 | "lwc1 $f15, 284(%[z1]) \n\t" |
| 234 | "add.s %[f7], $f12, $f13 \n\t" |
| 235 | "lwc1 $f0, 512(%[z1]) \n\t" |
| 236 | "lwc1 $f1, 516(%[z1]) \n\t" |
| 237 | "add.s %[f8], $f14, $f15 \n\t" |
| 238 | "lwc1 $f2, 520(%[z1]) \n\t" |
| 239 | "add.s %[f1], %[f1], $f0 \n\t" |
| 240 | "add.s %[f2], %[f2], $f1 \n\t" |
| 241 | "lwc1 $f3, 524(%[z1]) \n\t" |
| 242 | "add.s %[f3], %[f3], $f2 \n\t" |
| 243 | "lwc1 $f4, 528(%[z1]) \n\t" |
| 244 | "lwc1 $f5, 532(%[z1]) \n\t" |
| 245 | "add.s %[f4], %[f4], $f3 \n\t" |
| 246 | "lwc1 $f6, 536(%[z1]) \n\t" |
| 247 | "add.s %[f5], %[f5], $f4 \n\t" |
| 248 | "add.s %[f6], %[f6], $f5 \n\t" |
| 249 | "lwc1 $f7, 540(%[z1]) \n\t" |
| 250 | "add.s %[f7], %[f7], $f6 \n\t" |
| 251 | "lwc1 $f0, 768(%[z1]) \n\t" |
| 252 | "lwc1 $f1, 772(%[z1]) \n\t" |
| 253 | "add.s %[f8], %[f8], $f7 \n\t" |
| 254 | "lwc1 $f2, 776(%[z1]) \n\t" |
| 255 | "add.s %[f1], %[f1], $f0 \n\t" |
| 256 | "add.s %[f2], %[f2], $f1 \n\t" |
| 257 | "lwc1 $f3, 780(%[z1]) \n\t" |
| 258 | "add.s %[f3], %[f3], $f2 \n\t" |
| 259 | "lwc1 $f4, 784(%[z1]) \n\t" |
| 260 | "lwc1 $f5, 788(%[z1]) \n\t" |
| 261 | "add.s %[f4], %[f4], $f3 \n\t" |
| 262 | "lwc1 $f6, 792(%[z1]) \n\t" |
| 263 | "add.s %[f5], %[f5], $f4 \n\t" |
| 264 | "add.s %[f6], %[f6], $f5 \n\t" |
| 265 | "lwc1 $f7, 796(%[z1]) \n\t" |
| 266 | "add.s %[f7], %[f7], $f6 \n\t" |
| 267 | "lwc1 $f0, 1024(%[z1]) \n\t" |
| 268 | "lwc1 $f1, 1028(%[z1]) \n\t" |
| 269 | "add.s %[f8], %[f8], $f7 \n\t" |
| 270 | "lwc1 $f2, 1032(%[z1]) \n\t" |
| 271 | "add.s %[f1], %[f1], $f0 \n\t" |
| 272 | "add.s %[f2], %[f2], $f1 \n\t" |
| 273 | "lwc1 $f3, 1036(%[z1]) \n\t" |
| 274 | "add.s %[f3], %[f3], $f2 \n\t" |
| 275 | "lwc1 $f4, 1040(%[z1]) \n\t" |
| 276 | "lwc1 $f5, 1044(%[z1]) \n\t" |
| 277 | "add.s %[f4], %[f4], $f3 \n\t" |
| 278 | "lwc1 $f6, 1048(%[z1]) \n\t" |
| 279 | "add.s %[f5], %[f5], $f4 \n\t" |
| 280 | "add.s %[f6], %[f6], $f5 \n\t" |
| 281 | "lwc1 $f7, 1052(%[z1]) \n\t" |
| 282 | "add.s %[f7], %[f7], $f6 \n\t" |
| 283 | "swc1 %[f1], 0(%[z1]) \n\t" |
| 284 | "swc1 %[f2], 4(%[z1]) \n\t" |
| 285 | "add.s %[f8], %[f8], $f7 \n\t" |
| 286 | "swc1 %[f3], 8(%[z1]) \n\t" |
| 287 | "swc1 %[f4], 12(%[z1]) \n\t" |
| 288 | "swc1 %[f5], 16(%[z1]) \n\t" |
| 289 | "swc1 %[f6], 20(%[z1]) \n\t" |
| 290 | "swc1 %[f7], 24(%[z1]) \n\t" |
| 291 | "swc1 %[f8], 28(%[z1]) \n\t" |
| 292 | |
| 293 | : [f1]"=&f"(f1), [f2]"=&f"(f2), [f3]"=&f"(f3), |
| 294 | [f4]"=&f"(f4), [f5]"=&f"(f5), [f6]"=&f"(f6), |
| 295 | [f7]"=&f"(f7), [f8]"=&f"(f8) |
| 296 | : [z1]"r"(z1) |
| 297 | : "$f0", "$f1", "$f2", "$f3", "$f4", "$f5", |
| 298 | "$f6", "$f7", "$f8", "$f9", "$f10", "$f11", |
| 299 | "$f12", "$f13", "$f14", "$f15", |
| 300 | "memory" |
| 301 | ); |
| 302 | } |
| 303 | } |
| 304 | |
| 305 | static float sbr_sum_square_mips(float (*x)[2], int n) |
| 306 | { |
| 307 | float sum0 = 0.0f, sum1 = 0.0f; |
| 308 | float *p_x; |
| 309 | float temp0, temp1, temp2, temp3; |
| 310 | float *loop_end; |
| 311 | p_x = &x[0][0]; |
| 312 | loop_end = p_x + (n >> 1)*4 - 4; |
| 313 | |
| 314 | __asm__ volatile ( |
| 315 | ".set push \n\t" |
| 316 | ".set noreorder \n\t" |
| 317 | "lwc1 %[temp0], 0(%[p_x]) \n\t" |
| 318 | "lwc1 %[temp1], 4(%[p_x]) \n\t" |
| 319 | "lwc1 %[temp2], 8(%[p_x]) \n\t" |
| 320 | "lwc1 %[temp3], 12(%[p_x]) \n\t" |
| 321 | "1: \n\t" |
| 322 | "addiu %[p_x], %[p_x], 16 \n\t" |
| 323 | "madd.s %[sum0], %[sum0], %[temp0], %[temp0] \n\t" |
| 324 | "lwc1 %[temp0], 0(%[p_x]) \n\t" |
| 325 | "madd.s %[sum1], %[sum1], %[temp1], %[temp1] \n\t" |
| 326 | "lwc1 %[temp1], 4(%[p_x]) \n\t" |
| 327 | "madd.s %[sum0], %[sum0], %[temp2], %[temp2] \n\t" |
| 328 | "lwc1 %[temp2], 8(%[p_x]) \n\t" |
| 329 | "madd.s %[sum1], %[sum1], %[temp3], %[temp3] \n\t" |
| 330 | "bne %[p_x], %[loop_end], 1b \n\t" |
| 331 | " lwc1 %[temp3], 12(%[p_x]) \n\t" |
| 332 | "madd.s %[sum0], %[sum0], %[temp0], %[temp0] \n\t" |
| 333 | "madd.s %[sum1], %[sum1], %[temp1], %[temp1] \n\t" |
| 334 | "madd.s %[sum0], %[sum0], %[temp2], %[temp2] \n\t" |
| 335 | "madd.s %[sum1], %[sum1], %[temp3], %[temp3] \n\t" |
| 336 | ".set pop \n\t" |
| 337 | |
| 338 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 339 | [temp3]"=&f"(temp3), [sum0]"+f"(sum0), [sum1]"+f"(sum1), |
| 340 | [p_x]"+r"(p_x) |
| 341 | : [loop_end]"r"(loop_end) |
| 342 | : "memory" |
| 343 | ); |
| 344 | return sum0 + sum1; |
| 345 | } |
| 346 | |
| 347 | static void sbr_qmf_deint_bfly_mips(float *v, const float *src0, const float *src1) |
| 348 | { |
| 349 | int i; |
| 350 | float temp0, temp1, temp2, temp3, temp4, temp5; |
| 351 | float temp6, temp7, temp8, temp9, temp10, temp11; |
| 352 | float *v0 = v; |
| 353 | float *v1 = &v[127]; |
| 354 | float *psrc0 = (float*)src0; |
| 355 | float *psrc1 = (float*)&src1[63]; |
| 356 | |
| 357 | for (i = 0; i < 4; i++) { |
| 358 | |
| 359 | /* loop unrolled 16 times */ |
| 360 | __asm__ volatile( |
| 361 | "lwc1 %[temp0], 0(%[src0]) \n\t" |
| 362 | "lwc1 %[temp1], 0(%[src1]) \n\t" |
| 363 | "lwc1 %[temp3], 4(%[src0]) \n\t" |
| 364 | "lwc1 %[temp4], -4(%[src1]) \n\t" |
| 365 | "lwc1 %[temp6], 8(%[src0]) \n\t" |
| 366 | "lwc1 %[temp7], -8(%[src1]) \n\t" |
| 367 | "lwc1 %[temp9], 12(%[src0]) \n\t" |
| 368 | "lwc1 %[temp10], -12(%[src1]) \n\t" |
| 369 | "add.s %[temp2], %[temp0], %[temp1] \n\t" |
| 370 | "add.s %[temp5], %[temp3], %[temp4] \n\t" |
| 371 | "add.s %[temp8], %[temp6], %[temp7] \n\t" |
| 372 | "add.s %[temp11], %[temp9], %[temp10] \n\t" |
| 373 | "sub.s %[temp0], %[temp0], %[temp1] \n\t" |
| 374 | "sub.s %[temp3], %[temp3], %[temp4] \n\t" |
| 375 | "sub.s %[temp6], %[temp6], %[temp7] \n\t" |
| 376 | "sub.s %[temp9], %[temp9], %[temp10] \n\t" |
| 377 | "swc1 %[temp2], 0(%[v1]) \n\t" |
| 378 | "swc1 %[temp0], 0(%[v0]) \n\t" |
| 379 | "swc1 %[temp5], -4(%[v1]) \n\t" |
| 380 | "swc1 %[temp3], 4(%[v0]) \n\t" |
| 381 | "swc1 %[temp8], -8(%[v1]) \n\t" |
| 382 | "swc1 %[temp6], 8(%[v0]) \n\t" |
| 383 | "swc1 %[temp11], -12(%[v1]) \n\t" |
| 384 | "swc1 %[temp9], 12(%[v0]) \n\t" |
| 385 | "lwc1 %[temp0], 16(%[src0]) \n\t" |
| 386 | "lwc1 %[temp1], -16(%[src1]) \n\t" |
| 387 | "lwc1 %[temp3], 20(%[src0]) \n\t" |
| 388 | "lwc1 %[temp4], -20(%[src1]) \n\t" |
| 389 | "lwc1 %[temp6], 24(%[src0]) \n\t" |
| 390 | "lwc1 %[temp7], -24(%[src1]) \n\t" |
| 391 | "lwc1 %[temp9], 28(%[src0]) \n\t" |
| 392 | "lwc1 %[temp10], -28(%[src1]) \n\t" |
| 393 | "add.s %[temp2], %[temp0], %[temp1] \n\t" |
| 394 | "add.s %[temp5], %[temp3], %[temp4] \n\t" |
| 395 | "add.s %[temp8], %[temp6], %[temp7] \n\t" |
| 396 | "add.s %[temp11], %[temp9], %[temp10] \n\t" |
| 397 | "sub.s %[temp0], %[temp0], %[temp1] \n\t" |
| 398 | "sub.s %[temp3], %[temp3], %[temp4] \n\t" |
| 399 | "sub.s %[temp6], %[temp6], %[temp7] \n\t" |
| 400 | "sub.s %[temp9], %[temp9], %[temp10] \n\t" |
| 401 | "swc1 %[temp2], -16(%[v1]) \n\t" |
| 402 | "swc1 %[temp0], 16(%[v0]) \n\t" |
| 403 | "swc1 %[temp5], -20(%[v1]) \n\t" |
| 404 | "swc1 %[temp3], 20(%[v0]) \n\t" |
| 405 | "swc1 %[temp8], -24(%[v1]) \n\t" |
| 406 | "swc1 %[temp6], 24(%[v0]) \n\t" |
| 407 | "swc1 %[temp11], -28(%[v1]) \n\t" |
| 408 | "swc1 %[temp9], 28(%[v0]) \n\t" |
| 409 | "lwc1 %[temp0], 32(%[src0]) \n\t" |
| 410 | "lwc1 %[temp1], -32(%[src1]) \n\t" |
| 411 | "lwc1 %[temp3], 36(%[src0]) \n\t" |
| 412 | "lwc1 %[temp4], -36(%[src1]) \n\t" |
| 413 | "lwc1 %[temp6], 40(%[src0]) \n\t" |
| 414 | "lwc1 %[temp7], -40(%[src1]) \n\t" |
| 415 | "lwc1 %[temp9], 44(%[src0]) \n\t" |
| 416 | "lwc1 %[temp10], -44(%[src1]) \n\t" |
| 417 | "add.s %[temp2], %[temp0], %[temp1] \n\t" |
| 418 | "add.s %[temp5], %[temp3], %[temp4] \n\t" |
| 419 | "add.s %[temp8], %[temp6], %[temp7] \n\t" |
| 420 | "add.s %[temp11], %[temp9], %[temp10] \n\t" |
| 421 | "sub.s %[temp0], %[temp0], %[temp1] \n\t" |
| 422 | "sub.s %[temp3], %[temp3], %[temp4] \n\t" |
| 423 | "sub.s %[temp6], %[temp6], %[temp7] \n\t" |
| 424 | "sub.s %[temp9], %[temp9], %[temp10] \n\t" |
| 425 | "swc1 %[temp2], -32(%[v1]) \n\t" |
| 426 | "swc1 %[temp0], 32(%[v0]) \n\t" |
| 427 | "swc1 %[temp5], -36(%[v1]) \n\t" |
| 428 | "swc1 %[temp3], 36(%[v0]) \n\t" |
| 429 | "swc1 %[temp8], -40(%[v1]) \n\t" |
| 430 | "swc1 %[temp6], 40(%[v0]) \n\t" |
| 431 | "swc1 %[temp11], -44(%[v1]) \n\t" |
| 432 | "swc1 %[temp9], 44(%[v0]) \n\t" |
| 433 | "lwc1 %[temp0], 48(%[src0]) \n\t" |
| 434 | "lwc1 %[temp1], -48(%[src1]) \n\t" |
| 435 | "lwc1 %[temp3], 52(%[src0]) \n\t" |
| 436 | "lwc1 %[temp4], -52(%[src1]) \n\t" |
| 437 | "lwc1 %[temp6], 56(%[src0]) \n\t" |
| 438 | "lwc1 %[temp7], -56(%[src1]) \n\t" |
| 439 | "lwc1 %[temp9], 60(%[src0]) \n\t" |
| 440 | "lwc1 %[temp10], -60(%[src1]) \n\t" |
| 441 | "add.s %[temp2], %[temp0], %[temp1] \n\t" |
| 442 | "add.s %[temp5], %[temp3], %[temp4] \n\t" |
| 443 | "add.s %[temp8], %[temp6], %[temp7] \n\t" |
| 444 | "add.s %[temp11], %[temp9], %[temp10] \n\t" |
| 445 | "sub.s %[temp0], %[temp0], %[temp1] \n\t" |
| 446 | "sub.s %[temp3], %[temp3], %[temp4] \n\t" |
| 447 | "sub.s %[temp6], %[temp6], %[temp7] \n\t" |
| 448 | "sub.s %[temp9], %[temp9], %[temp10] \n\t" |
| 449 | "swc1 %[temp2], -48(%[v1]) \n\t" |
| 450 | "swc1 %[temp0], 48(%[v0]) \n\t" |
| 451 | "swc1 %[temp5], -52(%[v1]) \n\t" |
| 452 | "swc1 %[temp3], 52(%[v0]) \n\t" |
| 453 | "swc1 %[temp8], -56(%[v1]) \n\t" |
| 454 | "swc1 %[temp6], 56(%[v0]) \n\t" |
| 455 | "swc1 %[temp11], -60(%[v1]) \n\t" |
| 456 | "swc1 %[temp9], 60(%[v0]) \n\t" |
| 457 | "addiu %[src0], %[src0], 64 \n\t" |
| 458 | "addiu %[src1], %[src1], -64 \n\t" |
| 459 | "addiu %[v0], %[v0], 64 \n\t" |
| 460 | "addiu %[v1], %[v1], -64 \n\t" |
| 461 | |
| 462 | : [v0]"+r"(v0), [v1]"+r"(v1), [src0]"+r"(psrc0), [src1]"+r"(psrc1), |
| 463 | [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 464 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 465 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8), |
| 466 | [temp9]"=&f"(temp9), [temp10]"=&f"(temp10), [temp11]"=&f"(temp11) |
| 467 | : |
| 468 | :"memory" |
| 469 | ); |
| 470 | } |
| 471 | } |
| 472 | |
| 473 | static void sbr_autocorrelate_mips(const float x[40][2], float phi[3][2][2]) |
| 474 | { |
| 475 | int i; |
| 476 | float real_sum_0 = 0.0f; |
| 477 | float real_sum_1 = 0.0f; |
| 478 | float real_sum_2 = 0.0f; |
| 479 | float imag_sum_1 = 0.0f; |
| 480 | float imag_sum_2 = 0.0f; |
| 481 | float *p_x, *p_phi; |
| 482 | float temp0, temp1, temp2, temp3, temp4, temp5, temp6; |
| 483 | float temp7, temp_r, temp_r1, temp_r2, temp_r3, temp_r4; |
| 484 | p_x = (float*)&x[0][0]; |
| 485 | p_phi = &phi[0][0][0]; |
| 486 | |
| 487 | __asm__ volatile ( |
| 488 | "lwc1 %[temp0], 8(%[p_x]) \n\t" |
| 489 | "lwc1 %[temp1], 12(%[p_x]) \n\t" |
| 490 | "lwc1 %[temp2], 16(%[p_x]) \n\t" |
| 491 | "lwc1 %[temp3], 20(%[p_x]) \n\t" |
| 492 | "lwc1 %[temp4], 24(%[p_x]) \n\t" |
| 493 | "lwc1 %[temp5], 28(%[p_x]) \n\t" |
| 494 | "mul.s %[temp_r], %[temp1], %[temp1] \n\t" |
| 495 | "mul.s %[temp_r1], %[temp1], %[temp3] \n\t" |
| 496 | "mul.s %[temp_r2], %[temp1], %[temp2] \n\t" |
| 497 | "mul.s %[temp_r3], %[temp1], %[temp5] \n\t" |
| 498 | "mul.s %[temp_r4], %[temp1], %[temp4] \n\t" |
| 499 | "madd.s %[temp_r], %[temp_r], %[temp0], %[temp0] \n\t" |
| 500 | "madd.s %[temp_r1], %[temp_r1], %[temp0], %[temp2] \n\t" |
| 501 | "msub.s %[temp_r2], %[temp_r2], %[temp0], %[temp3] \n\t" |
| 502 | "madd.s %[temp_r3], %[temp_r3], %[temp0], %[temp4] \n\t" |
| 503 | "msub.s %[temp_r4], %[temp_r4], %[temp0], %[temp5] \n\t" |
| 504 | "add.s %[real_sum_0], %[real_sum_0], %[temp_r] \n\t" |
| 505 | "add.s %[real_sum_1], %[real_sum_1], %[temp_r1] \n\t" |
| 506 | "add.s %[imag_sum_1], %[imag_sum_1], %[temp_r2] \n\t" |
| 507 | "add.s %[real_sum_2], %[real_sum_2], %[temp_r3] \n\t" |
| 508 | "add.s %[imag_sum_2], %[imag_sum_2], %[temp_r4] \n\t" |
| 509 | "addiu %[p_x], %[p_x], 8 \n\t" |
| 510 | |
| 511 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 512 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 513 | [real_sum_0]"+f"(real_sum_0), [real_sum_1]"+f"(real_sum_1), |
| 514 | [imag_sum_1]"+f"(imag_sum_1), [real_sum_2]"+f"(real_sum_2), |
| 515 | [temp_r]"=&f"(temp_r), [temp_r1]"=&f"(temp_r1), [temp_r2]"=&f"(temp_r2), |
| 516 | [temp_r3]"=&f"(temp_r3), [temp_r4]"=&f"(temp_r4), |
| 517 | [p_x]"+r"(p_x), [imag_sum_2]"+f"(imag_sum_2) |
| 518 | : |
| 519 | : "memory" |
| 520 | ); |
| 521 | |
| 522 | for (i = 0; i < 12; i++) { |
| 523 | __asm__ volatile ( |
| 524 | "lwc1 %[temp0], 8(%[p_x]) \n\t" |
| 525 | "lwc1 %[temp1], 12(%[p_x]) \n\t" |
| 526 | "lwc1 %[temp2], 16(%[p_x]) \n\t" |
| 527 | "lwc1 %[temp3], 20(%[p_x]) \n\t" |
| 528 | "lwc1 %[temp4], 24(%[p_x]) \n\t" |
| 529 | "lwc1 %[temp5], 28(%[p_x]) \n\t" |
| 530 | "mul.s %[temp_r], %[temp1], %[temp1] \n\t" |
| 531 | "mul.s %[temp_r1], %[temp1], %[temp3] \n\t" |
| 532 | "mul.s %[temp_r2], %[temp1], %[temp2] \n\t" |
| 533 | "mul.s %[temp_r3], %[temp1], %[temp5] \n\t" |
| 534 | "mul.s %[temp_r4], %[temp1], %[temp4] \n\t" |
| 535 | "madd.s %[temp_r], %[temp_r], %[temp0], %[temp0] \n\t" |
| 536 | "madd.s %[temp_r1], %[temp_r1], %[temp0], %[temp2] \n\t" |
| 537 | "msub.s %[temp_r2], %[temp_r2], %[temp0], %[temp3] \n\t" |
| 538 | "madd.s %[temp_r3], %[temp_r3], %[temp0], %[temp4] \n\t" |
| 539 | "msub.s %[temp_r4], %[temp_r4], %[temp0], %[temp5] \n\t" |
| 540 | "add.s %[real_sum_0], %[real_sum_0], %[temp_r] \n\t" |
| 541 | "add.s %[real_sum_1], %[real_sum_1], %[temp_r1] \n\t" |
| 542 | "add.s %[imag_sum_1], %[imag_sum_1], %[temp_r2] \n\t" |
| 543 | "add.s %[real_sum_2], %[real_sum_2], %[temp_r3] \n\t" |
| 544 | "add.s %[imag_sum_2], %[imag_sum_2], %[temp_r4] \n\t" |
| 545 | "lwc1 %[temp0], 32(%[p_x]) \n\t" |
| 546 | "lwc1 %[temp1], 36(%[p_x]) \n\t" |
| 547 | "mul.s %[temp_r], %[temp3], %[temp3] \n\t" |
| 548 | "mul.s %[temp_r1], %[temp3], %[temp5] \n\t" |
| 549 | "mul.s %[temp_r2], %[temp3], %[temp4] \n\t" |
| 550 | "mul.s %[temp_r3], %[temp3], %[temp1] \n\t" |
| 551 | "mul.s %[temp_r4], %[temp3], %[temp0] \n\t" |
| 552 | "madd.s %[temp_r], %[temp_r], %[temp2], %[temp2] \n\t" |
| 553 | "madd.s %[temp_r1], %[temp_r1], %[temp2], %[temp4] \n\t" |
| 554 | "msub.s %[temp_r2], %[temp_r2], %[temp2], %[temp5] \n\t" |
| 555 | "madd.s %[temp_r3], %[temp_r3], %[temp2], %[temp0] \n\t" |
| 556 | "msub.s %[temp_r4], %[temp_r4], %[temp2], %[temp1] \n\t" |
| 557 | "add.s %[real_sum_0], %[real_sum_0], %[temp_r] \n\t" |
| 558 | "add.s %[real_sum_1], %[real_sum_1], %[temp_r1] \n\t" |
| 559 | "add.s %[imag_sum_1], %[imag_sum_1], %[temp_r2] \n\t" |
| 560 | "add.s %[real_sum_2], %[real_sum_2], %[temp_r3] \n\t" |
| 561 | "add.s %[imag_sum_2], %[imag_sum_2], %[temp_r4] \n\t" |
| 562 | "lwc1 %[temp2], 40(%[p_x]) \n\t" |
| 563 | "lwc1 %[temp3], 44(%[p_x]) \n\t" |
| 564 | "mul.s %[temp_r], %[temp5], %[temp5] \n\t" |
| 565 | "mul.s %[temp_r1], %[temp5], %[temp1] \n\t" |
| 566 | "mul.s %[temp_r2], %[temp5], %[temp0] \n\t" |
| 567 | "mul.s %[temp_r3], %[temp5], %[temp3] \n\t" |
| 568 | "mul.s %[temp_r4], %[temp5], %[temp2] \n\t" |
| 569 | "madd.s %[temp_r], %[temp_r], %[temp4], %[temp4] \n\t" |
| 570 | "madd.s %[temp_r1], %[temp_r1], %[temp4], %[temp0] \n\t" |
| 571 | "msub.s %[temp_r2], %[temp_r2], %[temp4], %[temp1] \n\t" |
| 572 | "madd.s %[temp_r3], %[temp_r3], %[temp4], %[temp2] \n\t" |
| 573 | "msub.s %[temp_r4], %[temp_r4], %[temp4], %[temp3] \n\t" |
| 574 | "add.s %[real_sum_0], %[real_sum_0], %[temp_r] \n\t" |
| 575 | "add.s %[real_sum_1], %[real_sum_1], %[temp_r1] \n\t" |
| 576 | "add.s %[imag_sum_1], %[imag_sum_1], %[temp_r2] \n\t" |
| 577 | "add.s %[real_sum_2], %[real_sum_2], %[temp_r3] \n\t" |
| 578 | "add.s %[imag_sum_2], %[imag_sum_2], %[temp_r4] \n\t" |
| 579 | "addiu %[p_x], %[p_x], 24 \n\t" |
| 580 | |
| 581 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 582 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 583 | [real_sum_0]"+f"(real_sum_0), [real_sum_1]"+f"(real_sum_1), |
| 584 | [imag_sum_1]"+f"(imag_sum_1), [real_sum_2]"+f"(real_sum_2), |
| 585 | [temp_r]"=&f"(temp_r), [temp_r1]"=&f"(temp_r1), |
| 586 | [temp_r2]"=&f"(temp_r2), [temp_r3]"=&f"(temp_r3), |
| 587 | [temp_r4]"=&f"(temp_r4), [p_x]"+r"(p_x), |
| 588 | [imag_sum_2]"+f"(imag_sum_2) |
| 589 | : |
| 590 | : "memory" |
| 591 | ); |
| 592 | } |
| 593 | __asm__ volatile ( |
| 594 | "lwc1 %[temp0], -296(%[p_x]) \n\t" |
| 595 | "lwc1 %[temp1], -292(%[p_x]) \n\t" |
| 596 | "lwc1 %[temp2], 8(%[p_x]) \n\t" |
| 597 | "lwc1 %[temp3], 12(%[p_x]) \n\t" |
| 598 | "lwc1 %[temp4], -288(%[p_x]) \n\t" |
| 599 | "lwc1 %[temp5], -284(%[p_x]) \n\t" |
| 600 | "lwc1 %[temp6], -280(%[p_x]) \n\t" |
| 601 | "lwc1 %[temp7], -276(%[p_x]) \n\t" |
| 602 | "madd.s %[temp_r], %[real_sum_0], %[temp0], %[temp0] \n\t" |
| 603 | "madd.s %[temp_r1], %[real_sum_0], %[temp2], %[temp2] \n\t" |
| 604 | "madd.s %[temp_r2], %[real_sum_1], %[temp0], %[temp4] \n\t" |
| 605 | "madd.s %[temp_r3], %[imag_sum_1], %[temp0], %[temp5] \n\t" |
| 606 | "madd.s %[temp_r], %[temp_r], %[temp1], %[temp1] \n\t" |
| 607 | "madd.s %[temp_r1], %[temp_r1], %[temp3], %[temp3] \n\t" |
| 608 | "madd.s %[temp_r2], %[temp_r2], %[temp1], %[temp5] \n\t" |
| 609 | "nmsub.s %[temp_r3], %[temp_r3], %[temp1], %[temp4] \n\t" |
| 610 | "lwc1 %[temp4], 16(%[p_x]) \n\t" |
| 611 | "lwc1 %[temp5], 20(%[p_x]) \n\t" |
| 612 | "swc1 %[temp_r], 40(%[p_phi]) \n\t" |
| 613 | "swc1 %[temp_r1], 16(%[p_phi]) \n\t" |
| 614 | "swc1 %[temp_r2], 24(%[p_phi]) \n\t" |
| 615 | "swc1 %[temp_r3], 28(%[p_phi]) \n\t" |
| 616 | "madd.s %[temp_r], %[real_sum_1], %[temp2], %[temp4] \n\t" |
| 617 | "madd.s %[temp_r1], %[imag_sum_1], %[temp2], %[temp5] \n\t" |
| 618 | "madd.s %[temp_r2], %[real_sum_2], %[temp0], %[temp6] \n\t" |
| 619 | "madd.s %[temp_r3], %[imag_sum_2], %[temp0], %[temp7] \n\t" |
| 620 | "madd.s %[temp_r], %[temp_r], %[temp3], %[temp5] \n\t" |
| 621 | "nmsub.s %[temp_r1], %[temp_r1], %[temp3], %[temp4] \n\t" |
| 622 | "madd.s %[temp_r2], %[temp_r2], %[temp1], %[temp7] \n\t" |
| 623 | "nmsub.s %[temp_r3], %[temp_r3], %[temp1], %[temp6] \n\t" |
| 624 | "swc1 %[temp_r], 0(%[p_phi]) \n\t" |
| 625 | "swc1 %[temp_r1], 4(%[p_phi]) \n\t" |
| 626 | "swc1 %[temp_r2], 8(%[p_phi]) \n\t" |
| 627 | "swc1 %[temp_r3], 12(%[p_phi]) \n\t" |
| 628 | |
| 629 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 630 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 631 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp_r]"=&f"(temp_r), |
| 632 | [real_sum_0]"+f"(real_sum_0), [real_sum_1]"+f"(real_sum_1), |
| 633 | [real_sum_2]"+f"(real_sum_2), [imag_sum_1]"+f"(imag_sum_1), |
| 634 | [temp_r2]"=&f"(temp_r2), [temp_r3]"=&f"(temp_r3), |
| 635 | [temp_r1]"=&f"(temp_r1), [p_phi]"+r"(p_phi), |
| 636 | [imag_sum_2]"+f"(imag_sum_2) |
| 637 | : [p_x]"r"(p_x) |
| 638 | : "memory" |
| 639 | ); |
| 640 | } |
| 641 | |
| 642 | static void sbr_hf_gen_mips(float (*X_high)[2], const float (*X_low)[2], |
| 643 | const float alpha0[2], const float alpha1[2], |
| 644 | float bw, int start, int end) |
| 645 | { |
| 646 | float alpha[4]; |
| 647 | int i; |
| 648 | float *p_x_low = (float*)&X_low[0][0] + 2*start; |
| 649 | float *p_x_high = &X_high[0][0] + 2*start; |
| 650 | float temp0, temp1, temp2, temp3, temp4, temp5, temp6; |
| 651 | float temp7, temp8, temp9, temp10, temp11, temp12; |
| 652 | |
| 653 | alpha[0] = alpha1[0] * bw * bw; |
| 654 | alpha[1] = alpha1[1] * bw * bw; |
| 655 | alpha[2] = alpha0[0] * bw; |
| 656 | alpha[3] = alpha0[1] * bw; |
| 657 | |
| 658 | for (i = start; i < end; i++) { |
| 659 | __asm__ volatile ( |
| 660 | "lwc1 %[temp0], -16(%[p_x_low]) \n\t" |
| 661 | "lwc1 %[temp1], -12(%[p_x_low]) \n\t" |
| 662 | "lwc1 %[temp2], -8(%[p_x_low]) \n\t" |
| 663 | "lwc1 %[temp3], -4(%[p_x_low]) \n\t" |
| 664 | "lwc1 %[temp5], 0(%[p_x_low]) \n\t" |
| 665 | "lwc1 %[temp6], 4(%[p_x_low]) \n\t" |
| 666 | "lwc1 %[temp7], 0(%[alpha]) \n\t" |
| 667 | "lwc1 %[temp8], 4(%[alpha]) \n\t" |
| 668 | "lwc1 %[temp9], 8(%[alpha]) \n\t" |
| 669 | "lwc1 %[temp10], 12(%[alpha]) \n\t" |
| 670 | "addiu %[p_x_high], %[p_x_high], 8 \n\t" |
| 671 | "addiu %[p_x_low], %[p_x_low], 8 \n\t" |
| 672 | "mul.s %[temp11], %[temp1], %[temp8] \n\t" |
| 673 | "msub.s %[temp11], %[temp11], %[temp0], %[temp7] \n\t" |
| 674 | "madd.s %[temp11], %[temp11], %[temp2], %[temp9] \n\t" |
| 675 | "nmsub.s %[temp11], %[temp11], %[temp3], %[temp10] \n\t" |
| 676 | "add.s %[temp11], %[temp11], %[temp5] \n\t" |
| 677 | "swc1 %[temp11], -8(%[p_x_high]) \n\t" |
| 678 | "mul.s %[temp12], %[temp1], %[temp7] \n\t" |
| 679 | "madd.s %[temp12], %[temp12], %[temp0], %[temp8] \n\t" |
| 680 | "madd.s %[temp12], %[temp12], %[temp3], %[temp9] \n\t" |
| 681 | "madd.s %[temp12], %[temp12], %[temp2], %[temp10] \n\t" |
| 682 | "add.s %[temp12], %[temp12], %[temp6] \n\t" |
| 683 | "swc1 %[temp12], -4(%[p_x_high]) \n\t" |
| 684 | |
| 685 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 686 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 687 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8), |
| 688 | [temp9]"=&f"(temp9), [temp10]"=&f"(temp10), [temp11]"=&f"(temp11), |
| 689 | [temp12]"=&f"(temp12), [p_x_high]"+r"(p_x_high), |
| 690 | [p_x_low]"+r"(p_x_low) |
| 691 | : [alpha]"r"(alpha) |
| 692 | : "memory" |
| 693 | ); |
| 694 | } |
| 695 | } |
| 696 | |
| 697 | static void sbr_hf_g_filt_mips(float (*Y)[2], const float (*X_high)[40][2], |
| 698 | const float *g_filt, int m_max, intptr_t ixh) |
| 699 | { |
| 700 | float *p_y, *p_x, *p_g; |
| 701 | float temp0, temp1, temp2; |
| 702 | int loop_end; |
| 703 | |
| 704 | p_g = (float*)&g_filt[0]; |
| 705 | p_y = &Y[0][0]; |
| 706 | p_x = (float*)&X_high[0][ixh][0]; |
| 707 | loop_end = (int)((int*)p_g + m_max); |
| 708 | |
| 709 | __asm__ volatile( |
| 710 | ".set push \n\t" |
| 711 | ".set noreorder \n\t" |
| 712 | "1: \n\t" |
| 713 | "lwc1 %[temp0], 0(%[p_g]) \n\t" |
| 714 | "lwc1 %[temp1], 0(%[p_x]) \n\t" |
| 715 | "lwc1 %[temp2], 4(%[p_x]) \n\t" |
| 716 | "mul.s %[temp1], %[temp1], %[temp0] \n\t" |
| 717 | "mul.s %[temp2], %[temp2], %[temp0] \n\t" |
| 718 | "addiu %[p_g], %[p_g], 4 \n\t" |
| 719 | "addiu %[p_x], %[p_x], 320 \n\t" |
| 720 | "swc1 %[temp1], 0(%[p_y]) \n\t" |
| 721 | "swc1 %[temp2], 4(%[p_y]) \n\t" |
| 722 | "bne %[p_g], %[loop_end], 1b \n\t" |
| 723 | " addiu %[p_y], %[p_y], 8 \n\t" |
| 724 | ".set pop \n\t" |
| 725 | |
| 726 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), |
| 727 | [temp2]"=&f"(temp2), [p_x]"+r"(p_x), |
| 728 | [p_y]"+r"(p_y), [p_g]"+r"(p_g) |
| 729 | : [loop_end]"r"(loop_end) |
| 730 | : "memory" |
| 731 | ); |
| 732 | } |
| 733 | |
| 734 | static void sbr_hf_apply_noise_0_mips(float (*Y)[2], const float *s_m, |
| 735 | const float *q_filt, int noise, |
| 736 | int kx, int m_max) |
| 737 | { |
| 738 | int m; |
| 739 | |
| 740 | for (m = 0; m < m_max; m++){ |
| 741 | |
| 742 | float *Y1=&Y[m][0]; |
| 743 | float *ff_table; |
| 744 | float y0,y1, temp1, temp2, temp4, temp5; |
| 745 | int temp0, temp3; |
| 746 | const float *s_m1=&s_m[m]; |
| 747 | const float *q_filt1= &q_filt[m]; |
| 748 | |
| 749 | __asm__ volatile( |
| 750 | "lwc1 %[y0], 0(%[Y1]) \n\t" |
| 751 | "lwc1 %[temp1], 0(%[s_m1]) \n\t" |
| 752 | "addiu %[noise], %[noise], 1 \n\t" |
| 753 | "andi %[noise], %[noise], 0x1ff \n\t" |
| 754 | "sll %[temp0], %[noise], 3 \n\t" |
| 755 | "addu %[ff_table], %[ff_sbr_noise_table], %[temp0] \n\t" |
| 756 | "add.s %[y0], %[y0], %[temp1] \n\t" |
| 757 | "mfc1 %[temp3], %[temp1] \n\t" |
| 758 | "bne %[temp3], $0, 1f \n\t" |
| 759 | "lwc1 %[y1], 4(%[Y1]) \n\t" |
| 760 | "lwc1 %[temp2], 0(%[q_filt1]) \n\t" |
| 761 | "lwc1 %[temp4], 0(%[ff_table]) \n\t" |
| 762 | "lwc1 %[temp5], 4(%[ff_table]) \n\t" |
| 763 | "madd.s %[y0], %[y0], %[temp2], %[temp4] \n\t" |
| 764 | "madd.s %[y1], %[y1], %[temp2], %[temp5] \n\t" |
| 765 | "swc1 %[y1], 4(%[Y1]) \n\t" |
| 766 | "1: \n\t" |
| 767 | "swc1 %[y0], 0(%[Y1]) \n\t" |
| 768 | |
| 769 | : [ff_table]"=&r"(ff_table), [y0]"=&f"(y0), [y1]"=&f"(y1), |
| 770 | [temp0]"=&r"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 771 | [temp3]"=&r"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5) |
| 772 | : [ff_sbr_noise_table]"r"(ff_sbr_noise_table), [noise]"r"(noise), |
| 773 | [Y1]"r"(Y1), [s_m1]"r"(s_m1), [q_filt1]"r"(q_filt1) |
| 774 | : "memory" |
| 775 | ); |
| 776 | } |
| 777 | } |
| 778 | |
| 779 | static void sbr_hf_apply_noise_1_mips(float (*Y)[2], const float *s_m, |
| 780 | const float *q_filt, int noise, |
| 781 | int kx, int m_max) |
| 782 | { |
| 783 | float y0,y1,temp1, temp2, temp4, temp5; |
| 784 | int temp0, temp3, m; |
| 785 | float phi_sign = 1 - 2 * (kx & 1); |
| 786 | |
| 787 | for (m = 0; m < m_max; m++) { |
| 788 | |
| 789 | float *ff_table; |
| 790 | float *Y1=&Y[m][0]; |
| 791 | const float *s_m1=&s_m[m]; |
| 792 | const float *q_filt1= &q_filt[m]; |
| 793 | |
| 794 | __asm__ volatile( |
| 795 | "lwc1 %[y1], 4(%[Y1]) \n\t" |
| 796 | "lwc1 %[temp1], 0(%[s_m1]) \n\t" |
| 797 | "lw %[temp3], 0(%[s_m1]) \n\t" |
| 798 | "addiu %[noise], %[noise], 1 \n\t" |
| 799 | "andi %[noise], %[noise], 0x1ff \n\t" |
| 800 | "sll %[temp0], %[noise], 3 \n\t" |
| 801 | "addu %[ff_table], %[ff_sbr_noise_table], %[temp0] \n\t" |
| 802 | "madd.s %[y1], %[y1], %[temp1], %[phi_sign] \n\t" |
| 803 | "bne %[temp3], $0, 1f \n\t" |
| 804 | "lwc1 %[y0], 0(%[Y1]) \n\t" |
| 805 | "lwc1 %[temp2], 0(%[q_filt1]) \n\t" |
| 806 | "lwc1 %[temp4], 0(%[ff_table]) \n\t" |
| 807 | "lwc1 %[temp5], 4(%[ff_table]) \n\t" |
| 808 | "madd.s %[y0], %[y0], %[temp2], %[temp4] \n\t" |
| 809 | "madd.s %[y1], %[y1], %[temp2], %[temp5] \n\t" |
| 810 | "swc1 %[y0], 0(%[Y1]) \n\t" |
| 811 | "1: \n\t" |
| 812 | "swc1 %[y1], 4(%[Y1]) \n\t" |
| 813 | |
| 814 | : [ff_table] "=&r" (ff_table), [y0] "=&f" (y0), [y1] "=&f" (y1), |
| 815 | [temp0] "=&r" (temp0), [temp1] "=&f" (temp1), [temp2] "=&f" (temp2), |
| 816 | [temp3] "=&r" (temp3), [temp4] "=&f" (temp4), [temp5] "=&f" (temp5) |
| 817 | : [ff_sbr_noise_table] "r" (ff_sbr_noise_table), [noise] "r" (noise), |
| 818 | [Y1] "r" (Y1), [s_m1] "r" (s_m1), [q_filt1] "r" (q_filt1), |
| 819 | [phi_sign] "f" (phi_sign) |
| 820 | : "memory" |
| 821 | ); |
| 822 | phi_sign = -phi_sign; |
| 823 | } |
| 824 | } |
| 825 | |
| 826 | static void sbr_hf_apply_noise_2_mips(float (*Y)[2], const float *s_m, |
| 827 | const float *q_filt, int noise, |
| 828 | int kx, int m_max) |
| 829 | { |
| 830 | int m; |
| 831 | float *ff_table; |
| 832 | float y0,y1, temp0, temp1, temp2, temp3, temp4, temp5; |
| 833 | |
| 834 | for (m = 0; m < m_max; m++) { |
| 835 | |
| 836 | float *Y1=&Y[m][0]; |
| 837 | const float *s_m1=&s_m[m]; |
| 838 | const float *q_filt1= &q_filt[m]; |
| 839 | |
| 840 | __asm__ volatile( |
| 841 | "lwc1 %[y0], 0(%[Y1]) \n\t" |
| 842 | "lwc1 %[temp1], 0(%[s_m1]) \n\t" |
| 843 | "addiu %[noise], %[noise], 1 \n\t" |
| 844 | "andi %[noise], %[noise], 0x1ff \n\t" |
| 845 | "sll %[temp0], %[noise], 3 \n\t" |
| 846 | "addu %[ff_table], %[ff_sbr_noise_table], %[temp0] \n\t" |
| 847 | "sub.s %[y0], %[y0], %[temp1] \n\t" |
| 848 | "mfc1 %[temp3], %[temp1] \n\t" |
| 849 | "bne %[temp3], $0, 1f \n\t" |
| 850 | "lwc1 %[y1], 4(%[Y1]) \n\t" |
| 851 | "lwc1 %[temp2], 0(%[q_filt1]) \n\t" |
| 852 | "lwc1 %[temp4], 0(%[ff_table]) \n\t" |
| 853 | "lwc1 %[temp5], 4(%[ff_table]) \n\t" |
| 854 | "madd.s %[y0], %[y0], %[temp2], %[temp4] \n\t" |
| 855 | "madd.s %[y1], %[y1], %[temp2], %[temp5] \n\t" |
| 856 | "swc1 %[y1], 4(%[Y1]) \n\t" |
| 857 | "1: \n\t" |
| 858 | "swc1 %[y0], 0(%[Y1]) \n\t" |
| 859 | |
| 860 | : [temp0]"=&r"(temp0), [ff_table]"=&r"(ff_table), [y0]"=&f"(y0), |
| 861 | [y1]"=&f"(y1), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 862 | [temp3]"=&r"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5) |
| 863 | : [ff_sbr_noise_table]"r"(ff_sbr_noise_table), [noise]"r"(noise), |
| 864 | [Y1]"r"(Y1), [s_m1]"r"(s_m1), [q_filt1]"r"(q_filt1) |
| 865 | : "memory" |
| 866 | ); |
| 867 | } |
| 868 | } |
| 869 | |
| 870 | static void sbr_hf_apply_noise_3_mips(float (*Y)[2], const float *s_m, |
| 871 | const float *q_filt, int noise, |
| 872 | int kx, int m_max) |
| 873 | { |
| 874 | float phi_sign = 1 - 2 * (kx & 1); |
| 875 | int m; |
| 876 | |
| 877 | for (m = 0; m < m_max; m++) { |
| 878 | |
| 879 | float *Y1=&Y[m][0]; |
| 880 | float *ff_table; |
| 881 | float y0,y1, temp1, temp2, temp4, temp5; |
| 882 | int temp0, temp3; |
| 883 | const float *s_m1=&s_m[m]; |
| 884 | const float *q_filt1= &q_filt[m]; |
| 885 | |
| 886 | __asm__ volatile( |
| 887 | "lwc1 %[y1], 4(%[Y1]) \n\t" |
| 888 | "lwc1 %[temp1], 0(%[s_m1]) \n\t" |
| 889 | "addiu %[noise], %[noise], 1 \n\t" |
| 890 | "andi %[noise], %[noise], 0x1ff \n\t" |
| 891 | "sll %[temp0], %[noise], 3 \n\t" |
| 892 | "addu %[ff_table], %[ff_sbr_noise_table], %[temp0] \n\t" |
| 893 | "nmsub.s %[y1], %[y1], %[temp1], %[phi_sign] \n\t" |
| 894 | "mfc1 %[temp3], %[temp1] \n\t" |
| 895 | "bne %[temp3], $0, 1f \n\t" |
| 896 | "lwc1 %[y0], 0(%[Y1]) \n\t" |
| 897 | "lwc1 %[temp2], 0(%[q_filt1]) \n\t" |
| 898 | "lwc1 %[temp4], 0(%[ff_table]) \n\t" |
| 899 | "lwc1 %[temp5], 4(%[ff_table]) \n\t" |
| 900 | "madd.s %[y0], %[y0], %[temp2], %[temp4] \n\t" |
| 901 | "madd.s %[y1], %[y1], %[temp2], %[temp5] \n\t" |
| 902 | "swc1 %[y0], 0(%[Y1]) \n\t" |
| 903 | "1: \n\t" |
| 904 | "swc1 %[y1], 4(%[Y1]) \n\t" |
| 905 | |
| 906 | : [ff_table]"=&r"(ff_table), [y0]"=&f"(y0), [y1]"=&f"(y1), |
| 907 | [temp0]"=&r"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 908 | [temp3]"=&r"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5) |
| 909 | : [ff_sbr_noise_table]"r"(ff_sbr_noise_table), [noise]"r"(noise), |
| 910 | [Y1]"r"(Y1), [s_m1]"r"(s_m1), [q_filt1]"r"(q_filt1), |
| 911 | [phi_sign]"f"(phi_sign) |
| 912 | : "memory" |
| 913 | ); |
| 914 | phi_sign = -phi_sign; |
| 915 | } |
| 916 | } |
| 917 | #endif /* HAVE_MIPSFPU */ |
| 918 | #endif /* HAVE_INLINE_ASM */ |
| 919 | |
| 920 | void ff_sbrdsp_init_mips(SBRDSPContext *s) |
| 921 | { |
| 922 | #if HAVE_INLINE_ASM |
| 923 | s->neg_odd_64 = sbr_neg_odd_64_mips; |
| 924 | s->qmf_pre_shuffle = sbr_qmf_pre_shuffle_mips; |
| 925 | s->qmf_post_shuffle = sbr_qmf_post_shuffle_mips; |
| 926 | #if HAVE_MIPSFPU |
| 927 | s->sum64x5 = sbr_sum64x5_mips; |
| 928 | s->sum_square = sbr_sum_square_mips; |
| 929 | s->qmf_deint_bfly = sbr_qmf_deint_bfly_mips; |
| 930 | s->autocorrelate = sbr_autocorrelate_mips; |
| 931 | s->hf_gen = sbr_hf_gen_mips; |
| 932 | s->hf_g_filt = sbr_hf_g_filt_mips; |
| 933 | |
| 934 | s->hf_apply_noise[0] = sbr_hf_apply_noise_0_mips; |
| 935 | s->hf_apply_noise[1] = sbr_hf_apply_noise_1_mips; |
| 936 | s->hf_apply_noise[2] = sbr_hf_apply_noise_2_mips; |
| 937 | s->hf_apply_noise[3] = sbr_hf_apply_noise_3_mips; |
| 938 | #endif /* HAVE_MIPSFPU */ |
| 939 | #endif /* HAVE_INLINE_ASM */ |
| 940 | } |