| 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 | * Author: Bojan Zivkovic (bojan@mips.com) |
| 30 | * |
| 31 | * MPEG Audio decoder optimized for MIPS floating-point architecture |
| 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/mpegaudiodsp_template.c |
| 53 | * libavcodec/dct32.c |
| 54 | */ |
| 55 | |
| 56 | #include <string.h> |
| 57 | |
| 58 | #include "libavcodec/mpegaudiodsp.h" |
| 59 | |
| 60 | static void ff_mpadsp_apply_window_mips_float(float *synth_buf, float *window, |
| 61 | int *dither_state, float *samples, int incr) |
| 62 | { |
| 63 | register const float *w, *w2, *p; |
| 64 | int j; |
| 65 | float *samples2; |
| 66 | float sum, sum2; |
| 67 | /* temporary variables */ |
| 68 | int incr1 = incr << 2; |
| 69 | int t_sample; |
| 70 | float in1, in2, in3, in4, in5, in6, in7, in8; |
| 71 | float *p2; |
| 72 | |
| 73 | /* copy to avoid wrap */ |
| 74 | memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf)); |
| 75 | |
| 76 | /** |
| 77 | * instructions are scheduled to minimize pipeline stall. |
| 78 | * use of round_sample function from the original code is |
| 79 | * changed with appropriate assembly instructions. |
| 80 | */ |
| 81 | |
| 82 | __asm__ volatile ( |
| 83 | "lwc1 %[sum], 0(%[dither_state]) \t\n" |
| 84 | "sll %[t_sample], %[incr1], 5 \t\n" |
| 85 | "sub %[t_sample], %[t_sample], %[incr1] \n\t" |
| 86 | "li %[j], 4 \t\n" |
| 87 | "lwc1 %[in1], 0(%[window]) \t\n" |
| 88 | "lwc1 %[in2], 16*4(%[synth_buf]) \t\n" |
| 89 | "sw $zero, 0(%[dither_state]) \t\n" |
| 90 | "lwc1 %[in3], 64*4(%[window]) \t\n" |
| 91 | "lwc1 %[in4], 80*4(%[synth_buf]) \t\n" |
| 92 | "addu %[samples2], %[samples], %[t_sample] \t\n" |
| 93 | "madd.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 94 | "lwc1 %[in5], 128*4(%[window]) \t\n" |
| 95 | "lwc1 %[in6], 144*4(%[synth_buf]) \t\n" |
| 96 | "lwc1 %[in7], 192*4(%[window]) \t\n" |
| 97 | "madd.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 98 | "lwc1 %[in8], 208*4(%[synth_buf]) \t\n" |
| 99 | "lwc1 %[in1], 256*4(%[window]) \t\n" |
| 100 | "lwc1 %[in2], 272*4(%[synth_buf]) \t\n" |
| 101 | "madd.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 102 | "lwc1 %[in3], 320*4(%[window]) \t\n" |
| 103 | "lwc1 %[in4], 336*4(%[synth_buf]) \t\n" |
| 104 | "lwc1 %[in5], 384*4(%[window]) \t\n" |
| 105 | "madd.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 106 | "lwc1 %[in6], 400*4(%[synth_buf]) \t\n" |
| 107 | "lwc1 %[in7], 448*4(%[window]) \t\n" |
| 108 | "lwc1 %[in8], 464*4(%[synth_buf]) \t\n" |
| 109 | "madd.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 110 | "lwc1 %[in1], 32*4(%[window]) \t\n" |
| 111 | "lwc1 %[in2], 48*4(%[synth_buf]) \t\n" |
| 112 | "madd.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 113 | "lwc1 %[in3], 96*4(%[window]) \t\n" |
| 114 | "lwc1 %[in4], 112*4(%[synth_buf]) \t\n" |
| 115 | "madd.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 116 | "lwc1 %[in5], 160*4(%[window]) \t\n" |
| 117 | "lwc1 %[in6], 176*4(%[synth_buf]) \t\n" |
| 118 | "madd.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 119 | "lwc1 %[in7], 224*4(%[window]) \t\n" |
| 120 | "lwc1 %[in8], 240*4(%[synth_buf]) \t\n" |
| 121 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 122 | "lwc1 %[in1], 288*4(%[window]) \t\n" |
| 123 | "lwc1 %[in2], 304*4(%[synth_buf]) \t\n" |
| 124 | "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 125 | "lwc1 %[in3], 352*4(%[window]) \t\n" |
| 126 | "lwc1 %[in4], 368*4(%[synth_buf]) \t\n" |
| 127 | "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 128 | "lwc1 %[in5], 416*4(%[window]) \t\n" |
| 129 | "lwc1 %[in6], 432*4(%[synth_buf]) \t\n" |
| 130 | "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 131 | "lwc1 %[in7], 480*4(%[window]) \t\n" |
| 132 | "lwc1 %[in8], 496*4(%[synth_buf]) \t\n" |
| 133 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 134 | "addu %[w], %[window], 4 \t\n" |
| 135 | "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 136 | "addu %[w2], %[window], 124 \t\n" |
| 137 | "addiu %[p], %[synth_buf], 68 \t\n" |
| 138 | "addiu %[p2], %[synth_buf], 188 \t\n" |
| 139 | "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 140 | "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 141 | "swc1 %[sum], 0(%[samples]) \t\n" |
| 142 | "addu %[samples], %[samples], %[incr1] \t\n" |
| 143 | |
| 144 | /* calculate two samples at the same time to avoid one memory |
| 145 | access per two sample */ |
| 146 | |
| 147 | "ff_mpadsp_apply_window_loop%=: \t\n" |
| 148 | "lwc1 %[in1], 0(%[w]) \t\n" |
| 149 | "lwc1 %[in2], 0(%[p]) \t\n" |
| 150 | "lwc1 %[in3], 0(%[w2]) \t\n" |
| 151 | "lwc1 %[in4], 64*4(%[w]) \t\n" |
| 152 | "lwc1 %[in5], 64*4(%[p]) \t\n" |
| 153 | "lwc1 %[in6], 64*4(%[w2]) \t\n" |
| 154 | "mul.s %[sum], %[in1], %[in2] \t\n" |
| 155 | "mul.s %[sum2], %[in2], %[in3] \t\n" |
| 156 | "lwc1 %[in1], 128*4(%[w]) \t\n" |
| 157 | "lwc1 %[in2], 128*4(%[p]) \t\n" |
| 158 | "madd.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 159 | "nmadd.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 160 | "lwc1 %[in3], 128*4(%[w2]) \t\n" |
| 161 | "lwc1 %[in4], 192*4(%[w]) \t\n" |
| 162 | "madd.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 163 | "lwc1 %[in5], 192*4(%[p]) \t\n" |
| 164 | "lwc1 %[in6], 192*4(%[w2]) \t\n" |
| 165 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 166 | "lwc1 %[in1], 256*4(%[w]) \t\n" |
| 167 | "lwc1 %[in2], 256*4(%[p]) \t\n" |
| 168 | "madd.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 169 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 170 | "lwc1 %[in3], 256*4(%[w2]) \t\n" |
| 171 | "lwc1 %[in4], 320*4(%[w]) \t\n" |
| 172 | "madd.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 173 | "lwc1 %[in5], 320*4(%[p]) \t\n" |
| 174 | "lwc1 %[in6], 320*4(%[w2]) \t\n" |
| 175 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 176 | "lwc1 %[in1], 384*4(%[w]) \t\n" |
| 177 | "lwc1 %[in2], 384*4(%[p]) \t\n" |
| 178 | "madd.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 179 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 180 | "lwc1 %[in3], 384*4(%[w2]) \t\n" |
| 181 | "lwc1 %[in4], 448*4(%[w]) \t\n" |
| 182 | "madd.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 183 | "lwc1 %[in5], 448*4(%[p]) \t\n" |
| 184 | "lwc1 %[in6], 448*4(%[w2]) \t\n" |
| 185 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 186 | "madd.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 187 | "lwc1 %[in1], 32*4(%[w]) \t\n" |
| 188 | "lwc1 %[in2], 0(%[p2]) \t\n" |
| 189 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 190 | "lwc1 %[in3], 32*4(%[w2]) \t\n" |
| 191 | "lwc1 %[in4], 96*4(%[w]) \t\n" |
| 192 | "lwc1 %[in5], 64*4(%[p2]) \t\n" |
| 193 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 194 | "lwc1 %[in6], 96*4(%[w2]) \t\n" |
| 195 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 196 | "lwc1 %[in1], 160*4(%[w]) \t\n" |
| 197 | "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 198 | "lwc1 %[in2], 128*4(%[p2]) \t\n" |
| 199 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 200 | "lwc1 %[in3], 160*4(%[w2]) \t\n" |
| 201 | "lwc1 %[in4], 224*4(%[w]) \t\n" |
| 202 | "lwc1 %[in5], 192*4(%[p2]) \t\n" |
| 203 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 204 | "lwc1 %[in6], 224*4(%[w2]) \t\n" |
| 205 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 206 | "lwc1 %[in1], 288*4(%[w]) \t\n" |
| 207 | "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 208 | "lwc1 %[in2], 256*4(%[p2]) \t\n" |
| 209 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 210 | "lwc1 %[in3], 288*4(%[w2]) \t\n" |
| 211 | "lwc1 %[in4], 352*4(%[w]) \t\n" |
| 212 | "lwc1 %[in5], 320*4(%[p2]) \t\n" |
| 213 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 214 | "lwc1 %[in6], 352*4(%[w2]) \t\n" |
| 215 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 216 | "lwc1 %[in1], 416*4(%[w]) \t\n" |
| 217 | "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 218 | "lwc1 %[in2], 384*4(%[p2]) \t\n" |
| 219 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 220 | "lwc1 %[in3], 416*4(%[w2]) \t\n" |
| 221 | "lwc1 %[in4], 480*4(%[w]) \t\n" |
| 222 | "lwc1 %[in5], 448*4(%[p2]) \t\n" |
| 223 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 224 | "lwc1 %[in6], 480*4(%[w2]) \t\n" |
| 225 | "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n" |
| 226 | "addiu %[w], %[w], 4 \t\n" |
| 227 | "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n" |
| 228 | "addiu %[w2], %[w2], -4 \t\n" |
| 229 | "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n" |
| 230 | "addu %[j], %[j], 4 \t\n" |
| 231 | "addiu %[p], 4 \t\n" |
| 232 | "swc1 %[sum], 0(%[samples]) \t\n" |
| 233 | "addiu %[p2], -4 \t\n" |
| 234 | "swc1 %[sum2], 0(%[samples2]) \t\n" |
| 235 | "addu %[samples], %[samples], %[incr1] \t\n" |
| 236 | "subu %[samples2], %[samples2], %[incr1] \t\n" |
| 237 | "bne %[j], 64, ff_mpadsp_apply_window_loop%= \t\n" |
| 238 | |
| 239 | "lwc1 %[in1], 48*4(%[window]) \t\n" |
| 240 | "lwc1 %[in2], 32*4(%[synth_buf]) \t\n" |
| 241 | "lwc1 %[in3], 112*4(%[window]) \t\n" |
| 242 | "lwc1 %[in4], 96*4(%[synth_buf]) \t\n" |
| 243 | "lwc1 %[in5], 176*4(%[window]) \t\n" |
| 244 | "lwc1 %[in6], 160*4(%[synth_buf]) \t\n" |
| 245 | "mul.s %[sum], %[in1], %[in2] \t\n" |
| 246 | "lwc1 %[in7], 240*4(%[window]) \t\n" |
| 247 | "lwc1 %[in8], 224*4(%[synth_buf]) \t\n" |
| 248 | "lwc1 %[in1], 304*4(%[window]) \t\n" |
| 249 | "nmadd.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 250 | "lwc1 %[in2], 288*4(%[synth_buf]) \t\n" |
| 251 | "lwc1 %[in3], 368*4(%[window]) \t\n" |
| 252 | "lwc1 %[in4], 352*4(%[synth_buf]) \t\n" |
| 253 | "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 254 | "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 255 | "lwc1 %[in5], 432*4(%[window]) \t\n" |
| 256 | "lwc1 %[in6], 416*4(%[synth_buf]) \t\n" |
| 257 | "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n" |
| 258 | "lwc1 %[in7], 496*4(%[window]) \t\n" |
| 259 | "lwc1 %[in8], 480*4(%[synth_buf]) \t\n" |
| 260 | "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n" |
| 261 | "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n" |
| 262 | "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n" |
| 263 | "swc1 %[sum], 0(%[samples]) \t\n" |
| 264 | |
| 265 | : [sum] "=&f" (sum), [sum2] "=&f" (sum2), |
| 266 | [w2] "=&r" (w2), [w] "=&r" (w), |
| 267 | [p] "=&r" (p), [p2] "=&r" (p2), [j] "=&r" (j), |
| 268 | [samples] "+r" (samples), [samples2] "=&r" (samples2), |
| 269 | [in1] "=&f" (in1), [in2] "=&f" (in2), |
| 270 | [in3] "=&f" (in3), [in4] "=&f" (in4), |
| 271 | [in5] "=&f" (in5), [in6] "=&f" (in6), |
| 272 | [in7] "=&f" (in7), [in8] "=&f" (in8), |
| 273 | [t_sample] "=&r" (t_sample) |
| 274 | : [synth_buf] "r" (synth_buf), [window] "r" (window), |
| 275 | [dither_state] "r" (dither_state), [incr1] "r" (incr1) |
| 276 | : "memory" |
| 277 | ); |
| 278 | } |
| 279 | |
| 280 | static void ff_dct32_mips_float(float *out, const float *tab) |
| 281 | { |
| 282 | float val0 , val1 , val2 , val3 , val4 , val5 , val6 , val7, |
| 283 | val8 , val9 , val10, val11, val12, val13, val14, val15, |
| 284 | val16, val17, val18, val19, val20, val21, val22, val23, |
| 285 | val24, val25, val26, val27, val28, val29, val30, val31; |
| 286 | float fTmp1, fTmp2, fTmp3, fTmp4, fTmp5, fTmp6, fTmp7, fTmp8, |
| 287 | fTmp9, fTmp10, fTmp11; |
| 288 | |
| 289 | /** |
| 290 | * instructions are scheduled to minimize pipeline stall. |
| 291 | */ |
| 292 | __asm__ volatile ( |
| 293 | "lwc1 %[fTmp1], 0*4(%[tab]) \n\t" |
| 294 | "lwc1 %[fTmp2], 31*4(%[tab]) \n\t" |
| 295 | "lwc1 %[fTmp3], 15*4(%[tab]) \n\t" |
| 296 | "lwc1 %[fTmp4], 16*4(%[tab]) \n\t" |
| 297 | "li.s %[fTmp7], 0.50241928618815570551 \n\t" |
| 298 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 299 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 300 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 301 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 302 | "li.s %[fTmp10], 0.50060299823519630134 \n\t" |
| 303 | "li.s %[fTmp11], 10.19000812354805681150 \n\t" |
| 304 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 305 | "add.s %[val0], %[fTmp5], %[fTmp6] \n\t" |
| 306 | "sub.s %[val15], %[fTmp5], %[fTmp6] \n\t" |
| 307 | "lwc1 %[fTmp1], 7*4(%[tab]) \n\t" |
| 308 | "lwc1 %[fTmp2], 24*4(%[tab]) \n\t" |
| 309 | "madd.s %[val16], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 310 | "nmsub.s %[val31], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 311 | "mul.s %[val15], %[val15], %[fTmp7] \n\t" |
| 312 | "lwc1 %[fTmp3], 8*4(%[tab]) \n\t" |
| 313 | "lwc1 %[fTmp4], 23*4(%[tab]) \n\t" |
| 314 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 315 | "mul.s %[val31], %[val31], %[fTmp7] \n\t" |
| 316 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 317 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 318 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 319 | "li.s %[fTmp7], 5.10114861868916385802 \n\t" |
| 320 | "li.s %[fTmp10], 0.67480834145500574602 \n\t" |
| 321 | "li.s %[fTmp11], 0.74453627100229844977 \n\t" |
| 322 | "add.s %[val7], %[fTmp5], %[fTmp6] \n\t" |
| 323 | "sub.s %[val8], %[fTmp5], %[fTmp6] \n\t" |
| 324 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 325 | "li.s %[fTmp1], 0.50979557910415916894 \n\t" |
| 326 | "sub.s %[fTmp2], %[val0], %[val7] \n\t" |
| 327 | "mul.s %[val8], %[val8], %[fTmp7] \n\t" |
| 328 | "madd.s %[val23], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 329 | "nmsub.s %[val24], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 330 | "add.s %[val0], %[val0], %[val7] \n\t" |
| 331 | "mul.s %[val7], %[fTmp1], %[fTmp2] \n\t" |
| 332 | "sub.s %[fTmp2], %[val15], %[val8] \n\t" |
| 333 | "add.s %[val8], %[val15], %[val8] \n\t" |
| 334 | "mul.s %[val24], %[val24], %[fTmp7] \n\t" |
| 335 | "sub.s %[fTmp3], %[val16], %[val23] \n\t" |
| 336 | "add.s %[val16], %[val16], %[val23] \n\t" |
| 337 | "mul.s %[val15], %[fTmp1], %[fTmp2] \n\t" |
| 338 | "sub.s %[fTmp4], %[val31], %[val24] \n\t" |
| 339 | "mul.s %[val23], %[fTmp1], %[fTmp3] \n\t" |
| 340 | "add.s %[val24], %[val31], %[val24] \n\t" |
| 341 | "mul.s %[val31], %[fTmp1], %[fTmp4] \n\t" |
| 342 | |
| 343 | : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), |
| 344 | [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6), |
| 345 | [fTmp7] "=&f" (fTmp7), [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9), |
| 346 | [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11), |
| 347 | [val0] "=f" (val0), [val7] "=f" (val7), |
| 348 | [val8] "=f" (val8), [val15] "=f" (val15), |
| 349 | [val16] "=f" (val16), [val23] "=f" (val23), |
| 350 | [val24] "=f" (val24), [val31] "=f" (val31) |
| 351 | : [tab] "r" (tab) |
| 352 | : "memory" |
| 353 | ); |
| 354 | |
| 355 | __asm__ volatile ( |
| 356 | "lwc1 %[fTmp1], 3*4(%[tab]) \n\t" |
| 357 | "lwc1 %[fTmp2], 28*4(%[tab]) \n\t" |
| 358 | "lwc1 %[fTmp3], 12*4(%[tab]) \n\t" |
| 359 | "lwc1 %[fTmp4], 19*4(%[tab]) \n\t" |
| 360 | "li.s %[fTmp7], 0.64682178335999012954 \n\t" |
| 361 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 362 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 363 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 364 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 365 | "li.s %[fTmp10], 0.53104259108978417447 \n\t" |
| 366 | "li.s %[fTmp11], 1.48416461631416627724 \n\t" |
| 367 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 368 | "add.s %[val3], %[fTmp5], %[fTmp6] \n\t" |
| 369 | "sub.s %[val12], %[fTmp5], %[fTmp6] \n\t" |
| 370 | "lwc1 %[fTmp1], 4*4(%[tab]) \n\t" |
| 371 | "lwc1 %[fTmp2], 27*4(%[tab]) \n\t" |
| 372 | "madd.s %[val19], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 373 | "nmsub.s %[val28], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 374 | "mul.s %[val12], %[val12], %[fTmp7] \n\t" |
| 375 | "lwc1 %[fTmp3], 11*4(%[tab]) \n\t" |
| 376 | "lwc1 %[fTmp4], 20*4(%[tab]) \n\t" |
| 377 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 378 | "mul.s %[val28], %[val28], %[fTmp7] \n\t" |
| 379 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 380 | "li.s %[fTmp7], 0.78815462345125022473 \n\t" |
| 381 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 382 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 383 | "li.s %[fTmp10], 0.55310389603444452782 \n\t" |
| 384 | "li.s %[fTmp11], 1.16943993343288495515 \n\t" |
| 385 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 386 | "add.s %[val4], %[fTmp5], %[fTmp6] \n\t" |
| 387 | "sub.s %[val11], %[fTmp5], %[fTmp6] \n\t" |
| 388 | "li.s %[fTmp1], 2.56291544774150617881 \n\t" |
| 389 | "madd.s %[val20], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 390 | "nmsub.s %[val27], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 391 | "mul.s %[val11], %[val11], %[fTmp7] \n\t" |
| 392 | "sub.s %[fTmp2], %[val3], %[val4] \n\t" |
| 393 | "add.s %[val3], %[val3], %[val4] \n\t" |
| 394 | "sub.s %[fTmp4], %[val19], %[val20] \n\t" |
| 395 | "mul.s %[val27], %[val27], %[fTmp7] \n\t" |
| 396 | "sub.s %[fTmp3], %[val12], %[val11] \n\t" |
| 397 | "mul.s %[val4], %[fTmp1], %[fTmp2] \n\t" |
| 398 | "add.s %[val11], %[val12], %[val11] \n\t" |
| 399 | "add.s %[val19], %[val19], %[val20] \n\t" |
| 400 | "mul.s %[val20], %[fTmp1], %[fTmp4] \n\t" |
| 401 | "mul.s %[val12], %[fTmp1], %[fTmp3] \n\t" |
| 402 | "sub.s %[fTmp2], %[val28], %[val27] \n\t" |
| 403 | "add.s %[val27], %[val28], %[val27] \n\t" |
| 404 | "mul.s %[val28], %[fTmp1], %[fTmp2] \n\t" |
| 405 | |
| 406 | : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), |
| 407 | [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6), |
| 408 | [fTmp7] "=&f" (fTmp7), [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9), |
| 409 | [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11), |
| 410 | [val3] "=f" (val3), [val4] "=f" (val4), |
| 411 | [val11] "=f" (val11), [val12] "=f" (val12), |
| 412 | [val19] "=f" (val19), [val20] "=f" (val20), |
| 413 | [val27] "=f" (val27), [val28] "=f" (val28) |
| 414 | : [tab] "r" (tab) |
| 415 | : "memory" |
| 416 | ); |
| 417 | |
| 418 | __asm__ volatile ( |
| 419 | "li.s %[fTmp1], 0.54119610014619698439 \n\t" |
| 420 | "sub.s %[fTmp2], %[val0], %[val3] \n\t" |
| 421 | "add.s %[val0], %[val0], %[val3] \n\t" |
| 422 | "sub.s %[fTmp3], %[val7], %[val4] \n\t" |
| 423 | "add.s %[val4], %[val7], %[val4] \n\t" |
| 424 | "sub.s %[fTmp4], %[val8], %[val11] \n\t" |
| 425 | "mul.s %[val3], %[fTmp1], %[fTmp2] \n\t" |
| 426 | "add.s %[val8], %[val8], %[val11] \n\t" |
| 427 | "mul.s %[val7], %[fTmp1], %[fTmp3] \n\t" |
| 428 | "sub.s %[fTmp2], %[val15], %[val12] \n\t" |
| 429 | "mul.s %[val11], %[fTmp1], %[fTmp4] \n\t" |
| 430 | "add.s %[val12], %[val15], %[val12] \n\t" |
| 431 | "mul.s %[val15], %[fTmp1], %[fTmp2] \n\t" |
| 432 | |
| 433 | : [val0] "+f" (val0), [val3] "+f" (val3), |
| 434 | [val4] "+f" (val4), [val7] "+f" (val7), |
| 435 | [val8] "+f" (val8), [val11] "+f" (val11), |
| 436 | [val12] "+f" (val12), [val15] "+f" (val15), |
| 437 | [fTmp1] "=f" (fTmp1), [fTmp2] "=&f" (fTmp2), |
| 438 | [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4) |
| 439 | : |
| 440 | ); |
| 441 | |
| 442 | __asm__ volatile ( |
| 443 | "sub.s %[fTmp2], %[val16], %[val19] \n\t" |
| 444 | "add.s %[val16], %[val16], %[val19] \n\t" |
| 445 | "sub.s %[fTmp3], %[val23], %[val20] \n\t" |
| 446 | "add.s %[val20], %[val23], %[val20] \n\t" |
| 447 | "sub.s %[fTmp4], %[val24], %[val27] \n\t" |
| 448 | "mul.s %[val19], %[fTmp1], %[fTmp2] \n\t" |
| 449 | "add.s %[val24], %[val24], %[val27] \n\t" |
| 450 | "mul.s %[val23], %[fTmp1], %[fTmp3] \n\t" |
| 451 | "sub.s %[fTmp2], %[val31], %[val28] \n\t" |
| 452 | "mul.s %[val27], %[fTmp1], %[fTmp4] \n\t" |
| 453 | "add.s %[val28], %[val31], %[val28] \n\t" |
| 454 | "mul.s %[val31], %[fTmp1], %[fTmp2] \n\t" |
| 455 | |
| 456 | : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 457 | [val16] "+f" (val16), [val19] "+f" (val19), [val20] "+f" (val20), |
| 458 | [val23] "+f" (val23), [val24] "+f" (val24), [val27] "+f" (val27), |
| 459 | [val28] "+f" (val28), [val31] "+f" (val31) |
| 460 | : [fTmp1] "f" (fTmp1) |
| 461 | ); |
| 462 | |
| 463 | __asm__ volatile ( |
| 464 | "lwc1 %[fTmp1], 1*4(%[tab]) \n\t" |
| 465 | "lwc1 %[fTmp2], 30*4(%[tab]) \n\t" |
| 466 | "lwc1 %[fTmp3], 14*4(%[tab]) \n\t" |
| 467 | "lwc1 %[fTmp4], 17*4(%[tab]) \n\t" |
| 468 | "li.s %[fTmp7], 0.52249861493968888062 \n\t" |
| 469 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 470 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 471 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 472 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 473 | "li.s %[fTmp10], 0.50547095989754365998 \n\t" |
| 474 | "li.s %[fTmp11], 3.40760841846871878570 \n\t" |
| 475 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 476 | "add.s %[val1], %[fTmp5], %[fTmp6] \n\t" |
| 477 | "sub.s %[val14], %[fTmp5], %[fTmp6] \n\t" |
| 478 | "lwc1 %[fTmp1], 6*4(%[tab]) \n\t" |
| 479 | "lwc1 %[fTmp2], 25*4(%[tab]) \n\t" |
| 480 | "madd.s %[val17], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 481 | "nmsub.s %[val30], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 482 | "mul.s %[val14], %[val14], %[fTmp7] \n\t" |
| 483 | "lwc1 %[fTmp3], 9*4(%[tab]) \n\t" |
| 484 | "lwc1 %[fTmp4], 22*4(%[tab]) \n\t" |
| 485 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 486 | "mul.s %[val30], %[val30], %[fTmp7] \n\t" |
| 487 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 488 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 489 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 490 | "li.s %[fTmp7], 1.72244709823833392782 \n\t" |
| 491 | "li.s %[fTmp10], 0.62250412303566481615 \n\t" |
| 492 | "li.s %[fTmp11], 0.83934964541552703873 \n\t" |
| 493 | "add.s %[val6], %[fTmp5], %[fTmp6] \n\t" |
| 494 | "sub.s %[val9], %[fTmp5], %[fTmp6] \n\t" |
| 495 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 496 | "li.s %[fTmp1], 0.60134488693504528054 \n\t" |
| 497 | "sub.s %[fTmp2], %[val1], %[val6] \n\t" |
| 498 | "add.s %[val1], %[val1], %[val6] \n\t" |
| 499 | "mul.s %[val9], %[val9], %[fTmp7] \n\t" |
| 500 | "madd.s %[val22], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 501 | "nmsub.s %[val25], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 502 | "mul.s %[val6], %[fTmp1], %[fTmp2] \n\t" |
| 503 | "sub.s %[fTmp2], %[val14], %[val9] \n\t" |
| 504 | "add.s %[val9], %[val14], %[val9] \n\t" |
| 505 | "mul.s %[val25], %[val25], %[fTmp7] \n\t" |
| 506 | "sub.s %[fTmp3], %[val17], %[val22] \n\t" |
| 507 | "add.s %[val17], %[val17], %[val22] \n\t" |
| 508 | "mul.s %[val14], %[fTmp1], %[fTmp2] \n\t" |
| 509 | "sub.s %[fTmp2], %[val30], %[val25] \n\t" |
| 510 | "mul.s %[val22], %[fTmp1], %[fTmp3] \n\t" |
| 511 | "add.s %[val25], %[val30], %[val25] \n\t" |
| 512 | "mul.s %[val30], %[fTmp1], %[fTmp2] \n\t" |
| 513 | |
| 514 | : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), |
| 515 | [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6), |
| 516 | [fTmp7] "=&f" (fTmp7), [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9), |
| 517 | [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11), |
| 518 | [val1] "=f" (val1), [val6] "=f" (val6), |
| 519 | [val9] "=f" (val9), [val14] "=f" (val14), |
| 520 | [val17] "=f" (val17), [val22] "=f" (val22), |
| 521 | [val25] "=f" (val25), [val30] "=f" (val30) |
| 522 | : [tab] "r" (tab) |
| 523 | : "memory" |
| 524 | ); |
| 525 | |
| 526 | __asm__ volatile ( |
| 527 | "lwc1 %[fTmp1], 2*4(%[tab]) \n\t" |
| 528 | "lwc1 %[fTmp2], 29*4(%[tab]) \n\t" |
| 529 | "lwc1 %[fTmp3], 13*4(%[tab]) \n\t" |
| 530 | "lwc1 %[fTmp4], 18*4(%[tab]) \n\t" |
| 531 | "li.s %[fTmp7], 0.56694403481635770368 \n\t" |
| 532 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 533 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 534 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 535 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 536 | "li.s %[fTmp10], 0.51544730992262454697 \n\t" |
| 537 | "li.s %[fTmp11], 2.05778100995341155085 \n\t" |
| 538 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 539 | "add.s %[val2], %[fTmp5], %[fTmp6] \n\t" |
| 540 | "sub.s %[val13], %[fTmp5], %[fTmp6] \n\t" |
| 541 | "lwc1 %[fTmp1], 5*4(%[tab]) \n\t" |
| 542 | "lwc1 %[fTmp2], 26*4(%[tab]) \n\t" |
| 543 | "madd.s %[val18], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 544 | "nmsub.s %[val29], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 545 | "mul.s %[val13], %[val13], %[fTmp7] \n\t" |
| 546 | "lwc1 %[fTmp3], 10*4(%[tab]) \n\t" |
| 547 | "lwc1 %[fTmp4], 21*4(%[tab]) \n\t" |
| 548 | "mul.s %[val29], %[val29], %[fTmp7] \n\t" |
| 549 | "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t" |
| 550 | "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t" |
| 551 | "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t" |
| 552 | "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t" |
| 553 | "li.s %[fTmp7], 1.06067768599034747134 \n\t" |
| 554 | "li.s %[fTmp10], 0.58293496820613387367 \n\t" |
| 555 | "li.s %[fTmp11], 0.97256823786196069369 \n\t" |
| 556 | "add.s %[val5], %[fTmp5], %[fTmp6] \n\t" |
| 557 | "sub.s %[val10], %[fTmp5], %[fTmp6] \n\t" |
| 558 | "mul.s %[fTmp8], %[fTmp8], %[fTmp10] \n\t" |
| 559 | "li.s %[fTmp1], 0.89997622313641570463 \n\t" |
| 560 | "sub.s %[fTmp2], %[val2], %[val5] \n\t" |
| 561 | "mul.s %[val10], %[val10], %[fTmp7] \n\t" |
| 562 | "madd.s %[val21], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 563 | "nmsub.s %[val26], %[fTmp8], %[fTmp9], %[fTmp11] \n\t" |
| 564 | "add.s %[val2], %[val2], %[val5] \n\t" |
| 565 | "mul.s %[val5], %[fTmp1], %[fTmp2] \n\t" |
| 566 | "sub.s %[fTmp3], %[val13], %[val10] \n\t" |
| 567 | "add.s %[val10], %[val13], %[val10] \n\t" |
| 568 | "mul.s %[val26], %[val26], %[fTmp7] \n\t" |
| 569 | "sub.s %[fTmp4], %[val18], %[val21] \n\t" |
| 570 | "add.s %[val18], %[val18], %[val21] \n\t" |
| 571 | "mul.s %[val13], %[fTmp1], %[fTmp3] \n\t" |
| 572 | "sub.s %[fTmp2], %[val29], %[val26] \n\t" |
| 573 | "add.s %[val26], %[val29], %[val26] \n\t" |
| 574 | "mul.s %[val21], %[fTmp1], %[fTmp4] \n\t" |
| 575 | "mul.s %[val29], %[fTmp1], %[fTmp2] \n\t" |
| 576 | |
| 577 | : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), |
| 578 | [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6), |
| 579 | [fTmp7] "=&f" (fTmp7), [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9), |
| 580 | [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11), |
| 581 | [val2] "=f" (val2), [val5] "=f" (val5), |
| 582 | [val10] "=f" (val10), [val13] "=f" (val13), |
| 583 | [val18] "=f" (val18), [val21] "=f" (val21), |
| 584 | [val26] "=f" (val26), [val29] "=f" (val29) |
| 585 | : [tab] "r" (tab) |
| 586 | : "memory" |
| 587 | ); |
| 588 | |
| 589 | __asm__ volatile ( |
| 590 | "li.s %[fTmp1], 1.30656296487637652785 \n\t" |
| 591 | "sub.s %[fTmp2], %[val1], %[val2] \n\t" |
| 592 | "add.s %[val1], %[val1], %[val2] \n\t" |
| 593 | "sub.s %[fTmp3], %[val6], %[val5] \n\t" |
| 594 | "add.s %[val5], %[val6], %[val5] \n\t" |
| 595 | "sub.s %[fTmp4], %[val9], %[val10] \n\t" |
| 596 | "mul.s %[val2], %[fTmp1], %[fTmp2] \n\t" |
| 597 | "add.s %[val9], %[val9], %[val10] \n\t" |
| 598 | "mul.s %[val6], %[fTmp1], %[fTmp3] \n\t" |
| 599 | "sub.s %[fTmp2], %[val14], %[val13] \n\t" |
| 600 | "mul.s %[val10], %[fTmp1], %[fTmp4] \n\t" |
| 601 | "add.s %[val13], %[val14], %[val13] \n\t" |
| 602 | "mul.s %[val14], %[fTmp1], %[fTmp2] \n\t" |
| 603 | |
| 604 | : [fTmp1] "=f" (fTmp1), [fTmp2] "=&f" (fTmp2), |
| 605 | [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 606 | [val1] "+f" (val1), [val2] "+f" (val2), |
| 607 | [val5] "+f" (val5), [val6] "+f" (val6), |
| 608 | [val9] "+f" (val9), [val10] "+f" (val10), |
| 609 | [val13] "+f" (val13), [val14] "+f" (val14) |
| 610 | : |
| 611 | ); |
| 612 | |
| 613 | __asm__ volatile ( |
| 614 | "sub.s %[fTmp2], %[val17], %[val18] \n\t" |
| 615 | "add.s %[val17], %[val17], %[val18] \n\t" |
| 616 | "sub.s %[fTmp3], %[val22], %[val21] \n\t" |
| 617 | "add.s %[val21], %[val22], %[val21] \n\t" |
| 618 | "sub.s %[fTmp4], %[val25], %[val26] \n\t" |
| 619 | "mul.s %[val18], %[fTmp1], %[fTmp2] \n\t" |
| 620 | "add.s %[val25], %[val25], %[val26] \n\t" |
| 621 | "mul.s %[val22], %[fTmp1], %[fTmp3] \n\t" |
| 622 | "sub.s %[fTmp2], %[val30], %[val29] \n\t" |
| 623 | "mul.s %[val26], %[fTmp1], %[fTmp4] \n\t" |
| 624 | "add.s %[val29], %[val30], %[val29] \n\t" |
| 625 | "mul.s %[val30], %[fTmp1], %[fTmp2] \n\t" |
| 626 | |
| 627 | : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 628 | [val17] "+f" (val17), [val18] "+f" (val18), [val21] "+f" (val21), |
| 629 | [val22] "+f" (val22), [val25] "+f" (val25), [val26] "+f" (val26), |
| 630 | [val29] "+f" (val29), [val30] "+f" (val30) |
| 631 | : [fTmp1] "f" (fTmp1) |
| 632 | ); |
| 633 | |
| 634 | __asm__ volatile ( |
| 635 | "li.s %[fTmp1], 0.70710678118654752439 \n\t" |
| 636 | "sub.s %[fTmp2], %[val0], %[val1] \n\t" |
| 637 | "add.s %[val0], %[val0], %[val1] \n\t" |
| 638 | "sub.s %[fTmp3], %[val3], %[val2] \n\t" |
| 639 | "add.s %[val2], %[val3], %[val2] \n\t" |
| 640 | "sub.s %[fTmp4], %[val4], %[val5] \n\t" |
| 641 | "mul.s %[val1], %[fTmp1], %[fTmp2] \n\t" |
| 642 | "swc1 %[val0], 0(%[out]) \n\t" |
| 643 | "mul.s %[val3], %[fTmp3], %[fTmp1] \n\t" |
| 644 | "add.s %[val4], %[val4], %[val5] \n\t" |
| 645 | "mul.s %[val5], %[fTmp1], %[fTmp4] \n\t" |
| 646 | "swc1 %[val1], 16*4(%[out]) \n\t" |
| 647 | "sub.s %[fTmp2], %[val7], %[val6] \n\t" |
| 648 | "add.s %[val2], %[val2], %[val3] \n\t" |
| 649 | "swc1 %[val3], 24*4(%[out]) \n\t" |
| 650 | "add.s %[val6], %[val7], %[val6] \n\t" |
| 651 | "mul.s %[val7], %[fTmp1], %[fTmp2] \n\t" |
| 652 | "swc1 %[val2], 8*4(%[out]) \n\t" |
| 653 | "add.s %[val6], %[val6], %[val7] \n\t" |
| 654 | "swc1 %[val7], 28*4(%[out]) \n\t" |
| 655 | "add.s %[val4], %[val4], %[val6] \n\t" |
| 656 | "add.s %[val6], %[val6], %[val5] \n\t" |
| 657 | "add.s %[val5], %[val5], %[val7] \n\t" |
| 658 | "swc1 %[val4], 4*4(%[out]) \n\t" |
| 659 | "swc1 %[val5], 20*4(%[out]) \n\t" |
| 660 | "swc1 %[val6], 12*4(%[out]) \n\t" |
| 661 | |
| 662 | : [fTmp1] "=f" (fTmp1), [fTmp2] "=&f" (fTmp2), |
| 663 | [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 664 | [val0] "+f" (val0), [val1] "+f" (val1), |
| 665 | [val2] "+f" (val2), [val3] "+f" (val3), |
| 666 | [val4] "+f" (val4), [val5] "+f" (val5), |
| 667 | [val6] "+f" (val6), [val7] "+f" (val7) |
| 668 | : [out] "r" (out) |
| 669 | ); |
| 670 | |
| 671 | __asm__ volatile ( |
| 672 | "sub.s %[fTmp2], %[val8], %[val9] \n\t" |
| 673 | "add.s %[val8], %[val8], %[val9] \n\t" |
| 674 | "sub.s %[fTmp3], %[val11], %[val10] \n\t" |
| 675 | "add.s %[val10], %[val11], %[val10] \n\t" |
| 676 | "sub.s %[fTmp4], %[val12], %[val13] \n\t" |
| 677 | "mul.s %[val9], %[fTmp1], %[fTmp2] \n\t" |
| 678 | "add.s %[val12], %[val12], %[val13] \n\t" |
| 679 | "mul.s %[val11], %[fTmp1], %[fTmp3] \n\t" |
| 680 | "sub.s %[fTmp2], %[val15], %[val14] \n\t" |
| 681 | "mul.s %[val13], %[fTmp1], %[fTmp4] \n\t" |
| 682 | "add.s %[val14], %[val15], %[val14] \n\t" |
| 683 | "add.s %[val10], %[val10], %[val11] \n\t" |
| 684 | "mul.s %[val15], %[fTmp1], %[fTmp2] \n\t" |
| 685 | "add.s %[val14], %[val14], %[val15] \n\t" |
| 686 | "add.s %[val12], %[val12], %[val14] \n\t" |
| 687 | "add.s %[val14], %[val14], %[val13] \n\t" |
| 688 | "add.s %[val13], %[val13], %[val15] \n\t" |
| 689 | "add.s %[val8], %[val8], %[val12] \n\t" |
| 690 | "add.s %[val12], %[val12], %[val10] \n\t" |
| 691 | "add.s %[val10], %[val10], %[val14] \n\t" |
| 692 | "add.s %[val14], %[val14], %[val9] \n\t" |
| 693 | "add.s %[val9], %[val9], %[val13] \n\t" |
| 694 | "add.s %[val13], %[val13], %[val11] \n\t" |
| 695 | "add.s %[val11], %[val11], %[val15] \n\t" |
| 696 | "swc1 %[val8], 2*4(%[out]) \n\t" |
| 697 | "swc1 %[val9], 18*4(%[out]) \n\t" |
| 698 | "swc1 %[val10], 10*4(%[out]) \n\t" |
| 699 | "swc1 %[val11], 26*4(%[out]) \n\t" |
| 700 | "swc1 %[val12], 6*4(%[out]) \n\t" |
| 701 | "swc1 %[val13], 22*4(%[out]) \n\t" |
| 702 | "swc1 %[val14], 14*4(%[out]) \n\t" |
| 703 | "swc1 %[val15], 30*4(%[out]) \n\t" |
| 704 | |
| 705 | : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 706 | [val8] "+f" (val8), [val9] "+f" (val9), [val10] "+f" (val10), |
| 707 | [val11] "+f" (val11), [val12] "+f" (val12), [val13] "+f" (val13), |
| 708 | [val14] "+f" (val14), [val15] "+f" (val15) |
| 709 | : [fTmp1] "f" (fTmp1), [out] "r" (out) |
| 710 | ); |
| 711 | |
| 712 | __asm__ volatile ( |
| 713 | "sub.s %[fTmp2], %[val16], %[val17] \n\t" |
| 714 | "add.s %[val16], %[val16], %[val17] \n\t" |
| 715 | "sub.s %[fTmp3], %[val19], %[val18] \n\t" |
| 716 | "add.s %[val18], %[val19], %[val18] \n\t" |
| 717 | "sub.s %[fTmp4], %[val20], %[val21] \n\t" |
| 718 | "mul.s %[val17], %[fTmp1], %[fTmp2] \n\t" |
| 719 | "add.s %[val20], %[val20], %[val21] \n\t" |
| 720 | "mul.s %[val19], %[fTmp1], %[fTmp3] \n\t" |
| 721 | "sub.s %[fTmp2], %[val23], %[val22] \n\t" |
| 722 | "mul.s %[val21], %[fTmp1], %[fTmp4] \n\t" |
| 723 | "add.s %[val22], %[val23], %[val22] \n\t" |
| 724 | "add.s %[val18], %[val18], %[val19] \n\t" |
| 725 | "mul.s %[val23], %[fTmp1], %[fTmp2] \n\t" |
| 726 | "add.s %[val22], %[val22], %[val23] \n\t" |
| 727 | "add.s %[val20], %[val20], %[val22] \n\t" |
| 728 | "add.s %[val22], %[val22], %[val21] \n\t" |
| 729 | "add.s %[val21], %[val21], %[val23] \n\t" |
| 730 | |
| 731 | : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 732 | [val16] "+f" (val16), [val17] "+f" (val17), [val18] "+f" (val18), |
| 733 | [val19] "+f" (val19), [val20] "+f" (val20), [val21] "+f" (val21), |
| 734 | [val22] "+f" (val22), [val23] "+f" (val23) |
| 735 | : [fTmp1] "f" (fTmp1) |
| 736 | ); |
| 737 | |
| 738 | __asm__ volatile ( |
| 739 | "sub.s %[fTmp2], %[val24], %[val25] \n\t" |
| 740 | "add.s %[val24], %[val24], %[val25] \n\t" |
| 741 | "sub.s %[fTmp3], %[val27], %[val26] \n\t" |
| 742 | "add.s %[val26], %[val27], %[val26] \n\t" |
| 743 | "sub.s %[fTmp4], %[val28], %[val29] \n\t" |
| 744 | "mul.s %[val25], %[fTmp1], %[fTmp2] \n\t" |
| 745 | "add.s %[val28], %[val28], %[val29] \n\t" |
| 746 | "mul.s %[val27], %[fTmp1], %[fTmp3] \n\t" |
| 747 | "sub.s %[fTmp2], %[val31], %[val30] \n\t" |
| 748 | "mul.s %[val29], %[fTmp1], %[fTmp4] \n\t" |
| 749 | "add.s %[val30], %[val31], %[val30] \n\t" |
| 750 | "add.s %[val26], %[val26], %[val27] \n\t" |
| 751 | "mul.s %[val31], %[fTmp1], %[fTmp2] \n\t" |
| 752 | "add.s %[val30], %[val30], %[val31] \n\t" |
| 753 | "add.s %[val28], %[val28], %[val30] \n\t" |
| 754 | "add.s %[val30], %[val30], %[val29] \n\t" |
| 755 | "add.s %[val29], %[val29], %[val31] \n\t" |
| 756 | "add.s %[val24], %[val24], %[val28] \n\t" |
| 757 | "add.s %[val28], %[val28], %[val26] \n\t" |
| 758 | "add.s %[val26], %[val26], %[val30] \n\t" |
| 759 | "add.s %[val30], %[val30], %[val25] \n\t" |
| 760 | "add.s %[val25], %[val25], %[val29] \n\t" |
| 761 | "add.s %[val29], %[val29], %[val27] \n\t" |
| 762 | "add.s %[val27], %[val27], %[val31] \n\t" |
| 763 | |
| 764 | : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4), |
| 765 | [val24] "+f" (val24), [val25] "+f" (val25), [val26] "+f" (val26), |
| 766 | [val27] "+f" (val27), [val28] "+f" (val28), [val29] "+f" (val29), |
| 767 | [val30] "+f" (val30), [val31] "+f" (val31) |
| 768 | : [fTmp1] "f" (fTmp1) |
| 769 | ); |
| 770 | |
| 771 | out[ 1] = val16 + val24; |
| 772 | out[17] = val17 + val25; |
| 773 | out[ 9] = val18 + val26; |
| 774 | out[25] = val19 + val27; |
| 775 | out[ 5] = val20 + val28; |
| 776 | out[21] = val21 + val29; |
| 777 | out[13] = val22 + val30; |
| 778 | out[29] = val23 + val31; |
| 779 | out[ 3] = val24 + val20; |
| 780 | out[19] = val25 + val21; |
| 781 | out[11] = val26 + val22; |
| 782 | out[27] = val27 + val23; |
| 783 | out[ 7] = val28 + val18; |
| 784 | out[23] = val29 + val19; |
| 785 | out[15] = val30 + val17; |
| 786 | out[31] = val31; |
| 787 | } |
| 788 | |
| 789 | static void imdct36_mips_float(float *out, float *buf, float *in, float *win) |
| 790 | { |
| 791 | float t0, t1, t2, t3, s0, s1, s2, s3; |
| 792 | float tmp[18]; |
| 793 | /* temporary variables */ |
| 794 | float in1, in2, in3, in4, in5, in6; |
| 795 | float out1, out2, out3, out4, out5; |
| 796 | float c1, c2, c3, c4, c5, c6, c7, c8, c9; |
| 797 | |
| 798 | /** |
| 799 | * all loops are unrolled totally, and instructions are scheduled to |
| 800 | * minimize pipeline stall. instructions of the first two loops are |
| 801 | * reorganized, in order to eliminate unnecessary readings and |
| 802 | * writings into array. values defined in macros and tables are |
| 803 | * eliminated - they are directly loaded in appropriate variables |
| 804 | */ |
| 805 | |
| 806 | /* loop 1 and 2 */ |
| 807 | __asm__ volatile ( |
| 808 | "lwc1 %[in1], 17*4(%[in]) \t\n" |
| 809 | "lwc1 %[in2], 16*4(%[in]) \t\n" |
| 810 | "lwc1 %[in3], 15*4(%[in]) \t\n" |
| 811 | "lwc1 %[in4], 14*4(%[in]) \t\n" |
| 812 | "lwc1 %[in5], 13*4(%[in]) \t\n" |
| 813 | "lwc1 %[in6], 12*4(%[in]) \t\n" |
| 814 | "add.s %[out1], %[in1], %[in2] \t\n" |
| 815 | "add.s %[out2], %[in2], %[in3] \t\n" |
| 816 | "add.s %[out3], %[in3], %[in4] \t\n" |
| 817 | "add.s %[out4], %[in4], %[in5] \t\n" |
| 818 | "add.s %[out5], %[in5], %[in6] \t\n" |
| 819 | "lwc1 %[in1], 11*4(%[in]) \t\n" |
| 820 | "swc1 %[out2], 16*4(%[in]) \t\n" |
| 821 | "add.s %[out1], %[out1], %[out3] \t\n" |
| 822 | "swc1 %[out4], 14*4(%[in]) \t\n" |
| 823 | "add.s %[out3], %[out3], %[out5] \t\n" |
| 824 | "lwc1 %[in2], 10*4(%[in]) \t\n" |
| 825 | "lwc1 %[in3], 9*4(%[in]) \t\n" |
| 826 | "swc1 %[out1], 17*4(%[in]) \t\n" |
| 827 | "lwc1 %[in4], 8*4(%[in]) \t\n" |
| 828 | "swc1 %[out3], 15*4(%[in]) \t\n" |
| 829 | "add.s %[out1], %[in6], %[in1] \t\n" |
| 830 | "add.s %[out2], %[in1], %[in2] \t\n" |
| 831 | "add.s %[out3], %[in2], %[in3] \t\n" |
| 832 | "add.s %[out4], %[in3], %[in4] \t\n" |
| 833 | "lwc1 %[in5], 7*4(%[in]) \t\n" |
| 834 | "swc1 %[out1], 12*4(%[in]) \t\n" |
| 835 | "add.s %[out5], %[out5], %[out2] \t\n" |
| 836 | "swc1 %[out3], 10*4(%[in]) \t\n" |
| 837 | "add.s %[out2], %[out2], %[out4] \t\n" |
| 838 | "lwc1 %[in6], 6*4(%[in]) \t\n" |
| 839 | "lwc1 %[in1], 5*4(%[in]) \t\n" |
| 840 | "swc1 %[out5], 13*4(%[in]) \t\n" |
| 841 | "lwc1 %[in2], 4*4(%[in]) \t\n" |
| 842 | "swc1 %[out2], 11*4(%[in]) \t\n" |
| 843 | "add.s %[out5], %[in4], %[in5] \t\n" |
| 844 | "add.s %[out1], %[in5], %[in6] \t\n" |
| 845 | "add.s %[out2], %[in6], %[in1] \t\n" |
| 846 | "add.s %[out3], %[in1], %[in2] \t\n" |
| 847 | "lwc1 %[in3], 3*4(%[in]) \t\n" |
| 848 | "swc1 %[out5], 8*4(%[in]) \t\n" |
| 849 | "add.s %[out4], %[out4], %[out1] \t\n" |
| 850 | "swc1 %[out2], 6*4(%[in]) \t\n" |
| 851 | "add.s %[out1], %[out1], %[out3] \t\n" |
| 852 | "lwc1 %[in4], 2*4(%[in]) \t\n" |
| 853 | "lwc1 %[in5], 1*4(%[in]) \t\n" |
| 854 | "swc1 %[out4], 9*4(%[in]) \t\n" |
| 855 | "lwc1 %[in6], 0(%[in]) \t\n" |
| 856 | "swc1 %[out1], 7*4(%[in]) \t\n" |
| 857 | "add.s %[out4], %[in2], %[in3] \t\n" |
| 858 | "add.s %[out5], %[in3], %[in4] \t\n" |
| 859 | "add.s %[out1], %[in4], %[in5] \t\n" |
| 860 | "add.s %[out2], %[in5], %[in6] \t\n" |
| 861 | "swc1 %[out4], 4*4(%[in]) \t\n" |
| 862 | "add.s %[out3], %[out3], %[out5] \t\n" |
| 863 | "swc1 %[out1], 2*4(%[in]) \t\n" |
| 864 | "add.s %[out5], %[out5], %[out2] \t\n" |
| 865 | "swc1 %[out2], 1*4(%[in]) \t\n" |
| 866 | "swc1 %[out3], 5*4(%[in]) \t\n" |
| 867 | "swc1 %[out5], 3*4(%[in]) \t\n" |
| 868 | |
| 869 | : [in1] "=&f" (in1), [in2] "=&f" (in2), |
| 870 | [in3] "=&f" (in3), [in4] "=&f" (in4), |
| 871 | [in5] "=&f" (in5), [in6] "=&f" (in6), |
| 872 | [out1] "=&f" (out1), [out2] "=&f" (out2), |
| 873 | [out3] "=&f" (out3), [out4] "=&f" (out4), |
| 874 | [out5] "=&f" (out5) |
| 875 | : [in] "r" (in) |
| 876 | : "memory" |
| 877 | ); |
| 878 | |
| 879 | /* loop 3 */ |
| 880 | __asm__ volatile ( |
| 881 | "li.s %[c1], 0.5 \t\n" |
| 882 | "lwc1 %[in1], 8*4(%[in]) \t\n" |
| 883 | "lwc1 %[in2], 16*4(%[in]) \t\n" |
| 884 | "lwc1 %[in3], 4*4(%[in]) \t\n" |
| 885 | "lwc1 %[in4], 0(%[in]) \t\n" |
| 886 | "lwc1 %[in5], 12*4(%[in]) \t\n" |
| 887 | "li.s %[c2], 0.93969262078590838405 \t\n" |
| 888 | "add.s %[t2], %[in1], %[in2] \t\n" |
| 889 | "add.s %[t0], %[in1], %[in3] \t\n" |
| 890 | "li.s %[c3], -0.76604444311897803520 \t\n" |
| 891 | "madd.s %[t3], %[in4], %[in5], %[c1] \t\n" |
| 892 | "sub.s %[t1], %[in4], %[in5] \t\n" |
| 893 | "sub.s %[t2], %[t2], %[in3] \t\n" |
| 894 | "mul.s %[t0], %[t0], %[c2] \t\n" |
| 895 | "li.s %[c4], -0.17364817766693034885 \t\n" |
| 896 | "li.s %[c5], -0.86602540378443864676 \t\n" |
| 897 | "li.s %[c6], 0.98480775301220805936 \t\n" |
| 898 | "nmsub.s %[out1], %[t1], %[t2], %[c1] \t\n" |
| 899 | "add.s %[out2], %[t1], %[t2] \t\n" |
| 900 | "add.s %[t2], %[in2], %[in3] \t\n" |
| 901 | "sub.s %[t1], %[in1], %[in2] \t\n" |
| 902 | "sub.s %[out3], %[t3], %[t0] \t\n" |
| 903 | "swc1 %[out1], 6*4(%[tmp]) \t\n" |
| 904 | "swc1 %[out2], 16*4(%[tmp]) \t\n" |
| 905 | "mul.s %[t2], %[t2], %[c3] \t\n" |
| 906 | "mul.s %[t1], %[t1], %[c4] \t\n" |
| 907 | "add.s %[out1], %[t3], %[t0] \t\n" |
| 908 | "lwc1 %[in1], 10*4(%[in]) \t\n" |
| 909 | "lwc1 %[in2], 14*4(%[in]) \t\n" |
| 910 | "sub.s %[out3], %[out3], %[t2] \t\n" |
| 911 | "add.s %[out2], %[t3], %[t2] \t\n" |
| 912 | "add.s %[out1], %[out1], %[t1] \t\n" |
| 913 | "lwc1 %[in3], 2*4(%[in]) \t\n" |
| 914 | "lwc1 %[in4], 6*4(%[in]) \t\n" |
| 915 | "swc1 %[out3], 10*4(%[tmp]) \t\n" |
| 916 | "sub.s %[out2], %[out2], %[t1] \t\n" |
| 917 | "swc1 %[out1], 2*4(%[tmp]) \t\n" |
| 918 | "add.s %[out1], %[in1], %[in2] \t\n" |
| 919 | "add.s %[t2], %[in1], %[in3] \t\n" |
| 920 | "sub.s %[t3], %[in1], %[in2] \t\n" |
| 921 | "swc1 %[out2], 14*4(%[tmp]) \t\n" |
| 922 | "li.s %[c7], -0.34202014332566873304 \t\n" |
| 923 | "sub.s %[out1], %[out1], %[in3] \t\n" |
| 924 | "mul.s %[t2], %[t2], %[c6] \t\n" |
| 925 | "mul.s %[t3], %[t3], %[c7] \t\n" |
| 926 | "li.s %[c8], 0.86602540378443864676 \t\n" |
| 927 | "mul.s %[t0], %[in4], %[c8] \t\n" |
| 928 | "mul.s %[out1], %[out1], %[c5] \t\n" |
| 929 | "add.s %[t1], %[in2], %[in3] \t\n" |
| 930 | "li.s %[c9], -0.64278760968653932632 \t\n" |
| 931 | "add.s %[out2], %[t2], %[t3] \t\n" |
| 932 | "lwc1 %[in1], 9*4(%[in]) \t\n" |
| 933 | "swc1 %[out1], 4*4(%[tmp]) \t\n" |
| 934 | "mul.s %[t1], %[t1], %[c9] \t\n" |
| 935 | "lwc1 %[in2], 17*4(%[in]) \t\n" |
| 936 | "add.s %[out2], %[out2], %[t0] \t\n" |
| 937 | "lwc1 %[in3], 5*4(%[in]) \t\n" |
| 938 | "lwc1 %[in4], 1*4(%[in]) \t\n" |
| 939 | "add.s %[out3], %[t2], %[t1] \t\n" |
| 940 | "sub.s %[out1], %[t3], %[t1] \t\n" |
| 941 | "swc1 %[out2], 0(%[tmp]) \t\n" |
| 942 | "lwc1 %[in5], 13*4(%[in]) \t\n" |
| 943 | "add.s %[t2], %[in1], %[in2] \t\n" |
| 944 | "sub.s %[out3], %[out3], %[t0] \t\n" |
| 945 | "sub.s %[out1], %[out1], %[t0] \t\n" |
| 946 | "add.s %[t0], %[in1], %[in3] \t\n" |
| 947 | "madd.s %[t3], %[in4], %[in5], %[c1] \t\n" |
| 948 | "sub.s %[t2], %[t2], %[in3] \t\n" |
| 949 | "swc1 %[out3], 12*4(%[tmp]) \t\n" |
| 950 | "swc1 %[out1], 8*4(%[tmp]) \t\n" |
| 951 | "sub.s %[t1], %[in4], %[in5] \t\n" |
| 952 | "mul.s %[t0], %[t0], %[c2] \t\n" |
| 953 | "nmsub.s %[out1], %[t1], %[t2], %[c1] \t\n" |
| 954 | "add.s %[out2], %[t1], %[t2] \t\n" |
| 955 | "add.s %[t2], %[in2], %[in3] \t\n" |
| 956 | "sub.s %[t1], %[in1], %[in2] \t\n" |
| 957 | "sub.s %[out3], %[t3], %[t0] \t\n" |
| 958 | "swc1 %[out1], 7*4(%[tmp]) \t\n" |
| 959 | "swc1 %[out2], 17*4(%[tmp]) \t\n" |
| 960 | "mul.s %[t2], %[t2], %[c3] \t\n" |
| 961 | "mul.s %[t1], %[t1], %[c4] \t\n" |
| 962 | "add.s %[out1], %[t3], %[t0] \t\n" |
| 963 | "lwc1 %[in1], 11*4(%[in]) \t\n" |
| 964 | "lwc1 %[in2], 15*4(%[in]) \t\n" |
| 965 | "sub.s %[out3], %[out3], %[t2] \t\n" |
| 966 | "add.s %[out2], %[t3], %[t2] \t\n" |
| 967 | "add.s %[out1], %[out1], %[t1] \t\n" |
| 968 | "lwc1 %[in3], 3*4(%[in]) \t\n" |
| 969 | "lwc1 %[in4], 7*4(%[in]) \t\n" |
| 970 | "swc1 %[out3], 11*4(%[tmp]) \t\n" |
| 971 | "sub.s %[out2], %[out2], %[t1] \t\n" |
| 972 | "swc1 %[out1], 3*4(%[tmp]) \t\n" |
| 973 | "add.s %[out3], %[in1], %[in2] \t\n" |
| 974 | "add.s %[t2], %[in1], %[in3] \t\n" |
| 975 | "sub.s %[t3], %[in1], %[in2] \t\n" |
| 976 | "swc1 %[out2], 15*4(%[tmp]) \t\n" |
| 977 | "mul.s %[t0], %[in4], %[c8] \t\n" |
| 978 | "sub.s %[out3], %[out3], %[in3] \t\n" |
| 979 | "mul.s %[t2], %[t2], %[c6] \t\n" |
| 980 | "mul.s %[t3], %[t3], %[c7] \t\n" |
| 981 | "add.s %[t1], %[in2], %[in3] \t\n" |
| 982 | "mul.s %[out3], %[out3], %[c5] \t\n" |
| 983 | "add.s %[out1], %[t2], %[t3] \t\n" |
| 984 | "mul.s %[t1], %[t1], %[c9] \t\n" |
| 985 | "swc1 %[out3], 5*4(%[tmp]) \t\n" |
| 986 | "add.s %[out1], %[out1], %[t0] \t\n" |
| 987 | "add.s %[out2], %[t2], %[t1] \t\n" |
| 988 | "sub.s %[out3], %[t3], %[t1] \t\n" |
| 989 | "swc1 %[out1], 1*4(%[tmp]) \t\n" |
| 990 | "sub.s %[out2], %[out2], %[t0] \t\n" |
| 991 | "sub.s %[out3], %[out3], %[t0] \t\n" |
| 992 | "swc1 %[out2], 13*4(%[tmp]) \t\n" |
| 993 | "swc1 %[out3], 9*4(%[tmp]) \t\n" |
| 994 | |
| 995 | : [t0] "=&f" (t0), [t1] "=&f" (t1), |
| 996 | [t2] "=&f" (t2), [t3] "=&f" (t3), |
| 997 | [in1] "=&f" (in1), [in2] "=&f" (in2), |
| 998 | [in3] "=&f" (in3), [in4] "=&f" (in4), |
| 999 | [in5] "=&f" (in5), |
| 1000 | [out1] "=&f" (out1), [out2] "=&f" (out2), |
| 1001 | [out3] "=&f" (out3), |
| 1002 | [c1] "=&f" (c1), [c2] "=&f" (c2), |
| 1003 | [c3] "=&f" (c3), [c4] "=&f" (c4), |
| 1004 | [c5] "=&f" (c5), [c6] "=&f" (c6), |
| 1005 | [c7] "=&f" (c7), [c8] "=&f" (c8), |
| 1006 | [c9] "=&f" (c9) |
| 1007 | : [in] "r" (in), [tmp] "r" (tmp) |
| 1008 | : "memory" |
| 1009 | ); |
| 1010 | |
| 1011 | /* loop 4 */ |
| 1012 | __asm__ volatile ( |
| 1013 | "lwc1 %[in1], 2*4(%[tmp]) \t\n" |
| 1014 | "lwc1 %[in2], 0(%[tmp]) \t\n" |
| 1015 | "lwc1 %[in3], 3*4(%[tmp]) \t\n" |
| 1016 | "lwc1 %[in4], 1*4(%[tmp]) \t\n" |
| 1017 | "li.s %[c1], 0.50190991877167369479 \t\n" |
| 1018 | "li.s %[c2], 5.73685662283492756461 \t\n" |
| 1019 | "add.s %[s0], %[in1], %[in2] \t\n" |
| 1020 | "sub.s %[s2], %[in1], %[in2] \t\n" |
| 1021 | "add.s %[s1], %[in3], %[in4] \t\n" |
| 1022 | "sub.s %[s3], %[in3], %[in4] \t\n" |
| 1023 | "lwc1 %[in1], 9*4(%[win]) \t\n" |
| 1024 | "lwc1 %[in2], 4*9*4(%[buf]) \t\n" |
| 1025 | "lwc1 %[in3], 8*4(%[win]) \t\n" |
| 1026 | "mul.s %[s1], %[s1], %[c1] \t\n" |
| 1027 | "mul.s %[s3], %[s3], %[c2] \t\n" |
| 1028 | "lwc1 %[in4], 4*8*4(%[buf]) \t\n" |
| 1029 | "lwc1 %[in5], 29*4(%[win]) \t\n" |
| 1030 | "lwc1 %[in6], 28*4(%[win]) \t\n" |
| 1031 | "add.s %[t0], %[s0], %[s1] \t\n" |
| 1032 | "sub.s %[t1], %[s0], %[s1] \t\n" |
| 1033 | "li.s %[c1], 0.51763809020504152469 \t\n" |
| 1034 | "li.s %[c2], 1.93185165257813657349 \t\n" |
| 1035 | "mul.s %[out3], %[in5], %[t0] \t\n" |
| 1036 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1037 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1038 | "mul.s %[out4], %[in6], %[t0] \t\n" |
| 1039 | "add.s %[t0], %[s2], %[s3] \t\n" |
| 1040 | "swc1 %[out3], 4*9*4(%[buf]) \t\n" |
| 1041 | "swc1 %[out1], 288*4(%[out]) \t\n" |
| 1042 | "swc1 %[out2], 256*4(%[out]) \t\n" |
| 1043 | "swc1 %[out4], 4*8*4(%[buf]) \t\n" |
| 1044 | "sub.s %[t1], %[s2], %[s3] \t\n" |
| 1045 | "lwc1 %[in1], 17*4(%[win]) \t\n" |
| 1046 | "lwc1 %[in2], 4*17*4(%[buf]) \t\n" |
| 1047 | "lwc1 %[in3], 0(%[win]) \t\n" |
| 1048 | "lwc1 %[in4], 0(%[buf]) \t\n" |
| 1049 | "lwc1 %[in5], 37*4(%[win]) \t\n" |
| 1050 | "lwc1 %[in6], 20*4(%[win]) \t\n" |
| 1051 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1052 | "lwc1 %[in1], 6*4(%[tmp]) \t\n" |
| 1053 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1054 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1055 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1056 | "swc1 %[out1], 544*4(%[out]) \t\n" |
| 1057 | "lwc1 %[in2], 4*4(%[tmp]) \t\n" |
| 1058 | "swc1 %[out2], 0(%[out]) \t\n" |
| 1059 | "swc1 %[out3], 4*17*4(%[buf]) \t\n" |
| 1060 | "swc1 %[out4], 0(%[buf]) \t\n" |
| 1061 | "lwc1 %[in3], 7*4(%[tmp]) \t\n" |
| 1062 | "add.s %[s0], %[in1], %[in2] \t\n" |
| 1063 | "sub.s %[s2], %[in1], %[in2] \t\n" |
| 1064 | "lwc1 %[in4], 5*4(%[tmp]) \t\n" |
| 1065 | "add.s %[s1], %[in3], %[in4] \t\n" |
| 1066 | "sub.s %[s3], %[in3], %[in4] \t\n" |
| 1067 | "lwc1 %[in1], 10*4(%[win]) \t\n" |
| 1068 | "lwc1 %[in2], 4*10*4(%[buf]) \t\n" |
| 1069 | "lwc1 %[in3], 7*4(%[win]) \t\n" |
| 1070 | "mul.s %[s1], %[s1], %[c1] \t\n" |
| 1071 | "mul.s %[s3], %[s3], %[c2] \t\n" |
| 1072 | "add.s %[t0], %[s0], %[s1] \t\n" |
| 1073 | "sub.s %[t1], %[s0], %[s1] \t\n" |
| 1074 | "lwc1 %[in4], 4*7*4(%[buf]) \t\n" |
| 1075 | "lwc1 %[in5], 30*4(%[win]) \t\n" |
| 1076 | "lwc1 %[in6], 27*4(%[win]) \t\n" |
| 1077 | "li.s %[c1], 0.55168895948124587824 \t\n" |
| 1078 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1079 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1080 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1081 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1082 | "add.s %[t0], %[s2], %[s3] \t\n" |
| 1083 | "swc1 %[out1], 320*4(%[out]) \t\n" |
| 1084 | "swc1 %[out2], 224*4(%[out]) \t\n" |
| 1085 | "swc1 %[out3], 4*10*4(%[buf]) \t\n" |
| 1086 | "swc1 %[out4], 4*7*4(%[buf]) \t\n" |
| 1087 | "sub.s %[t1], %[s2], %[s3] \t\n" |
| 1088 | "lwc1 %[in1], 16*4(%[win]) \t\n" |
| 1089 | "lwc1 %[in2], 4*16*4(%[buf]) \t\n" |
| 1090 | "lwc1 %[in3], 1*4(%[win]) \t\n" |
| 1091 | "lwc1 %[in4], 4*1*4(%[buf]) \t\n" |
| 1092 | "lwc1 %[in5], 36*4(%[win]) \t\n" |
| 1093 | "lwc1 %[in6], 21*4(%[win]) \t\n" |
| 1094 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1095 | "lwc1 %[in1], 10*4(%[tmp]) \t\n" |
| 1096 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1097 | "mul.s %[out3], %[in5], %[t0] \t\n" |
| 1098 | "mul.s %[out4], %[in6], %[t0] \t\n" |
| 1099 | "swc1 %[out1], 512*4(%[out]) \t\n" |
| 1100 | "lwc1 %[in2], 8*4(%[tmp]) \t\n" |
| 1101 | "swc1 %[out2], 32*4(%[out]) \t\n" |
| 1102 | "swc1 %[out3], 4*16*4(%[buf]) \t\n" |
| 1103 | "swc1 %[out4], 4*1*4(%[buf]) \t\n" |
| 1104 | "li.s %[c2], 1.18310079157624925896 \t\n" |
| 1105 | "add.s %[s0], %[in1], %[in2] \t\n" |
| 1106 | "sub.s %[s2], %[in1], %[in2] \t\n" |
| 1107 | "lwc1 %[in3], 11*4(%[tmp]) \t\n" |
| 1108 | "lwc1 %[in4], 9*4(%[tmp]) \t\n" |
| 1109 | "add.s %[s1], %[in3], %[in4] \t\n" |
| 1110 | "sub.s %[s3], %[in3], %[in4] \t\n" |
| 1111 | "lwc1 %[in1], 11*4(%[win]) \t\n" |
| 1112 | "lwc1 %[in2], 4*11*4(%[buf]) \t\n" |
| 1113 | "lwc1 %[in3], 6*4(%[win]) \t\n" |
| 1114 | "mul.s %[s1], %[s1], %[c1] \t\n" |
| 1115 | "mul.s %[s3], %[s3], %[c2] \t\n" |
| 1116 | "lwc1 %[in4], 4*6*4(%[buf]) \t\n" |
| 1117 | "lwc1 %[in5], 31*4(%[win]) \t\n" |
| 1118 | "lwc1 %[in6], 26*4(%[win]) \t\n" |
| 1119 | "add.s %[t0], %[s0], %[s1] \t\n" |
| 1120 | "sub.s %[t1], %[s0], %[s1] \t\n" |
| 1121 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1122 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1123 | "add.s %[t0], %[s2], %[s3] \t\n" |
| 1124 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1125 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1126 | "swc1 %[out3], 4*11*4(%[buf]) \t\n" |
| 1127 | "swc1 %[out4], 4*6*4(%[buf]) \t\n" |
| 1128 | "sub.s %[t1], %[s2], %[s3] \t\n" |
| 1129 | "swc1 %[out1], 352*4(%[out]) \t\n" |
| 1130 | "swc1 %[out2], 192*4(%[out]) \t\n" |
| 1131 | "lwc1 %[in1], 15*4(%[win]) \t\n" |
| 1132 | "lwc1 %[in2], 4*15*4(%[buf]) \t\n" |
| 1133 | "lwc1 %[in3], 2*4(%[win]) \t\n" |
| 1134 | "lwc1 %[in4], 4*2*4(%[buf]) \t\n" |
| 1135 | "lwc1 %[in5], 35*4(%[win]) \t\n" |
| 1136 | "lwc1 %[in6], 22*4(%[win]) \t\n" |
| 1137 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1138 | "lwc1 %[in1], 14*4(%[tmp]) \t\n" |
| 1139 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1140 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1141 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1142 | "swc1 %[out1], 480*4(%[out]) \t\n" |
| 1143 | "lwc1 %[in2], 12*4(%[tmp]) \t\n" |
| 1144 | "swc1 %[out2], 64*4(%[out]) \t\n" |
| 1145 | "swc1 %[out3], 4*15*4(%[buf]) \t\n" |
| 1146 | "swc1 %[out4], 4*2*4(%[buf]) \t\n" |
| 1147 | "lwc1 %[in3], 15*4(%[tmp]) \t\n" |
| 1148 | "add.s %[s0], %[in1], %[in2] \t\n" |
| 1149 | "sub.s %[s2], %[in1], %[in2] \t\n" |
| 1150 | "lwc1 %[in4], 13*4(%[tmp]) \t\n" |
| 1151 | "li.s %[c1], 0.61038729438072803416 \t\n" |
| 1152 | "li.s %[c2], 0.87172339781054900991 \t\n" |
| 1153 | "add.s %[s1], %[in3], %[in4] \t\n" |
| 1154 | "sub.s %[s3], %[in3], %[in4] \t\n" |
| 1155 | "lwc1 %[in1], 12*4(%[win]) \t\n" |
| 1156 | "lwc1 %[in2], 4*12*4(%[buf]) \t\n" |
| 1157 | "lwc1 %[in3], 5*4(%[win]) \t\n" |
| 1158 | "mul.s %[s1], %[s1], %[c1] \t\n" |
| 1159 | "mul.s %[s3], %[s3], %[c2] \t\n" |
| 1160 | "lwc1 %[in4], 4*5*4(%[buf]) \t\n" |
| 1161 | "lwc1 %[in5], 32*4(%[win]) \t\n" |
| 1162 | "lwc1 %[in6], 25*4(%[win]) \t\n" |
| 1163 | "add.s %[t0], %[s0], %[s1] \t\n" |
| 1164 | "sub.s %[t1], %[s0], %[s1] \t\n" |
| 1165 | "lwc1 %[s0], 16*4(%[tmp]) \t\n" |
| 1166 | "lwc1 %[s1], 17*4(%[tmp]) \t\n" |
| 1167 | "li.s %[c1], 0.70710678118654752439 \t\n" |
| 1168 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1169 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1170 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1171 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1172 | "add.s %[t0], %[s2], %[s3] \t\n" |
| 1173 | "swc1 %[out3], 4*12*4(%[buf]) \t\n" |
| 1174 | "swc1 %[out1], 384*4(%[out]) \t\n" |
| 1175 | "swc1 %[out2], 160*4(%[out]) \t\n" |
| 1176 | "swc1 %[out4], 4*5*4(%[buf]) \t\n" |
| 1177 | "sub.s %[t1], %[s2], %[s3] \t\n" |
| 1178 | "lwc1 %[in1], 14*4(%[win]) \t\n" |
| 1179 | "lwc1 %[in2], 4*14*4(%[buf]) \t\n" |
| 1180 | "lwc1 %[in3], 3*4(%[win]) \t\n" |
| 1181 | "lwc1 %[in4], 4*3*4(%[buf]) \t\n" |
| 1182 | "lwc1 %[in5], 34*4(%[win]) \t\n" |
| 1183 | "lwc1 %[in6], 23*4(%[win]) \t\n" |
| 1184 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1185 | "mul.s %[s1], %[s1], %[c1] \t\n" |
| 1186 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1187 | "mul.s %[out3], %[in5], %[t0] \t\n" |
| 1188 | "mul.s %[out4], %[in6], %[t0] \t\n" |
| 1189 | "swc1 %[out1], 448*4(%[out]) \t\n" |
| 1190 | "add.s %[t0], %[s0], %[s1] \t\n" |
| 1191 | "swc1 %[out2], 96*4(%[out]) \t\n" |
| 1192 | "swc1 %[out3], 4*14*4(%[buf]) \t\n" |
| 1193 | "swc1 %[out4], 4*3*4(%[buf]) \t\n" |
| 1194 | "sub.s %[t1], %[s0], %[s1] \t\n" |
| 1195 | "lwc1 %[in1], 13*4(%[win]) \t\n" |
| 1196 | "lwc1 %[in2], 4*13*4(%[buf]) \t\n" |
| 1197 | "lwc1 %[in3], 4*4(%[win]) \t\n" |
| 1198 | "lwc1 %[in4], 4*4*4(%[buf]) \t\n" |
| 1199 | "lwc1 %[in5], 33*4(%[win]) \t\n" |
| 1200 | "lwc1 %[in6], 24*4(%[win]) \t\n" |
| 1201 | "madd.s %[out1], %[in2], %[in1], %[t1] \t\n" |
| 1202 | "madd.s %[out2], %[in4], %[in3], %[t1] \t\n" |
| 1203 | "mul.s %[out3], %[t0], %[in5] \t\n" |
| 1204 | "mul.s %[out4], %[t0], %[in6] \t\n" |
| 1205 | "swc1 %[out1], 416*4(%[out]) \t\n" |
| 1206 | "swc1 %[out2], 128*4(%[out]) \t\n" |
| 1207 | "swc1 %[out3], 4*13*4(%[buf]) \t\n" |
| 1208 | "swc1 %[out4], 4*4*4(%[buf]) \t\n" |
| 1209 | |
| 1210 | : [c1] "=&f" (c1), [c2] "=&f" (c2), |
| 1211 | [in1] "=&f" (in1), [in2] "=&f" (in2), |
| 1212 | [in3] "=&f" (in3), [in4] "=&f" (in4), |
| 1213 | [in5] "=&f" (in5), [in6] "=&f" (in6), |
| 1214 | [out1] "=&f" (out1), [out2] "=&f" (out2), |
| 1215 | [out3] "=&f" (out3), [out4] "=&f" (out4), |
| 1216 | [t0] "=&f" (t0), [t1] "=&f" (t1), |
| 1217 | [t2] "=&f" (t2), [t3] "=&f" (t3), |
| 1218 | [s0] "=&f" (s0), [s1] "=&f" (s1), |
| 1219 | [s2] "=&f" (s2), [s3] "=&f" (s3) |
| 1220 | : [tmp] "r" (tmp), [win] "r" (win), |
| 1221 | [buf] "r" (buf), [out] "r" (out) |
| 1222 | : "memory" |
| 1223 | ); |
| 1224 | } |
| 1225 | |
| 1226 | static void ff_imdct36_blocks_mips_float(float *out, float *buf, float *in, |
| 1227 | int count, int switch_point, int block_type) |
| 1228 | { |
| 1229 | int j; |
| 1230 | for (j=0 ; j < count; j++) { |
| 1231 | /* apply window & overlap with previous buffer */ |
| 1232 | |
| 1233 | /* select window */ |
| 1234 | int win_idx = (switch_point && j < 2) ? 0 : block_type; |
| 1235 | float *win = ff_mdct_win_float[win_idx + (4 & -(j & 1))]; |
| 1236 | |
| 1237 | imdct36_mips_float(out, buf, in, win); |
| 1238 | |
| 1239 | in += 18; |
| 1240 | buf += ((j&3) != 3 ? 1 : (72-3)); |
| 1241 | out++; |
| 1242 | } |
| 1243 | } |
| 1244 | |
| 1245 | void ff_mpadsp_init_mipsfpu(MPADSPContext *s) |
| 1246 | { |
| 1247 | s->apply_window_float = ff_mpadsp_apply_window_mips_float; |
| 1248 | s->imdct36_blocks_float = ff_imdct36_blocks_mips_float; |
| 1249 | s->dct32_float = ff_dct32_mips_float; |
| 1250 | } |