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 "config.h"
59 #include "libavutil/mips/asmdefs.h"
60 #include "libavcodec/mpegaudiodsp.h"
61
62 #if HAVE_INLINE_ASM && HAVE_MIPSFPU
63 #if !HAVE_MIPS32R6 && !HAVE_MIPS64R6
64
ff_mpadsp_apply_window_mips_float(float * synth_buf,float * window,int * dither_state,float * samples,ptrdiff_t incr)65 static void ff_mpadsp_apply_window_mips_float(float *synth_buf, float *window,
66 int *dither_state, float *samples, ptrdiff_t incr)
67 {
68 register const float *w, *w2, *p;
69 int j;
70 float *samples2;
71 float sum, sum2;
72 /* temporary variables */
73 int incr1 = incr << 2;
74 int t_sample;
75 float in1, in2, in3, in4, in5, in6, in7, in8;
76 float *p2;
77
78 /* copy to avoid wrap */
79 memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));
80
81 /**
82 * instructions are scheduled to minimize pipeline stall.
83 * use of round_sample function from the original code is
84 * changed with appropriate assembly instructions.
85 */
86
87 __asm__ volatile (
88 "lwc1 %[sum], 0(%[dither_state]) \t\n"
89 "sll %[t_sample], %[incr1], 5 \t\n"
90 "sub %[t_sample], %[t_sample], %[incr1] \n\t"
91 "li %[j], 4 \t\n"
92 "lwc1 %[in1], 0(%[window]) \t\n"
93 "lwc1 %[in2], 16*4(%[synth_buf]) \t\n"
94 "sw $zero, 0(%[dither_state]) \t\n"
95 "lwc1 %[in3], 64*4(%[window]) \t\n"
96 "lwc1 %[in4], 80*4(%[synth_buf]) \t\n"
97 PTR_ADDU "%[samples2],%[samples], %[t_sample] \t\n"
98 "madd.s %[sum], %[sum], %[in1], %[in2] \t\n"
99 "lwc1 %[in5], 128*4(%[window]) \t\n"
100 "lwc1 %[in6], 144*4(%[synth_buf]) \t\n"
101 "lwc1 %[in7], 192*4(%[window]) \t\n"
102 "madd.s %[sum], %[sum], %[in3], %[in4] \t\n"
103 "lwc1 %[in8], 208*4(%[synth_buf]) \t\n"
104 "lwc1 %[in1], 256*4(%[window]) \t\n"
105 "lwc1 %[in2], 272*4(%[synth_buf]) \t\n"
106 "madd.s %[sum], %[sum], %[in5], %[in6] \t\n"
107 "lwc1 %[in3], 320*4(%[window]) \t\n"
108 "lwc1 %[in4], 336*4(%[synth_buf]) \t\n"
109 "lwc1 %[in5], 384*4(%[window]) \t\n"
110 "madd.s %[sum], %[sum], %[in7], %[in8] \t\n"
111 "lwc1 %[in6], 400*4(%[synth_buf]) \t\n"
112 "lwc1 %[in7], 448*4(%[window]) \t\n"
113 "lwc1 %[in8], 464*4(%[synth_buf]) \t\n"
114 "madd.s %[sum], %[sum], %[in1], %[in2] \t\n"
115 "lwc1 %[in1], 32*4(%[window]) \t\n"
116 "lwc1 %[in2], 48*4(%[synth_buf]) \t\n"
117 "madd.s %[sum], %[sum], %[in3], %[in4] \t\n"
118 "lwc1 %[in3], 96*4(%[window]) \t\n"
119 "lwc1 %[in4], 112*4(%[synth_buf]) \t\n"
120 "madd.s %[sum], %[sum], %[in5], %[in6] \t\n"
121 "lwc1 %[in5], 160*4(%[window]) \t\n"
122 "lwc1 %[in6], 176*4(%[synth_buf]) \t\n"
123 "madd.s %[sum], %[sum], %[in7], %[in8] \t\n"
124 "lwc1 %[in7], 224*4(%[window]) \t\n"
125 "lwc1 %[in8], 240*4(%[synth_buf]) \t\n"
126 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
127 "lwc1 %[in1], 288*4(%[window]) \t\n"
128 "lwc1 %[in2], 304*4(%[synth_buf]) \t\n"
129 "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n"
130 "lwc1 %[in3], 352*4(%[window]) \t\n"
131 "lwc1 %[in4], 368*4(%[synth_buf]) \t\n"
132 "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n"
133 "lwc1 %[in5], 416*4(%[window]) \t\n"
134 "lwc1 %[in6], 432*4(%[synth_buf]) \t\n"
135 "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n"
136 "lwc1 %[in7], 480*4(%[window]) \t\n"
137 "lwc1 %[in8], 496*4(%[synth_buf]) \t\n"
138 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
139 PTR_ADDU "%[w], %[window], 4 \t\n"
140 "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n"
141 PTR_ADDU "%[w2], %[window], 124 \t\n"
142 PTR_ADDIU "%[p], %[synth_buf], 68 \t\n"
143 PTR_ADDIU "%[p2], %[synth_buf], 188 \t\n"
144 "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n"
145 "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n"
146 "swc1 %[sum], 0(%[samples]) \t\n"
147 PTR_ADDU "%[samples], %[samples], %[incr1] \t\n"
148
149 /* calculate two samples at the same time to avoid one memory
150 access per two sample */
151
152 "ff_mpadsp_apply_window_loop%=: \t\n"
153 "lwc1 %[in1], 0(%[w]) \t\n"
154 "lwc1 %[in2], 0(%[p]) \t\n"
155 "lwc1 %[in3], 0(%[w2]) \t\n"
156 "lwc1 %[in4], 64*4(%[w]) \t\n"
157 "lwc1 %[in5], 64*4(%[p]) \t\n"
158 "lwc1 %[in6], 64*4(%[w2]) \t\n"
159 "mul.s %[sum], %[in1], %[in2] \t\n"
160 "mul.s %[sum2], %[in2], %[in3] \t\n"
161 "lwc1 %[in1], 128*4(%[w]) \t\n"
162 "lwc1 %[in2], 128*4(%[p]) \t\n"
163 "madd.s %[sum], %[sum], %[in4], %[in5] \t\n"
164 "nmadd.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
165 "lwc1 %[in3], 128*4(%[w2]) \t\n"
166 "lwc1 %[in4], 192*4(%[w]) \t\n"
167 "madd.s %[sum], %[sum], %[in1], %[in2] \t\n"
168 "lwc1 %[in5], 192*4(%[p]) \t\n"
169 "lwc1 %[in6], 192*4(%[w2]) \t\n"
170 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
171 "lwc1 %[in1], 256*4(%[w]) \t\n"
172 "lwc1 %[in2], 256*4(%[p]) \t\n"
173 "madd.s %[sum], %[sum], %[in4], %[in5] \t\n"
174 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
175 "lwc1 %[in3], 256*4(%[w2]) \t\n"
176 "lwc1 %[in4], 320*4(%[w]) \t\n"
177 "madd.s %[sum], %[sum], %[in1], %[in2] \t\n"
178 "lwc1 %[in5], 320*4(%[p]) \t\n"
179 "lwc1 %[in6], 320*4(%[w2]) \t\n"
180 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
181 "lwc1 %[in1], 384*4(%[w]) \t\n"
182 "lwc1 %[in2], 384*4(%[p]) \t\n"
183 "madd.s %[sum], %[sum], %[in4], %[in5] \t\n"
184 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
185 "lwc1 %[in3], 384*4(%[w2]) \t\n"
186 "lwc1 %[in4], 448*4(%[w]) \t\n"
187 "madd.s %[sum], %[sum], %[in1], %[in2] \t\n"
188 "lwc1 %[in5], 448*4(%[p]) \t\n"
189 "lwc1 %[in6], 448*4(%[w2]) \t\n"
190 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
191 "madd.s %[sum], %[sum], %[in4], %[in5] \t\n"
192 "lwc1 %[in1], 32*4(%[w]) \t\n"
193 "lwc1 %[in2], 0(%[p2]) \t\n"
194 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
195 "lwc1 %[in3], 32*4(%[w2]) \t\n"
196 "lwc1 %[in4], 96*4(%[w]) \t\n"
197 "lwc1 %[in5], 64*4(%[p2]) \t\n"
198 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
199 "lwc1 %[in6], 96*4(%[w2]) \t\n"
200 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
201 "lwc1 %[in1], 160*4(%[w]) \t\n"
202 "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n"
203 "lwc1 %[in2], 128*4(%[p2]) \t\n"
204 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
205 "lwc1 %[in3], 160*4(%[w2]) \t\n"
206 "lwc1 %[in4], 224*4(%[w]) \t\n"
207 "lwc1 %[in5], 192*4(%[p2]) \t\n"
208 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
209 "lwc1 %[in6], 224*4(%[w2]) \t\n"
210 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
211 "lwc1 %[in1], 288*4(%[w]) \t\n"
212 "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n"
213 "lwc1 %[in2], 256*4(%[p2]) \t\n"
214 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
215 "lwc1 %[in3], 288*4(%[w2]) \t\n"
216 "lwc1 %[in4], 352*4(%[w]) \t\n"
217 "lwc1 %[in5], 320*4(%[p2]) \t\n"
218 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
219 "lwc1 %[in6], 352*4(%[w2]) \t\n"
220 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
221 "lwc1 %[in1], 416*4(%[w]) \t\n"
222 "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n"
223 "lwc1 %[in2], 384*4(%[p2]) \t\n"
224 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
225 "lwc1 %[in3], 416*4(%[w2]) \t\n"
226 "lwc1 %[in4], 480*4(%[w]) \t\n"
227 "lwc1 %[in5], 448*4(%[p2]) \t\n"
228 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
229 "lwc1 %[in6], 480*4(%[w2]) \t\n"
230 "nmsub.s %[sum2], %[sum2], %[in2], %[in3] \t\n"
231 PTR_ADDIU "%[w], %[w], 4 \t\n"
232 "nmsub.s %[sum], %[sum], %[in4], %[in5] \t\n"
233 PTR_ADDIU "%[w2], %[w2], -4 \t\n"
234 "nmsub.s %[sum2], %[sum2], %[in5], %[in6] \t\n"
235 "addu %[j], %[j], 4 \t\n"
236 PTR_ADDIU "%[p], 4 \t\n"
237 "swc1 %[sum], 0(%[samples]) \t\n"
238 PTR_ADDIU "%[p2], -4 \t\n"
239 "swc1 %[sum2], 0(%[samples2]) \t\n"
240 PTR_ADDU "%[samples], %[samples], %[incr1] \t\n"
241 PTR_SUBU "%[samples2],%[samples2], %[incr1] \t\n"
242 "bne %[j], 64, ff_mpadsp_apply_window_loop%= \t\n"
243
244 "lwc1 %[in1], 48*4(%[window]) \t\n"
245 "lwc1 %[in2], 32*4(%[synth_buf]) \t\n"
246 "lwc1 %[in3], 112*4(%[window]) \t\n"
247 "lwc1 %[in4], 96*4(%[synth_buf]) \t\n"
248 "lwc1 %[in5], 176*4(%[window]) \t\n"
249 "lwc1 %[in6], 160*4(%[synth_buf]) \t\n"
250 "mul.s %[sum], %[in1], %[in2] \t\n"
251 "lwc1 %[in7], 240*4(%[window]) \t\n"
252 "lwc1 %[in8], 224*4(%[synth_buf]) \t\n"
253 "lwc1 %[in1], 304*4(%[window]) \t\n"
254 "nmadd.s %[sum], %[sum], %[in3], %[in4] \t\n"
255 "lwc1 %[in2], 288*4(%[synth_buf]) \t\n"
256 "lwc1 %[in3], 368*4(%[window]) \t\n"
257 "lwc1 %[in4], 352*4(%[synth_buf]) \t\n"
258 "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n"
259 "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n"
260 "lwc1 %[in5], 432*4(%[window]) \t\n"
261 "lwc1 %[in6], 416*4(%[synth_buf]) \t\n"
262 "nmsub.s %[sum], %[sum], %[in1], %[in2] \t\n"
263 "lwc1 %[in7], 496*4(%[window]) \t\n"
264 "lwc1 %[in8], 480*4(%[synth_buf]) \t\n"
265 "nmsub.s %[sum], %[sum], %[in3], %[in4] \t\n"
266 "nmsub.s %[sum], %[sum], %[in5], %[in6] \t\n"
267 "nmsub.s %[sum], %[sum], %[in7], %[in8] \t\n"
268 "swc1 %[sum], 0(%[samples]) \t\n"
269
270 : [sum] "=&f" (sum), [sum2] "=&f" (sum2),
271 [w2] "=&r" (w2), [w] "=&r" (w),
272 [p] "=&r" (p), [p2] "=&r" (p2), [j] "=&r" (j),
273 [samples] "+r" (samples), [samples2] "=&r" (samples2),
274 [in1] "=&f" (in1), [in2] "=&f" (in2),
275 [in3] "=&f" (in3), [in4] "=&f" (in4),
276 [in5] "=&f" (in5), [in6] "=&f" (in6),
277 [in7] "=&f" (in7), [in8] "=&f" (in8),
278 [t_sample] "=&r" (t_sample)
279 : [synth_buf] "r" (synth_buf), [window] "r" (window),
280 [dither_state] "r" (dither_state), [incr1] "r" (incr1)
281 : "memory"
282 );
283 }
284
ff_dct32_mips_float(float * out,const float * tab)285 static void ff_dct32_mips_float(float *out, const float *tab)
286 {
287 float val0 , val1 , val2 , val3 , val4 , val5 , val6 , val7,
288 val8 , val9 , val10, val11, val12, val13, val14, val15,
289 val16, val17, val18, val19, val20, val21, val22, val23,
290 val24, val25, val26, val27, val28, val29, val30, val31;
291 float fTmp1, fTmp2, fTmp3, fTmp4, fTmp5, fTmp6, fTmp8, fTmp9;
292 float f1, f2, f3, f4, f5, f6, f7;
293
294 f1 = 0.50241928618815570551;
295 f2 = 0.50060299823519630134;
296 f3 = 10.19000812354805681150;
297 f4 = 5.10114861868916385802;
298 f5 = 0.67480834145500574602;
299 f6 = 0.74453627100229844977;
300 f7 = 0.50979557910415916894;
301 /**
302 * instructions are scheduled to minimize pipeline stall.
303 */
304 __asm__ volatile (
305 "lwc1 %[fTmp1], 0*4(%[tab]) \n\t"
306 "lwc1 %[fTmp2], 31*4(%[tab]) \n\t"
307 "lwc1 %[fTmp3], 15*4(%[tab]) \n\t"
308 "lwc1 %[fTmp4], 16*4(%[tab]) \n\t"
309 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
310 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
311 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
312 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
313 "mul.s %[fTmp8], %[fTmp8], %[f2] \n\t"
314 "add.s %[val0], %[fTmp5], %[fTmp6] \n\t"
315 "sub.s %[val15], %[fTmp5], %[fTmp6] \n\t"
316 "lwc1 %[fTmp1], 7*4(%[tab]) \n\t"
317 "lwc1 %[fTmp2], 24*4(%[tab]) \n\t"
318 "madd.s %[val16], %[fTmp8], %[fTmp9], %[f3] \n\t"
319 "nmsub.s %[val31], %[fTmp8], %[fTmp9], %[f3] \n\t"
320 "mul.s %[val15], %[val15], %[f1] \n\t"
321 "lwc1 %[fTmp3], 8*4(%[tab]) \n\t"
322 "lwc1 %[fTmp4], 23*4(%[tab]) \n\t"
323 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
324 "mul.s %[val31], %[val31], %[f1] \n\t"
325 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
326 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
327 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
328 "add.s %[val7], %[fTmp5], %[fTmp6] \n\t"
329 "sub.s %[val8], %[fTmp5], %[fTmp6] \n\t"
330 "mul.s %[fTmp8], %[fTmp8], %[f5] \n\t"
331 "sub.s %[fTmp2], %[val0], %[val7] \n\t"
332 "mul.s %[val8], %[val8], %[f4] \n\t"
333 "madd.s %[val23], %[fTmp8], %[fTmp9], %[f6] \n\t"
334 "nmsub.s %[val24], %[fTmp8], %[fTmp9], %[f6] \n\t"
335 "add.s %[val0], %[val0], %[val7] \n\t"
336 "mul.s %[val7], %[f7], %[fTmp2] \n\t"
337 "sub.s %[fTmp2], %[val15], %[val8] \n\t"
338 "add.s %[val8], %[val15], %[val8] \n\t"
339 "mul.s %[val24], %[val24], %[f4] \n\t"
340 "sub.s %[fTmp3], %[val16], %[val23] \n\t"
341 "add.s %[val16], %[val16], %[val23] \n\t"
342 "mul.s %[val15], %[f7], %[fTmp2] \n\t"
343 "sub.s %[fTmp4], %[val31], %[val24] \n\t"
344 "mul.s %[val23], %[f7], %[fTmp3] \n\t"
345 "add.s %[val24], %[val31], %[val24] \n\t"
346 "mul.s %[val31], %[f7], %[fTmp4] \n\t"
347
348 : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
349 [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
350 [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
351 [val0] "=&f" (val0), [val7] "=&f" (val7),
352 [val8] "=&f" (val8), [val15] "=&f" (val15),
353 [val16] "=&f" (val16), [val23] "=&f" (val23),
354 [val24] "=&f" (val24), [val31] "=&f" (val31)
355 : [tab] "r" (tab), [f1]"f"(f1), [f2]"f"(f2), [f3]"f"(f3),
356 [f4]"f"(f4), [f5]"f"(f5), [f6]"f"(f6), [f7]"f"(f7)
357 : "memory"
358 );
359
360 f1 = 0.64682178335999012954;
361 f2 = 0.53104259108978417447;
362 f3 = 1.48416461631416627724;
363 f4 = 0.78815462345125022473;
364 f5 = 0.55310389603444452782;
365 f6 = 1.16943993343288495515;
366 f7 = 2.56291544774150617881;
367 __asm__ volatile (
368 "lwc1 %[fTmp1], 3*4(%[tab]) \n\t"
369 "lwc1 %[fTmp2], 28*4(%[tab]) \n\t"
370 "lwc1 %[fTmp3], 12*4(%[tab]) \n\t"
371 "lwc1 %[fTmp4], 19*4(%[tab]) \n\t"
372 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
373 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
374 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
375 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
376 "mul.s %[fTmp8], %[fTmp8], %[f2] \n\t"
377 "add.s %[val3], %[fTmp5], %[fTmp6] \n\t"
378 "sub.s %[val12], %[fTmp5], %[fTmp6] \n\t"
379 "lwc1 %[fTmp1], 4*4(%[tab]) \n\t"
380 "lwc1 %[fTmp2], 27*4(%[tab]) \n\t"
381 "madd.s %[val19], %[fTmp8], %[fTmp9], %[f3] \n\t"
382 "nmsub.s %[val28], %[fTmp8], %[fTmp9], %[f3] \n\t"
383 "mul.s %[val12], %[val12], %[f1] \n\t"
384 "lwc1 %[fTmp3], 11*4(%[tab]) \n\t"
385 "lwc1 %[fTmp4], 20*4(%[tab]) \n\t"
386 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
387 "mul.s %[val28], %[val28], %[f1] \n\t"
388 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
389 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
390 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
391 "mul.s %[fTmp8], %[fTmp8], %[f5] \n\t"
392 "add.s %[val4], %[fTmp5], %[fTmp6] \n\t"
393 "sub.s %[val11], %[fTmp5], %[fTmp6] \n\t"
394 "madd.s %[val20], %[fTmp8], %[fTmp9], %[f6] \n\t"
395 "nmsub.s %[val27], %[fTmp8], %[fTmp9], %[f6] \n\t"
396 "mul.s %[val11], %[val11], %[f4] \n\t"
397 "sub.s %[fTmp2], %[val3], %[val4] \n\t"
398 "add.s %[val3], %[val3], %[val4] \n\t"
399 "sub.s %[fTmp4], %[val19], %[val20] \n\t"
400 "mul.s %[val27], %[val27], %[f4] \n\t"
401 "sub.s %[fTmp3], %[val12], %[val11] \n\t"
402 "mul.s %[val4], %[f7], %[fTmp2] \n\t"
403 "add.s %[val11], %[val12], %[val11] \n\t"
404 "add.s %[val19], %[val19], %[val20] \n\t"
405 "mul.s %[val20], %[f7], %[fTmp4] \n\t"
406 "mul.s %[val12], %[f7], %[fTmp3] \n\t"
407 "sub.s %[fTmp2], %[val28], %[val27] \n\t"
408 "add.s %[val27], %[val28], %[val27] \n\t"
409 "mul.s %[val28], %[f7], %[fTmp2] \n\t"
410
411 : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
412 [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
413 [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
414 [val3] "=&f" (val3), [val4] "=&f" (val4),
415 [val11] "=&f" (val11), [val12] "=&f" (val12),
416 [val19] "=&f" (val19), [val20] "=&f" (val20),
417 [val27] "=&f" (val27), [val28] "=&f" (val28)
418 : [tab] "r" (tab), [f1]"f"(f1), [f2]"f"(f2), [f3]"f"(f3),
419 [f4]"f"(f4), [f5]"f"(f5), [f6]"f"(f6), [f7]"f"(f7)
420 : "memory"
421 );
422
423 f1 = 0.54119610014619698439;
424 __asm__ volatile (
425 "sub.s %[fTmp2], %[val0], %[val3] \n\t"
426 "add.s %[val0], %[val0], %[val3] \n\t"
427 "sub.s %[fTmp3], %[val7], %[val4] \n\t"
428 "add.s %[val4], %[val7], %[val4] \n\t"
429 "sub.s %[fTmp4], %[val8], %[val11] \n\t"
430 "mul.s %[val3], %[f1], %[fTmp2] \n\t"
431 "add.s %[val8], %[val8], %[val11] \n\t"
432 "mul.s %[val7], %[f1], %[fTmp3] \n\t"
433 "sub.s %[fTmp2], %[val15], %[val12] \n\t"
434 "mul.s %[val11], %[f1], %[fTmp4] \n\t"
435 "add.s %[val12], %[val15], %[val12] \n\t"
436 "mul.s %[val15], %[f1], %[fTmp2] \n\t"
437
438 : [val0] "+&f" (val0), [val3] "+&f" (val3),
439 [val4] "+&f" (val4), [val7] "+&f" (val7),
440 [val8] "+&f" (val8), [val11] "+&f" (val11),
441 [val12] "+&f" (val12), [val15] "+&f" (val15),
442 [fTmp2] "=&f" (fTmp2),
443 [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4)
444 : [f1] "f" (f1)
445 );
446
447 __asm__ volatile (
448 "sub.s %[fTmp2], %[val16], %[val19] \n\t"
449 "add.s %[val16], %[val16], %[val19] \n\t"
450 "sub.s %[fTmp3], %[val23], %[val20] \n\t"
451 "add.s %[val20], %[val23], %[val20] \n\t"
452 "sub.s %[fTmp4], %[val24], %[val27] \n\t"
453 "mul.s %[val19], %[f1], %[fTmp2] \n\t"
454 "add.s %[val24], %[val24], %[val27] \n\t"
455 "mul.s %[val23], %[f1], %[fTmp3] \n\t"
456 "sub.s %[fTmp2], %[val31], %[val28] \n\t"
457 "mul.s %[val27], %[f1], %[fTmp4] \n\t"
458 "add.s %[val28], %[val31], %[val28] \n\t"
459 "mul.s %[val31], %[f1], %[fTmp2] \n\t"
460
461 : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
462 [val16] "+&f" (val16), [val19] "+&f" (val19), [val20] "+&f" (val20),
463 [val23] "+&f" (val23), [val24] "+&f" (val24), [val27] "+&f" (val27),
464 [val28] "+&f" (val28), [val31] "+&f" (val31)
465 : [f1] "f" (f1)
466 );
467
468 f1 = 0.52249861493968888062;
469 f2 = 0.50547095989754365998;
470 f3 = 3.40760841846871878570;
471 f4 = 1.72244709823833392782;
472 f5 = 0.62250412303566481615;
473 f6 = 0.83934964541552703873;
474 f7 = 0.60134488693504528054;
475 __asm__ volatile (
476 "lwc1 %[fTmp1], 1*4(%[tab]) \n\t"
477 "lwc1 %[fTmp2], 30*4(%[tab]) \n\t"
478 "lwc1 %[fTmp3], 14*4(%[tab]) \n\t"
479 "lwc1 %[fTmp4], 17*4(%[tab]) \n\t"
480 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
481 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
482 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
483 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
484 "mul.s %[fTmp8], %[fTmp8], %[f2] \n\t"
485 "add.s %[val1], %[fTmp5], %[fTmp6] \n\t"
486 "sub.s %[val14], %[fTmp5], %[fTmp6] \n\t"
487 "lwc1 %[fTmp1], 6*4(%[tab]) \n\t"
488 "lwc1 %[fTmp2], 25*4(%[tab]) \n\t"
489 "madd.s %[val17], %[fTmp8], %[fTmp9], %[f3] \n\t"
490 "nmsub.s %[val30], %[fTmp8], %[fTmp9], %[f3] \n\t"
491 "mul.s %[val14], %[val14], %[f1] \n\t"
492 "lwc1 %[fTmp3], 9*4(%[tab]) \n\t"
493 "lwc1 %[fTmp4], 22*4(%[tab]) \n\t"
494 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
495 "mul.s %[val30], %[val30], %[f1] \n\t"
496 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
497 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
498 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
499 "add.s %[val6], %[fTmp5], %[fTmp6] \n\t"
500 "sub.s %[val9], %[fTmp5], %[fTmp6] \n\t"
501 "mul.s %[fTmp8], %[fTmp8], %[f5] \n\t"
502 "sub.s %[fTmp2], %[val1], %[val6] \n\t"
503 "add.s %[val1], %[val1], %[val6] \n\t"
504 "mul.s %[val9], %[val9], %[f4] \n\t"
505 "madd.s %[val22], %[fTmp8], %[fTmp9], %[f6] \n\t"
506 "nmsub.s %[val25], %[fTmp8], %[fTmp9], %[f6] \n\t"
507 "mul.s %[val6], %[f7], %[fTmp2] \n\t"
508 "sub.s %[fTmp2], %[val14], %[val9] \n\t"
509 "add.s %[val9], %[val14], %[val9] \n\t"
510 "mul.s %[val25], %[val25], %[f4] \n\t"
511 "sub.s %[fTmp3], %[val17], %[val22] \n\t"
512 "add.s %[val17], %[val17], %[val22] \n\t"
513 "mul.s %[val14], %[f7], %[fTmp2] \n\t"
514 "sub.s %[fTmp2], %[val30], %[val25] \n\t"
515 "mul.s %[val22], %[f7], %[fTmp3] \n\t"
516 "add.s %[val25], %[val30], %[val25] \n\t"
517 "mul.s %[val30], %[f7], %[fTmp2] \n\t"
518
519 : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
520 [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
521 [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
522 [val1] "=&f" (val1), [val6] "=&f" (val6),
523 [val9] "=&f" (val9), [val14] "=&f" (val14),
524 [val17] "=&f" (val17), [val22] "=&f" (val22),
525 [val25] "=&f" (val25), [val30] "=&f" (val30)
526 : [tab] "r" (tab), [f1]"f"(f1), [f2]"f"(f2), [f3]"f"(f3),
527 [f4]"f"(f4), [f5]"f"(f5), [f6]"f"(f6), [f7]"f"(f7)
528 : "memory"
529 );
530
531 f1 = 0.56694403481635770368;
532 f2 = 0.51544730992262454697;
533 f3 = 2.05778100995341155085;
534 f4 = 1.06067768599034747134;
535 f5 = 0.58293496820613387367;
536 f6 = 0.97256823786196069369;
537 f7 = 0.89997622313641570463;
538 __asm__ volatile (
539 "lwc1 %[fTmp1], 2*4(%[tab]) \n\t"
540 "lwc1 %[fTmp2], 29*4(%[tab]) \n\t"
541 "lwc1 %[fTmp3], 13*4(%[tab]) \n\t"
542 "lwc1 %[fTmp4], 18*4(%[tab]) \n\t"
543 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
544 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
545 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
546 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
547 "mul.s %[fTmp8], %[fTmp8], %[f2] \n\t"
548 "add.s %[val2], %[fTmp5], %[fTmp6] \n\t"
549 "sub.s %[val13], %[fTmp5], %[fTmp6] \n\t"
550 "lwc1 %[fTmp1], 5*4(%[tab]) \n\t"
551 "lwc1 %[fTmp2], 26*4(%[tab]) \n\t"
552 "madd.s %[val18], %[fTmp8], %[fTmp9], %[f3] \n\t"
553 "nmsub.s %[val29], %[fTmp8], %[fTmp9], %[f3] \n\t"
554 "mul.s %[val13], %[val13], %[f1] \n\t"
555 "lwc1 %[fTmp3], 10*4(%[tab]) \n\t"
556 "lwc1 %[fTmp4], 21*4(%[tab]) \n\t"
557 "mul.s %[val29], %[val29], %[f1] \n\t"
558 "add.s %[fTmp5], %[fTmp1], %[fTmp2] \n\t"
559 "sub.s %[fTmp8], %[fTmp1], %[fTmp2] \n\t"
560 "add.s %[fTmp6], %[fTmp3], %[fTmp4] \n\t"
561 "sub.s %[fTmp9], %[fTmp3], %[fTmp4] \n\t"
562 "add.s %[val5], %[fTmp5], %[fTmp6] \n\t"
563 "sub.s %[val10], %[fTmp5], %[fTmp6] \n\t"
564 "mul.s %[fTmp8], %[fTmp8], %[f5] \n\t"
565 "sub.s %[fTmp2], %[val2], %[val5] \n\t"
566 "mul.s %[val10], %[val10], %[f4] \n\t"
567 "madd.s %[val21], %[fTmp8], %[fTmp9], %[f6] \n\t"
568 "nmsub.s %[val26], %[fTmp8], %[fTmp9], %[f6] \n\t"
569 "add.s %[val2], %[val2], %[val5] \n\t"
570 "mul.s %[val5], %[f7], %[fTmp2] \n\t"
571 "sub.s %[fTmp3], %[val13], %[val10] \n\t"
572 "add.s %[val10], %[val13], %[val10] \n\t"
573 "mul.s %[val26], %[val26], %[f4] \n\t"
574 "sub.s %[fTmp4], %[val18], %[val21] \n\t"
575 "add.s %[val18], %[val18], %[val21] \n\t"
576 "mul.s %[val13], %[f7], %[fTmp3] \n\t"
577 "sub.s %[fTmp2], %[val29], %[val26] \n\t"
578 "add.s %[val26], %[val29], %[val26] \n\t"
579 "mul.s %[val21], %[f7], %[fTmp4] \n\t"
580 "mul.s %[val29], %[f7], %[fTmp2] \n\t"
581
582 : [fTmp1] "=&f" (fTmp1), [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
583 [fTmp4] "=&f" (fTmp4), [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
584 [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
585 [val2] "=&f" (val2), [val5] "=&f" (val5),
586 [val10] "=&f" (val10), [val13] "=&f" (val13),
587 [val18] "=&f" (val18), [val21] "=&f" (val21),
588 [val26] "=&f" (val26), [val29] "=&f" (val29)
589 : [tab] "r" (tab), [f1]"f"(f1), [f2]"f"(f2), [f3]"f"(f3),
590 [f4]"f"(f4), [f5]"f"(f5), [f6]"f"(f6), [f7]"f"(f7)
591 : "memory"
592 );
593
594 f1 = 1.30656296487637652785;
595 __asm__ volatile (
596 "sub.s %[fTmp2], %[val1], %[val2] \n\t"
597 "add.s %[val1], %[val1], %[val2] \n\t"
598 "sub.s %[fTmp3], %[val6], %[val5] \n\t"
599 "add.s %[val5], %[val6], %[val5] \n\t"
600 "sub.s %[fTmp4], %[val9], %[val10] \n\t"
601 "mul.s %[val2], %[f1], %[fTmp2] \n\t"
602 "add.s %[val9], %[val9], %[val10] \n\t"
603 "mul.s %[val6], %[f1], %[fTmp3] \n\t"
604 "sub.s %[fTmp2], %[val14], %[val13] \n\t"
605 "mul.s %[val10], %[f1], %[fTmp4] \n\t"
606 "add.s %[val13], %[val14], %[val13] \n\t"
607 "mul.s %[val14], %[f1], %[fTmp2] \n\t"
608
609 : [fTmp2] "=&f" (fTmp2),
610 [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
611 [val1] "+&f" (val1), [val2] "+&f" (val2),
612 [val5] "+&f" (val5), [val6] "+&f" (val6),
613 [val9] "+&f" (val9), [val10] "+&f" (val10),
614 [val13] "+&f" (val13), [val14] "+&f" (val14)
615 : [f1]"f"(f1)
616 );
617
618 __asm__ volatile (
619 "sub.s %[fTmp2], %[val17], %[val18] \n\t"
620 "add.s %[val17], %[val17], %[val18] \n\t"
621 "sub.s %[fTmp3], %[val22], %[val21] \n\t"
622 "add.s %[val21], %[val22], %[val21] \n\t"
623 "sub.s %[fTmp4], %[val25], %[val26] \n\t"
624 "mul.s %[val18], %[f1], %[fTmp2] \n\t"
625 "add.s %[val25], %[val25], %[val26] \n\t"
626 "mul.s %[val22], %[f1], %[fTmp3] \n\t"
627 "sub.s %[fTmp2], %[val30], %[val29] \n\t"
628 "mul.s %[val26], %[f1], %[fTmp4] \n\t"
629 "add.s %[val29], %[val30], %[val29] \n\t"
630 "mul.s %[val30], %[f1], %[fTmp2] \n\t"
631
632 : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
633 [val17] "+&f" (val17), [val18] "+&f" (val18), [val21] "+&f" (val21),
634 [val22] "+&f" (val22), [val25] "+&f" (val25), [val26] "+&f" (val26),
635 [val29] "+&f" (val29), [val30] "+&f" (val30)
636 : [f1] "f" (f1)
637 );
638
639 f1 = 0.70710678118654752439;
640 __asm__ volatile (
641 "sub.s %[fTmp2], %[val0], %[val1] \n\t"
642 "add.s %[val0], %[val0], %[val1] \n\t"
643 "sub.s %[fTmp3], %[val3], %[val2] \n\t"
644 "add.s %[val2], %[val3], %[val2] \n\t"
645 "sub.s %[fTmp4], %[val4], %[val5] \n\t"
646 "mul.s %[val1], %[f1], %[fTmp2] \n\t"
647 "swc1 %[val0], 0(%[out]) \n\t"
648 "mul.s %[val3], %[fTmp3], %[f1] \n\t"
649 "add.s %[val4], %[val4], %[val5] \n\t"
650 "mul.s %[val5], %[f1], %[fTmp4] \n\t"
651 "swc1 %[val1], 16*4(%[out]) \n\t"
652 "sub.s %[fTmp2], %[val7], %[val6] \n\t"
653 "add.s %[val2], %[val2], %[val3] \n\t"
654 "swc1 %[val3], 24*4(%[out]) \n\t"
655 "add.s %[val6], %[val7], %[val6] \n\t"
656 "mul.s %[val7], %[f1], %[fTmp2] \n\t"
657 "swc1 %[val2], 8*4(%[out]) \n\t"
658 "add.s %[val6], %[val6], %[val7] \n\t"
659 "swc1 %[val7], 28*4(%[out]) \n\t"
660 "add.s %[val4], %[val4], %[val6] \n\t"
661 "add.s %[val6], %[val6], %[val5] \n\t"
662 "add.s %[val5], %[val5], %[val7] \n\t"
663 "swc1 %[val4], 4*4(%[out]) \n\t"
664 "swc1 %[val5], 20*4(%[out]) \n\t"
665 "swc1 %[val6], 12*4(%[out]) \n\t"
666
667 : [fTmp2] "=&f" (fTmp2),
668 [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
669 [val0] "+&f" (val0), [val1] "+&f" (val1),
670 [val2] "+&f" (val2), [val3] "+&f" (val3),
671 [val4] "+&f" (val4), [val5] "+&f" (val5),
672 [val6] "+&f" (val6), [val7] "+&f" (val7)
673 : [out] "r" (out), [f1]"f"(f1)
674 );
675
676 __asm__ volatile (
677 "sub.s %[fTmp2], %[val8], %[val9] \n\t"
678 "add.s %[val8], %[val8], %[val9] \n\t"
679 "sub.s %[fTmp3], %[val11], %[val10] \n\t"
680 "add.s %[val10], %[val11], %[val10] \n\t"
681 "sub.s %[fTmp4], %[val12], %[val13] \n\t"
682 "mul.s %[val9], %[f1], %[fTmp2] \n\t"
683 "add.s %[val12], %[val12], %[val13] \n\t"
684 "mul.s %[val11], %[f1], %[fTmp3] \n\t"
685 "sub.s %[fTmp2], %[val15], %[val14] \n\t"
686 "mul.s %[val13], %[f1], %[fTmp4] \n\t"
687 "add.s %[val14], %[val15], %[val14] \n\t"
688 "add.s %[val10], %[val10], %[val11] \n\t"
689 "mul.s %[val15], %[f1], %[fTmp2] \n\t"
690 "add.s %[val14], %[val14], %[val15] \n\t"
691 "add.s %[val12], %[val12], %[val14] \n\t"
692 "add.s %[val14], %[val14], %[val13] \n\t"
693 "add.s %[val13], %[val13], %[val15] \n\t"
694 "add.s %[val8], %[val8], %[val12] \n\t"
695 "add.s %[val12], %[val12], %[val10] \n\t"
696 "add.s %[val10], %[val10], %[val14] \n\t"
697 "add.s %[val14], %[val14], %[val9] \n\t"
698 "add.s %[val9], %[val9], %[val13] \n\t"
699 "add.s %[val13], %[val13], %[val11] \n\t"
700 "add.s %[val11], %[val11], %[val15] \n\t"
701 "swc1 %[val8], 2*4(%[out]) \n\t"
702 "swc1 %[val9], 18*4(%[out]) \n\t"
703 "swc1 %[val10], 10*4(%[out]) \n\t"
704 "swc1 %[val11], 26*4(%[out]) \n\t"
705 "swc1 %[val12], 6*4(%[out]) \n\t"
706 "swc1 %[val13], 22*4(%[out]) \n\t"
707 "swc1 %[val14], 14*4(%[out]) \n\t"
708 "swc1 %[val15], 30*4(%[out]) \n\t"
709
710 : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
711 [val8] "+&f" (val8), [val9] "+&f" (val9), [val10] "+&f" (val10),
712 [val11] "+&f" (val11), [val12] "+&f" (val12), [val13] "+&f" (val13),
713 [val14] "+&f" (val14), [val15] "+&f" (val15)
714 : [f1] "f" (f1), [out] "r" (out)
715 );
716
717 __asm__ volatile (
718 "sub.s %[fTmp2], %[val16], %[val17] \n\t"
719 "add.s %[val16], %[val16], %[val17] \n\t"
720 "sub.s %[fTmp3], %[val19], %[val18] \n\t"
721 "add.s %[val18], %[val19], %[val18] \n\t"
722 "sub.s %[fTmp4], %[val20], %[val21] \n\t"
723 "mul.s %[val17], %[f1], %[fTmp2] \n\t"
724 "add.s %[val20], %[val20], %[val21] \n\t"
725 "mul.s %[val19], %[f1], %[fTmp3] \n\t"
726 "sub.s %[fTmp2], %[val23], %[val22] \n\t"
727 "mul.s %[val21], %[f1], %[fTmp4] \n\t"
728 "add.s %[val22], %[val23], %[val22] \n\t"
729 "add.s %[val18], %[val18], %[val19] \n\t"
730 "mul.s %[val23], %[f1], %[fTmp2] \n\t"
731 "add.s %[val22], %[val22], %[val23] \n\t"
732 "add.s %[val20], %[val20], %[val22] \n\t"
733 "add.s %[val22], %[val22], %[val21] \n\t"
734 "add.s %[val21], %[val21], %[val23] \n\t"
735
736 : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
737 [val16] "+&f" (val16), [val17] "+&f" (val17), [val18] "+&f" (val18),
738 [val19] "+&f" (val19), [val20] "+&f" (val20), [val21] "+&f" (val21),
739 [val22] "+&f" (val22), [val23] "+&f" (val23)
740 : [f1] "f" (f1)
741 );
742
743 __asm__ volatile (
744 "sub.s %[fTmp2], %[val24], %[val25] \n\t"
745 "add.s %[val24], %[val24], %[val25] \n\t"
746 "sub.s %[fTmp3], %[val27], %[val26] \n\t"
747 "add.s %[val26], %[val27], %[val26] \n\t"
748 "sub.s %[fTmp4], %[val28], %[val29] \n\t"
749 "mul.s %[val25], %[f1], %[fTmp2] \n\t"
750 "add.s %[val28], %[val28], %[val29] \n\t"
751 "mul.s %[val27], %[f1], %[fTmp3] \n\t"
752 "sub.s %[fTmp2], %[val31], %[val30] \n\t"
753 "mul.s %[val29], %[f1], %[fTmp4] \n\t"
754 "add.s %[val30], %[val31], %[val30] \n\t"
755 "add.s %[val26], %[val26], %[val27] \n\t"
756 "mul.s %[val31], %[f1], %[fTmp2] \n\t"
757 "add.s %[val30], %[val30], %[val31] \n\t"
758 "add.s %[val28], %[val28], %[val30] \n\t"
759 "add.s %[val30], %[val30], %[val29] \n\t"
760 "add.s %[val29], %[val29], %[val31] \n\t"
761 "add.s %[val24], %[val24], %[val28] \n\t"
762 "add.s %[val28], %[val28], %[val26] \n\t"
763 "add.s %[val26], %[val26], %[val30] \n\t"
764 "add.s %[val30], %[val30], %[val25] \n\t"
765 "add.s %[val25], %[val25], %[val29] \n\t"
766 "add.s %[val29], %[val29], %[val27] \n\t"
767 "add.s %[val27], %[val27], %[val31] \n\t"
768
769 : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
770 [val24] "+&f" (val24), [val25] "+&f" (val25), [val26] "+&f" (val26),
771 [val27] "+&f" (val27), [val28] "+&f" (val28), [val29] "+&f" (val29),
772 [val30] "+&f" (val30), [val31] "+&f" (val31)
773 : [f1] "f" (f1)
774 );
775
776 out[ 1] = val16 + val24;
777 out[17] = val17 + val25;
778 out[ 9] = val18 + val26;
779 out[25] = val19 + val27;
780 out[ 5] = val20 + val28;
781 out[21] = val21 + val29;
782 out[13] = val22 + val30;
783 out[29] = val23 + val31;
784 out[ 3] = val24 + val20;
785 out[19] = val25 + val21;
786 out[11] = val26 + val22;
787 out[27] = val27 + val23;
788 out[ 7] = val28 + val18;
789 out[23] = val29 + val19;
790 out[15] = val30 + val17;
791 out[31] = val31;
792 }
793
imdct36_mips_float(float * out,float * buf,float * in,float * win)794 static void imdct36_mips_float(float *out, float *buf, float *in, float *win)
795 {
796 float t0, t1, t2, t3, s0, s1, s2, s3;
797 float tmp[18];
798 /* temporary variables */
799 float in1, in2, in3, in4, in5, in6;
800 float out1, out2, out3, out4, out5;
801 float f1, f2, f3, f4, f5, f6, f7, f8, f9;
802
803 /**
804 * all loops are unrolled totally, and instructions are scheduled to
805 * minimize pipeline stall. instructions of the first two loops are
806 * reorganized, in order to eliminate unnecessary readings and
807 * writings into array. values defined in macros and tables are
808 * eliminated - they are directly loaded in appropriate variables
809 */
810
811 /* loop 1 and 2 */
812 __asm__ volatile (
813 "lwc1 %[in1], 17*4(%[in]) \t\n"
814 "lwc1 %[in2], 16*4(%[in]) \t\n"
815 "lwc1 %[in3], 15*4(%[in]) \t\n"
816 "lwc1 %[in4], 14*4(%[in]) \t\n"
817 "lwc1 %[in5], 13*4(%[in]) \t\n"
818 "lwc1 %[in6], 12*4(%[in]) \t\n"
819 "add.s %[out1], %[in1], %[in2] \t\n"
820 "add.s %[out2], %[in2], %[in3] \t\n"
821 "add.s %[out3], %[in3], %[in4] \t\n"
822 "add.s %[out4], %[in4], %[in5] \t\n"
823 "add.s %[out5], %[in5], %[in6] \t\n"
824 "lwc1 %[in1], 11*4(%[in]) \t\n"
825 "swc1 %[out2], 16*4(%[in]) \t\n"
826 "add.s %[out1], %[out1], %[out3] \t\n"
827 "swc1 %[out4], 14*4(%[in]) \t\n"
828 "add.s %[out3], %[out3], %[out5] \t\n"
829 "lwc1 %[in2], 10*4(%[in]) \t\n"
830 "lwc1 %[in3], 9*4(%[in]) \t\n"
831 "swc1 %[out1], 17*4(%[in]) \t\n"
832 "lwc1 %[in4], 8*4(%[in]) \t\n"
833 "swc1 %[out3], 15*4(%[in]) \t\n"
834 "add.s %[out1], %[in6], %[in1] \t\n"
835 "add.s %[out2], %[in1], %[in2] \t\n"
836 "add.s %[out3], %[in2], %[in3] \t\n"
837 "add.s %[out4], %[in3], %[in4] \t\n"
838 "lwc1 %[in5], 7*4(%[in]) \t\n"
839 "swc1 %[out1], 12*4(%[in]) \t\n"
840 "add.s %[out5], %[out5], %[out2] \t\n"
841 "swc1 %[out3], 10*4(%[in]) \t\n"
842 "add.s %[out2], %[out2], %[out4] \t\n"
843 "lwc1 %[in6], 6*4(%[in]) \t\n"
844 "lwc1 %[in1], 5*4(%[in]) \t\n"
845 "swc1 %[out5], 13*4(%[in]) \t\n"
846 "lwc1 %[in2], 4*4(%[in]) \t\n"
847 "swc1 %[out2], 11*4(%[in]) \t\n"
848 "add.s %[out5], %[in4], %[in5] \t\n"
849 "add.s %[out1], %[in5], %[in6] \t\n"
850 "add.s %[out2], %[in6], %[in1] \t\n"
851 "add.s %[out3], %[in1], %[in2] \t\n"
852 "lwc1 %[in3], 3*4(%[in]) \t\n"
853 "swc1 %[out5], 8*4(%[in]) \t\n"
854 "add.s %[out4], %[out4], %[out1] \t\n"
855 "swc1 %[out2], 6*4(%[in]) \t\n"
856 "add.s %[out1], %[out1], %[out3] \t\n"
857 "lwc1 %[in4], 2*4(%[in]) \t\n"
858 "lwc1 %[in5], 1*4(%[in]) \t\n"
859 "swc1 %[out4], 9*4(%[in]) \t\n"
860 "lwc1 %[in6], 0(%[in]) \t\n"
861 "swc1 %[out1], 7*4(%[in]) \t\n"
862 "add.s %[out4], %[in2], %[in3] \t\n"
863 "add.s %[out5], %[in3], %[in4] \t\n"
864 "add.s %[out1], %[in4], %[in5] \t\n"
865 "add.s %[out2], %[in5], %[in6] \t\n"
866 "swc1 %[out4], 4*4(%[in]) \t\n"
867 "add.s %[out3], %[out3], %[out5] \t\n"
868 "swc1 %[out1], 2*4(%[in]) \t\n"
869 "add.s %[out5], %[out5], %[out2] \t\n"
870 "swc1 %[out2], 1*4(%[in]) \t\n"
871 "swc1 %[out3], 5*4(%[in]) \t\n"
872 "swc1 %[out5], 3*4(%[in]) \t\n"
873
874 : [in1] "=&f" (in1), [in2] "=&f" (in2),
875 [in3] "=&f" (in3), [in4] "=&f" (in4),
876 [in5] "=&f" (in5), [in6] "=&f" (in6),
877 [out1] "=&f" (out1), [out2] "=&f" (out2),
878 [out3] "=&f" (out3), [out4] "=&f" (out4),
879 [out5] "=&f" (out5)
880 : [in] "r" (in)
881 : "memory"
882 );
883
884 /* loop 3 */
885 f1 = 0.5;
886 f2 = 0.93969262078590838405;
887 f3 = -0.76604444311897803520;
888 f4 = -0.17364817766693034885;
889 f5 = -0.86602540378443864676;
890 f6 = 0.98480775301220805936;
891 f7 = -0.34202014332566873304;
892 f8 = 0.86602540378443864676;
893 f9 = -0.64278760968653932632;
894 __asm__ volatile (
895 "lwc1 %[in1], 8*4(%[in]) \t\n"
896 "lwc1 %[in2], 16*4(%[in]) \t\n"
897 "lwc1 %[in3], 4*4(%[in]) \t\n"
898 "lwc1 %[in4], 0(%[in]) \t\n"
899 "lwc1 %[in5], 12*4(%[in]) \t\n"
900 "add.s %[t2], %[in1], %[in2] \t\n"
901 "add.s %[t0], %[in1], %[in3] \t\n"
902 "madd.s %[t3], %[in4], %[in5], %[f1] \t\n"
903 "sub.s %[t1], %[in4], %[in5] \t\n"
904 "sub.s %[t2], %[t2], %[in3] \t\n"
905 "mul.s %[t0], %[t0], %[f2] \t\n"
906 "nmsub.s %[out1], %[t1], %[t2], %[f1] \t\n"
907 "add.s %[out2], %[t1], %[t2] \t\n"
908 "add.s %[t2], %[in2], %[in3] \t\n"
909 "sub.s %[t1], %[in1], %[in2] \t\n"
910 "sub.s %[out3], %[t3], %[t0] \t\n"
911 "swc1 %[out1], 6*4(%[tmp]) \t\n"
912 "swc1 %[out2], 16*4(%[tmp]) \t\n"
913 "mul.s %[t2], %[t2], %[f3] \t\n"
914 "mul.s %[t1], %[t1], %[f4] \t\n"
915 "add.s %[out1], %[t3], %[t0] \t\n"
916 "lwc1 %[in1], 10*4(%[in]) \t\n"
917 "lwc1 %[in2], 14*4(%[in]) \t\n"
918 "sub.s %[out3], %[out3], %[t2] \t\n"
919 "add.s %[out2], %[t3], %[t2] \t\n"
920 "add.s %[out1], %[out1], %[t1] \t\n"
921 "lwc1 %[in3], 2*4(%[in]) \t\n"
922 "lwc1 %[in4], 6*4(%[in]) \t\n"
923 "swc1 %[out3], 10*4(%[tmp]) \t\n"
924 "sub.s %[out2], %[out2], %[t1] \t\n"
925 "swc1 %[out1], 2*4(%[tmp]) \t\n"
926 "add.s %[out1], %[in1], %[in2] \t\n"
927 "add.s %[t2], %[in1], %[in3] \t\n"
928 "sub.s %[t3], %[in1], %[in2] \t\n"
929 "swc1 %[out2], 14*4(%[tmp]) \t\n"
930 "sub.s %[out1], %[out1], %[in3] \t\n"
931 "mul.s %[t2], %[t2], %[f6] \t\n"
932 "mul.s %[t3], %[t3], %[f7] \t\n"
933 "mul.s %[t0], %[in4], %[f8] \t\n"
934 "mul.s %[out1], %[out1], %[f5] \t\n"
935 "add.s %[t1], %[in2], %[in3] \t\n"
936 "add.s %[out2], %[t2], %[t3] \t\n"
937 "lwc1 %[in1], 9*4(%[in]) \t\n"
938 "swc1 %[out1], 4*4(%[tmp]) \t\n"
939 "mul.s %[t1], %[t1], %[f9] \t\n"
940 "lwc1 %[in2], 17*4(%[in]) \t\n"
941 "add.s %[out2], %[out2], %[t0] \t\n"
942 "lwc1 %[in3], 5*4(%[in]) \t\n"
943 "lwc1 %[in4], 1*4(%[in]) \t\n"
944 "add.s %[out3], %[t2], %[t1] \t\n"
945 "sub.s %[out1], %[t3], %[t1] \t\n"
946 "swc1 %[out2], 0(%[tmp]) \t\n"
947 "lwc1 %[in5], 13*4(%[in]) \t\n"
948 "add.s %[t2], %[in1], %[in2] \t\n"
949 "sub.s %[out3], %[out3], %[t0] \t\n"
950 "sub.s %[out1], %[out1], %[t0] \t\n"
951 "add.s %[t0], %[in1], %[in3] \t\n"
952 "madd.s %[t3], %[in4], %[in5], %[f1] \t\n"
953 "sub.s %[t2], %[t2], %[in3] \t\n"
954 "swc1 %[out3], 12*4(%[tmp]) \t\n"
955 "swc1 %[out1], 8*4(%[tmp]) \t\n"
956 "sub.s %[t1], %[in4], %[in5] \t\n"
957 "mul.s %[t0], %[t0], %[f2] \t\n"
958 "nmsub.s %[out1], %[t1], %[t2], %[f1] \t\n"
959 "add.s %[out2], %[t1], %[t2] \t\n"
960 "add.s %[t2], %[in2], %[in3] \t\n"
961 "sub.s %[t1], %[in1], %[in2] \t\n"
962 "sub.s %[out3], %[t3], %[t0] \t\n"
963 "swc1 %[out1], 7*4(%[tmp]) \t\n"
964 "swc1 %[out2], 17*4(%[tmp]) \t\n"
965 "mul.s %[t2], %[t2], %[f3] \t\n"
966 "mul.s %[t1], %[t1], %[f4] \t\n"
967 "add.s %[out1], %[t3], %[t0] \t\n"
968 "lwc1 %[in1], 11*4(%[in]) \t\n"
969 "lwc1 %[in2], 15*4(%[in]) \t\n"
970 "sub.s %[out3], %[out3], %[t2] \t\n"
971 "add.s %[out2], %[t3], %[t2] \t\n"
972 "add.s %[out1], %[out1], %[t1] \t\n"
973 "lwc1 %[in3], 3*4(%[in]) \t\n"
974 "lwc1 %[in4], 7*4(%[in]) \t\n"
975 "swc1 %[out3], 11*4(%[tmp]) \t\n"
976 "sub.s %[out2], %[out2], %[t1] \t\n"
977 "swc1 %[out1], 3*4(%[tmp]) \t\n"
978 "add.s %[out3], %[in1], %[in2] \t\n"
979 "add.s %[t2], %[in1], %[in3] \t\n"
980 "sub.s %[t3], %[in1], %[in2] \t\n"
981 "swc1 %[out2], 15*4(%[tmp]) \t\n"
982 "mul.s %[t0], %[in4], %[f8] \t\n"
983 "sub.s %[out3], %[out3], %[in3] \t\n"
984 "mul.s %[t2], %[t2], %[f6] \t\n"
985 "mul.s %[t3], %[t3], %[f7] \t\n"
986 "add.s %[t1], %[in2], %[in3] \t\n"
987 "mul.s %[out3], %[out3], %[f5] \t\n"
988 "add.s %[out1], %[t2], %[t3] \t\n"
989 "mul.s %[t1], %[t1], %[f9] \t\n"
990 "swc1 %[out3], 5*4(%[tmp]) \t\n"
991 "add.s %[out1], %[out1], %[t0] \t\n"
992 "add.s %[out2], %[t2], %[t1] \t\n"
993 "sub.s %[out3], %[t3], %[t1] \t\n"
994 "swc1 %[out1], 1*4(%[tmp]) \t\n"
995 "sub.s %[out2], %[out2], %[t0] \t\n"
996 "sub.s %[out3], %[out3], %[t0] \t\n"
997 "swc1 %[out2], 13*4(%[tmp]) \t\n"
998 "swc1 %[out3], 9*4(%[tmp]) \t\n"
999
1000 : [t0] "=&f" (t0), [t1] "=&f" (t1),
1001 [t2] "=&f" (t2), [t3] "=&f" (t3),
1002 [in1] "=&f" (in1), [in2] "=&f" (in2),
1003 [in3] "=&f" (in3), [in4] "=&f" (in4),
1004 [in5] "=&f" (in5), [out1] "=&f" (out1),
1005 [out2] "=&f" (out2), [out3] "=&f" (out3)
1006 : [in] "r" (in), [tmp] "r" (tmp), [f1]"f"(f1), [f2]"f"(f2),
1007 [f3]"f"(f3), [f4]"f"(f4), [f5]"f"(f5), [f6]"f"(f6),
1008 [f7]"f"(f7), [f8]"f"(f8), [f9]"f"(f9)
1009 : "memory"
1010 );
1011
1012 /* loop 4 */
1013 f1 = 0.50190991877167369479;
1014 f2 = 5.73685662283492756461;
1015 f3 = 0.51763809020504152469;
1016 f4 = 1.93185165257813657349;
1017 f5 = 0.55168895948124587824;
1018 f6 = 1.18310079157624925896;
1019 f7 = 0.61038729438072803416;
1020 f8 = 0.87172339781054900991;
1021 f9 = 0.70710678118654752439;
1022 __asm__ volatile (
1023 "lwc1 %[in1], 2*4(%[tmp]) \t\n"
1024 "lwc1 %[in2], 0(%[tmp]) \t\n"
1025 "lwc1 %[in3], 3*4(%[tmp]) \t\n"
1026 "lwc1 %[in4], 1*4(%[tmp]) \t\n"
1027 "add.s %[s0], %[in1], %[in2] \t\n"
1028 "sub.s %[s2], %[in1], %[in2] \t\n"
1029 "add.s %[s1], %[in3], %[in4] \t\n"
1030 "sub.s %[s3], %[in3], %[in4] \t\n"
1031 "lwc1 %[in1], 9*4(%[win]) \t\n"
1032 "lwc1 %[in2], 4*9*4(%[buf]) \t\n"
1033 "lwc1 %[in3], 8*4(%[win]) \t\n"
1034 "mul.s %[s1], %[s1], %[f1] \t\n"
1035 "mul.s %[s3], %[s3], %[f2] \t\n"
1036 "lwc1 %[in4], 4*8*4(%[buf]) \t\n"
1037 "lwc1 %[in5], 29*4(%[win]) \t\n"
1038 "lwc1 %[in6], 28*4(%[win]) \t\n"
1039 "add.s %[t0], %[s0], %[s1] \t\n"
1040 "sub.s %[t1], %[s0], %[s1] \t\n"
1041 "mul.s %[out3], %[in5], %[t0] \t\n"
1042 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1043 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1044 "mul.s %[out4], %[in6], %[t0] \t\n"
1045 "add.s %[t0], %[s2], %[s3] \t\n"
1046 "swc1 %[out3], 4*9*4(%[buf]) \t\n"
1047 "swc1 %[out1], 288*4(%[out]) \t\n"
1048 "swc1 %[out2], 256*4(%[out]) \t\n"
1049 "swc1 %[out4], 4*8*4(%[buf]) \t\n"
1050 "sub.s %[t1], %[s2], %[s3] \t\n"
1051 "lwc1 %[in1], 17*4(%[win]) \t\n"
1052 "lwc1 %[in2], 4*17*4(%[buf]) \t\n"
1053 "lwc1 %[in3], 0(%[win]) \t\n"
1054 "lwc1 %[in4], 0(%[buf]) \t\n"
1055 "lwc1 %[in5], 37*4(%[win]) \t\n"
1056 "lwc1 %[in6], 20*4(%[win]) \t\n"
1057 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1058 "lwc1 %[in1], 6*4(%[tmp]) \t\n"
1059 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1060 "mul.s %[out3], %[t0], %[in5] \t\n"
1061 "mul.s %[out4], %[t0], %[in6] \t\n"
1062 "swc1 %[out1], 544*4(%[out]) \t\n"
1063 "lwc1 %[in2], 4*4(%[tmp]) \t\n"
1064 "swc1 %[out2], 0(%[out]) \t\n"
1065 "swc1 %[out3], 4*17*4(%[buf]) \t\n"
1066 "swc1 %[out4], 0(%[buf]) \t\n"
1067 "lwc1 %[in3], 7*4(%[tmp]) \t\n"
1068 "add.s %[s0], %[in1], %[in2] \t\n"
1069 "sub.s %[s2], %[in1], %[in2] \t\n"
1070 "lwc1 %[in4], 5*4(%[tmp]) \t\n"
1071 "add.s %[s1], %[in3], %[in4] \t\n"
1072 "sub.s %[s3], %[in3], %[in4] \t\n"
1073 "lwc1 %[in1], 10*4(%[win]) \t\n"
1074 "lwc1 %[in2], 4*10*4(%[buf]) \t\n"
1075 "lwc1 %[in3], 7*4(%[win]) \t\n"
1076 "mul.s %[s1], %[s1], %[f3] \t\n"
1077 "mul.s %[s3], %[s3], %[f4] \t\n"
1078 "add.s %[t0], %[s0], %[s1] \t\n"
1079 "sub.s %[t1], %[s0], %[s1] \t\n"
1080 "lwc1 %[in4], 4*7*4(%[buf]) \t\n"
1081 "lwc1 %[in5], 30*4(%[win]) \t\n"
1082 "lwc1 %[in6], 27*4(%[win]) \t\n"
1083 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1084 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1085 "mul.s %[out3], %[t0], %[in5] \t\n"
1086 "mul.s %[out4], %[t0], %[in6] \t\n"
1087 "add.s %[t0], %[s2], %[s3] \t\n"
1088 "swc1 %[out1], 320*4(%[out]) \t\n"
1089 "swc1 %[out2], 224*4(%[out]) \t\n"
1090 "swc1 %[out3], 4*10*4(%[buf]) \t\n"
1091 "swc1 %[out4], 4*7*4(%[buf]) \t\n"
1092 "sub.s %[t1], %[s2], %[s3] \t\n"
1093 "lwc1 %[in1], 16*4(%[win]) \t\n"
1094 "lwc1 %[in2], 4*16*4(%[buf]) \t\n"
1095 "lwc1 %[in3], 1*4(%[win]) \t\n"
1096 "lwc1 %[in4], 4*1*4(%[buf]) \t\n"
1097 "lwc1 %[in5], 36*4(%[win]) \t\n"
1098 "lwc1 %[in6], 21*4(%[win]) \t\n"
1099 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1100 "lwc1 %[in1], 10*4(%[tmp]) \t\n"
1101 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1102 "mul.s %[out3], %[in5], %[t0] \t\n"
1103 "mul.s %[out4], %[in6], %[t0] \t\n"
1104 "swc1 %[out1], 512*4(%[out]) \t\n"
1105 "lwc1 %[in2], 8*4(%[tmp]) \t\n"
1106 "swc1 %[out2], 32*4(%[out]) \t\n"
1107 "swc1 %[out3], 4*16*4(%[buf]) \t\n"
1108 "swc1 %[out4], 4*1*4(%[buf]) \t\n"
1109 "add.s %[s0], %[in1], %[in2] \t\n"
1110 "sub.s %[s2], %[in1], %[in2] \t\n"
1111 "lwc1 %[in3], 11*4(%[tmp]) \t\n"
1112 "lwc1 %[in4], 9*4(%[tmp]) \t\n"
1113 "add.s %[s1], %[in3], %[in4] \t\n"
1114 "sub.s %[s3], %[in3], %[in4] \t\n"
1115 "lwc1 %[in1], 11*4(%[win]) \t\n"
1116 "lwc1 %[in2], 4*11*4(%[buf]) \t\n"
1117 "lwc1 %[in3], 6*4(%[win]) \t\n"
1118 "mul.s %[s1], %[s1], %[f5] \t\n"
1119 "mul.s %[s3], %[s3], %[f6] \t\n"
1120 "lwc1 %[in4], 4*6*4(%[buf]) \t\n"
1121 "lwc1 %[in5], 31*4(%[win]) \t\n"
1122 "lwc1 %[in6], 26*4(%[win]) \t\n"
1123 "add.s %[t0], %[s0], %[s1] \t\n"
1124 "sub.s %[t1], %[s0], %[s1] \t\n"
1125 "mul.s %[out3], %[t0], %[in5] \t\n"
1126 "mul.s %[out4], %[t0], %[in6] \t\n"
1127 "add.s %[t0], %[s2], %[s3] \t\n"
1128 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1129 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1130 "swc1 %[out3], 4*11*4(%[buf]) \t\n"
1131 "swc1 %[out4], 4*6*4(%[buf]) \t\n"
1132 "sub.s %[t1], %[s2], %[s3] \t\n"
1133 "swc1 %[out1], 352*4(%[out]) \t\n"
1134 "swc1 %[out2], 192*4(%[out]) \t\n"
1135 "lwc1 %[in1], 15*4(%[win]) \t\n"
1136 "lwc1 %[in2], 4*15*4(%[buf]) \t\n"
1137 "lwc1 %[in3], 2*4(%[win]) \t\n"
1138 "lwc1 %[in4], 4*2*4(%[buf]) \t\n"
1139 "lwc1 %[in5], 35*4(%[win]) \t\n"
1140 "lwc1 %[in6], 22*4(%[win]) \t\n"
1141 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1142 "lwc1 %[in1], 14*4(%[tmp]) \t\n"
1143 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1144 "mul.s %[out3], %[t0], %[in5] \t\n"
1145 "mul.s %[out4], %[t0], %[in6] \t\n"
1146 "swc1 %[out1], 480*4(%[out]) \t\n"
1147 "lwc1 %[in2], 12*4(%[tmp]) \t\n"
1148 "swc1 %[out2], 64*4(%[out]) \t\n"
1149 "swc1 %[out3], 4*15*4(%[buf]) \t\n"
1150 "swc1 %[out4], 4*2*4(%[buf]) \t\n"
1151 "lwc1 %[in3], 15*4(%[tmp]) \t\n"
1152 "add.s %[s0], %[in1], %[in2] \t\n"
1153 "sub.s %[s2], %[in1], %[in2] \t\n"
1154 "lwc1 %[in4], 13*4(%[tmp]) \t\n"
1155 "add.s %[s1], %[in3], %[in4] \t\n"
1156 "sub.s %[s3], %[in3], %[in4] \t\n"
1157 "lwc1 %[in1], 12*4(%[win]) \t\n"
1158 "lwc1 %[in2], 4*12*4(%[buf]) \t\n"
1159 "lwc1 %[in3], 5*4(%[win]) \t\n"
1160 "mul.s %[s1], %[s1], %[f7] \t\n"
1161 "mul.s %[s3], %[s3], %[f8] \t\n"
1162 "lwc1 %[in4], 4*5*4(%[buf]) \t\n"
1163 "lwc1 %[in5], 32*4(%[win]) \t\n"
1164 "lwc1 %[in6], 25*4(%[win]) \t\n"
1165 "add.s %[t0], %[s0], %[s1] \t\n"
1166 "sub.s %[t1], %[s0], %[s1] \t\n"
1167 "lwc1 %[s0], 16*4(%[tmp]) \t\n"
1168 "lwc1 %[s1], 17*4(%[tmp]) \t\n"
1169 "mul.s %[out3], %[t0], %[in5] \t\n"
1170 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1171 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1172 "mul.s %[out4], %[t0], %[in6] \t\n"
1173 "add.s %[t0], %[s2], %[s3] \t\n"
1174 "swc1 %[out3], 4*12*4(%[buf]) \t\n"
1175 "swc1 %[out1], 384*4(%[out]) \t\n"
1176 "swc1 %[out2], 160*4(%[out]) \t\n"
1177 "swc1 %[out4], 4*5*4(%[buf]) \t\n"
1178 "sub.s %[t1], %[s2], %[s3] \t\n"
1179 "lwc1 %[in1], 14*4(%[win]) \t\n"
1180 "lwc1 %[in2], 4*14*4(%[buf]) \t\n"
1181 "lwc1 %[in3], 3*4(%[win]) \t\n"
1182 "lwc1 %[in4], 4*3*4(%[buf]) \t\n"
1183 "lwc1 %[in5], 34*4(%[win]) \t\n"
1184 "lwc1 %[in6], 23*4(%[win]) \t\n"
1185 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1186 "mul.s %[s1], %[s1], %[f9] \t\n"
1187 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1188 "mul.s %[out3], %[in5], %[t0] \t\n"
1189 "mul.s %[out4], %[in6], %[t0] \t\n"
1190 "swc1 %[out1], 448*4(%[out]) \t\n"
1191 "add.s %[t0], %[s0], %[s1] \t\n"
1192 "swc1 %[out2], 96*4(%[out]) \t\n"
1193 "swc1 %[out3], 4*14*4(%[buf]) \t\n"
1194 "swc1 %[out4], 4*3*4(%[buf]) \t\n"
1195 "sub.s %[t1], %[s0], %[s1] \t\n"
1196 "lwc1 %[in1], 13*4(%[win]) \t\n"
1197 "lwc1 %[in2], 4*13*4(%[buf]) \t\n"
1198 "lwc1 %[in3], 4*4(%[win]) \t\n"
1199 "lwc1 %[in4], 4*4*4(%[buf]) \t\n"
1200 "lwc1 %[in5], 33*4(%[win]) \t\n"
1201 "lwc1 %[in6], 24*4(%[win]) \t\n"
1202 "madd.s %[out1], %[in2], %[in1], %[t1] \t\n"
1203 "madd.s %[out2], %[in4], %[in3], %[t1] \t\n"
1204 "mul.s %[out3], %[t0], %[in5] \t\n"
1205 "mul.s %[out4], %[t0], %[in6] \t\n"
1206 "swc1 %[out1], 416*4(%[out]) \t\n"
1207 "swc1 %[out2], 128*4(%[out]) \t\n"
1208 "swc1 %[out3], 4*13*4(%[buf]) \t\n"
1209 "swc1 %[out4], 4*4*4(%[buf]) \t\n"
1210
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 [s0] "=&f" (s0), [s1] "=&f" (s1),
1218 [s2] "=&f" (s2), [s3] "=&f" (s3)
1219 : [tmp] "r" (tmp), [win] "r" (win),
1220 [buf] "r" (buf), [out] "r" (out),
1221 [f1]"f"(f1), [f2]"f"(f2), [f3]"f"(f3), [f4]"f"(f4),
1222 [f5]"f"(f5), [f6]"f"(f6), [f7]"f"(f7), [f8]"f"(f8), [f9]"f"(f9)
1223 : "memory"
1224 );
1225 }
1226
ff_imdct36_blocks_mips_float(float * out,float * buf,float * in,int count,int switch_point,int block_type)1227 static void ff_imdct36_blocks_mips_float(float *out, float *buf, float *in,
1228 int count, int switch_point, int block_type)
1229 {
1230 int j;
1231 for (j=0 ; j < count; j++) {
1232 /* apply window & overlap with previous buffer */
1233
1234 /* select window */
1235 int win_idx = (switch_point && j < 2) ? 0 : block_type;
1236 float *win = ff_mdct_win_float[win_idx + (4 & -(j & 1))];
1237
1238 imdct36_mips_float(out, buf, in, win);
1239
1240 in += 18;
1241 buf += ((j&3) != 3 ? 1 : (72-3));
1242 out++;
1243 }
1244 }
1245
1246 #endif /* !HAVE_MIPS32R6 && !HAVE_MIPS64R6 */
1247 #endif /* HAVE_INLINE_ASM && HAVE_MIPSFPU */
1248
ff_mpadsp_init_mipsfpu(MPADSPContext * s)1249 void ff_mpadsp_init_mipsfpu(MPADSPContext *s)
1250 {
1251 #if HAVE_INLINE_ASM && HAVE_MIPSFPU
1252 #if !HAVE_MIPS32R6 && !HAVE_MIPS64R6
1253 s->apply_window_float = ff_mpadsp_apply_window_mips_float;
1254 s->imdct36_blocks_float = ff_imdct36_blocks_mips_float;
1255 s->dct32_float = ff_dct32_mips_float;
1256 #endif
1257 #endif
1258 }
1259