avcodec: loongson remove useless macros in mipsfpu optimization
[ffmpeg.git] / libavcodec / mips / mpegaudiodsp_mips_float.c
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 "libavutil/mips/asmdefs.h"
59 #include "libavcodec/mpegaudiodsp.h"
60
61 static void ff_mpadsp_apply_window_mips_float(float *synth_buf, float *window,
62                                int *dither_state, float *samples, int incr)
63 {
64     register const float *w, *w2, *p;
65     int j;
66     float *samples2;
67     float sum, sum2;
68     /* temporary variables */
69     int incr1 = incr << 2;
70     int t_sample;
71     float in1, in2, in3, in4, in5, in6, in7, in8;
72     float *p2;
73
74     /* copy to avoid wrap */
75     memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));
76
77     /**
78     * instructions are scheduled to minimize pipeline stall.
79     * use of round_sample function from the original code is
80     * changed with appropriate assembly instructions.
81     */
82
83     __asm__ volatile (
84         "lwc1    %[sum],      0(%[dither_state])                            \t\n"
85         "sll     %[t_sample], %[incr1],     5                               \t\n"
86         "sub     %[t_sample], %[t_sample],  %[incr1]                        \n\t"
87         "li      %[j],        4                                             \t\n"
88         "lwc1    %[in1],      0(%[window])                                  \t\n"
89         "lwc1    %[in2],      16*4(%[synth_buf])                            \t\n"
90         "sw      $zero,       0(%[dither_state])                            \t\n"
91         "lwc1    %[in3],      64*4(%[window])                               \t\n"
92         "lwc1    %[in4],      80*4(%[synth_buf])                            \t\n"
93         PTR_ADDU "%[samples2],%[samples],   %[t_sample]                     \t\n"
94         "madd.s  %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
95         "lwc1    %[in5],      128*4(%[window])                              \t\n"
96         "lwc1    %[in6],      144*4(%[synth_buf])                           \t\n"
97         "lwc1    %[in7],      192*4(%[window])                              \t\n"
98         "madd.s  %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
99         "lwc1    %[in8],      208*4(%[synth_buf])                           \t\n"
100         "lwc1    %[in1],      256*4(%[window])                              \t\n"
101         "lwc1    %[in2],      272*4(%[synth_buf])                           \t\n"
102         "madd.s  %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
103         "lwc1    %[in3],      320*4(%[window])                              \t\n"
104         "lwc1    %[in4],      336*4(%[synth_buf])                           \t\n"
105         "lwc1    %[in5],      384*4(%[window])                              \t\n"
106         "madd.s  %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
107         "lwc1    %[in6],      400*4(%[synth_buf])                           \t\n"
108         "lwc1    %[in7],      448*4(%[window])                              \t\n"
109         "lwc1    %[in8],      464*4(%[synth_buf])                           \t\n"
110         "madd.s  %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
111         "lwc1    %[in1],      32*4(%[window])                               \t\n"
112         "lwc1    %[in2],      48*4(%[synth_buf])                            \t\n"
113         "madd.s  %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
114         "lwc1    %[in3],      96*4(%[window])                               \t\n"
115         "lwc1    %[in4],      112*4(%[synth_buf])                           \t\n"
116         "madd.s  %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
117         "lwc1    %[in5],      160*4(%[window])                              \t\n"
118         "lwc1    %[in6],      176*4(%[synth_buf])                           \t\n"
119         "madd.s  %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
120         "lwc1    %[in7],      224*4(%[window])                              \t\n"
121         "lwc1    %[in8],      240*4(%[synth_buf])                           \t\n"
122         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
123         "lwc1    %[in1],      288*4(%[window])                              \t\n"
124         "lwc1    %[in2],      304*4(%[synth_buf])                           \t\n"
125         "nmsub.s %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
126         "lwc1    %[in3],      352*4(%[window])                              \t\n"
127         "lwc1    %[in4],      368*4(%[synth_buf])                           \t\n"
128         "nmsub.s %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
129         "lwc1    %[in5],      416*4(%[window])                              \t\n"
130         "lwc1    %[in6],      432*4(%[synth_buf])                           \t\n"
131         "nmsub.s %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
132         "lwc1    %[in7],      480*4(%[window])                              \t\n"
133         "lwc1    %[in8],      496*4(%[synth_buf])                           \t\n"
134         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
135         PTR_ADDU "%[w],       %[window],    4                               \t\n"
136         "nmsub.s %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
137         PTR_ADDU "%[w2],      %[window],    124                             \t\n"
138         PTR_ADDIU "%[p],      %[synth_buf], 68                              \t\n"
139         PTR_ADDIU "%[p2],     %[synth_buf], 188                             \t\n"
140         "nmsub.s %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
141         "nmsub.s %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
142         "swc1    %[sum],      0(%[samples])                                 \t\n"
143         PTR_ADDU "%[samples], %[samples],   %[incr1]                        \t\n"
144
145         /* calculate two samples at the same time to avoid one memory
146            access per two sample */
147
148         "ff_mpadsp_apply_window_loop%=:                                     \t\n"
149         "lwc1    %[in1],      0(%[w])                                       \t\n"
150         "lwc1    %[in2],      0(%[p])                                       \t\n"
151         "lwc1    %[in3],      0(%[w2])                                      \t\n"
152         "lwc1    %[in4],      64*4(%[w])                                    \t\n"
153         "lwc1    %[in5],      64*4(%[p])                                    \t\n"
154         "lwc1    %[in6],      64*4(%[w2])                                   \t\n"
155         "mul.s   %[sum],      %[in1],       %[in2]                          \t\n"
156         "mul.s   %[sum2],     %[in2],       %[in3]                          \t\n"
157         "lwc1    %[in1],      128*4(%[w])                                   \t\n"
158         "lwc1    %[in2],      128*4(%[p])                                   \t\n"
159         "madd.s  %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
160         "nmadd.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
161         "lwc1    %[in3],      128*4(%[w2])                                  \t\n"
162         "lwc1    %[in4],      192*4(%[w])                                   \t\n"
163         "madd.s  %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
164         "lwc1    %[in5],      192*4(%[p])                                   \t\n"
165         "lwc1    %[in6],      192*4(%[w2])                                  \t\n"
166         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
167         "lwc1    %[in1],      256*4(%[w])                                   \t\n"
168         "lwc1    %[in2],      256*4(%[p])                                   \t\n"
169         "madd.s  %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
170         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
171         "lwc1    %[in3],      256*4(%[w2])                                  \t\n"
172         "lwc1    %[in4],      320*4(%[w])                                   \t\n"
173         "madd.s  %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
174         "lwc1    %[in5],      320*4(%[p])                                   \t\n"
175         "lwc1    %[in6],      320*4(%[w2])                                  \t\n"
176         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
177         "lwc1    %[in1],      384*4(%[w])                                   \t\n"
178         "lwc1    %[in2],      384*4(%[p])                                   \t\n"
179         "madd.s  %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
180         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
181         "lwc1    %[in3],      384*4(%[w2])                                  \t\n"
182         "lwc1    %[in4],      448*4(%[w])                                   \t\n"
183         "madd.s  %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
184         "lwc1    %[in5],      448*4(%[p])                                   \t\n"
185         "lwc1    %[in6],      448*4(%[w2])                                  \t\n"
186         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
187         "madd.s  %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
188         "lwc1    %[in1],      32*4(%[w])                                    \t\n"
189         "lwc1    %[in2],      0(%[p2])                                      \t\n"
190         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
191         "lwc1    %[in3],      32*4(%[w2])                                   \t\n"
192         "lwc1    %[in4],      96*4(%[w])                                    \t\n"
193         "lwc1    %[in5],      64*4(%[p2])                                   \t\n"
194         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
195         "lwc1    %[in6],      96*4(%[w2])                                   \t\n"
196         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
197         "lwc1    %[in1],      160*4(%[w])                                   \t\n"
198         "nmsub.s %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
199         "lwc1    %[in2],      128*4(%[p2])                                  \t\n"
200         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
201         "lwc1    %[in3],      160*4(%[w2])                                  \t\n"
202         "lwc1    %[in4],      224*4(%[w])                                   \t\n"
203         "lwc1    %[in5],      192*4(%[p2])                                  \t\n"
204         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
205         "lwc1    %[in6],      224*4(%[w2])                                  \t\n"
206         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
207         "lwc1    %[in1],      288*4(%[w])                                   \t\n"
208         "nmsub.s %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
209         "lwc1    %[in2],      256*4(%[p2])                                  \t\n"
210         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
211         "lwc1    %[in3],      288*4(%[w2])                                  \t\n"
212         "lwc1    %[in4],      352*4(%[w])                                   \t\n"
213         "lwc1    %[in5],      320*4(%[p2])                                  \t\n"
214         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
215         "lwc1    %[in6],      352*4(%[w2])                                  \t\n"
216         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
217         "lwc1    %[in1],      416*4(%[w])                                   \t\n"
218         "nmsub.s %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
219         "lwc1    %[in2],      384*4(%[p2])                                  \t\n"
220         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
221         "lwc1    %[in3],      416*4(%[w2])                                  \t\n"
222         "lwc1    %[in4],      480*4(%[w])                                   \t\n"
223         "lwc1    %[in5],      448*4(%[p2])                                  \t\n"
224         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
225         "lwc1    %[in6],      480*4(%[w2])                                  \t\n"
226         "nmsub.s %[sum2],     %[sum2],      %[in2], %[in3]                  \t\n"
227         PTR_ADDIU "%[w],      %[w],         4                               \t\n"
228         "nmsub.s %[sum],      %[sum],       %[in4], %[in5]                  \t\n"
229         PTR_ADDIU "%[w2],     %[w2],        -4                              \t\n"
230         "nmsub.s %[sum2],     %[sum2],      %[in5], %[in6]                  \t\n"
231         "addu    %[j],        %[j],         4                               \t\n"
232         PTR_ADDIU "%[p],      4                                             \t\n"
233         "swc1    %[sum],      0(%[samples])                                 \t\n"
234         PTR_ADDIU "%[p2],     -4                                            \t\n"
235         "swc1    %[sum2],     0(%[samples2])                                \t\n"
236         PTR_ADDU "%[samples], %[samples],   %[incr1]                        \t\n"
237         PTR_SUBU "%[samples2],%[samples2],  %[incr1]                        \t\n"
238         "bne     %[j],        64,           ff_mpadsp_apply_window_loop%=   \t\n"
239
240         "lwc1    %[in1],      48*4(%[window])                               \t\n"
241         "lwc1    %[in2],      32*4(%[synth_buf])                            \t\n"
242         "lwc1    %[in3],      112*4(%[window])                              \t\n"
243         "lwc1    %[in4],      96*4(%[synth_buf])                            \t\n"
244         "lwc1    %[in5],      176*4(%[window])                              \t\n"
245         "lwc1    %[in6],      160*4(%[synth_buf])                           \t\n"
246         "mul.s   %[sum],      %[in1],       %[in2]                          \t\n"
247         "lwc1    %[in7],      240*4(%[window])                              \t\n"
248         "lwc1    %[in8],      224*4(%[synth_buf])                           \t\n"
249         "lwc1    %[in1],      304*4(%[window])                              \t\n"
250         "nmadd.s %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
251         "lwc1    %[in2],      288*4(%[synth_buf])                           \t\n"
252         "lwc1    %[in3],      368*4(%[window])                              \t\n"
253         "lwc1    %[in4],      352*4(%[synth_buf])                           \t\n"
254         "nmsub.s %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
255         "nmsub.s %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
256         "lwc1    %[in5],      432*4(%[window])                              \t\n"
257         "lwc1    %[in6],      416*4(%[synth_buf])                           \t\n"
258         "nmsub.s %[sum],      %[sum],       %[in1], %[in2]                  \t\n"
259         "lwc1    %[in7],      496*4(%[window])                              \t\n"
260         "lwc1    %[in8],      480*4(%[synth_buf])                           \t\n"
261         "nmsub.s %[sum],      %[sum],       %[in3], %[in4]                  \t\n"
262         "nmsub.s %[sum],      %[sum],       %[in5], %[in6]                  \t\n"
263         "nmsub.s %[sum],      %[sum],       %[in7], %[in8]                  \t\n"
264         "swc1    %[sum],      0(%[samples])                                 \t\n"
265
266         : [sum] "=&f" (sum), [sum2] "=&f" (sum2),
267           [w2] "=&r" (w2),   [w] "=&r" (w),
268           [p] "=&r" (p), [p2] "=&r" (p2), [j] "=&r" (j),
269           [samples] "+r" (samples), [samples2] "=&r" (samples2),
270           [in1] "=&f" (in1), [in2] "=&f" (in2),
271           [in3] "=&f" (in3), [in4] "=&f" (in4),
272           [in5] "=&f" (in5), [in6] "=&f" (in6),
273           [in7] "=&f" (in7), [in8] "=&f" (in8),
274           [t_sample] "=&r" (t_sample)
275         : [synth_buf] "r" (synth_buf), [window] "r" (window),
276           [dither_state] "r" (dither_state), [incr1] "r" (incr1)
277         : "memory"
278     );
279 }
280
281 static void ff_dct32_mips_float(float *out, const float *tab)
282 {
283     float val0 , val1 , val2 , val3 , val4 , val5 , val6 , val7,
284           val8 , val9 , val10, val11, val12, val13, val14, val15,
285           val16, val17, val18, val19, val20, val21, val22, val23,
286           val24, val25, val26, val27, val28, val29, val30, val31;
287     float fTmp1, fTmp2, fTmp3, fTmp4, fTmp5, fTmp6, fTmp7, fTmp8,
288           fTmp9, fTmp10, fTmp11;
289
290     /**
291     * instructions are scheduled to minimize pipeline stall.
292     */
293     __asm__ volatile (
294         "lwc1       %[fTmp1],       0*4(%[tab])                             \n\t"
295         "lwc1       %[fTmp2],       31*4(%[tab])                            \n\t"
296         "lwc1       %[fTmp3],       15*4(%[tab])                            \n\t"
297         "lwc1       %[fTmp4],       16*4(%[tab])                            \n\t"
298         "li.s       %[fTmp7],       0.50241928618815570551                  \n\t"
299         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
300         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
301         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
302         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
303         "li.s       %[fTmp10],      0.50060299823519630134                  \n\t"
304         "li.s       %[fTmp11],      10.19000812354805681150                 \n\t"
305         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
306         "add.s      %[val0],        %[fTmp5],       %[fTmp6]                \n\t"
307         "sub.s      %[val15],       %[fTmp5],       %[fTmp6]                \n\t"
308         "lwc1       %[fTmp1],       7*4(%[tab])                             \n\t"
309         "lwc1       %[fTmp2],       24*4(%[tab])                            \n\t"
310         "madd.s     %[val16],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
311         "nmsub.s    %[val31],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
312         "mul.s      %[val15],       %[val15],       %[fTmp7]                \n\t"
313         "lwc1       %[fTmp3],       8*4(%[tab])                             \n\t"
314         "lwc1       %[fTmp4],       23*4(%[tab])                            \n\t"
315         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
316         "mul.s      %[val31],       %[val31],       %[fTmp7]                \n\t"
317         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
318         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
319         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
320         "li.s       %[fTmp7],       5.10114861868916385802                  \n\t"
321         "li.s       %[fTmp10],      0.67480834145500574602                  \n\t"
322         "li.s       %[fTmp11],      0.74453627100229844977                  \n\t"
323         "add.s      %[val7],        %[fTmp5],       %[fTmp6]                \n\t"
324         "sub.s      %[val8],        %[fTmp5],       %[fTmp6]                \n\t"
325         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
326         "li.s       %[fTmp1],       0.50979557910415916894                  \n\t"
327         "sub.s      %[fTmp2],       %[val0],        %[val7]                 \n\t"
328         "mul.s      %[val8],        %[val8],        %[fTmp7]                \n\t"
329         "madd.s     %[val23],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
330         "nmsub.s    %[val24],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
331         "add.s      %[val0],        %[val0],        %[val7]                 \n\t"
332         "mul.s      %[val7],        %[fTmp1],       %[fTmp2]                \n\t"
333         "sub.s      %[fTmp2],       %[val15],       %[val8]                 \n\t"
334         "add.s      %[val8],        %[val15],       %[val8]                 \n\t"
335         "mul.s      %[val24],       %[val24],       %[fTmp7]                \n\t"
336         "sub.s      %[fTmp3],       %[val16],       %[val23]                \n\t"
337         "add.s      %[val16],       %[val16],       %[val23]                \n\t"
338         "mul.s      %[val15],       %[fTmp1],       %[fTmp2]                \n\t"
339         "sub.s      %[fTmp4],       %[val31],       %[val24]                \n\t"
340         "mul.s      %[val23],       %[fTmp1],       %[fTmp3]                \n\t"
341         "add.s      %[val24],       %[val31],       %[val24]                \n\t"
342         "mul.s      %[val31],       %[fTmp1],       %[fTmp4]                \n\t"
343
344         : [fTmp1]  "=&f" (fTmp1),  [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
345           [fTmp4]  "=&f" (fTmp4),  [fTmp5] "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
346           [fTmp7]  "=&f" (fTmp7),  [fTmp8] "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
347           [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11),
348           [val0]  "=f" (val0),  [val7]  "=f" (val7),
349           [val8]  "=f" (val8),  [val15] "=f" (val15),
350           [val16] "=f" (val16), [val23] "=f" (val23),
351           [val24] "=f" (val24), [val31] "=f" (val31)
352         : [tab] "r" (tab)
353         : "memory"
354     );
355
356     __asm__ volatile (
357         "lwc1       %[fTmp1],       3*4(%[tab])                             \n\t"
358         "lwc1       %[fTmp2],       28*4(%[tab])                            \n\t"
359         "lwc1       %[fTmp3],       12*4(%[tab])                            \n\t"
360         "lwc1       %[fTmp4],       19*4(%[tab])                            \n\t"
361         "li.s       %[fTmp7],       0.64682178335999012954                  \n\t"
362         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
363         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
364         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
365         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
366         "li.s       %[fTmp10],      0.53104259108978417447                  \n\t"
367         "li.s       %[fTmp11],      1.48416461631416627724                  \n\t"
368         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
369         "add.s      %[val3],        %[fTmp5],       %[fTmp6]                \n\t"
370         "sub.s      %[val12],       %[fTmp5],       %[fTmp6]                \n\t"
371         "lwc1       %[fTmp1],       4*4(%[tab])                             \n\t"
372         "lwc1       %[fTmp2],       27*4(%[tab])                            \n\t"
373         "madd.s     %[val19],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
374         "nmsub.s    %[val28],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
375         "mul.s      %[val12],       %[val12],       %[fTmp7]                \n\t"
376         "lwc1       %[fTmp3],       11*4(%[tab])                            \n\t"
377         "lwc1       %[fTmp4],       20*4(%[tab])                            \n\t"
378         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
379         "mul.s      %[val28],       %[val28],       %[fTmp7]                \n\t"
380         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
381         "li.s       %[fTmp7],       0.78815462345125022473                  \n\t"
382         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
383         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
384         "li.s       %[fTmp10],      0.55310389603444452782                  \n\t"
385         "li.s       %[fTmp11],      1.16943993343288495515                  \n\t"
386         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
387         "add.s      %[val4],        %[fTmp5],       %[fTmp6]                \n\t"
388         "sub.s      %[val11],       %[fTmp5],       %[fTmp6]                \n\t"
389         "li.s       %[fTmp1],       2.56291544774150617881                  \n\t"
390         "madd.s     %[val20],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
391         "nmsub.s    %[val27],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
392         "mul.s      %[val11],       %[val11],       %[fTmp7]                \n\t"
393         "sub.s      %[fTmp2],       %[val3],        %[val4]                 \n\t"
394         "add.s      %[val3],        %[val3],        %[val4]                 \n\t"
395         "sub.s      %[fTmp4],       %[val19],       %[val20]                \n\t"
396         "mul.s      %[val27],       %[val27],       %[fTmp7]                \n\t"
397         "sub.s      %[fTmp3],       %[val12],       %[val11]                \n\t"
398         "mul.s      %[val4],        %[fTmp1],       %[fTmp2]                \n\t"
399         "add.s      %[val11],       %[val12],       %[val11]                \n\t"
400         "add.s      %[val19],       %[val19],       %[val20]                \n\t"
401         "mul.s      %[val20],       %[fTmp1],       %[fTmp4]                \n\t"
402         "mul.s      %[val12],       %[fTmp1],       %[fTmp3]                \n\t"
403         "sub.s      %[fTmp2],       %[val28],       %[val27]                \n\t"
404         "add.s      %[val27],       %[val28],       %[val27]                \n\t"
405         "mul.s      %[val28],       %[fTmp1],       %[fTmp2]                \n\t"
406
407         : [fTmp1]  "=&f" (fTmp1),  [fTmp2]  "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
408           [fTmp4]  "=&f" (fTmp4),  [fTmp5]  "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
409           [fTmp7]  "=&f" (fTmp7),  [fTmp8]  "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
410           [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11),
411           [val3]  "=f" (val3),  [val4]  "=f" (val4),
412           [val11] "=f" (val11), [val12] "=f" (val12),
413           [val19] "=f" (val19), [val20] "=f" (val20),
414           [val27] "=f" (val27), [val28] "=f" (val28)
415         : [tab] "r" (tab)
416         : "memory"
417     );
418
419     __asm__ volatile (
420         "li.s       %[fTmp1],       0.54119610014619698439                  \n\t"
421         "sub.s      %[fTmp2],       %[val0],        %[val3]                 \n\t"
422         "add.s      %[val0],        %[val0],        %[val3]                 \n\t"
423         "sub.s      %[fTmp3],       %[val7],        %[val4]                 \n\t"
424         "add.s      %[val4],        %[val7],        %[val4]                 \n\t"
425         "sub.s      %[fTmp4],       %[val8],        %[val11]                \n\t"
426         "mul.s      %[val3],        %[fTmp1],       %[fTmp2]                \n\t"
427         "add.s      %[val8],        %[val8],        %[val11]                \n\t"
428         "mul.s      %[val7],        %[fTmp1],       %[fTmp3]                \n\t"
429         "sub.s      %[fTmp2],       %[val15],       %[val12]                \n\t"
430         "mul.s      %[val11],       %[fTmp1],       %[fTmp4]                \n\t"
431         "add.s      %[val12],       %[val15],       %[val12]                \n\t"
432         "mul.s      %[val15],       %[fTmp1],       %[fTmp2]                \n\t"
433
434         : [val0]  "+f" (val0),   [val3] "+f" (val3),
435           [val4]  "+f" (val4),   [val7] "+f" (val7),
436           [val8]  "+f" (val8),   [val11] "+f" (val11),
437           [val12] "+f" (val12),  [val15] "+f" (val15),
438           [fTmp1] "=f"  (fTmp1), [fTmp2] "=&f" (fTmp2),
439           [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4)
440         :
441     );
442
443     __asm__ volatile (
444         "sub.s      %[fTmp2],       %[val16],       %[val19]                \n\t"
445         "add.s      %[val16],       %[val16],       %[val19]                \n\t"
446         "sub.s      %[fTmp3],       %[val23],       %[val20]                \n\t"
447         "add.s      %[val20],       %[val23],       %[val20]                \n\t"
448         "sub.s      %[fTmp4],       %[val24],       %[val27]                \n\t"
449         "mul.s      %[val19],       %[fTmp1],       %[fTmp2]                \n\t"
450         "add.s      %[val24],       %[val24],       %[val27]                \n\t"
451         "mul.s      %[val23],       %[fTmp1],       %[fTmp3]                \n\t"
452         "sub.s      %[fTmp2],       %[val31],       %[val28]                \n\t"
453         "mul.s      %[val27],       %[fTmp1],       %[fTmp4]                \n\t"
454         "add.s      %[val28],       %[val31],       %[val28]                \n\t"
455         "mul.s      %[val31],       %[fTmp1],       %[fTmp2]                \n\t"
456
457         : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
458           [val16] "+f" (val16), [val19] "+f" (val19), [val20] "+f" (val20),
459           [val23] "+f" (val23), [val24] "+f" (val24), [val27] "+f" (val27),
460           [val28] "+f" (val28), [val31] "+f" (val31)
461         : [fTmp1] "f" (fTmp1)
462     );
463
464     __asm__ volatile (
465         "lwc1       %[fTmp1],       1*4(%[tab])                             \n\t"
466         "lwc1       %[fTmp2],       30*4(%[tab])                            \n\t"
467         "lwc1       %[fTmp3],       14*4(%[tab])                            \n\t"
468         "lwc1       %[fTmp4],       17*4(%[tab])                            \n\t"
469         "li.s       %[fTmp7],       0.52249861493968888062                  \n\t"
470         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
471         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
472         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
473         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
474         "li.s       %[fTmp10],      0.50547095989754365998                  \n\t"
475         "li.s       %[fTmp11],      3.40760841846871878570                  \n\t"
476         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
477         "add.s      %[val1],        %[fTmp5],       %[fTmp6]                \n\t"
478         "sub.s      %[val14],       %[fTmp5],       %[fTmp6]                \n\t"
479         "lwc1       %[fTmp1],       6*4(%[tab])                             \n\t"
480         "lwc1       %[fTmp2],       25*4(%[tab])                            \n\t"
481         "madd.s     %[val17],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
482         "nmsub.s    %[val30],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
483         "mul.s      %[val14],       %[val14],       %[fTmp7]                \n\t"
484         "lwc1       %[fTmp3],       9*4(%[tab])                             \n\t"
485         "lwc1       %[fTmp4],       22*4(%[tab])                            \n\t"
486         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
487         "mul.s      %[val30],       %[val30],       %[fTmp7]                \n\t"
488         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
489         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
490         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
491         "li.s       %[fTmp7],       1.72244709823833392782                  \n\t"
492         "li.s       %[fTmp10],      0.62250412303566481615                  \n\t"
493         "li.s       %[fTmp11],      0.83934964541552703873                  \n\t"
494         "add.s      %[val6],        %[fTmp5],       %[fTmp6]                \n\t"
495         "sub.s      %[val9],        %[fTmp5],       %[fTmp6]                \n\t"
496         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
497         "li.s       %[fTmp1],       0.60134488693504528054                  \n\t"
498         "sub.s      %[fTmp2],       %[val1],        %[val6]                 \n\t"
499         "add.s      %[val1],        %[val1],        %[val6]                 \n\t"
500         "mul.s      %[val9],        %[val9],        %[fTmp7]                \n\t"
501         "madd.s     %[val22],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
502         "nmsub.s    %[val25],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
503         "mul.s      %[val6],        %[fTmp1],       %[fTmp2]                \n\t"
504         "sub.s      %[fTmp2],       %[val14],       %[val9]                 \n\t"
505         "add.s      %[val9],        %[val14],       %[val9]                 \n\t"
506         "mul.s      %[val25],       %[val25],       %[fTmp7]                \n\t"
507         "sub.s      %[fTmp3],       %[val17],       %[val22]                \n\t"
508         "add.s      %[val17],       %[val17],       %[val22]                \n\t"
509         "mul.s      %[val14],       %[fTmp1],       %[fTmp2]                \n\t"
510         "sub.s      %[fTmp2],       %[val30],       %[val25]                \n\t"
511         "mul.s      %[val22],       %[fTmp1],       %[fTmp3]                \n\t"
512         "add.s      %[val25],       %[val30],       %[val25]                \n\t"
513         "mul.s      %[val30],       %[fTmp1],       %[fTmp2]                \n\t"
514
515         : [fTmp1]  "=&f" (fTmp1),  [fTmp2]  "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
516           [fTmp4]  "=&f" (fTmp4),  [fTmp5]  "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
517           [fTmp7]  "=&f" (fTmp7),  [fTmp8]  "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
518           [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11),
519           [val1]  "=f" (val1),  [val6]  "=f" (val6),
520           [val9]  "=f" (val9),  [val14] "=f" (val14),
521           [val17] "=f" (val17), [val22] "=f" (val22),
522           [val25] "=f" (val25), [val30] "=f" (val30)
523         : [tab] "r" (tab)
524         : "memory"
525     );
526
527     __asm__ volatile (
528         "lwc1       %[fTmp1],       2*4(%[tab])                             \n\t"
529         "lwc1       %[fTmp2],       29*4(%[tab])                            \n\t"
530         "lwc1       %[fTmp3],       13*4(%[tab])                            \n\t"
531         "lwc1       %[fTmp4],       18*4(%[tab])                            \n\t"
532         "li.s       %[fTmp7],       0.56694403481635770368                  \n\t"
533         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
534         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
535         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
536         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
537         "li.s       %[fTmp10],      0.51544730992262454697                  \n\t"
538         "li.s       %[fTmp11],      2.05778100995341155085                  \n\t"
539         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
540         "add.s      %[val2],        %[fTmp5],       %[fTmp6]                \n\t"
541         "sub.s      %[val13],       %[fTmp5],       %[fTmp6]                \n\t"
542         "lwc1       %[fTmp1],       5*4(%[tab])                             \n\t"
543         "lwc1       %[fTmp2],       26*4(%[tab])                            \n\t"
544         "madd.s     %[val18],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
545         "nmsub.s    %[val29],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
546         "mul.s      %[val13],       %[val13],       %[fTmp7]                \n\t"
547         "lwc1       %[fTmp3],       10*4(%[tab])                            \n\t"
548         "lwc1       %[fTmp4],       21*4(%[tab])                            \n\t"
549         "mul.s      %[val29],       %[val29],       %[fTmp7]                \n\t"
550         "add.s      %[fTmp5],       %[fTmp1],       %[fTmp2]                \n\t"
551         "sub.s      %[fTmp8],       %[fTmp1],       %[fTmp2]                \n\t"
552         "add.s      %[fTmp6],       %[fTmp3],       %[fTmp4]                \n\t"
553         "sub.s      %[fTmp9],       %[fTmp3],       %[fTmp4]                \n\t"
554         "li.s       %[fTmp7],       1.06067768599034747134                  \n\t"
555         "li.s       %[fTmp10],      0.58293496820613387367                  \n\t"
556         "li.s       %[fTmp11],      0.97256823786196069369                  \n\t"
557         "add.s      %[val5],        %[fTmp5],       %[fTmp6]                \n\t"
558         "sub.s      %[val10],       %[fTmp5],       %[fTmp6]                \n\t"
559         "mul.s      %[fTmp8],       %[fTmp8],       %[fTmp10]               \n\t"
560         "li.s       %[fTmp1],       0.89997622313641570463                  \n\t"
561         "sub.s      %[fTmp2],       %[val2],        %[val5]                 \n\t"
562         "mul.s      %[val10],       %[val10],       %[fTmp7]                \n\t"
563         "madd.s     %[val21],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
564         "nmsub.s    %[val26],       %[fTmp8],       %[fTmp9],   %[fTmp11]   \n\t"
565         "add.s      %[val2],        %[val2],        %[val5]                 \n\t"
566         "mul.s      %[val5],        %[fTmp1],       %[fTmp2]                \n\t"
567         "sub.s      %[fTmp3],       %[val13],       %[val10]                \n\t"
568         "add.s      %[val10],       %[val13],       %[val10]                \n\t"
569         "mul.s      %[val26],       %[val26],       %[fTmp7]                \n\t"
570         "sub.s      %[fTmp4],       %[val18],       %[val21]                \n\t"
571         "add.s      %[val18],       %[val18],       %[val21]                \n\t"
572         "mul.s      %[val13],       %[fTmp1],       %[fTmp3]                \n\t"
573         "sub.s      %[fTmp2],       %[val29],       %[val26]                \n\t"
574         "add.s      %[val26],       %[val29],       %[val26]                \n\t"
575         "mul.s      %[val21],       %[fTmp1],       %[fTmp4]                \n\t"
576         "mul.s      %[val29],       %[fTmp1],       %[fTmp2]                \n\t"
577
578         : [fTmp1]  "=&f" (fTmp1),  [fTmp2]  "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3),
579           [fTmp4]  "=&f" (fTmp4),  [fTmp5]  "=&f" (fTmp5), [fTmp6] "=&f" (fTmp6),
580           [fTmp7]  "=&f" (fTmp7),  [fTmp8]  "=&f" (fTmp8), [fTmp9] "=&f" (fTmp9),
581           [fTmp10] "=&f" (fTmp10), [fTmp11] "=&f" (fTmp11),
582           [val2]  "=f" (val2),  [val5]  "=f" (val5),
583           [val10] "=f" (val10), [val13] "=f" (val13),
584           [val18] "=f" (val18), [val21] "=f" (val21),
585           [val26] "=f" (val26), [val29] "=f" (val29)
586         : [tab] "r" (tab)
587         : "memory"
588     );
589
590     __asm__ volatile (
591         "li.s       %[fTmp1],       1.30656296487637652785                  \n\t"
592         "sub.s      %[fTmp2],       %[val1],        %[val2]                 \n\t"
593         "add.s      %[val1],        %[val1],        %[val2]                 \n\t"
594         "sub.s      %[fTmp3],       %[val6],        %[val5]                 \n\t"
595         "add.s      %[val5],        %[val6],        %[val5]                 \n\t"
596         "sub.s      %[fTmp4],       %[val9],        %[val10]                \n\t"
597         "mul.s      %[val2],        %[fTmp1],       %[fTmp2]                \n\t"
598         "add.s      %[val9],        %[val9],        %[val10]                \n\t"
599         "mul.s      %[val6],        %[fTmp1],       %[fTmp3]                \n\t"
600         "sub.s      %[fTmp2],       %[val14],       %[val13]                \n\t"
601         "mul.s      %[val10],       %[fTmp1],       %[fTmp4]                \n\t"
602         "add.s      %[val13],       %[val14],       %[val13]                \n\t"
603         "mul.s      %[val14],       %[fTmp1],       %[fTmp2]                \n\t"
604
605         : [fTmp1] "=f"  (fTmp1), [fTmp2] "=&f" (fTmp2),
606           [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
607           [val1]  "+f" (val1),  [val2]  "+f" (val2),
608           [val5]  "+f" (val5),  [val6]  "+f" (val6),
609           [val9]  "+f" (val9),  [val10] "+f" (val10),
610           [val13] "+f" (val13), [val14] "+f" (val14)
611         :
612     );
613
614     __asm__ volatile (
615         "sub.s      %[fTmp2],       %[val17],       %[val18]                \n\t"
616         "add.s      %[val17],       %[val17],       %[val18]                \n\t"
617         "sub.s      %[fTmp3],       %[val22],       %[val21]                \n\t"
618         "add.s      %[val21],       %[val22],       %[val21]                \n\t"
619         "sub.s      %[fTmp4],       %[val25],       %[val26]                \n\t"
620         "mul.s      %[val18],       %[fTmp1],       %[fTmp2]                \n\t"
621         "add.s      %[val25],       %[val25],       %[val26]                \n\t"
622         "mul.s      %[val22],       %[fTmp1],       %[fTmp3]                \n\t"
623         "sub.s      %[fTmp2],       %[val30],       %[val29]                \n\t"
624         "mul.s      %[val26],       %[fTmp1],       %[fTmp4]                \n\t"
625         "add.s      %[val29],       %[val30],       %[val29]                \n\t"
626         "mul.s      %[val30],       %[fTmp1],       %[fTmp2]                \n\t"
627
628         : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
629           [val17] "+f" (val17), [val18] "+f" (val18), [val21] "+f" (val21),
630           [val22] "+f" (val22), [val25] "+f" (val25), [val26] "+f" (val26),
631           [val29] "+f" (val29), [val30] "+f" (val30)
632         : [fTmp1] "f" (fTmp1)
633     );
634
635     __asm__ volatile (
636         "li.s       %[fTmp1],       0.70710678118654752439                  \n\t"
637         "sub.s      %[fTmp2],       %[val0],        %[val1]                 \n\t"
638         "add.s      %[val0],        %[val0],        %[val1]                 \n\t"
639         "sub.s      %[fTmp3],       %[val3],        %[val2]                 \n\t"
640         "add.s      %[val2],        %[val3],        %[val2]                 \n\t"
641         "sub.s      %[fTmp4],       %[val4],        %[val5]                 \n\t"
642         "mul.s      %[val1],        %[fTmp1],       %[fTmp2]                \n\t"
643         "swc1       %[val0],        0(%[out])                               \n\t"
644         "mul.s      %[val3],        %[fTmp3],       %[fTmp1]                \n\t"
645         "add.s      %[val4],        %[val4],        %[val5]                 \n\t"
646         "mul.s      %[val5],        %[fTmp1],       %[fTmp4]                \n\t"
647         "swc1       %[val1],        16*4(%[out])                            \n\t"
648         "sub.s      %[fTmp2],       %[val7],        %[val6]                 \n\t"
649         "add.s      %[val2],        %[val2],        %[val3]                 \n\t"
650         "swc1       %[val3],        24*4(%[out])                            \n\t"
651         "add.s      %[val6],        %[val7],        %[val6]                 \n\t"
652         "mul.s      %[val7],        %[fTmp1],       %[fTmp2]                \n\t"
653         "swc1       %[val2],        8*4(%[out])                             \n\t"
654         "add.s      %[val6],        %[val6],        %[val7]                 \n\t"
655         "swc1       %[val7],        28*4(%[out])                            \n\t"
656         "add.s      %[val4],        %[val4],        %[val6]                 \n\t"
657         "add.s      %[val6],        %[val6],        %[val5]                 \n\t"
658         "add.s      %[val5],        %[val5],        %[val7]                 \n\t"
659         "swc1       %[val4],        4*4(%[out])                             \n\t"
660         "swc1       %[val5],        20*4(%[out])                            \n\t"
661         "swc1       %[val6],        12*4(%[out])                            \n\t"
662
663         : [fTmp1] "=f"  (fTmp1), [fTmp2] "=&f" (fTmp2),
664           [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
665           [val0] "+f" (val0), [val1] "+f" (val1),
666           [val2] "+f" (val2), [val3] "+f" (val3),
667           [val4] "+f" (val4), [val5] "+f" (val5),
668           [val6] "+f" (val6), [val7] "+f" (val7)
669         : [out] "r" (out)
670     );
671
672     __asm__ volatile (
673         "sub.s      %[fTmp2],       %[val8],        %[val9]                 \n\t"
674         "add.s      %[val8],        %[val8],        %[val9]                 \n\t"
675         "sub.s      %[fTmp3],       %[val11],       %[val10]                \n\t"
676         "add.s      %[val10],       %[val11],       %[val10]                \n\t"
677         "sub.s      %[fTmp4],       %[val12],       %[val13]                \n\t"
678         "mul.s      %[val9],        %[fTmp1],       %[fTmp2]                \n\t"
679         "add.s      %[val12],       %[val12],       %[val13]                \n\t"
680         "mul.s      %[val11],       %[fTmp1],       %[fTmp3]                \n\t"
681         "sub.s      %[fTmp2],       %[val15],       %[val14]                \n\t"
682         "mul.s      %[val13],       %[fTmp1],       %[fTmp4]                \n\t"
683         "add.s      %[val14],       %[val15],       %[val14]                \n\t"
684         "add.s      %[val10],       %[val10],       %[val11]                \n\t"
685         "mul.s      %[val15],       %[fTmp1],       %[fTmp2]                \n\t"
686         "add.s      %[val14],       %[val14],       %[val15]                \n\t"
687         "add.s      %[val12],       %[val12],       %[val14]                \n\t"
688         "add.s      %[val14],       %[val14],       %[val13]                \n\t"
689         "add.s      %[val13],       %[val13],       %[val15]                \n\t"
690         "add.s      %[val8],        %[val8],        %[val12]                \n\t"
691         "add.s      %[val12],       %[val12],       %[val10]                \n\t"
692         "add.s      %[val10],       %[val10],       %[val14]                \n\t"
693         "add.s      %[val14],       %[val14],       %[val9]                 \n\t"
694         "add.s      %[val9],        %[val9],        %[val13]                \n\t"
695         "add.s      %[val13],       %[val13],       %[val11]                \n\t"
696         "add.s      %[val11],       %[val11],       %[val15]                \n\t"
697         "swc1       %[val8],         2*4(%[out])                            \n\t"
698         "swc1       %[val9],        18*4(%[out])                            \n\t"
699         "swc1       %[val10],       10*4(%[out])                            \n\t"
700         "swc1       %[val11],       26*4(%[out])                            \n\t"
701         "swc1       %[val12],        6*4(%[out])                            \n\t"
702         "swc1       %[val13],       22*4(%[out])                            \n\t"
703         "swc1       %[val14],       14*4(%[out])                            \n\t"
704         "swc1       %[val15],       30*4(%[out])                            \n\t"
705
706         : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
707           [val8]  "+f" (val8),  [val9]  "+f" (val9),  [val10] "+f" (val10),
708           [val11] "+f" (val11), [val12] "+f" (val12), [val13] "+f" (val13),
709           [val14] "+f" (val14), [val15] "+f" (val15)
710         : [fTmp1] "f" (fTmp1), [out] "r" (out)
711     );
712
713     __asm__ volatile (
714         "sub.s      %[fTmp2],       %[val16],       %[val17]                \n\t"
715         "add.s      %[val16],       %[val16],       %[val17]                \n\t"
716         "sub.s      %[fTmp3],       %[val19],       %[val18]                \n\t"
717         "add.s      %[val18],       %[val19],       %[val18]                \n\t"
718         "sub.s      %[fTmp4],       %[val20],       %[val21]                \n\t"
719         "mul.s      %[val17],       %[fTmp1],       %[fTmp2]                \n\t"
720         "add.s      %[val20],       %[val20],       %[val21]                \n\t"
721         "mul.s      %[val19],       %[fTmp1],       %[fTmp3]                \n\t"
722         "sub.s      %[fTmp2],       %[val23],       %[val22]                \n\t"
723         "mul.s      %[val21],       %[fTmp1],       %[fTmp4]                \n\t"
724         "add.s      %[val22],       %[val23],       %[val22]                \n\t"
725         "add.s      %[val18],       %[val18],       %[val19]                \n\t"
726         "mul.s      %[val23],       %[fTmp1],       %[fTmp2]                \n\t"
727         "add.s      %[val22],       %[val22],       %[val23]                \n\t"
728         "add.s      %[val20],       %[val20],       %[val22]                \n\t"
729         "add.s      %[val22],       %[val22],       %[val21]                \n\t"
730         "add.s      %[val21],       %[val21],       %[val23]                \n\t"
731
732         : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
733           [val16] "+f" (val16), [val17] "+f" (val17), [val18] "+f" (val18),
734           [val19] "+f" (val19), [val20] "+f" (val20), [val21] "+f" (val21),
735           [val22] "+f" (val22), [val23] "+f" (val23)
736         : [fTmp1] "f" (fTmp1)
737     );
738
739     __asm__ volatile (
740         "sub.s      %[fTmp2],       %[val24],       %[val25]                \n\t"
741         "add.s      %[val24],       %[val24],       %[val25]                \n\t"
742         "sub.s      %[fTmp3],       %[val27],       %[val26]                \n\t"
743         "add.s      %[val26],       %[val27],       %[val26]                \n\t"
744         "sub.s      %[fTmp4],       %[val28],       %[val29]                \n\t"
745         "mul.s      %[val25],       %[fTmp1],       %[fTmp2]                \n\t"
746         "add.s      %[val28],       %[val28],       %[val29]                \n\t"
747         "mul.s      %[val27],       %[fTmp1],       %[fTmp3]                \n\t"
748         "sub.s      %[fTmp2],       %[val31],       %[val30]                \n\t"
749         "mul.s      %[val29],       %[fTmp1],       %[fTmp4]                \n\t"
750         "add.s      %[val30],       %[val31],       %[val30]                \n\t"
751         "add.s      %[val26],       %[val26],       %[val27]                \n\t"
752         "mul.s      %[val31],       %[fTmp1],       %[fTmp2]                \n\t"
753         "add.s      %[val30],       %[val30],       %[val31]                \n\t"
754         "add.s      %[val28],       %[val28],       %[val30]                \n\t"
755         "add.s      %[val30],       %[val30],       %[val29]                \n\t"
756         "add.s      %[val29],       %[val29],       %[val31]                \n\t"
757         "add.s      %[val24],       %[val24],       %[val28]                \n\t"
758         "add.s      %[val28],       %[val28],       %[val26]                \n\t"
759         "add.s      %[val26],       %[val26],       %[val30]                \n\t"
760         "add.s      %[val30],       %[val30],       %[val25]                \n\t"
761         "add.s      %[val25],       %[val25],       %[val29]                \n\t"
762         "add.s      %[val29],       %[val29],       %[val27]                \n\t"
763         "add.s      %[val27],       %[val27],       %[val31]                \n\t"
764
765         : [fTmp2] "=&f" (fTmp2), [fTmp3] "=&f" (fTmp3), [fTmp4] "=&f" (fTmp4),
766           [val24] "+f" (val24), [val25] "+f" (val25), [val26] "+f" (val26),
767           [val27] "+f" (val27), [val28] "+f" (val28), [val29] "+f" (val29),
768           [val30] "+f" (val30), [val31] "+f" (val31)
769         : [fTmp1] "f" (fTmp1)
770     );
771
772     out[ 1] = val16 + val24;
773     out[17] = val17 + val25;
774     out[ 9] = val18 + val26;
775     out[25] = val19 + val27;
776     out[ 5] = val20 + val28;
777     out[21] = val21 + val29;
778     out[13] = val22 + val30;
779     out[29] = val23 + val31;
780     out[ 3] = val24 + val20;
781     out[19] = val25 + val21;
782     out[11] = val26 + val22;
783     out[27] = val27 + val23;
784     out[ 7] = val28 + val18;
785     out[23] = val29 + val19;
786     out[15] = val30 + val17;
787     out[31] = val31;
788 }
789
790 static void imdct36_mips_float(float *out, float *buf, float *in, float *win)
791 {
792     float t0, t1, t2, t3, s0, s1, s2, s3;
793     float tmp[18];
794     /* temporary variables */
795     float in1, in2, in3, in4, in5, in6;
796     float out1, out2, out3, out4, out5;
797     float c1, c2, c3, c4, c5, c6, c7, c8, c9;
798
799     /**
800     * all loops are unrolled totally, and instructions are scheduled to
801     * minimize pipeline stall. instructions of the first two loops are
802     * reorganized, in order to eliminate unnecessary readings and
803     * writings into array. values defined in macros and tables are
804     * eliminated - they are directly loaded in appropriate variables
805     */
806
807     /* loop 1 and 2 */
808     __asm__ volatile (
809         "lwc1   %[in1],  17*4(%[in])                                    \t\n"
810         "lwc1   %[in2],  16*4(%[in])                                    \t\n"
811         "lwc1   %[in3],  15*4(%[in])                                    \t\n"
812         "lwc1   %[in4],  14*4(%[in])                                    \t\n"
813         "lwc1   %[in5],  13*4(%[in])                                    \t\n"
814         "lwc1   %[in6],  12*4(%[in])                                    \t\n"
815         "add.s  %[out1], %[in1],  %[in2]                                \t\n"
816         "add.s  %[out2], %[in2],  %[in3]                                \t\n"
817         "add.s  %[out3], %[in3],  %[in4]                                \t\n"
818         "add.s  %[out4], %[in4],  %[in5]                                \t\n"
819         "add.s  %[out5], %[in5],  %[in6]                                \t\n"
820         "lwc1   %[in1],  11*4(%[in])                                    \t\n"
821         "swc1   %[out2], 16*4(%[in])                                    \t\n"
822         "add.s  %[out1], %[out1], %[out3]                               \t\n"
823         "swc1   %[out4], 14*4(%[in])                                    \t\n"
824         "add.s  %[out3], %[out3], %[out5]                               \t\n"
825         "lwc1   %[in2],  10*4(%[in])                                    \t\n"
826         "lwc1   %[in3],  9*4(%[in])                                     \t\n"
827         "swc1   %[out1], 17*4(%[in])                                    \t\n"
828         "lwc1   %[in4],  8*4(%[in])                                     \t\n"
829         "swc1   %[out3], 15*4(%[in])                                    \t\n"
830         "add.s  %[out1], %[in6],  %[in1]                                \t\n"
831         "add.s  %[out2], %[in1],  %[in2]                                \t\n"
832         "add.s  %[out3], %[in2],  %[in3]                                \t\n"
833         "add.s  %[out4], %[in3],  %[in4]                                \t\n"
834         "lwc1   %[in5],  7*4(%[in])                                     \t\n"
835         "swc1   %[out1], 12*4(%[in])                                    \t\n"
836         "add.s  %[out5], %[out5], %[out2]                               \t\n"
837         "swc1   %[out3], 10*4(%[in])                                    \t\n"
838         "add.s  %[out2], %[out2], %[out4]                               \t\n"
839         "lwc1   %[in6],  6*4(%[in])                                     \t\n"
840         "lwc1   %[in1],  5*4(%[in])                                     \t\n"
841         "swc1   %[out5], 13*4(%[in])                                    \t\n"
842         "lwc1   %[in2],  4*4(%[in])                                     \t\n"
843         "swc1   %[out2], 11*4(%[in])                                    \t\n"
844         "add.s  %[out5], %[in4],  %[in5]                                \t\n"
845         "add.s  %[out1], %[in5],  %[in6]                                \t\n"
846         "add.s  %[out2], %[in6],  %[in1]                                \t\n"
847         "add.s  %[out3], %[in1],  %[in2]                                \t\n"
848         "lwc1   %[in3],  3*4(%[in])                                     \t\n"
849         "swc1   %[out5], 8*4(%[in])                                     \t\n"
850         "add.s  %[out4], %[out4], %[out1]                               \t\n"
851         "swc1   %[out2], 6*4(%[in])                                     \t\n"
852         "add.s  %[out1], %[out1], %[out3]                               \t\n"
853         "lwc1   %[in4],  2*4(%[in])                                     \t\n"
854         "lwc1   %[in5],  1*4(%[in])                                     \t\n"
855         "swc1   %[out4], 9*4(%[in])                                     \t\n"
856         "lwc1   %[in6],  0(%[in])                                       \t\n"
857         "swc1   %[out1], 7*4(%[in])                                     \t\n"
858         "add.s  %[out4], %[in2],  %[in3]                                \t\n"
859         "add.s  %[out5], %[in3],  %[in4]                                \t\n"
860         "add.s  %[out1], %[in4],  %[in5]                                \t\n"
861         "add.s  %[out2], %[in5],  %[in6]                                \t\n"
862         "swc1   %[out4], 4*4(%[in])                                     \t\n"
863         "add.s  %[out3], %[out3], %[out5]                               \t\n"
864         "swc1   %[out1], 2*4(%[in])                                     \t\n"
865         "add.s  %[out5], %[out5], %[out2]                               \t\n"
866         "swc1   %[out2], 1*4(%[in])                                     \t\n"
867         "swc1   %[out3], 5*4(%[in])                                     \t\n"
868         "swc1   %[out5], 3*4(%[in])                                     \t\n"
869
870         : [in1] "=&f" (in1), [in2] "=&f" (in2),
871           [in3] "=&f" (in3), [in4] "=&f" (in4),
872           [in5] "=&f" (in5), [in6] "=&f" (in6),
873           [out1] "=&f" (out1), [out2] "=&f" (out2),
874           [out3] "=&f" (out3), [out4] "=&f" (out4),
875           [out5] "=&f" (out5)
876         : [in] "r" (in)
877         : "memory"
878     );
879
880     /* loop 3 */
881     __asm__ volatile (
882         "li.s    %[c1],   0.5                                           \t\n"
883         "lwc1    %[in1],  8*4(%[in])                                    \t\n"
884         "lwc1    %[in2],  16*4(%[in])                                   \t\n"
885         "lwc1    %[in3],  4*4(%[in])                                    \t\n"
886         "lwc1    %[in4],  0(%[in])                                      \t\n"
887         "lwc1    %[in5],  12*4(%[in])                                   \t\n"
888         "li.s    %[c2],   0.93969262078590838405                        \t\n"
889         "add.s   %[t2],   %[in1],  %[in2]                               \t\n"
890         "add.s   %[t0],   %[in1],  %[in3]                               \t\n"
891         "li.s    %[c3],   -0.76604444311897803520                       \t\n"
892         "madd.s  %[t3],   %[in4],  %[in5], %[c1]                        \t\n"
893         "sub.s   %[t1],   %[in4],  %[in5]                               \t\n"
894         "sub.s   %[t2],   %[t2],   %[in3]                               \t\n"
895         "mul.s   %[t0],   %[t0],   %[c2]                                \t\n"
896         "li.s    %[c4],   -0.17364817766693034885                       \t\n"
897         "li.s    %[c5],   -0.86602540378443864676                       \t\n"
898         "li.s    %[c6],   0.98480775301220805936                        \t\n"
899         "nmsub.s %[out1], %[t1],   %[t2],  %[c1]                        \t\n"
900         "add.s   %[out2], %[t1],   %[t2]                                \t\n"
901         "add.s   %[t2],   %[in2],  %[in3]                               \t\n"
902         "sub.s   %[t1],   %[in1],  %[in2]                               \t\n"
903         "sub.s   %[out3], %[t3],   %[t0]                                \t\n"
904         "swc1    %[out1], 6*4(%[tmp])                                   \t\n"
905         "swc1    %[out2], 16*4(%[tmp])                                  \t\n"
906         "mul.s   %[t2],   %[t2],   %[c3]                                \t\n"
907         "mul.s   %[t1],   %[t1],   %[c4]                                \t\n"
908         "add.s   %[out1], %[t3],   %[t0]                                \t\n"
909         "lwc1    %[in1],  10*4(%[in])                                   \t\n"
910         "lwc1    %[in2],  14*4(%[in])                                   \t\n"
911         "sub.s   %[out3], %[out3], %[t2]                                \t\n"
912         "add.s   %[out2], %[t3],   %[t2]                                \t\n"
913         "add.s   %[out1], %[out1], %[t1]                                \t\n"
914         "lwc1    %[in3],  2*4(%[in])                                    \t\n"
915         "lwc1    %[in4],  6*4(%[in])                                    \t\n"
916         "swc1    %[out3], 10*4(%[tmp])                                  \t\n"
917         "sub.s   %[out2], %[out2], %[t1]                                \t\n"
918         "swc1    %[out1], 2*4(%[tmp])                                   \t\n"
919         "add.s   %[out1], %[in1],  %[in2]                               \t\n"
920         "add.s   %[t2],   %[in1],  %[in3]                               \t\n"
921         "sub.s   %[t3],   %[in1],  %[in2]                               \t\n"
922         "swc1    %[out2], 14*4(%[tmp])                                  \t\n"
923         "li.s    %[c7],   -0.34202014332566873304                       \t\n"
924         "sub.s   %[out1], %[out1], %[in3]                               \t\n"
925         "mul.s   %[t2],   %[t2],   %[c6]                                \t\n"
926         "mul.s   %[t3],   %[t3],   %[c7]                                \t\n"
927         "li.s    %[c8],   0.86602540378443864676                        \t\n"
928         "mul.s   %[t0],   %[in4],  %[c8]                                \t\n"
929         "mul.s   %[out1], %[out1], %[c5]                                \t\n"
930         "add.s   %[t1],   %[in2],  %[in3]                               \t\n"
931         "li.s    %[c9],   -0.64278760968653932632                       \t\n"
932         "add.s   %[out2], %[t2],   %[t3]                                \t\n"
933         "lwc1    %[in1],  9*4(%[in])                                    \t\n"
934         "swc1    %[out1], 4*4(%[tmp])                                   \t\n"
935         "mul.s   %[t1],   %[t1],   %[c9]                                \t\n"
936         "lwc1    %[in2],  17*4(%[in])                                   \t\n"
937         "add.s   %[out2], %[out2], %[t0]                                \t\n"
938         "lwc1    %[in3],  5*4(%[in])                                    \t\n"
939         "lwc1    %[in4],  1*4(%[in])                                    \t\n"
940         "add.s   %[out3], %[t2],   %[t1]                                \t\n"
941         "sub.s   %[out1], %[t3],   %[t1]                                \t\n"
942         "swc1    %[out2], 0(%[tmp])                                     \t\n"
943         "lwc1    %[in5],  13*4(%[in])                                   \t\n"
944         "add.s   %[t2],   %[in1],  %[in2]                               \t\n"
945         "sub.s   %[out3], %[out3], %[t0]                                \t\n"
946         "sub.s   %[out1], %[out1], %[t0]                                \t\n"
947         "add.s   %[t0],   %[in1],  %[in3]                               \t\n"
948         "madd.s  %[t3],   %[in4],  %[in5], %[c1]                        \t\n"
949         "sub.s   %[t2],   %[t2],   %[in3]                               \t\n"
950         "swc1    %[out3], 12*4(%[tmp])                                  \t\n"
951         "swc1    %[out1], 8*4(%[tmp])                                   \t\n"
952         "sub.s   %[t1],   %[in4],  %[in5]                               \t\n"
953         "mul.s   %[t0],   %[t0],   %[c2]                                \t\n"
954         "nmsub.s %[out1], %[t1],   %[t2],  %[c1]                        \t\n"
955         "add.s   %[out2], %[t1],   %[t2]                                \t\n"
956         "add.s   %[t2],   %[in2],  %[in3]                               \t\n"
957         "sub.s   %[t1],   %[in1],  %[in2]                               \t\n"
958         "sub.s   %[out3], %[t3],   %[t0]                                \t\n"
959         "swc1    %[out1], 7*4(%[tmp])                                   \t\n"
960         "swc1    %[out2], 17*4(%[tmp])                                  \t\n"
961         "mul.s   %[t2],   %[t2],   %[c3]                                \t\n"
962         "mul.s   %[t1],   %[t1],   %[c4]                                \t\n"
963         "add.s   %[out1], %[t3],   %[t0]                                \t\n"
964         "lwc1    %[in1],  11*4(%[in])                                   \t\n"
965         "lwc1    %[in2],  15*4(%[in])                                   \t\n"
966         "sub.s   %[out3], %[out3], %[t2]                                \t\n"
967         "add.s   %[out2], %[t3],   %[t2]                                \t\n"
968         "add.s   %[out1], %[out1], %[t1]                                \t\n"
969         "lwc1    %[in3],  3*4(%[in])                                    \t\n"
970         "lwc1    %[in4],  7*4(%[in])                                    \t\n"
971         "swc1    %[out3], 11*4(%[tmp])                                  \t\n"
972         "sub.s   %[out2], %[out2], %[t1]                                \t\n"
973         "swc1    %[out1], 3*4(%[tmp])                                   \t\n"
974         "add.s   %[out3], %[in1],  %[in2]                               \t\n"
975         "add.s   %[t2],   %[in1],  %[in3]                               \t\n"
976         "sub.s   %[t3],   %[in1],  %[in2]                               \t\n"
977         "swc1    %[out2], 15*4(%[tmp])                                  \t\n"
978         "mul.s   %[t0],   %[in4],  %[c8]                                \t\n"
979         "sub.s   %[out3], %[out3], %[in3]                               \t\n"
980         "mul.s   %[t2],   %[t2],   %[c6]                                \t\n"
981         "mul.s   %[t3],   %[t3],   %[c7]                                \t\n"
982         "add.s   %[t1],   %[in2],  %[in3]                               \t\n"
983         "mul.s   %[out3], %[out3], %[c5]                                \t\n"
984         "add.s   %[out1], %[t2],   %[t3]                                \t\n"
985         "mul.s   %[t1],   %[t1],   %[c9]                                \t\n"
986         "swc1    %[out3], 5*4(%[tmp])                                   \t\n"
987         "add.s   %[out1], %[out1], %[t0]                                \t\n"
988         "add.s   %[out2], %[t2],   %[t1]                                \t\n"
989         "sub.s   %[out3], %[t3],   %[t1]                                \t\n"
990         "swc1    %[out1], 1*4(%[tmp])                                   \t\n"
991         "sub.s   %[out2], %[out2], %[t0]                                \t\n"
992         "sub.s   %[out3], %[out3], %[t0]                                \t\n"
993         "swc1    %[out2], 13*4(%[tmp])                                  \t\n"
994         "swc1    %[out3], 9*4(%[tmp])                                   \t\n"
995
996         : [t0] "=&f" (t0), [t1] "=&f" (t1),
997           [t2] "=&f" (t2), [t3] "=&f" (t3),
998           [in1] "=&f" (in1), [in2] "=&f" (in2),
999           [in3] "=&f" (in3), [in4] "=&f" (in4),
1000           [in5] "=&f" (in5),
1001           [out1] "=&f" (out1), [out2] "=&f" (out2),
1002           [out3] "=&f" (out3),
1003           [c1] "=&f" (c1), [c2] "=&f" (c2),
1004           [c3] "=&f" (c3), [c4] "=&f" (c4),
1005           [c5] "=&f" (c5), [c6] "=&f" (c6),
1006           [c7] "=&f" (c7), [c8] "=&f" (c8),
1007           [c9] "=&f" (c9)
1008         : [in] "r" (in), [tmp] "r" (tmp)
1009         : "memory"
1010     );
1011
1012     /* loop 4 */
1013     __asm__ volatile (
1014         "lwc1   %[in1],  2*4(%[tmp])                                    \t\n"
1015         "lwc1   %[in2],  0(%[tmp])                                      \t\n"
1016         "lwc1   %[in3],  3*4(%[tmp])                                    \t\n"
1017         "lwc1   %[in4],  1*4(%[tmp])                                    \t\n"
1018         "li.s   %[c1],   0.50190991877167369479                         \t\n"
1019         "li.s   %[c2],   5.73685662283492756461                         \t\n"
1020         "add.s  %[s0],   %[in1], %[in2]                                 \t\n"
1021         "sub.s  %[s2],   %[in1], %[in2]                                 \t\n"
1022         "add.s  %[s1],   %[in3], %[in4]                                 \t\n"
1023         "sub.s  %[s3],   %[in3], %[in4]                                 \t\n"
1024         "lwc1   %[in1],  9*4(%[win])                                    \t\n"
1025         "lwc1   %[in2],  4*9*4(%[buf])                                  \t\n"
1026         "lwc1   %[in3],  8*4(%[win])                                    \t\n"
1027         "mul.s  %[s1],   %[s1],  %[c1]                                  \t\n"
1028         "mul.s  %[s3],   %[s3],  %[c2]                                  \t\n"
1029         "lwc1   %[in4],  4*8*4(%[buf])                                  \t\n"
1030         "lwc1   %[in5],  29*4(%[win])                                   \t\n"
1031         "lwc1   %[in6],  28*4(%[win])                                   \t\n"
1032         "add.s  %[t0],   %[s0],  %[s1]                                  \t\n"
1033         "sub.s  %[t1],   %[s0],  %[s1]                                  \t\n"
1034         "li.s   %[c1],   0.51763809020504152469                         \t\n"
1035         "li.s   %[c2],   1.93185165257813657349                         \t\n"
1036         "mul.s  %[out3], %[in5], %[t0]                                  \t\n"
1037         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1038         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1039         "mul.s  %[out4], %[in6], %[t0]                                  \t\n"
1040         "add.s  %[t0],   %[s2],  %[s3]                                  \t\n"
1041         "swc1   %[out3], 4*9*4(%[buf])                                  \t\n"
1042         "swc1   %[out1], 288*4(%[out])                                  \t\n"
1043         "swc1   %[out2], 256*4(%[out])                                  \t\n"
1044         "swc1   %[out4], 4*8*4(%[buf])                                  \t\n"
1045         "sub.s  %[t1],   %[s2],  %[s3]                                  \t\n"
1046         "lwc1   %[in1],  17*4(%[win])                                   \t\n"
1047         "lwc1   %[in2],  4*17*4(%[buf])                                 \t\n"
1048         "lwc1   %[in3],  0(%[win])                                      \t\n"
1049         "lwc1   %[in4],  0(%[buf])                                      \t\n"
1050         "lwc1   %[in5],  37*4(%[win])                                   \t\n"
1051         "lwc1   %[in6],  20*4(%[win])                                   \t\n"
1052         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1053         "lwc1   %[in1],  6*4(%[tmp])                                    \t\n"
1054         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1055         "mul.s  %[out3], %[t0],  %[in5]                                 \t\n"
1056         "mul.s  %[out4], %[t0],  %[in6]                                 \t\n"
1057         "swc1   %[out1], 544*4(%[out])                                  \t\n"
1058         "lwc1   %[in2],  4*4(%[tmp])                                    \t\n"
1059         "swc1   %[out2], 0(%[out])                                      \t\n"
1060         "swc1   %[out3], 4*17*4(%[buf])                                 \t\n"
1061         "swc1   %[out4], 0(%[buf])                                      \t\n"
1062         "lwc1   %[in3],  7*4(%[tmp])                                    \t\n"
1063         "add.s  %[s0],   %[in1], %[in2]                                 \t\n"
1064         "sub.s  %[s2],   %[in1], %[in2]                                 \t\n"
1065         "lwc1   %[in4],  5*4(%[tmp])                                    \t\n"
1066         "add.s  %[s1],   %[in3], %[in4]                                 \t\n"
1067         "sub.s  %[s3],   %[in3], %[in4]                                 \t\n"
1068         "lwc1   %[in1],  10*4(%[win])                                   \t\n"
1069         "lwc1   %[in2],  4*10*4(%[buf])                                 \t\n"
1070         "lwc1   %[in3],  7*4(%[win])                                    \t\n"
1071         "mul.s  %[s1],   %[s1],  %[c1]                                  \t\n"
1072         "mul.s  %[s3],   %[s3],  %[c2]                                  \t\n"
1073         "add.s  %[t0],   %[s0],  %[s1]                                  \t\n"
1074         "sub.s  %[t1],   %[s0],  %[s1]                                  \t\n"
1075         "lwc1   %[in4],  4*7*4(%[buf])                                  \t\n"
1076         "lwc1   %[in5],  30*4(%[win])                                   \t\n"
1077         "lwc1   %[in6],  27*4(%[win])                                   \t\n"
1078         "li.s   %[c1],   0.55168895948124587824                         \t\n"
1079         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1080         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1081         "mul.s  %[out3], %[t0],  %[in5]                                 \t\n"
1082         "mul.s  %[out4], %[t0],  %[in6]                                 \t\n"
1083         "add.s  %[t0],   %[s2],  %[s3]                                  \t\n"
1084         "swc1   %[out1], 320*4(%[out])                                  \t\n"
1085         "swc1   %[out2], 224*4(%[out])                                  \t\n"
1086         "swc1   %[out3], 4*10*4(%[buf])                                 \t\n"
1087         "swc1   %[out4], 4*7*4(%[buf])                                  \t\n"
1088         "sub.s  %[t1],   %[s2],  %[s3]                                  \t\n"
1089         "lwc1   %[in1],  16*4(%[win])                                   \t\n"
1090         "lwc1   %[in2],  4*16*4(%[buf])                                 \t\n"
1091         "lwc1   %[in3],  1*4(%[win])                                    \t\n"
1092         "lwc1   %[in4],  4*1*4(%[buf])                                  \t\n"
1093         "lwc1   %[in5],  36*4(%[win])                                   \t\n"
1094         "lwc1   %[in6],  21*4(%[win])                                   \t\n"
1095         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1096         "lwc1   %[in1],  10*4(%[tmp])                                   \t\n"
1097         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1098         "mul.s  %[out3], %[in5], %[t0]                                  \t\n"
1099         "mul.s  %[out4], %[in6], %[t0]                                  \t\n"
1100         "swc1   %[out1], 512*4(%[out])                                  \t\n"
1101         "lwc1   %[in2],  8*4(%[tmp])                                    \t\n"
1102         "swc1   %[out2], 32*4(%[out])                                   \t\n"
1103         "swc1   %[out3], 4*16*4(%[buf])                                 \t\n"
1104         "swc1   %[out4], 4*1*4(%[buf])                                  \t\n"
1105         "li.s   %[c2],   1.18310079157624925896                         \t\n"
1106         "add.s  %[s0],   %[in1], %[in2]                                 \t\n"
1107         "sub.s  %[s2],   %[in1], %[in2]                                 \t\n"
1108         "lwc1   %[in3],  11*4(%[tmp])                                   \t\n"
1109         "lwc1   %[in4],  9*4(%[tmp])                                    \t\n"
1110         "add.s  %[s1],   %[in3], %[in4]                                 \t\n"
1111         "sub.s  %[s3],   %[in3], %[in4]                                 \t\n"
1112         "lwc1   %[in1],  11*4(%[win])                                   \t\n"
1113         "lwc1   %[in2],  4*11*4(%[buf])                                 \t\n"
1114         "lwc1   %[in3],  6*4(%[win])                                    \t\n"
1115         "mul.s  %[s1],   %[s1],  %[c1]                                  \t\n"
1116         "mul.s  %[s3],   %[s3],  %[c2]                                  \t\n"
1117         "lwc1   %[in4],  4*6*4(%[buf])                                  \t\n"
1118         "lwc1   %[in5],  31*4(%[win])                                   \t\n"
1119         "lwc1   %[in6],  26*4(%[win])                                   \t\n"
1120         "add.s  %[t0],   %[s0],  %[s1]                                  \t\n"
1121         "sub.s  %[t1],   %[s0],  %[s1]                                  \t\n"
1122         "mul.s  %[out3], %[t0],  %[in5]                                 \t\n"
1123         "mul.s  %[out4], %[t0],  %[in6]                                 \t\n"
1124         "add.s  %[t0],   %[s2],  %[s3]                                  \t\n"
1125         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1126         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1127         "swc1   %[out3], 4*11*4(%[buf])                                 \t\n"
1128         "swc1   %[out4], 4*6*4(%[buf])                                  \t\n"
1129         "sub.s  %[t1],   %[s2],  %[s3]                                  \t\n"
1130         "swc1   %[out1], 352*4(%[out])                                  \t\n"
1131         "swc1   %[out2], 192*4(%[out])                                  \t\n"
1132         "lwc1   %[in1],  15*4(%[win])                                   \t\n"
1133         "lwc1   %[in2],  4*15*4(%[buf])                                 \t\n"
1134         "lwc1   %[in3],  2*4(%[win])                                    \t\n"
1135         "lwc1   %[in4],  4*2*4(%[buf])                                  \t\n"
1136         "lwc1   %[in5],  35*4(%[win])                                   \t\n"
1137         "lwc1   %[in6],  22*4(%[win])                                   \t\n"
1138         "madd.s %[out1], %[in2], %[in1], %[t1]                          \t\n"
1139         "lwc1   %[in1],  14*4(%[tmp])                                   \t\n"
1140         "madd.s %[out2], %[in4], %[in3], %[t1]                          \t\n"
1141         "mul.s  %[out3], %[t0],  %[in5]                                 \t\n"
1142         "mul.s  %[out4], %[t0],  %[in6]                                 \t\n"
1143         "swc1   %[out1], 480*4(%[out])                                  \t\n"
1144         "lwc1   %[in2],  12*4(%[tmp])                                   \t\n"
1145         "swc1   %[out2], 64*4(%[out])                                   \t\n"
1146         "swc1   %[out3], 4*15*4(%[buf])                                 \t\n"
1147         "swc1   %[out4], 4*2*4(%[buf])                                  \t\n"
1148         "lwc1   %[in3],  15*4(%[tmp])                                   \t\n"
1149         "add.s  %[s0],   %[in1], %[in2]                                 \t\n"
1150         "sub.s  %[s2],   %[in1], %[in2]                                 \t\n"
1151         "lwc1   %[in4],  13*4(%[tmp])                                   \t\n"
1152         "li.s   %[c1],   0.61038729438072803416                         \t\n"
1153         "li.s   %[c2],   0.87172339781054900991                         \t\n"
1154         "add.s  %[s1],   %[in3], %[in4]                                 \t\n"
1155         "sub.s  %[s3],   %[in3], %[in4]                                 \t\n"
1156         "lwc1   %[in1],  12*4(%[win])                                   \t\n"
1157         "lwc1   %[in2],  4*12*4(%[buf])                                 \t\n"
1158         "lwc1   %[in3],  5*4(%[win])                                    \t\n"
1159         "mul.s  %[s1],   %[s1],  %[c1]                                  \t\n"
1160         "mul.s  %[s3],   %[s3],  %[c2]                                  \t\n"
1161         "lwc1   %[in4],  4*5*4(%[buf])                                  \t\n"
1162         "lwc1   %[in5],  32*4(%[win])                                   \t\n"
1163         "lwc1   %[in6],  25*4(%[win])                                   \t\n"
1164         "add.s  %[t0],   %[s0],  %[s1]                                  \t\n"
1165         "sub.s  %[t1],   %[s0],  %[s1]                                  \t\n"
1166         "lwc1   %[s0],   16*4(%[tmp])                                   \t\n"
1167         "lwc1   %[s1],   17*4(%[tmp])                                   \t\n"
1168         "li.s   %[c1],   0.70710678118654752439                         \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],  %[c1]                                  \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         : [c1] "=&f" (c1), [c2] "=&f" (c2),
1212           [in1] "=&f" (in1), [in2] "=&f" (in2),
1213           [in3] "=&f" (in3), [in4] "=&f" (in4),
1214           [in5] "=&f" (in5), [in6] "=&f" (in6),
1215           [out1] "=&f" (out1), [out2] "=&f" (out2),
1216           [out3] "=&f" (out3), [out4] "=&f" (out4),
1217           [t0] "=&f" (t0), [t1] "=&f" (t1),
1218           [t2] "=&f" (t2), [t3] "=&f" (t3),
1219           [s0] "=&f" (s0), [s1] "=&f" (s1),
1220           [s2] "=&f" (s2), [s3] "=&f" (s3)
1221         : [tmp] "r" (tmp), [win] "r" (win),
1222           [buf] "r" (buf), [out] "r" (out)
1223         : "memory"
1224     );
1225 }
1226
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 void ff_mpadsp_init_mipsfpu(MPADSPContext *s)
1247 {
1248     s->apply_window_float   = ff_mpadsp_apply_window_mips_float;
1249     s->imdct36_blocks_float = ff_imdct36_blocks_mips_float;
1250     s->dct32_float          = ff_dct32_mips_float;
1251 }