Mercurial > hg > pmhd

annotate ffmpeg/libavcodec/mips/mpegaudiodsp_mips_fixed.c @ 13:844d341cf643 tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Back up before ISMIR
author Yading Song <yading.song@eecs.qmul.ac.uk>
date Thu, 31 Oct 2013 13:17:06 +0000
parents 6840f77b83aa
children
rev   line source
yading@10 1 /*
yading@10 2 * Copyright (c) 2012
yading@10 3 * MIPS Technologies, Inc., California.
yading@10 4 *
yading@10 5 * Redistribution and use in source and binary forms, with or without
yading@10 6 * modification, are permitted provided that the following conditions
yading@10 7 * are met:
yading@10 8 * 1. Redistributions of source code must retain the above copyright
yading@10 9 * notice, this list of conditions and the following disclaimer.
yading@10 10 * 2. Redistributions in binary form must reproduce the above copyright
yading@10 11 * notice, this list of conditions and the following disclaimer in the
yading@10 12 * documentation and/or other materials provided with the distribution.
yading@10 13 * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its
yading@10 14 * contributors may be used to endorse or promote products derived from
yading@10 15 * this software without specific prior written permission.
yading@10 16 *
yading@10 17 * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
yading@10 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
yading@10 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
yading@10 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
yading@10 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
yading@10 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
yading@10 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
yading@10 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
yading@10 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
yading@10 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
yading@10 27 * SUCH DAMAGE.
yading@10 28 *
yading@10 29 * Author: Bojan Zivkovic (bojan@mips.com)
yading@10 30 *
yading@10 31 * MPEG Audio decoder optimized for MIPS fixed-point architecture
yading@10 32 *
yading@10 33 * This file is part of FFmpeg.
yading@10 34 *
yading@10 35 * FFmpeg is free software; you can redistribute it and/or
yading@10 36 * modify it under the terms of the GNU Lesser General Public
yading@10 37 * License as published by the Free Software Foundation; either
yading@10 38 * version 2.1 of the License, or (at your option) any later version.
yading@10 39 *
yading@10 40 * FFmpeg is distributed in the hope that it will be useful,
yading@10 41 * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10 42 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
yading@10 43 * Lesser General Public License for more details.
yading@10 44 *
yading@10 45 * You should have received a copy of the GNU Lesser General Public
yading@10 46 * License along with FFmpeg; if not, write to the Free Software
yading@10 47 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10 48 */
yading@10 49
yading@10 50 /**
yading@10 51 * @file
yading@10 52 * Reference: libavcodec/mpegaudiodsp_template.c
yading@10 53 */
yading@10 54
yading@10 55 #include <string.h>
yading@10 56
yading@10 57 #include "libavcodec/mpegaudiodsp.h"
yading@10 58
yading@10 59 static void ff_mpadsp_apply_window_mips_fixed(int32_t *synth_buf, int32_t *window,
yading@10 60 int *dither_state, int16_t *samples, int incr)
yading@10 61 {
yading@10 62 register const int32_t *w, *w2, *p;
yading@10 63 int j;
yading@10 64 int16_t *samples2;
yading@10 65 int w_asm, p_asm, w_asm1, p_asm1, w_asm2, p_asm2;
yading@10 66 int w2_asm, w2_asm1, *p_temp1, *p_temp2;
yading@10 67 int sum1 = 0;
yading@10 68 int const min_asm = -32768, max_asm = 32767;
yading@10 69 int temp1, temp2 = 0, temp3 = 0;
yading@10 70 int64_t sum;
yading@10 71
yading@10 72 /* copy to avoid wrap */
yading@10 73 memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));
yading@10 74 samples2 = samples + 31 * incr;
yading@10 75 w = window;
yading@10 76 w2 = window + 31;
yading@10 77 sum = *dither_state;
yading@10 78 p = synth_buf + 16;
yading@10 79 p_temp1 = synth_buf + 16;
yading@10 80 p_temp2 = synth_buf + 48;
yading@10 81 temp1 = sum;
yading@10 82
yading@10 83 /**
yading@10 84 * use of round_sample function from the original code is eliminated,
yading@10 85 * changed with appropriate assembly instructions.
yading@10 86 */
yading@10 87 __asm__ volatile (
yading@10 88 "mthi $zero \n\t"
yading@10 89 "mtlo %[temp1] \n\t"
yading@10 90 "lw %[w_asm], 0(%[w]) \n\t"
yading@10 91 "lw %[p_asm], 0(%[p]) \n\t"
yading@10 92 "lw %[w_asm1], 64*4(%[w]) \n\t"
yading@10 93 "lw %[p_asm1], 64*4(%[p]) \n\t"
yading@10 94 "lw %[w_asm2], 128*4(%[w]) \n\t"
yading@10 95 "lw %[p_asm2], 128*4(%[p]) \n\t"
yading@10 96 "madd %[w_asm], %[p_asm] \n\t"
yading@10 97 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 98 "madd %[w_asm2], %[p_asm2] \n\t"
yading@10 99 "lw %[w_asm], 192*4(%[w]) \n\t"
yading@10 100 "lw %[p_asm], 192*4(%[p]) \n\t"
yading@10 101 "lw %[w_asm1], 256*4(%[w]) \n\t"
yading@10 102 "lw %[p_asm1], 256*4(%[p]) \n\t"
yading@10 103 "lw %[w_asm2], 320*4(%[w]) \n\t"
yading@10 104 "lw %[p_asm2], 320*4(%[p]) \n\t"
yading@10 105 "madd %[w_asm], %[p_asm] \n\t"
yading@10 106 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 107 "madd %[w_asm2], %[p_asm2] \n\t"
yading@10 108 "lw %[w_asm], 384*4(%[w]) \n\t"
yading@10 109 "lw %[p_asm], 384*4(%[p]) \n\t"
yading@10 110 "lw %[w_asm1], 448*4(%[w]) \n\t"
yading@10 111 "lw %[p_asm1], 448*4(%[p]) \n\t"
yading@10 112 "lw %[w_asm2], 32*4(%[w]) \n\t"
yading@10 113 "lw %[p_asm2], 32*4(%[p]) \n\t"
yading@10 114 "madd %[w_asm], %[p_asm] \n\t"
yading@10 115 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 116 "msub %[w_asm2], %[p_asm2] \n\t"
yading@10 117 "lw %[w_asm], 96*4(%[w]) \n\t"
yading@10 118 "lw %[p_asm], 96*4(%[p]) \n\t"
yading@10 119 "lw %[w_asm1], 160*4(%[w]) \n\t"
yading@10 120 "lw %[p_asm1], 160*4(%[p]) \n\t"
yading@10 121 "lw %[w_asm2], 224*4(%[w]) \n\t"
yading@10 122 "lw %[p_asm2], 224*4(%[p]) \n\t"
yading@10 123 "msub %[w_asm], %[p_asm] \n\t"
yading@10 124 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 125 "msub %[w_asm2], %[p_asm2] \n\t"
yading@10 126 "lw %[w_asm], 288*4(%[w]) \n\t"
yading@10 127 "lw %[p_asm], 288*4(%[p]) \n\t"
yading@10 128 "lw %[w_asm1], 352*4(%[w]) \n\t"
yading@10 129 "lw %[p_asm1], 352*4(%[p]) \n\t"
yading@10 130 "msub %[w_asm], %[p_asm] \n\t"
yading@10 131 "lw %[w_asm], 480*4(%[w]) \n\t"
yading@10 132 "lw %[p_asm], 480*4(%[p]) \n\t"
yading@10 133 "lw %[w_asm2], 416*4(%[w]) \n\t"
yading@10 134 "lw %[p_asm2], 416*4(%[p]) \n\t"
yading@10 135 "msub %[w_asm], %[p_asm] \n\t"
yading@10 136 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 137 "msub %[w_asm2], %[p_asm2] \n\t"
yading@10 138
yading@10 139 /*round_sample function from the original code is eliminated,
yading@10 140 * changed with appropriate assembly instructions
yading@10 141 * code example:
yading@10 142
yading@10 143 "extr.w %[sum1],$ac0,24 \n\t"
yading@10 144 "mflo %[temp3], $ac0 \n\t"
yading@10 145 "and %[temp1], %[temp3], 0x00ffffff \n\t"
yading@10 146 "slt %[temp2], %[sum1], %[min_asm] \n\t"
yading@10 147 "movn %[sum1], %[min_asm],%[temp2] \n\t"
yading@10 148 "slt %[temp2], %[max_asm],%[sum1] \n\t"
yading@10 149 "movn %[sum1], %[max_asm],%[temp2] \n\t"
yading@10 150 "sh %[sum1], 0(%[samples]) \n\t"
yading@10 151 */
yading@10 152
yading@10 153 "extr.w %[sum1], $ac0, 24 \n\t"
yading@10 154 "mflo %[temp3] \n\t"
yading@10 155 "addi %[w], %[w], 4 \n\t"
yading@10 156 "and %[temp1], %[temp3], 0x00ffffff \n\t"
yading@10 157 "slt %[temp2], %[sum1], %[min_asm] \n\t"
yading@10 158 "movn %[sum1], %[min_asm], %[temp2] \n\t"
yading@10 159 "slt %[temp2], %[max_asm], %[sum1] \n\t"
yading@10 160 "movn %[sum1], %[max_asm], %[temp2] \n\t"
yading@10 161 "sh %[sum1], 0(%[samples]) \n\t"
yading@10 162
yading@10 163 : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
yading@10 164 [p_asm1] "=&r" (p_asm1), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
yading@10 165 [w_asm2] "=&r" (w_asm2), [p_asm2] "=&r" (p_asm2),
yading@10 166 [sum1] "+r" (sum1), [w] "+r" (w), [temp3] "+r" (temp3)
yading@10 167 : [p] "r" (p), [samples] "r" (samples), [min_asm] "r" (min_asm),
yading@10 168 [max_asm] "r" (max_asm)
yading@10 169 : "memory", "hi","lo"
yading@10 170 );
yading@10 171
yading@10 172 samples += incr;
yading@10 173
yading@10 174 /* we calculate two samples at the same time to avoid one memory
yading@10 175 access per two sample */
yading@10 176
yading@10 177 for(j = 1; j < 16; j++) {
yading@10 178 __asm__ volatile (
yading@10 179 "mthi $0, $ac1 \n\t"
yading@10 180 "mtlo $0, $ac1 \n\t"
yading@10 181 "mthi $0 \n\t"
yading@10 182 "mtlo %[temp1] \n\t"
yading@10 183 "addi %[p_temp1], %[p_temp1], 4 \n\t"
yading@10 184 "lw %[w_asm], 0(%[w]) \n\t"
yading@10 185 "lw %[p_asm], 0(%[p_temp1]) \n\t"
yading@10 186 "lw %[w2_asm], 0(%[w2]) \n\t"
yading@10 187 "lw %[w_asm1], 64*4(%[w]) \n\t"
yading@10 188 "lw %[p_asm1], 64*4(%[p_temp1]) \n\t"
yading@10 189 "lw %[w2_asm1], 64*4(%[w2]) \n\t"
yading@10 190 "madd %[w_asm], %[p_asm] \n\t"
yading@10 191 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 192 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 193 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 194 "lw %[w_asm], 128*4(%[w]) \n\t"
yading@10 195 "lw %[p_asm], 128*4(%[p_temp1]) \n\t"
yading@10 196 "lw %[w2_asm], 128*4(%[w2]) \n\t"
yading@10 197 "lw %[w_asm1], 192*4(%[w]) \n\t"
yading@10 198 "lw %[p_asm1], 192*4(%[p_temp1]) \n\t"
yading@10 199 "lw %[w2_asm1], 192*4(%[w2]) \n\t"
yading@10 200 "madd %[w_asm], %[p_asm] \n\t"
yading@10 201 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 202 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 203 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 204 "lw %[w_asm], 256*4(%[w]) \n\t"
yading@10 205 "lw %[p_asm], 256*4(%[p_temp1]) \n\t"
yading@10 206 "lw %[w2_asm], 256*4(%[w2]) \n\t"
yading@10 207 "lw %[w_asm1], 320*4(%[w]) \n\t"
yading@10 208 "lw %[p_asm1], 320*4(%[p_temp1]) \n\t"
yading@10 209 "lw %[w2_asm1], 320*4(%[w2]) \n\t"
yading@10 210 "madd %[w_asm], %[p_asm] \n\t"
yading@10 211 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 212 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 213 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 214 "lw %[w_asm], 384*4(%[w]) \n\t"
yading@10 215 "lw %[p_asm], 384*4(%[p_temp1]) \n\t"
yading@10 216 "lw %[w2_asm], 384*4(%[w2]) \n\t"
yading@10 217 "lw %[w_asm1], 448*4(%[w]) \n\t"
yading@10 218 "lw %[p_asm1], 448*4(%[p_temp1]) \n\t"
yading@10 219 "lw %[w2_asm1], 448*4(%[w2]) \n\t"
yading@10 220 "madd %[w_asm], %[p_asm] \n\t"
yading@10 221 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 222 "madd %[w_asm1], %[p_asm1] \n\t"
yading@10 223 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 224 "addi %[p_temp2], %[p_temp2], -4 \n\t"
yading@10 225 "lw %[w_asm], 32*4(%[w]) \n\t"
yading@10 226 "lw %[p_asm], 0(%[p_temp2]) \n\t"
yading@10 227 "lw %[w2_asm], 32*4(%[w2]) \n\t"
yading@10 228 "lw %[w_asm1], 96*4(%[w]) \n\t"
yading@10 229 "lw %[p_asm1], 64*4(%[p_temp2]) \n\t"
yading@10 230 "lw %[w2_asm1], 96*4(%[w2]) \n\t"
yading@10 231 "msub %[w_asm], %[p_asm] \n\t"
yading@10 232 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 233 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 234 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 235 "lw %[w_asm], 160*4(%[w]) \n\t"
yading@10 236 "lw %[p_asm], 128*4(%[p_temp2]) \n\t"
yading@10 237 "lw %[w2_asm], 160*4(%[w2]) \n\t"
yading@10 238 "lw %[w_asm1], 224*4(%[w]) \n\t"
yading@10 239 "lw %[p_asm1], 192*4(%[p_temp2]) \n\t"
yading@10 240 "lw %[w2_asm1], 224*4(%[w2]) \n\t"
yading@10 241 "msub %[w_asm], %[p_asm] \n\t"
yading@10 242 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 243 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 244 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 245 "lw %[w_asm], 288*4(%[w]) \n\t"
yading@10 246 "lw %[p_asm], 256*4(%[p_temp2]) \n\t"
yading@10 247 "lw %[w2_asm], 288*4(%[w2]) \n\t"
yading@10 248 "lw %[w_asm1], 352*4(%[w]) \n\t"
yading@10 249 "lw %[p_asm1], 320*4(%[p_temp2]) \n\t"
yading@10 250 "lw %[w2_asm1], 352*4(%[w2]) \n\t"
yading@10 251 "msub %[w_asm], %[p_asm] \n\t"
yading@10 252 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 253 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 254 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 255 "lw %[w_asm], 416*4(%[w]) \n\t"
yading@10 256 "lw %[p_asm], 384*4(%[p_temp2]) \n\t"
yading@10 257 "lw %[w2_asm], 416*4(%[w2]) \n\t"
yading@10 258 "lw %[w_asm1], 480*4(%[w]) \n\t"
yading@10 259 "lw %[p_asm1], 448*4(%[p_temp2]) \n\t"
yading@10 260 "lw %[w2_asm1], 480*4(%[w2]) \n\t"
yading@10 261 "msub %[w_asm], %[p_asm] \n\t"
yading@10 262 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 263 "msub $ac1, %[w2_asm], %[p_asm] \n\t"
yading@10 264 "msub $ac1, %[w2_asm1], %[p_asm1] \n\t"
yading@10 265 "addi %[w], %[w], 4 \n\t"
yading@10 266 "addi %[w2], %[w2], -4 \n\t"
yading@10 267 "mflo %[temp2] \n\t"
yading@10 268 "extr.w %[sum1], $ac0, 24 \n\t"
yading@10 269 "li %[temp3], 1 \n\t"
yading@10 270 "and %[temp1], %[temp2], 0x00ffffff \n\t"
yading@10 271 "madd $ac1, %[temp1], %[temp3] \n\t"
yading@10 272 "slt %[temp2], %[sum1], %[min_asm] \n\t"
yading@10 273 "movn %[sum1], %[min_asm], %[temp2] \n\t"
yading@10 274 "slt %[temp2], %[max_asm], %[sum1] \n\t"
yading@10 275 "movn %[sum1], %[max_asm], %[temp2] \n\t"
yading@10 276 "sh %[sum1], 0(%[samples]) \n\t"
yading@10 277 "mflo %[temp3], $ac1 \n\t"
yading@10 278 "extr.w %[sum1], $ac1, 24 \n\t"
yading@10 279 "and %[temp1], %[temp3], 0x00ffffff \n\t"
yading@10 280 "slt %[temp2], %[sum1], %[min_asm] \n\t"
yading@10 281 "movn %[sum1], %[min_asm], %[temp2] \n\t"
yading@10 282 "slt %[temp2], %[max_asm], %[sum1] \n\t"
yading@10 283 "movn %[sum1], %[max_asm], %[temp2] \n\t"
yading@10 284 "sh %[sum1], 0(%[samples2]) \n\t"
yading@10 285
yading@10 286 : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
yading@10 287 [p_asm1] "=&r" (p_asm1), [w2_asm1] "=&r" (w2_asm1),
yading@10 288 [w2_asm] "=&r" (w2_asm), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
yading@10 289 [p_temp1] "+r" (p_temp1), [p_temp2] "+r" (p_temp2), [sum1] "+r" (sum1),
yading@10 290 [w] "+r" (w), [w2] "+r" (w2), [samples] "+r" (samples),
yading@10 291 [samples2] "+r" (samples2), [temp3] "+r" (temp3)
yading@10 292 : [min_asm] "r" (min_asm), [max_asm] "r" (max_asm)
yading@10 293 : "memory", "hi", "lo", "$ac1hi", "$ac1lo"
yading@10 294 );
yading@10 295
yading@10 296 samples += incr;
yading@10 297 samples2 -= incr;
yading@10 298 }
yading@10 299
yading@10 300 p = synth_buf + 32;
yading@10 301
yading@10 302 __asm__ volatile (
yading@10 303 "mthi $0 \n\t"
yading@10 304 "mtlo %[temp1] \n\t"
yading@10 305 "lw %[w_asm], 32*4(%[w]) \n\t"
yading@10 306 "lw %[p_asm], 0(%[p]) \n\t"
yading@10 307 "lw %[w_asm1], 96*4(%[w]) \n\t"
yading@10 308 "lw %[p_asm1], 64*4(%[p]) \n\t"
yading@10 309 "lw %[w_asm2], 160*4(%[w]) \n\t"
yading@10 310 "lw %[p_asm2], 128*4(%[p]) \n\t"
yading@10 311 "msub %[w_asm], %[p_asm] \n\t"
yading@10 312 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 313 "msub %[w_asm2], %[p_asm2] \n\t"
yading@10 314 "lw %[w_asm], 224*4(%[w]) \n\t"
yading@10 315 "lw %[p_asm], 192*4(%[p]) \n\t"
yading@10 316 "lw %[w_asm1], 288*4(%[w]) \n\t"
yading@10 317 "lw %[p_asm1], 256*4(%[p]) \n\t"
yading@10 318 "lw %[w_asm2], 352*4(%[w]) \n\t"
yading@10 319 "lw %[p_asm2], 320*4(%[p]) \n\t"
yading@10 320 "msub %[w_asm], %[p_asm] \n\t"
yading@10 321 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 322 "msub %[w_asm2], %[p_asm2] \n\t"
yading@10 323 "lw %[w_asm], 416*4(%[w]) \n\t"
yading@10 324 "lw %[p_asm], 384*4(%[p]) \n\t"
yading@10 325 "lw %[w_asm1], 480*4(%[w]) \n\t"
yading@10 326 "lw %[p_asm1], 448*4(%[p]) \n\t"
yading@10 327 "msub %[w_asm], %[p_asm] \n\t"
yading@10 328 "msub %[w_asm1], %[p_asm1] \n\t"
yading@10 329 "extr.w %[sum1], $ac0, 24 \n\t"
yading@10 330 "mflo %[temp2] \n\t"
yading@10 331 "and %[temp1], %[temp2], 0x00ffffff \n\t"
yading@10 332 "slt %[temp2], %[sum1], %[min_asm] \n\t"
yading@10 333 "movn %[sum1], %[min_asm], %[temp2] \n\t"
yading@10 334 "slt %[temp2], %[max_asm], %[sum1] \n\t"
yading@10 335 "movn %[sum1], %[max_asm], %[temp2] \n\t"
yading@10 336 "sh %[sum1], 0(%[samples]) \n\t"
yading@10 337
yading@10 338 : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
yading@10 339 [p_asm1] "=&r" (p_asm1), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
yading@10 340 [w_asm2] "=&r" (w_asm2), [p_asm2] "=&r" (p_asm2), [sum1] "+r" (sum1)
yading@10 341 : [w] "r" (w), [p] "r" (p), [samples] "r" (samples), [min_asm] "r" (min_asm),
yading@10 342 [max_asm] "r" (max_asm)
yading@10 343 : "memory", "hi", "lo", "$ac1hi", "$ac1lo"
yading@10 344 );
yading@10 345
yading@10 346 *dither_state= temp1;
yading@10 347 }
yading@10 348
yading@10 349 static void imdct36_mips_fixed(int *out, int *buf, int *in, int *win)
yading@10 350 {
yading@10 351 int j;
yading@10 352 int t0, t1, t2, t3, s0, s1, s2, s3;
yading@10 353 int tmp[18], *tmp1, *in1;
yading@10 354 /* temporary variables */
yading@10 355 int temp_reg1, temp_reg2, temp_reg3, temp_reg4, temp_reg5, temp_reg6;
yading@10 356 int t4, t5, t6, t8, t7;
yading@10 357
yading@10 358 /* values defined in macros and tables are
yading@10 359 * eliminated - they are directly loaded in appropriate variables
yading@10 360 */
yading@10 361 int const C_1 = 4229717092; /* cos(pi*1/18)*2 */
yading@10 362 int const C_2 = 4035949074; /* cos(pi*2/18)*2 */
yading@10 363 int const C_3 = 575416510; /* -cos(pi*3/18)*2 */
yading@10 364 int const C_3A = 3719550786; /* cos(pi*3/18)*2 */
yading@10 365 int const C_4 = 1004831466; /* -cos(pi*4/18)*2 */
yading@10 366 int const C_5 = 1534215534; /* -cos(pi*5/18)*2 */
yading@10 367 int const C_7 = -1468965330; /* -cos(pi*7/18)*2 */
yading@10 368 int const C_8 = -745813244; /* -cos(pi*8/18)*2 */
yading@10 369
yading@10 370 /*
yading@10 371 * instructions of the first two loops are reorganized and loops are unrolled,
yading@10 372 * in order to eliminate unnecessary readings and writings in array
yading@10 373 */
yading@10 374
yading@10 375 __asm__ volatile (
yading@10 376 "lw %[t1], 17*4(%[in]) \n\t"
yading@10 377 "lw %[t2], 16*4(%[in]) \n\t"
yading@10 378 "lw %[t3], 15*4(%[in]) \n\t"
yading@10 379 "lw %[t4], 14*4(%[in]) \n\t"
yading@10 380 "addu %[t1], %[t1], %[t2] \n\t"
yading@10 381 "addu %[t2], %[t2], %[t3] \n\t"
yading@10 382 "addu %[t3], %[t3], %[t4] \n\t"
yading@10 383 "lw %[t5], 13*4(%[in]) \n\t"
yading@10 384 "addu %[t1], %[t1], %[t3] \n\t"
yading@10 385 "sw %[t2], 16*4(%[in]) \n\t"
yading@10 386 "lw %[t6], 12*4(%[in]) \n\t"
yading@10 387 "sw %[t1], 17*4(%[in]) \n\t"
yading@10 388 "addu %[t4], %[t4], %[t5] \n\t"
yading@10 389 "addu %[t5], %[t5], %[t6] \n\t"
yading@10 390 "lw %[t7], 11*4(%[in]) \n\t"
yading@10 391 "addu %[t3], %[t3], %[t5] \n\t"
yading@10 392 "sw %[t4], 14*4(%[in]) \n\t"
yading@10 393 "lw %[t8], 10*4(%[in]) \n\t"
yading@10 394 "sw %[t3], 15*4(%[in]) \n\t"
yading@10 395 "addu %[t6], %[t6], %[t7] \n\t"
yading@10 396 "addu %[t7], %[t7], %[t8] \n\t"
yading@10 397 "sw %[t6], 12*4(%[in]) \n\t"
yading@10 398 "addu %[t5], %[t5], %[t7] \n\t"
yading@10 399 "lw %[t1], 9*4(%[in]) \n\t"
yading@10 400 "lw %[t2], 8*4(%[in]) \n\t"
yading@10 401 "sw %[t5], 13*4(%[in]) \n\t"
yading@10 402 "addu %[t8], %[t8], %[t1] \n\t"
yading@10 403 "addu %[t1], %[t1], %[t2] \n\t"
yading@10 404 "sw %[t8], 10*4(%[in]) \n\t"
yading@10 405 "addu %[t7], %[t7], %[t1] \n\t"
yading@10 406 "lw %[t3], 7*4(%[in]) \n\t"
yading@10 407 "lw %[t4], 6*4(%[in]) \n\t"
yading@10 408 "sw %[t7], 11*4(%[in]) \n\t"
yading@10 409 "addu %[t2], %[t2], %[t3] \n\t"
yading@10 410 "addu %[t3], %[t3], %[t4] \n\t"
yading@10 411 "sw %[t2], 8*4(%[in]) \n\t"
yading@10 412 "addu %[t1], %[t1], %[t3] \n\t"
yading@10 413 "lw %[t5], 5*4(%[in]) \n\t"
yading@10 414 "lw %[t6], 4*4(%[in]) \n\t"
yading@10 415 "sw %[t1], 9*4(%[in]) \n\t"
yading@10 416 "addu %[t4], %[t4], %[t5] \n\t"
yading@10 417 "addu %[t5], %[t5], %[t6] \n\t"
yading@10 418 "sw %[t4], 6*4(%[in]) \n\t"
yading@10 419 "addu %[t3], %[t3], %[t5] \n\t"
yading@10 420 "lw %[t7], 3*4(%[in]) \n\t"
yading@10 421 "lw %[t8], 2*4(%[in]) \n\t"
yading@10 422 "sw %[t3], 7*4(%[in]) \n\t"
yading@10 423 "addu %[t6], %[t6], %[t7] \n\t"
yading@10 424 "addu %[t7], %[t7], %[t8] \n\t"
yading@10 425 "sw %[t6], 4*4(%[in]) \n\t"
yading@10 426 "addu %[t5], %[t5], %[t7] \n\t"
yading@10 427 "lw %[t1], 1*4(%[in]) \n\t"
yading@10 428 "lw %[t2], 0*4(%[in]) \n\t"
yading@10 429 "sw %[t5], 5*4(%[in]) \n\t"
yading@10 430 "addu %[t8], %[t8], %[t1] \n\t"
yading@10 431 "addu %[t1], %[t1], %[t2] \n\t"
yading@10 432 "sw %[t8], 2*4(%[in]) \n\t"
yading@10 433 "addu %[t7], %[t7], %[t1] \n\t"
yading@10 434 "sw %[t7], 3*4(%[in]) \n\t"
yading@10 435 "sw %[t1], 1*4(%[in]) \n\t"
yading@10 436
yading@10 437 : [in] "+r" (in), [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3),
yading@10 438 [t4] "=&r" (t4), [t5] "=&r" (t5), [t6] "=&r" (t6),
yading@10 439 [t7] "=&r" (t7), [t8] "=&r" (t8)
yading@10 440 :
yading@10 441 : "memory"
yading@10 442 );
yading@10 443
yading@10 444 for(j = 0; j < 2; j++) {
yading@10 445
yading@10 446 tmp1 = tmp + j;
yading@10 447 in1 = in + j;
yading@10 448
yading@10 449 /**
yading@10 450 * Original constants are multiplied by two in advanced
yading@10 451 * for assembly optimization (e.g. C_2 = 2 * C2).
yading@10 452 * That can lead to overflow in operations where they are used.
yading@10 453 *
yading@10 454 * Example of the solution:
yading@10 455 *
yading@10 456 * in original code:
yading@10 457 * t0 = ((int64_t)(in1[2*2] + in1[2*4]) * (int64_t)(2*C2))>>32
yading@10 458 *
yading@10 459 * in assembly:
yading@10 460 * C_2 = 2 * C2;
yading@10 461 * .
yading@10 462 * .
yading@10 463 * "lw %[t7], 4*4(%[in1]) \n\t"
yading@10 464 * "lw %[t8], 8*4(%[in1]) \n\t"
yading@10 465 * "addu %[temp_reg2],%[t7], %[t8] \n\t"
yading@10 466 * "multu %[C_2], %[temp_reg2] \n\t"
yading@10 467 * "mfhi %[temp_reg1] \n\t"
yading@10 468 * "sra %[temp_reg2],%[temp_reg2],31 \n\t"
yading@10 469 * "move %[t0], $0 \n\t"
yading@10 470 * "movn %[t0], %[C_2], %[temp_reg2] \n\t"
yading@10 471 * "sub %[t0], %[temp_reg1],%[t0] \n\t"
yading@10 472 */
yading@10 473
yading@10 474 __asm__ volatile (
yading@10 475 "lw %[t7], 4*4(%[in1]) \n\t"
yading@10 476 "lw %[t8], 8*4(%[in1]) \n\t"
yading@10 477 "lw %[t6], 16*4(%[in1]) \n\t"
yading@10 478 "lw %[t4], 0*4(%[in1]) \n\t"
yading@10 479 "addu %[temp_reg2], %[t7], %[t8] \n\t"
yading@10 480 "addu %[t2], %[t6], %[t8] \n\t"
yading@10 481 "multu %[C_2], %[temp_reg2] \n\t"
yading@10 482 "lw %[t5], 12*4(%[in1]) \n\t"
yading@10 483 "sub %[t2], %[t2], %[t7] \n\t"
yading@10 484 "sub %[t1], %[t4], %[t5] \n\t"
yading@10 485 "sra %[t3], %[t5], 1 \n\t"
yading@10 486 "sra %[temp_reg1], %[t2], 1 \n\t"
yading@10 487 "addu %[t3], %[t3], %[t4] \n\t"
yading@10 488 "sub %[temp_reg1], %[t1], %[temp_reg1] \n\t"
yading@10 489 "sra %[temp_reg2], %[temp_reg2], 31 \n\t"
yading@10 490 "sw %[temp_reg1], 6*4(%[tmp1]) \n\t"
yading@10 491 "move %[t0], $0 \n\t"
yading@10 492 "movn %[t0], %[C_2], %[temp_reg2] \n\t"
yading@10 493 "mfhi %[temp_reg1] \n\t"
yading@10 494 "addu %[t1], %[t1], %[t2] \n\t"
yading@10 495 "sw %[t1], 16*4(%[tmp1]) \n\t"
yading@10 496 "sub %[temp_reg4], %[t8], %[t6] \n\t"
yading@10 497 "add %[temp_reg2], %[t7], %[t6] \n\t"
yading@10 498 "mult $ac1, %[C_8], %[temp_reg4] \n\t"
yading@10 499 "multu $ac2, %[C_4], %[temp_reg2] \n\t"
yading@10 500 "sub %[t0], %[temp_reg1], %[t0] \n\t"
yading@10 501 "sra %[temp_reg1], %[temp_reg2], 31 \n\t"
yading@10 502 "move %[t2], $0 \n\t"
yading@10 503 "movn %[t2], %[C_4], %[temp_reg1] \n\t"
yading@10 504 "mfhi %[t1], $ac1 \n\t"
yading@10 505 "mfhi %[temp_reg1], $ac2 \n\t"
yading@10 506 "lw %[t6], 10*4(%[in1]) \n\t"
yading@10 507 "lw %[t8], 14*4(%[in1]) \n\t"
yading@10 508 "lw %[t7], 2*4(%[in1]) \n\t"
yading@10 509 "lw %[t4], 6*4(%[in1]) \n\t"
yading@10 510 "sub %[temp_reg3], %[t3], %[t0] \n\t"
yading@10 511 "add %[temp_reg4], %[t3], %[t0] \n\t"
yading@10 512 "sub %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 513 "add %[temp_reg4], %[temp_reg4], %[t1] \n\t"
yading@10 514 "sub %[t2], %[temp_reg1], %[t2] \n\t"
yading@10 515 "sw %[temp_reg4], 2*4(%[tmp1]) \n\t"
yading@10 516 "sub %[temp_reg3], %[temp_reg3], %[t2] \n\t"
yading@10 517 "add %[temp_reg1], %[t3], %[t2] \n\t"
yading@10 518 "sw %[temp_reg3], 10*4(%[tmp1]) \n\t"
yading@10 519 "sub %[temp_reg1], %[temp_reg1], %[t1] \n\t"
yading@10 520 "addu %[temp_reg2], %[t6], %[t8] \n\t"
yading@10 521 "sw %[temp_reg1], 14*4(%[tmp1]) \n\t"
yading@10 522 "sub %[temp_reg2], %[temp_reg2], %[t7] \n\t"
yading@10 523 "addu %[temp_reg3], %[t7], %[t6] \n\t"
yading@10 524 "multu $ac3, %[C_3], %[temp_reg2] \n\t"
yading@10 525 "multu %[C_1], %[temp_reg3] \n\t"
yading@10 526 "sra %[temp_reg1], %[temp_reg2], 31 \n\t"
yading@10 527 "move %[t1], $0 \n\t"
yading@10 528 "sra %[temp_reg3], %[temp_reg3], 31 \n\t"
yading@10 529 "movn %[t1], %[C_3], %[temp_reg1] \n\t"
yading@10 530 "mfhi %[temp_reg1], $ac3 \n\t"
yading@10 531 "mfhi %[temp_reg4] \n\t"
yading@10 532 "move %[t2], $0 \n\t"
yading@10 533 "movn %[t2], %[C_1], %[temp_reg3] \n\t"
yading@10 534 "sub %[temp_reg3], %[t6], %[t8] \n\t"
yading@10 535 "sub %[t2], %[temp_reg4], %[t2] \n\t"
yading@10 536 "multu $ac1, %[C_7], %[temp_reg3] \n\t"
yading@10 537 "sub %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 538 "sra %[temp_reg4], %[temp_reg3], 31 \n\t"
yading@10 539 "sub %[t1], %[temp_reg1], %[t1] \n\t"
yading@10 540 "move %[t3], $0 \n\t"
yading@10 541 "sw %[t1], 4*4(%[tmp1]) \n\t"
yading@10 542 "movn %[t3], %[C_7], %[temp_reg4] \n\t"
yading@10 543 "multu $ac2, %[C_3A], %[t4] \n\t"
yading@10 544 "add %[temp_reg2], %[t7], %[t8] \n\t"
yading@10 545 "move %[t1], $0 \n\t"
yading@10 546 "mfhi %[temp_reg4], $ac1 \n\t"
yading@10 547 "multu $ac3,%[C_5], %[temp_reg2] \n\t"
yading@10 548 "move %[t0], $0 \n\t"
yading@10 549 "sra %[temp_reg1], %[temp_reg2], 31 \n\t"
yading@10 550 "movn %[t1],%[C_5], %[temp_reg1] \n\t"
yading@10 551 "sub %[temp_reg4], %[temp_reg4], %[temp_reg3] \n\t"
yading@10 552 "mfhi %[temp_reg1], $ac3 \n\t"
yading@10 553 "sra %[temp_reg3], %[t4], 31 \n\t"
yading@10 554 "movn %[t0], %[C_3A], %[temp_reg3] \n\t"
yading@10 555 "mfhi %[temp_reg3], $ac2 \n\t"
yading@10 556 "sub %[t3], %[temp_reg4], %[t3] \n\t"
yading@10 557 "add %[temp_reg4], %[t3], %[t2] \n\t"
yading@10 558 "sub %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 559 "sub %[t1], %[temp_reg1], %[t1] \n\t"
yading@10 560 "sub %[t0], %[temp_reg3], %[t0] \n\t"
yading@10 561 "add %[temp_reg1], %[t2], %[t1] \n\t"
yading@10 562 "add %[temp_reg4], %[temp_reg4], %[t0] \n\t"
yading@10 563 "sub %[temp_reg2], %[t3], %[t1] \n\t"
yading@10 564 "sw %[temp_reg4], 0*4(%[tmp1]) \n\t"
yading@10 565 "sub %[temp_reg1], %[temp_reg1], %[t0] \n\t"
yading@10 566 "sub %[temp_reg2], %[temp_reg2], %[t0] \n\t"
yading@10 567 "sw %[temp_reg1], 12*4(%[tmp1]) \n\t"
yading@10 568 "sw %[temp_reg2], 8*4(%[tmp1]) \n\t"
yading@10 569
yading@10 570 : [t7] "=&r" (t7), [temp_reg1] "=&r" (temp_reg1),
yading@10 571 [temp_reg2] "=&r" (temp_reg2), [temp_reg4] "=&r" (temp_reg4),
yading@10 572 [temp_reg3] "=&r" (temp_reg3), [t8] "=&r" (t8), [t0] "=&r" (t0),
yading@10 573 [t4] "=&r" (t4), [t5] "=&r" (t5), [t6] "=&r"(t6), [t2] "=&r" (t2),
yading@10 574 [t3] "=&r" (t3), [t1] "=&r" (t1)
yading@10 575 : [C_2] "r" (C_2), [in1] "r" (in1), [tmp1] "r" (tmp1), [C_8] "r" (C_8),
yading@10 576 [C_4] "r" (C_4), [C_3] "r" (C_3), [C_1] "r" (C_1), [C_7] "r" (C_7),
yading@10 577 [C_3A] "r" (C_3A), [C_5] "r" (C_5)
yading@10 578 : "memory", "hi", "lo", "$ac1hi", "$ac1lo", "$ac2hi", "$ac2lo",
yading@10 579 "$ac3hi", "$ac3lo"
yading@10 580 );
yading@10 581 }
yading@10 582
yading@10 583 /**
yading@10 584 * loop is unrolled four times
yading@10 585 *
yading@10 586 * values defined in tables(icos36[] and icos36h[]) are not loaded from
yading@10 587 * these tables - they are directly loaded in appropriate registers
yading@10 588 *
yading@10 589 */
yading@10 590
yading@10 591 __asm__ volatile (
yading@10 592 "lw %[t2], 1*4(%[tmp]) \n\t"
yading@10 593 "lw %[t3], 3*4(%[tmp]) \n\t"
yading@10 594 "lw %[t0], 0*4(%[tmp]) \n\t"
yading@10 595 "lw %[t1], 2*4(%[tmp]) \n\t"
yading@10 596 "addu %[temp_reg1], %[t3], %[t2] \n\t"
yading@10 597 "li %[temp_reg2], 0x807D2B1E \n\t"
yading@10 598 "move %[s1], $0 \n\t"
yading@10 599 "multu %[temp_reg2], %[temp_reg1] \n\t"
yading@10 600 "sra %[temp_reg1], %[temp_reg1], 31 \n\t"
yading@10 601 "movn %[s1], %[temp_reg2], %[temp_reg1] \n\t"
yading@10 602 "sub %[temp_reg3], %[t3], %[t2] \n\t"
yading@10 603 "li %[temp_reg4], 0x2de5151 \n\t"
yading@10 604 "mfhi %[temp_reg2] \n\t"
yading@10 605 "addu %[s0], %[t1], %[t0] \n\t"
yading@10 606 "lw %[temp_reg5], 9*4(%[win]) \n\t"
yading@10 607 "mult $ac1, %[temp_reg4], %[temp_reg3] \n\t"
yading@10 608 "lw %[temp_reg6], 4*9*4(%[buf]) \n\t"
yading@10 609 "sub %[s2], %[t1], %[t0] \n\t"
yading@10 610 "lw %[temp_reg3], 29*4(%[win]) \n\t"
yading@10 611 "subu %[s1], %[temp_reg2], %[s1] \n\t"
yading@10 612 "lw %[temp_reg4], 28*4(%[win]) \n\t"
yading@10 613 "add %[t0], %[s0], %[s1] \n\t"
yading@10 614 "extr.w %[s3], $ac1,23 \n\t"
yading@10 615 "mult $ac2, %[t0], %[temp_reg3] \n\t"
yading@10 616 "sub %[t1], %[s0], %[s1] \n\t"
yading@10 617 "lw %[temp_reg1], 4*8*4(%[buf]) \n\t"
yading@10 618 "mult %[t1], %[temp_reg5] \n\t"
yading@10 619 "lw %[temp_reg2], 8*4(%[win]) \n\t"
yading@10 620 "mfhi %[temp_reg3], $ac2 \n\t"
yading@10 621 "mult $ac3, %[t0], %[temp_reg4] \n\t"
yading@10 622 "add %[t0], %[s2], %[s3] \n\t"
yading@10 623 "mfhi %[temp_reg5] \n\t"
yading@10 624 "mult $ac1, %[t1], %[temp_reg2] \n\t"
yading@10 625 "sub %[t1], %[s2], %[s3] \n\t"
yading@10 626 "sw %[temp_reg3], 4*9*4(%[buf]) \n\t"
yading@10 627 "mfhi %[temp_reg4], $ac3 \n\t"
yading@10 628 "lw %[temp_reg3], 37*4(%[win]) \n\t"
yading@10 629 "mfhi %[temp_reg2], $ac1 \n\t"
yading@10 630 "add %[temp_reg5], %[temp_reg5], %[temp_reg6] \n\t"
yading@10 631 "lw %[temp_reg6], 17*4(%[win]) \n\t"
yading@10 632 "sw %[temp_reg5], 32*9*4(%[out]) \n\t"
yading@10 633 "sw %[temp_reg4], 4*8*4(%[buf]) \n\t"
yading@10 634 "mult %[t1], %[temp_reg6] \n\t"
yading@10 635 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 636 "lw %[temp_reg2], 0*4(%[win]) \n\t"
yading@10 637 "lw %[temp_reg5], 4*17*4(%[buf]) \n\t"
yading@10 638 "sw %[temp_reg1], 8*32*4(%[out]) \n\t"
yading@10 639 "mfhi %[temp_reg6] \n\t"
yading@10 640 "mult $ac1, %[t1], %[temp_reg2] \n\t"
yading@10 641 "lw %[temp_reg4], 20*4(%[win]) \n\t"
yading@10 642 "lw %[temp_reg1], 0(%[buf]) \n\t"
yading@10 643 "mult $ac2, %[t0], %[temp_reg3] \n\t"
yading@10 644 "mult %[t0], %[temp_reg4] \n\t"
yading@10 645 "mfhi %[temp_reg2], $ac1 \n\t"
yading@10 646 "lw %[t0], 4*4(%[tmp]) \n\t"
yading@10 647 "add %[temp_reg5], %[temp_reg5], %[temp_reg6] \n\t"
yading@10 648 "mfhi %[temp_reg3], $ac2 \n\t"
yading@10 649 "mfhi %[temp_reg4] \n\t"
yading@10 650 "sw %[temp_reg5], 17*32*4(%[out]) \n\t"
yading@10 651 "lw %[t1], 6*4(%[tmp]) \n\t"
yading@10 652 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 653 "lw %[t2], 5*4(%[tmp]) \n\t"
yading@10 654 "sw %[temp_reg1], 0*32*4(%[out]) \n\t"
yading@10 655 "addu %[s0], %[t1], %[t0] \n\t"
yading@10 656 "sw %[temp_reg3], 4*17*4(%[buf]) \n\t"
yading@10 657 "lw %[t3], 7*4(%[tmp]) \n\t"
yading@10 658 "sub %[s2], %[t1], %[t0] \n\t"
yading@10 659 "sw %[temp_reg4], 0(%[buf]) \n\t"
yading@10 660 "addu %[temp_reg5], %[t3], %[t2] \n\t"
yading@10 661 "li %[temp_reg6], 0x8483EE0C \n\t"
yading@10 662 "move %[s1], $0 \n\t"
yading@10 663 "multu %[temp_reg6], %[temp_reg5] \n\t"
yading@10 664 "sub %[temp_reg1], %[t3], %[t2] \n\t"
yading@10 665 "li %[temp_reg2], 0xf746ea \n\t"
yading@10 666 "sra %[temp_reg5], %[temp_reg5], 31 \n\t"
yading@10 667 "mult $ac1, %[temp_reg2], %[temp_reg1] \n\t"
yading@10 668 "movn %[s1], %[temp_reg6], %[temp_reg5] \n\t"
yading@10 669 "mfhi %[temp_reg5] \n\t"
yading@10 670 "lw %[temp_reg3], 10*4(%[win]) \n\t"
yading@10 671 "lw %[temp_reg4], 4*10*4(%[buf]) \n\t"
yading@10 672 "extr.w %[s3], $ac1, 23 \n\t"
yading@10 673 "lw %[temp_reg1], 4*7*4(%[buf]) \n\t"
yading@10 674 "lw %[temp_reg2], 7*4(%[win]) \n\t"
yading@10 675 "lw %[temp_reg6], 30*4(%[win]) \n\t"
yading@10 676 "subu %[s1], %[temp_reg5], %[s1] \n\t"
yading@10 677 "sub %[t1], %[s0], %[s1] \n\t"
yading@10 678 "add %[t0], %[s0], %[s1] \n\t"
yading@10 679 "mult $ac2, %[t1], %[temp_reg3] \n\t"
yading@10 680 "mult $ac3, %[t1], %[temp_reg2] \n\t"
yading@10 681 "mult %[t0], %[temp_reg6] \n\t"
yading@10 682 "lw %[temp_reg5], 27*4(%[win]) \n\t"
yading@10 683 "mult $ac1, %[t0], %[temp_reg5] \n\t"
yading@10 684 "mfhi %[temp_reg3], $ac2 \n\t"
yading@10 685 "mfhi %[temp_reg2], $ac3 \n\t"
yading@10 686 "mfhi %[temp_reg6] \n\t"
yading@10 687 "add %[t0], %[s2], %[s3] \n\t"
yading@10 688 "sub %[t1], %[s2], %[s3] \n\t"
yading@10 689 "add %[temp_reg3], %[temp_reg3], %[temp_reg4] \n\t"
yading@10 690 "lw %[temp_reg4], 16*4(%[win]) \n\t"
yading@10 691 "mfhi %[temp_reg5], $ac1 \n\t"
yading@10 692 "sw %[temp_reg3], 32*10*4(%[out]) \n\t"
yading@10 693 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 694 "lw %[temp_reg3], 4*16*4(%[buf]) \n\t"
yading@10 695 "sw %[temp_reg6], 4*10*4(%[buf]) \n\t"
yading@10 696 "sw %[temp_reg1], 7*32*4(%[out]) \n\t"
yading@10 697 "mult $ac2, %[t1], %[temp_reg4] \n\t"
yading@10 698 "sw %[temp_reg5], 4*7*4(%[buf]) \n\t"
yading@10 699 "lw %[temp_reg6], 1*4(%[win]) \n\t"
yading@10 700 "lw %[temp_reg5], 4*1*4(%[buf]) \n\t"
yading@10 701 "lw %[temp_reg1], 36*4(%[win]) \n\t"
yading@10 702 "mult $ac3, %[t1], %[temp_reg6] \n\t"
yading@10 703 "lw %[temp_reg2], 21*4(%[win]) \n\t"
yading@10 704 "mfhi %[temp_reg4], $ac2 \n\t"
yading@10 705 "mult %[t0], %[temp_reg1] \n\t"
yading@10 706 "mult $ac1, %[t0],%[temp_reg2] \n\t"
yading@10 707 "lw %[t0], 8*4(%[tmp]) \n\t"
yading@10 708 "mfhi %[temp_reg6], $ac3 \n\t"
yading@10 709 "lw %[t1], 10*4(%[tmp]) \n\t"
yading@10 710 "lw %[t3], 11*4(%[tmp]) \n\t"
yading@10 711 "mfhi %[temp_reg1] \n\t"
yading@10 712 "add %[temp_reg3], %[temp_reg3], %[temp_reg4] \n\t"
yading@10 713 "lw %[t2], 9*4(%[tmp]) \n\t"
yading@10 714 "mfhi %[temp_reg2], $ac1 \n\t"
yading@10 715 "add %[temp_reg5], %[temp_reg5], %[temp_reg6] \n\t"
yading@10 716 "sw %[temp_reg3], 16*32*4(%[out]) \n\t"
yading@10 717 "sw %[temp_reg5], 1*32*4(%[out]) \n\t"
yading@10 718 "sw %[temp_reg1], 4*16*4(%[buf]) \n\t"
yading@10 719 "addu %[temp_reg3], %[t3], %[t2] \n\t"
yading@10 720 "li %[temp_reg4], 0x8D3B7CD6 \n\t"
yading@10 721 "sw %[temp_reg2], 4*1*4(%[buf]) \n\t"
yading@10 722 "multu %[temp_reg4],%[temp_reg3] \n\t"
yading@10 723 "sra %[temp_reg3], %[temp_reg3], 31 \n\t"
yading@10 724 "move %[s1], $0 \n\t"
yading@10 725 "movn %[s1], %[temp_reg4], %[temp_reg3] \n\t"
yading@10 726 "addu %[s0], %[t1], %[t0] \n\t"
yading@10 727 "mfhi %[temp_reg3] \n\t"
yading@10 728 "sub %[s2], %[t1], %[t0] \n\t"
yading@10 729 "sub %[temp_reg5], %[t3], %[t2] \n\t"
yading@10 730 "li %[temp_reg6], 0x976fd9 \n\t"
yading@10 731 "lw %[temp_reg2], 11*4(%[win]) \n\t"
yading@10 732 "lw %[temp_reg1], 4*11*4(%[buf]) \n\t"
yading@10 733 "mult $ac1, %[temp_reg6], %[temp_reg5] \n\t"
yading@10 734 "subu %[s1], %[temp_reg3], %[s1] \n\t"
yading@10 735 "lw %[temp_reg5], 31*4(%[win]) \n\t"
yading@10 736 "sub %[t1], %[s0], %[s1] \n\t"
yading@10 737 "add %[t0], %[s0], %[s1] \n\t"
yading@10 738 "mult $ac2, %[t1], %[temp_reg2] \n\t"
yading@10 739 "mult %[t0], %[temp_reg5] \n\t"
yading@10 740 "lw %[temp_reg4], 6*4(%[win]) \n\t"
yading@10 741 "extr.w %[s3], $ac1, 23 \n\t"
yading@10 742 "lw %[temp_reg3], 4*6*4(%[buf]) \n\t"
yading@10 743 "mfhi %[temp_reg2], $ac2 \n\t"
yading@10 744 "lw %[temp_reg6], 26*4(%[win]) \n\t"
yading@10 745 "mfhi %[temp_reg5] \n\t"
yading@10 746 "mult $ac3, %[t1], %[temp_reg4] \n\t"
yading@10 747 "mult $ac1, %[t0], %[temp_reg6] \n\t"
yading@10 748 "add %[t0], %[s2], %[s3] \n\t"
yading@10 749 "sub %[t1], %[s2], %[s3] \n\t"
yading@10 750 "add %[temp_reg2], %[temp_reg2], %[temp_reg1] \n\t"
yading@10 751 "mfhi %[temp_reg4], $ac3 \n\t"
yading@10 752 "mfhi %[temp_reg6], $ac1 \n\t"
yading@10 753 "sw %[temp_reg5], 4*11*4(%[buf]) \n\t"
yading@10 754 "sw %[temp_reg2], 32*11*4(%[out]) \n\t"
yading@10 755 "lw %[temp_reg1], 4*15*4(%[buf]) \n\t"
yading@10 756 "add %[temp_reg3], %[temp_reg3], %[temp_reg4] \n\t"
yading@10 757 "lw %[temp_reg2], 15*4(%[win]) \n\t"
yading@10 758 "sw %[temp_reg3], 6*32*4(%[out]) \n\t"
yading@10 759 "sw %[temp_reg6], 4*6*4(%[buf]) \n\t"
yading@10 760 "mult %[t1], %[temp_reg2] \n\t"
yading@10 761 "lw %[temp_reg3], 2*4(%[win]) \n\t"
yading@10 762 "lw %[temp_reg4], 4*2*4(%[buf]) \n\t"
yading@10 763 "lw %[temp_reg5], 35*4(%[win]) \n\t"
yading@10 764 "mult $ac1, %[t1], %[temp_reg3] \n\t"
yading@10 765 "mfhi %[temp_reg2] \n\t"
yading@10 766 "lw %[temp_reg6], 22*4(%[win]) \n\t"
yading@10 767 "mult $ac2, %[t0], %[temp_reg5] \n\t"
yading@10 768 "lw %[t1], 14*4(%[tmp]) \n\t"
yading@10 769 "mult $ac3, %[t0], %[temp_reg6] \n\t"
yading@10 770 "lw %[t0], 12*4(%[tmp]) \n\t"
yading@10 771 "mfhi %[temp_reg3], $ac1 \n\t"
yading@10 772 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 773 "mfhi %[temp_reg5], $ac2 \n\t"
yading@10 774 "sw %[temp_reg1], 15*32*4(%[out]) \n\t"
yading@10 775 "mfhi %[temp_reg6], $ac3 \n\t"
yading@10 776 "lw %[t2], 13*4(%[tmp]) \n\t"
yading@10 777 "lw %[t3], 15*4(%[tmp]) \n\t"
yading@10 778 "add %[temp_reg4], %[temp_reg4], %[temp_reg3] \n\t"
yading@10 779 "sw %[temp_reg5], 4*15*4(%[buf]) \n\t"
yading@10 780 "addu %[temp_reg1], %[t3], %[t2] \n\t"
yading@10 781 "li %[temp_reg2], 0x9C42577C \n\t"
yading@10 782 "move %[s1], $0 \n\t"
yading@10 783 "multu %[temp_reg2], %[temp_reg1] \n\t"
yading@10 784 "sw %[temp_reg4], 2*32*4(%[out]) \n\t"
yading@10 785 "sra %[temp_reg1], %[temp_reg1], 31 \n\t"
yading@10 786 "movn %[s1], %[temp_reg2], %[temp_reg1] \n\t"
yading@10 787 "sub %[temp_reg3], %[t3], %[t2] \n\t"
yading@10 788 "li %[temp_reg4], 0x6f94a2 \n\t"
yading@10 789 "mfhi %[temp_reg1] \n\t"
yading@10 790 "addu %[s0], %[t1], %[t0] \n\t"
yading@10 791 "sw %[temp_reg6], 4*2*4(%[buf]) \n\t"
yading@10 792 "mult $ac1, %[temp_reg4], %[temp_reg3] \n\t"
yading@10 793 "sub %[s2], %[t1], %[t0] \n\t"
yading@10 794 "lw %[temp_reg5], 12*4(%[win]) \n\t"
yading@10 795 "lw %[temp_reg6], 4*12*4(%[buf]) \n\t"
yading@10 796 "subu %[s1], %[temp_reg1], %[s1] \n\t"
yading@10 797 "sub %[t1], %[s0], %[s1] \n\t"
yading@10 798 "lw %[temp_reg3], 32*4(%[win]) \n\t"
yading@10 799 "mult $ac2, %[t1], %[temp_reg5] \n\t"
yading@10 800 "add %[t0], %[s0], %[s1] \n\t"
yading@10 801 "extr.w %[s3], $ac1, 23 \n\t"
yading@10 802 "lw %[temp_reg2], 5*4(%[win]) \n\t"
yading@10 803 "mult %[t0], %[temp_reg3] \n\t"
yading@10 804 "mfhi %[temp_reg5], $ac2 \n\t"
yading@10 805 "lw %[temp_reg4], 25*4(%[win]) \n\t"
yading@10 806 "lw %[temp_reg1], 4*5*4(%[buf]) \n\t"
yading@10 807 "mult $ac3, %[t1], %[temp_reg2] \n\t"
yading@10 808 "mult $ac1, %[t0], %[temp_reg4] \n\t"
yading@10 809 "mfhi %[temp_reg3] \n\t"
yading@10 810 "add %[t0], %[s2], %[s3] \n\t"
yading@10 811 "add %[temp_reg5], %[temp_reg5], %[temp_reg6] \n\t"
yading@10 812 "mfhi %[temp_reg2], $ac3 \n\t"
yading@10 813 "mfhi %[temp_reg4], $ac1 \n\t"
yading@10 814 "sub %[t1], %[s2], %[s3] \n\t"
yading@10 815 "sw %[temp_reg5], 32*12*4(%[out]) \n\t"
yading@10 816 "sw %[temp_reg3], 4*12*4(%[buf]) \n\t"
yading@10 817 "lw %[temp_reg6], 14*4(%[win]) \n\t"
yading@10 818 "lw %[temp_reg5], 4*14*4(%[buf]) \n\t"
yading@10 819 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 820 "sw %[temp_reg4], 4*5*4(%[buf]) \n\t"
yading@10 821 "sw %[temp_reg1], 5*32*4(%[out]) \n\t"
yading@10 822 "mult %[t1], %[temp_reg6] \n\t"
yading@10 823 "lw %[temp_reg4], 34*4(%[win]) \n\t"
yading@10 824 "lw %[temp_reg2], 3*4(%[win]) \n\t"
yading@10 825 "lw %[temp_reg1], 4*3*4(%[buf]) \n\t"
yading@10 826 "mult $ac2, %[t0], %[temp_reg4] \n\t"
yading@10 827 "mfhi %[temp_reg6] \n\t"
yading@10 828 "mult $ac1, %[t1], %[temp_reg2] \n\t"
yading@10 829 "lw %[temp_reg3], 23*4(%[win]) \n\t"
yading@10 830 "lw %[s0], 16*4(%[tmp]) \n\t"
yading@10 831 "mfhi %[temp_reg4], $ac2 \n\t"
yading@10 832 "lw %[t1], 17*4(%[tmp]) \n\t"
yading@10 833 "mult $ac3, %[t0], %[temp_reg3] \n\t"
yading@10 834 "move %[s1], $0 \n\t"
yading@10 835 "add %[temp_reg5], %[temp_reg5], %[temp_reg6] \n\t"
yading@10 836 "mfhi %[temp_reg2], $ac1 \n\t"
yading@10 837 "sw %[temp_reg5], 14*32*4(%[out]) \n\t"
yading@10 838 "sw %[temp_reg4], 4*14*4(%[buf]) \n\t"
yading@10 839 "mfhi %[temp_reg3], $ac3 \n\t"
yading@10 840 "li %[temp_reg5], 0xB504F334 \n\t"
yading@10 841 "add %[temp_reg1], %[temp_reg1], %[temp_reg2] \n\t"
yading@10 842 "multu %[temp_reg5], %[t1] \n\t"
yading@10 843 "lw %[temp_reg2], 4*13*4(%[buf]) \n\t"
yading@10 844 "sw %[temp_reg1], 3*32*4(%[out]) \n\t"
yading@10 845 "sra %[t1], %[t1], 31 \n\t"
yading@10 846 "mfhi %[temp_reg6] \n\t"
yading@10 847 "movn %[s1], %[temp_reg5], %[t1] \n\t"
yading@10 848 "sw %[temp_reg3], 4*3*4(%[buf]) \n\t"
yading@10 849 "lw %[temp_reg1], 13*4(%[win]) \n\t"
yading@10 850 "lw %[temp_reg4], 4*4*4(%[buf]) \n\t"
yading@10 851 "lw %[temp_reg3], 4*4(%[win]) \n\t"
yading@10 852 "lw %[temp_reg5], 33*4(%[win]) \n\t"
yading@10 853 "subu %[s1], %[temp_reg6], %[s1] \n\t"
yading@10 854 "lw %[temp_reg6], 24*4(%[win]) \n\t"
yading@10 855 "sub %[t1], %[s0], %[s1] \n\t"
yading@10 856 "add %[t0], %[s0], %[s1] \n\t"
yading@10 857 "mult $ac1, %[t1], %[temp_reg1] \n\t"
yading@10 858 "mult $ac2, %[t1], %[temp_reg3] \n\t"
yading@10 859 "mult $ac3, %[t0], %[temp_reg5] \n\t"
yading@10 860 "mult %[t0], %[temp_reg6] \n\t"
yading@10 861 "mfhi %[temp_reg1], $ac1 \n\t"
yading@10 862 "mfhi %[temp_reg3], $ac2 \n\t"
yading@10 863 "mfhi %[temp_reg5], $ac3 \n\t"
yading@10 864 "mfhi %[temp_reg6] \n\t"
yading@10 865 "add %[temp_reg2], %[temp_reg2], %[temp_reg1] \n\t"
yading@10 866 "add %[temp_reg4], %[temp_reg4], %[temp_reg3] \n\t"
yading@10 867 "sw %[temp_reg2], 13*32*4(%[out]) \n\t"
yading@10 868 "sw %[temp_reg4], 4*32*4(%[out]) \n\t"
yading@10 869 "sw %[temp_reg5], 4*13*4(%[buf]) \n\t"
yading@10 870 "sw %[temp_reg6], 4*4*4(%[buf]) \n\t"
yading@10 871
yading@10 872 : [t0] "=&r" (t0), [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3),
yading@10 873 [s0] "=&r" (s0), [s2] "=&r" (s2), [temp_reg1] "=&r" (temp_reg1),
yading@10 874 [temp_reg2] "=&r" (temp_reg2), [s1] "=&r" (s1), [s3] "=&r" (s3),
yading@10 875 [temp_reg3] "=&r" (temp_reg3), [temp_reg4] "=&r" (temp_reg4),
yading@10 876 [temp_reg5] "=&r" (temp_reg5), [temp_reg6] "=&r" (temp_reg6),
yading@10 877 [out] "+r" (out)
yading@10 878 : [tmp] "r" (tmp), [win] "r" (win), [buf] "r" (buf)
yading@10 879 : "memory", "hi", "lo", "$ac1hi", "$ac1lo", "$ac2hi", "$ac2lo",
yading@10 880 "$ac3hi", "$ac3lo"
yading@10 881 );
yading@10 882 }
yading@10 883
yading@10 884 static void ff_imdct36_blocks_mips_fixed(int *out, int *buf, int *in,
yading@10 885 int count, int switch_point, int block_type)
yading@10 886 {
yading@10 887 int j;
yading@10 888 for (j=0 ; j < count; j++) {
yading@10 889 /* apply window & overlap with previous buffer */
yading@10 890
yading@10 891 /* select window */
yading@10 892 int win_idx = (switch_point && j < 2) ? 0 : block_type;
yading@10 893 int *win = ff_mdct_win_fixed[win_idx + (4 & -(j & 1))];
yading@10 894
yading@10 895 imdct36_mips_fixed(out, buf, in, win);
yading@10 896
yading@10 897 in += 18;
yading@10 898 buf += ((j&3) != 3 ? 1 : (72-3));
yading@10 899 out++;
yading@10 900 }
yading@10 901 }
yading@10 902
yading@10 903 void ff_mpadsp_init_mipsdspr1(MPADSPContext *s)
yading@10 904 {
yading@10 905 s->apply_window_fixed = ff_mpadsp_apply_window_mips_fixed;
yading@10 906 s->imdct36_blocks_fixed = ff_imdct36_blocks_mips_fixed;
yading@10 907 }