Merge commit '2caa93b813adc5dbb7771dfe615da826a2947d18'
[ffmpeg.git] / libavcodec / mips / mpegaudiodsp_mips_fixed.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 fixed-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  */
54
55 #include <string.h>
56
57 #include "libavutil/mips/asmdefs.h"
58 #include "libavcodec/mpegaudiodsp.h"
59
60 #if HAVE_INLINE_ASM
61 #if !HAVE_MIPS32R6 && !HAVE_MIPS64R6
62
63 static void ff_mpadsp_apply_window_mips_fixed(int32_t *synth_buf, int32_t *window,
64                                int *dither_state, int16_t *samples, ptrdiff_t incr)
65 {
66     register const int32_t *w, *w2, *p;
67     int j;
68     int16_t *samples2;
69     int w_asm, p_asm, w_asm1, p_asm1, w_asm2, p_asm2;
70     int w2_asm, w2_asm1, *p_temp1, *p_temp2;
71     int sum1 = 0;
72     int const min_asm = -32768, max_asm = 32767;
73     int temp1, temp2 = 0, temp3 = 0;
74     int64_t sum;
75
76     /* copy to avoid wrap */
77     memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));
78     samples2 = samples + 31 * incr;
79     w = window;
80     w2 = window + 31;
81     sum = *dither_state;
82     p = synth_buf + 16;
83     p_temp1 = synth_buf + 16;
84     p_temp2 = synth_buf + 48;
85     temp1 = sum;
86
87     /**
88     * use of round_sample function from the original code is eliminated,
89     * changed with appropriate assembly instructions.
90     */
91     __asm__ volatile (
92          "mthi   $zero                                                    \n\t"
93          "mtlo   %[temp1]                                                 \n\t"
94          "lw     %[w_asm],  0(%[w])                                       \n\t"
95          "lw     %[p_asm],  0(%[p])                                       \n\t"
96          "lw     %[w_asm1], 64*4(%[w])                                    \n\t"
97          "lw     %[p_asm1], 64*4(%[p])                                    \n\t"
98          "lw     %[w_asm2], 128*4(%[w])                                   \n\t"
99          "lw     %[p_asm2], 128*4(%[p])                                   \n\t"
100          "madd   %[w_asm],  %[p_asm]                                      \n\t"
101          "madd   %[w_asm1], %[p_asm1]                                     \n\t"
102          "madd   %[w_asm2], %[p_asm2]                                     \n\t"
103          "lw     %[w_asm],  192*4(%[w])                                   \n\t"
104          "lw     %[p_asm],  192*4(%[p])                                   \n\t"
105          "lw     %[w_asm1], 256*4(%[w])                                   \n\t"
106          "lw     %[p_asm1], 256*4(%[p])                                   \n\t"
107          "lw     %[w_asm2], 320*4(%[w])                                   \n\t"
108          "lw     %[p_asm2], 320*4(%[p])                                   \n\t"
109          "madd   %[w_asm],  %[p_asm]                                      \n\t"
110          "madd   %[w_asm1], %[p_asm1]                                     \n\t"
111          "madd   %[w_asm2], %[p_asm2]                                     \n\t"
112          "lw     %[w_asm],  384*4(%[w])                                   \n\t"
113          "lw     %[p_asm],  384*4(%[p])                                   \n\t"
114          "lw     %[w_asm1], 448*4(%[w])                                   \n\t"
115          "lw     %[p_asm1], 448*4(%[p])                                   \n\t"
116          "lw     %[w_asm2], 32*4(%[w])                                    \n\t"
117          "lw     %[p_asm2], 32*4(%[p])                                    \n\t"
118          "madd   %[w_asm],  %[p_asm]                                      \n\t"
119          "madd   %[w_asm1], %[p_asm1]                                     \n\t"
120          "msub   %[w_asm2], %[p_asm2]                                     \n\t"
121          "lw     %[w_asm],  96*4(%[w])                                    \n\t"
122          "lw     %[p_asm],  96*4(%[p])                                    \n\t"
123          "lw     %[w_asm1], 160*4(%[w])                                   \n\t"
124          "lw     %[p_asm1], 160*4(%[p])                                   \n\t"
125          "lw     %[w_asm2], 224*4(%[w])                                   \n\t"
126          "lw     %[p_asm2], 224*4(%[p])                                   \n\t"
127          "msub   %[w_asm],  %[p_asm]                                      \n\t"
128          "msub   %[w_asm1], %[p_asm1]                                     \n\t"
129          "msub   %[w_asm2], %[p_asm2]                                     \n\t"
130          "lw     %[w_asm],  288*4(%[w])                                   \n\t"
131          "lw     %[p_asm],  288*4(%[p])                                   \n\t"
132          "lw     %[w_asm1], 352*4(%[w])                                   \n\t"
133          "lw     %[p_asm1], 352*4(%[p])                                   \n\t"
134          "msub   %[w_asm],  %[p_asm]                                      \n\t"
135          "lw     %[w_asm],  480*4(%[w])                                   \n\t"
136          "lw     %[p_asm],  480*4(%[p])                                   \n\t"
137          "lw     %[w_asm2], 416*4(%[w])                                   \n\t"
138          "lw     %[p_asm2], 416*4(%[p])                                   \n\t"
139          "msub   %[w_asm],  %[p_asm]                                      \n\t"
140          "msub   %[w_asm1], %[p_asm1]                                     \n\t"
141          "msub   %[w_asm2], %[p_asm2]                                     \n\t"
142
143          /*round_sample function from the original code is eliminated,
144           * changed with appropriate assembly instructions
145           * code example:
146
147          "extr.w  %[sum1],$ac0,24                                       \n\t"
148          "mflo %[temp3],  $ac0                                          \n\t"
149          "and  %[temp1],  %[temp3],  0x00ffffff                         \n\t"
150          "slt  %[temp2],  %[sum1],   %[min_asm]                         \n\t"
151          "movn %[sum1],   %[min_asm],%[temp2]                           \n\t"
152          "slt  %[temp2],  %[max_asm],%[sum1]                            \n\t"
153          "movn %[sum1],   %[max_asm],%[temp2]                           \n\t"
154          "sh   %[sum1],   0(%[samples])                                 \n\t"
155          */
156
157          "extr.w %[sum1],   $ac0,       24                                \n\t"
158          "mflo   %[temp3]                                                 \n\t"
159          PTR_ADDIU "%[w],   %[w],       4                                 \n\t"
160          "and    %[temp1],  %[temp3],   0x00ffffff                        \n\t"
161          "slt    %[temp2],  %[sum1],    %[min_asm]                        \n\t"
162          "movn   %[sum1],   %[min_asm], %[temp2]                          \n\t"
163          "slt    %[temp2],  %[max_asm], %[sum1]                           \n\t"
164          "movn   %[sum1],   %[max_asm], %[temp2]                          \n\t"
165          "sh     %[sum1],   0(%[samples])                                 \n\t"
166
167         : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
168           [p_asm1] "=&r" (p_asm1), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
169           [w_asm2] "=&r" (w_asm2), [p_asm2] "=&r" (p_asm2),
170           [sum1] "+r" (sum1), [w] "+r" (w), [temp3] "+r" (temp3)
171         : [p] "r" (p), [samples] "r" (samples), [min_asm] "r" (min_asm),
172           [max_asm] "r" (max_asm)
173         : "memory", "hi","lo"
174      );
175
176      samples += incr;
177
178     /* we calculate two samples at the same time to avoid one memory
179        access per two sample */
180
181     for(j = 1; j < 16; j++) {
182         __asm__ volatile (
183              "mthi   $0,         $ac1                                      \n\t"
184              "mtlo   $0,         $ac1                                      \n\t"
185              "mthi   $0                                                    \n\t"
186              "mtlo   %[temp1]                                              \n\t"
187              PTR_ADDIU "%[p_temp1], %[p_temp1],    4                       \n\t"
188              "lw     %[w_asm],   0(%[w])                                   \n\t"
189              "lw     %[p_asm],   0(%[p_temp1])                             \n\t"
190              "lw     %[w2_asm],  0(%[w2])                                  \n\t"
191              "lw     %[w_asm1],  64*4(%[w])                                \n\t"
192              "lw     %[p_asm1],  64*4(%[p_temp1])                          \n\t"
193              "lw     %[w2_asm1], 64*4(%[w2])                               \n\t"
194              "madd   %[w_asm],   %[p_asm]                                  \n\t"
195              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
196              "madd   %[w_asm1],  %[p_asm1]                                 \n\t"
197              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
198              "lw     %[w_asm],   128*4(%[w])                               \n\t"
199              "lw     %[p_asm],   128*4(%[p_temp1])                         \n\t"
200              "lw     %[w2_asm],  128*4(%[w2])                              \n\t"
201              "lw     %[w_asm1],  192*4(%[w])                               \n\t"
202              "lw     %[p_asm1],  192*4(%[p_temp1])                         \n\t"
203              "lw     %[w2_asm1], 192*4(%[w2])                              \n\t"
204              "madd   %[w_asm],   %[p_asm]                                  \n\t"
205              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
206              "madd   %[w_asm1],  %[p_asm1]                                 \n\t"
207              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
208              "lw     %[w_asm],   256*4(%[w])                               \n\t"
209              "lw     %[p_asm],   256*4(%[p_temp1])                         \n\t"
210              "lw     %[w2_asm],  256*4(%[w2])                              \n\t"
211              "lw     %[w_asm1],  320*4(%[w])                               \n\t"
212              "lw     %[p_asm1],  320*4(%[p_temp1])                         \n\t"
213              "lw     %[w2_asm1], 320*4(%[w2])                              \n\t"
214              "madd   %[w_asm],   %[p_asm]                                  \n\t"
215              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
216              "madd   %[w_asm1],  %[p_asm1]                                 \n\t"
217              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
218              "lw     %[w_asm],   384*4(%[w])                               \n\t"
219              "lw     %[p_asm],   384*4(%[p_temp1])                         \n\t"
220              "lw     %[w2_asm],  384*4(%[w2])                              \n\t"
221              "lw     %[w_asm1],  448*4(%[w])                               \n\t"
222              "lw     %[p_asm1],  448*4(%[p_temp1])                         \n\t"
223              "lw     %[w2_asm1], 448*4(%[w2])                              \n\t"
224              "madd   %[w_asm],   %[p_asm]                                  \n\t"
225              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
226              "madd   %[w_asm1],  %[p_asm1]                                 \n\t"
227              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
228              PTR_ADDIU "%[p_temp2], %[p_temp2],   -4                      \n\t"
229              "lw     %[w_asm],   32*4(%[w])                                \n\t"
230              "lw     %[p_asm],   0(%[p_temp2])                             \n\t"
231              "lw     %[w2_asm],  32*4(%[w2])                               \n\t"
232              "lw     %[w_asm1],  96*4(%[w])                                \n\t"
233              "lw     %[p_asm1],  64*4(%[p_temp2])                          \n\t"
234              "lw     %[w2_asm1], 96*4(%[w2])                               \n\t"
235              "msub   %[w_asm],   %[p_asm]                                  \n\t"
236              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
237              "msub   %[w_asm1],  %[p_asm1]                                 \n\t"
238              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
239              "lw     %[w_asm],   160*4(%[w])                               \n\t"
240              "lw     %[p_asm],   128*4(%[p_temp2])                         \n\t"
241              "lw     %[w2_asm],  160*4(%[w2])                              \n\t"
242              "lw     %[w_asm1],  224*4(%[w])                               \n\t"
243              "lw     %[p_asm1],  192*4(%[p_temp2])                         \n\t"
244              "lw     %[w2_asm1], 224*4(%[w2])                              \n\t"
245              "msub   %[w_asm],   %[p_asm]                                  \n\t"
246              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
247              "msub   %[w_asm1],  %[p_asm1]                                 \n\t"
248              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
249              "lw     %[w_asm],   288*4(%[w])                               \n\t"
250              "lw     %[p_asm],   256*4(%[p_temp2])                         \n\t"
251              "lw     %[w2_asm],  288*4(%[w2])                              \n\t"
252              "lw     %[w_asm1],  352*4(%[w])                               \n\t"
253              "lw     %[p_asm1],  320*4(%[p_temp2])                         \n\t"
254              "lw     %[w2_asm1], 352*4(%[w2])                              \n\t"
255              "msub   %[w_asm],   %[p_asm]                                  \n\t"
256              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
257              "msub   %[w_asm1],  %[p_asm1]                                 \n\t"
258              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
259              "lw     %[w_asm],   416*4(%[w])                               \n\t"
260              "lw     %[p_asm],   384*4(%[p_temp2])                         \n\t"
261              "lw     %[w2_asm],  416*4(%[w2])                              \n\t"
262              "lw     %[w_asm1],  480*4(%[w])                               \n\t"
263              "lw     %[p_asm1],  448*4(%[p_temp2])                         \n\t"
264              "lw     %[w2_asm1], 480*4(%[w2])                              \n\t"
265              "msub   %[w_asm],   %[p_asm]                                  \n\t"
266              "msub   %[w_asm1],  %[p_asm1]                                 \n\t"
267              "msub   $ac1,       %[w2_asm],        %[p_asm]                \n\t"
268              "msub   $ac1,       %[w2_asm1],       %[p_asm1]               \n\t"
269              PTR_ADDIU "%[w],    %[w],             4                       \n\t"
270              PTR_ADDIU "%[w2],   %[w2],            -4                      \n\t"
271              "mflo   %[temp2]                                              \n\t"
272              "extr.w %[sum1],    $ac0,             24                      \n\t"
273              "li     %[temp3],   1                                         \n\t"
274              "and    %[temp1],   %[temp2],         0x00ffffff              \n\t"
275              "madd   $ac1,       %[temp1],         %[temp3]                \n\t"
276              "slt    %[temp2],   %[sum1],          %[min_asm]              \n\t"
277              "movn   %[sum1],    %[min_asm],       %[temp2]                \n\t"
278              "slt    %[temp2],   %[max_asm],       %[sum1]                 \n\t"
279              "movn   %[sum1],    %[max_asm],       %[temp2]                \n\t"
280              "sh     %[sum1],    0(%[samples])                             \n\t"
281              "mflo   %[temp3],   $ac1                                      \n\t"
282              "extr.w %[sum1],    $ac1,             24                      \n\t"
283              "and    %[temp1],   %[temp3],         0x00ffffff              \n\t"
284              "slt    %[temp2],   %[sum1],          %[min_asm]              \n\t"
285              "movn   %[sum1],    %[min_asm],       %[temp2]                \n\t"
286              "slt    %[temp2],   %[max_asm],       %[sum1]                 \n\t"
287              "movn   %[sum1],    %[max_asm],       %[temp2]                \n\t"
288              "sh     %[sum1],    0(%[samples2])                            \n\t"
289
290             : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
291               [p_asm1] "=&r" (p_asm1), [w2_asm1] "=&r" (w2_asm1),
292               [w2_asm] "=&r" (w2_asm), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
293               [p_temp1] "+r" (p_temp1), [p_temp2] "+r" (p_temp2), [sum1] "+r" (sum1),
294               [w] "+r" (w), [w2] "+r" (w2), [samples] "+r" (samples),
295               [samples2] "+r" (samples2), [temp3] "+r" (temp3)
296             : [min_asm] "r" (min_asm), [max_asm] "r" (max_asm)
297             : "memory", "hi", "lo", "$ac1hi", "$ac1lo"
298         );
299
300         samples += incr;
301         samples2 -= incr;
302     }
303
304     p = synth_buf + 32;
305
306     __asm__ volatile (
307         "mthi   $0                                                        \n\t"
308         "mtlo   %[temp1]                                                  \n\t"
309         "lw     %[w_asm],  32*4(%[w])                                     \n\t"
310         "lw     %[p_asm],  0(%[p])                                        \n\t"
311         "lw     %[w_asm1], 96*4(%[w])                                     \n\t"
312         "lw     %[p_asm1], 64*4(%[p])                                     \n\t"
313         "lw     %[w_asm2], 160*4(%[w])                                    \n\t"
314         "lw     %[p_asm2], 128*4(%[p])                                    \n\t"
315         "msub   %[w_asm],  %[p_asm]                                       \n\t"
316         "msub   %[w_asm1], %[p_asm1]                                      \n\t"
317         "msub   %[w_asm2], %[p_asm2]                                      \n\t"
318         "lw     %[w_asm],  224*4(%[w])                                    \n\t"
319         "lw     %[p_asm],  192*4(%[p])                                    \n\t"
320         "lw     %[w_asm1], 288*4(%[w])                                    \n\t"
321         "lw     %[p_asm1], 256*4(%[p])                                    \n\t"
322         "lw     %[w_asm2], 352*4(%[w])                                    \n\t"
323         "lw     %[p_asm2], 320*4(%[p])                                    \n\t"
324         "msub   %[w_asm],  %[p_asm]                                       \n\t"
325         "msub   %[w_asm1], %[p_asm1]                                      \n\t"
326         "msub   %[w_asm2], %[p_asm2]                                      \n\t"
327         "lw     %[w_asm],  416*4(%[w])                                    \n\t"
328         "lw     %[p_asm],  384*4(%[p])                                    \n\t"
329         "lw     %[w_asm1], 480*4(%[w])                                    \n\t"
330         "lw     %[p_asm1], 448*4(%[p])                                    \n\t"
331         "msub   %[w_asm],  %[p_asm]                                       \n\t"
332         "msub   %[w_asm1], %[p_asm1]                                      \n\t"
333         "extr.w %[sum1],   $ac0,       24                                 \n\t"
334         "mflo   %[temp2]                                                  \n\t"
335         "and    %[temp1],  %[temp2],   0x00ffffff                         \n\t"
336         "slt    %[temp2],  %[sum1],    %[min_asm]                         \n\t"
337         "movn   %[sum1],   %[min_asm], %[temp2]                           \n\t"
338         "slt    %[temp2],  %[max_asm], %[sum1]                            \n\t"
339         "movn   %[sum1],   %[max_asm], %[temp2]                           \n\t"
340         "sh     %[sum1],   0(%[samples])                                  \n\t"
341
342         : [w_asm] "=&r" (w_asm), [p_asm] "=&r" (p_asm), [w_asm1] "=&r" (w_asm1),
343           [p_asm1] "=&r" (p_asm1), [temp1] "+r" (temp1), [temp2] "+r" (temp2),
344           [w_asm2] "=&r" (w_asm2), [p_asm2] "=&r" (p_asm2), [sum1] "+r" (sum1)
345         : [w] "r" (w), [p] "r" (p), [samples] "r" (samples), [min_asm] "r" (min_asm),
346           [max_asm] "r" (max_asm)
347         : "memory", "hi", "lo", "$ac1hi", "$ac1lo"
348      );
349
350     *dither_state= temp1;
351 }
352
353 static void imdct36_mips_fixed(int *out, int *buf, int *in, int *win)
354 {
355     int j;
356     int t0, t1, t2, t3, s0, s1, s2, s3;
357     int tmp[18], *tmp1, *in1;
358     /* temporary variables */
359     int temp_reg1, temp_reg2, temp_reg3, temp_reg4, temp_reg5, temp_reg6;
360     int t4, t5, t6, t8, t7;
361
362    /* values defined in macros and tables are
363     * eliminated - they are directly loaded in appropriate variables
364     */
365     int const C_1  =  4229717092; /* cos(pi*1/18)*2  */
366     int const C_2  =  4035949074; /* cos(pi*2/18)*2  */
367     int const C_3  =  575416510;  /* -cos(pi*3/18)*2 */
368     int const C_3A =  3719550786; /* cos(pi*3/18)*2  */
369     int const C_4  =  1004831466; /* -cos(pi*4/18)*2 */
370     int const C_5  =  1534215534; /* -cos(pi*5/18)*2 */
371     int const C_7  = -1468965330; /* -cos(pi*7/18)*2 */
372     int const C_8  = -745813244;  /* -cos(pi*8/18)*2 */
373
374    /*
375     * instructions of the first two loops are reorganized and loops are unrolled,
376     * in order to eliminate unnecessary readings and writings in array
377     */
378
379     __asm__ volatile (
380         "lw   %[t1], 17*4(%[in])                                         \n\t"
381         "lw   %[t2], 16*4(%[in])                                         \n\t"
382         "lw   %[t3], 15*4(%[in])                                         \n\t"
383         "lw   %[t4], 14*4(%[in])                                         \n\t"
384         "addu %[t1], %[t1],      %[t2]                                   \n\t"
385         "addu %[t2], %[t2],      %[t3]                                   \n\t"
386         "addu %[t3], %[t3],      %[t4]                                   \n\t"
387         "lw   %[t5], 13*4(%[in])                                         \n\t"
388         "addu %[t1], %[t1],      %[t3]                                   \n\t"
389         "sw   %[t2], 16*4(%[in])                                         \n\t"
390         "lw   %[t6], 12*4(%[in])                                         \n\t"
391         "sw   %[t1], 17*4(%[in])                                         \n\t"
392         "addu %[t4], %[t4],      %[t5]                                   \n\t"
393         "addu %[t5], %[t5],      %[t6]                                   \n\t"
394         "lw   %[t7], 11*4(%[in])                                         \n\t"
395         "addu %[t3], %[t3],      %[t5]                                   \n\t"
396         "sw   %[t4], 14*4(%[in])                                         \n\t"
397         "lw   %[t8], 10*4(%[in])                                         \n\t"
398         "sw   %[t3], 15*4(%[in])                                         \n\t"
399         "addu %[t6], %[t6],      %[t7]                                   \n\t"
400         "addu %[t7], %[t7],      %[t8]                                   \n\t"
401         "sw   %[t6], 12*4(%[in])                                         \n\t"
402         "addu %[t5], %[t5],      %[t7]                                   \n\t"
403         "lw   %[t1], 9*4(%[in])                                          \n\t"
404         "lw   %[t2], 8*4(%[in])                                          \n\t"
405         "sw   %[t5], 13*4(%[in])                                         \n\t"
406         "addu %[t8], %[t8],      %[t1]                                   \n\t"
407         "addu %[t1], %[t1],      %[t2]                                   \n\t"
408         "sw   %[t8], 10*4(%[in])                                         \n\t"
409         "addu %[t7], %[t7],      %[t1]                                   \n\t"
410         "lw   %[t3], 7*4(%[in])                                          \n\t"
411         "lw   %[t4], 6*4(%[in])                                          \n\t"
412         "sw   %[t7], 11*4(%[in])                                         \n\t"
413         "addu %[t2], %[t2],      %[t3]                                   \n\t"
414         "addu %[t3], %[t3],      %[t4]                                   \n\t"
415         "sw   %[t2], 8*4(%[in])                                          \n\t"
416         "addu %[t1], %[t1],      %[t3]                                   \n\t"
417         "lw   %[t5], 5*4(%[in])                                          \n\t"
418         "lw   %[t6], 4*4(%[in])                                          \n\t"
419         "sw   %[t1], 9*4(%[in])                                          \n\t"
420         "addu %[t4], %[t4],      %[t5]                                   \n\t"
421         "addu %[t5], %[t5],      %[t6]                                   \n\t"
422         "sw   %[t4], 6*4(%[in])                                          \n\t"
423         "addu %[t3], %[t3],      %[t5]                                   \n\t"
424         "lw   %[t7], 3*4(%[in])                                          \n\t"
425         "lw   %[t8], 2*4(%[in])                                          \n\t"
426         "sw   %[t3], 7*4(%[in])                                          \n\t"
427         "addu %[t6], %[t6],      %[t7]                                   \n\t"
428         "addu %[t7], %[t7],      %[t8]                                   \n\t"
429         "sw   %[t6], 4*4(%[in])                                          \n\t"
430         "addu %[t5], %[t5],      %[t7]                                   \n\t"
431         "lw   %[t1], 1*4(%[in])                                          \n\t"
432         "lw   %[t2], 0*4(%[in])                                          \n\t"
433         "sw   %[t5], 5*4(%[in])                                          \n\t"
434         "addu %[t8], %[t8],      %[t1]                                   \n\t"
435         "addu %[t1], %[t1],      %[t2]                                   \n\t"
436         "sw   %[t8], 2*4(%[in])                                          \n\t"
437         "addu %[t7], %[t7],      %[t1]                                   \n\t"
438         "sw   %[t7], 3*4(%[in])                                          \n\t"
439         "sw   %[t1], 1*4(%[in])                                          \n\t"
440
441         : [in] "+r" (in), [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3),
442           [t4] "=&r" (t4), [t5] "=&r" (t5), [t6] "=&r" (t6),
443           [t7] "=&r" (t7), [t8] "=&r" (t8)
444         :
445         : "memory"
446     );
447
448     for(j = 0; j < 2; j++) {
449
450         tmp1 = tmp + j;
451         in1 = in + j;
452
453          /**
454          *  Original constants are multiplied by two in advanced
455          *  for assembly optimization (e.g. C_2 = 2 * C2).
456          *  That can lead to overflow in operations where they are used.
457          *
458          *  Example of the solution:
459          *
460          *  in original code:
461          *  t0 = ((int64_t)(in1[2*2] + in1[2*4]) * (int64_t)(2*C2))>>32
462          *
463          *  in assembly:
464          *  C_2 = 2 * C2;
465          *   .
466          *   .
467          *  "lw   %[t7],       4*4(%[in1])                               \n\t"
468          *  "lw   %[t8],       8*4(%[in1])                               \n\t"
469          *  "addu %[temp_reg2],%[t7],       %[t8]                        \n\t"
470          *  "multu %[C_2],     %[temp_reg2]                              \n\t"
471          *  "mfhi %[temp_reg1]                                           \n\t"
472          *  "sra  %[temp_reg2],%[temp_reg2],31                           \n\t"
473          *  "move %[t0],       $0                                        \n\t"
474          *  "movn %[t0],       %[C_2],      %[temp_reg2]                 \n\t"
475          *  "sub  %[t0],       %[temp_reg1],%[t0]                        \n\t"
476          */
477
478         __asm__ volatile (
479             "lw    %[t7],        4*4(%[in1])                               \n\t"
480             "lw    %[t8],        8*4(%[in1])                               \n\t"
481             "lw    %[t6],        16*4(%[in1])                              \n\t"
482             "lw    %[t4],        0*4(%[in1])                               \n\t"
483             "addu  %[temp_reg2], %[t7],        %[t8]                       \n\t"
484             "addu  %[t2],        %[t6],        %[t8]                       \n\t"
485             "multu %[C_2],       %[temp_reg2]                              \n\t"
486             "lw    %[t5],        12*4(%[in1])                              \n\t"
487             "sub   %[t2],        %[t2],        %[t7]                       \n\t"
488             "sub   %[t1],        %[t4],        %[t5]                       \n\t"
489             "sra   %[t3],        %[t5],        1                           \n\t"
490             "sra   %[temp_reg1], %[t2],        1                           \n\t"
491             "addu  %[t3],        %[t3],        %[t4]                       \n\t"
492             "sub   %[temp_reg1], %[t1],        %[temp_reg1]                \n\t"
493             "sra   %[temp_reg2], %[temp_reg2], 31                          \n\t"
494             "sw    %[temp_reg1], 6*4(%[tmp1])                              \n\t"
495             "move  %[t0],        $0                                        \n\t"
496             "movn  %[t0],        %[C_2],       %[temp_reg2]                \n\t"
497             "mfhi  %[temp_reg1]                                            \n\t"
498             "addu  %[t1],        %[t1],        %[t2]                       \n\t"
499             "sw    %[t1],        16*4(%[tmp1])                             \n\t"
500             "sub   %[temp_reg4], %[t8],        %[t6]                       \n\t"
501             "add   %[temp_reg2], %[t7],        %[t6]                       \n\t"
502             "mult  $ac1,         %[C_8],       %[temp_reg4]                \n\t"
503             "multu $ac2,         %[C_4],       %[temp_reg2]                \n\t"
504             "sub   %[t0],        %[temp_reg1], %[t0]                       \n\t"
505             "sra   %[temp_reg1], %[temp_reg2], 31                          \n\t"
506             "move  %[t2],        $0                                        \n\t"
507             "movn  %[t2],        %[C_4],       %[temp_reg1]                \n\t"
508             "mfhi  %[t1],        $ac1                                      \n\t"
509             "mfhi  %[temp_reg1], $ac2                                      \n\t"
510             "lw    %[t6],        10*4(%[in1])                              \n\t"
511             "lw    %[t8],        14*4(%[in1])                              \n\t"
512             "lw    %[t7],        2*4(%[in1])                               \n\t"
513             "lw    %[t4],        6*4(%[in1])                               \n\t"
514             "sub   %[temp_reg3], %[t3],        %[t0]                       \n\t"
515             "add   %[temp_reg4], %[t3],        %[t0]                       \n\t"
516             "sub   %[temp_reg1], %[temp_reg1], %[temp_reg2]                \n\t"
517             "add   %[temp_reg4], %[temp_reg4], %[t1]                       \n\t"
518             "sub   %[t2],        %[temp_reg1], %[t2]                       \n\t"
519             "sw    %[temp_reg4], 2*4(%[tmp1])                              \n\t"
520             "sub   %[temp_reg3], %[temp_reg3], %[t2]                       \n\t"
521             "add   %[temp_reg1], %[t3],        %[t2]                       \n\t"
522             "sw    %[temp_reg3], 10*4(%[tmp1])                             \n\t"
523             "sub   %[temp_reg1], %[temp_reg1], %[t1]                       \n\t"
524             "addu  %[temp_reg2], %[t6],        %[t8]                       \n\t"
525             "sw    %[temp_reg1], 14*4(%[tmp1])                             \n\t"
526             "sub   %[temp_reg2], %[temp_reg2], %[t7]                       \n\t"
527             "addu  %[temp_reg3], %[t7],        %[t6]                       \n\t"
528             "multu $ac3,         %[C_3],       %[temp_reg2]                \n\t"
529             "multu %[C_1],       %[temp_reg3]                              \n\t"
530             "sra   %[temp_reg1], %[temp_reg2], 31                          \n\t"
531             "move  %[t1],        $0                                        \n\t"
532             "sra   %[temp_reg3], %[temp_reg3], 31                          \n\t"
533             "movn  %[t1],        %[C_3],       %[temp_reg1]                \n\t"
534             "mfhi  %[temp_reg1], $ac3                                      \n\t"
535             "mfhi  %[temp_reg4]                                            \n\t"
536             "move  %[t2],        $0                                        \n\t"
537             "movn  %[t2],        %[C_1],       %[temp_reg3]                \n\t"
538             "sub   %[temp_reg3], %[t6],        %[t8]                       \n\t"
539             "sub   %[t2],        %[temp_reg4], %[t2]                       \n\t"
540             "multu $ac1,         %[C_7],       %[temp_reg3]                \n\t"
541             "sub   %[temp_reg1], %[temp_reg1], %[temp_reg2]                \n\t"
542             "sra   %[temp_reg4], %[temp_reg3], 31                          \n\t"
543             "sub   %[t1],        %[temp_reg1], %[t1]                       \n\t"
544             "move  %[t3],        $0                                        \n\t"
545             "sw    %[t1],        4*4(%[tmp1])                              \n\t"
546             "movn  %[t3],        %[C_7],       %[temp_reg4]                \n\t"
547             "multu $ac2,         %[C_3A],      %[t4]                       \n\t"
548             "add   %[temp_reg2], %[t7],        %[t8]                       \n\t"
549             "move  %[t1],        $0                                        \n\t"
550             "mfhi  %[temp_reg4], $ac1                                      \n\t"
551             "multu $ac3,%[C_5],  %[temp_reg2]                              \n\t"
552             "move  %[t0],        $0                                        \n\t"
553             "sra   %[temp_reg1], %[temp_reg2], 31                          \n\t"
554             "movn  %[t1],%[C_5], %[temp_reg1]                              \n\t"
555             "sub   %[temp_reg4], %[temp_reg4], %[temp_reg3]                \n\t"
556             "mfhi  %[temp_reg1], $ac3                                      \n\t"
557             "sra   %[temp_reg3], %[t4],        31                          \n\t"
558             "movn  %[t0],        %[C_3A],      %[temp_reg3]                \n\t"
559             "mfhi  %[temp_reg3], $ac2                                      \n\t"
560             "sub   %[t3],        %[temp_reg4], %[t3]                       \n\t"
561             "add   %[temp_reg4], %[t3],        %[t2]                       \n\t"
562             "sub   %[temp_reg1], %[temp_reg1], %[temp_reg2]                \n\t"
563             "sub   %[t1],        %[temp_reg1], %[t1]                       \n\t"
564             "sub   %[t0],        %[temp_reg3], %[t0]                       \n\t"
565             "add   %[temp_reg1], %[t2],        %[t1]                       \n\t"
566             "add   %[temp_reg4], %[temp_reg4], %[t0]                       \n\t"
567             "sub   %[temp_reg2], %[t3],        %[t1]                       \n\t"
568             "sw    %[temp_reg4], 0*4(%[tmp1])                              \n\t"
569             "sub   %[temp_reg1], %[temp_reg1], %[t0]                       \n\t"
570             "sub   %[temp_reg2], %[temp_reg2], %[t0]                       \n\t"
571             "sw    %[temp_reg1], 12*4(%[tmp1])                             \n\t"
572             "sw    %[temp_reg2], 8*4(%[tmp1])                              \n\t"
573
574             : [t7] "=&r" (t7), [temp_reg1] "=&r" (temp_reg1),
575               [temp_reg2] "=&r" (temp_reg2), [temp_reg4] "=&r" (temp_reg4),
576               [temp_reg3] "=&r" (temp_reg3), [t8] "=&r" (t8), [t0] "=&r" (t0),
577               [t4] "=&r" (t4), [t5] "=&r" (t5), [t6] "=&r"(t6), [t2] "=&r" (t2),
578               [t3] "=&r" (t3), [t1] "=&r" (t1)
579             : [C_2] "r" (C_2), [in1] "r" (in1), [tmp1] "r" (tmp1), [C_8] "r" (C_8),
580               [C_4] "r" (C_4), [C_3] "r" (C_3), [C_1] "r" (C_1), [C_7] "r" (C_7),
581               [C_3A] "r" (C_3A), [C_5] "r" (C_5)
582             : "memory", "hi", "lo", "$ac1hi", "$ac1lo", "$ac2hi", "$ac2lo",
583               "$ac3hi", "$ac3lo"
584          );
585     }
586
587     /**
588     * loop is unrolled four times
589     *
590     * values defined in tables(icos36[] and icos36h[]) are not loaded from
591     * these tables - they are directly loaded in appropriate registers
592     *
593     */
594
595     __asm__ volatile (
596         "lw     %[t2],        1*4(%[tmp])                                  \n\t"
597         "lw     %[t3],        3*4(%[tmp])                                  \n\t"
598         "lw     %[t0],        0*4(%[tmp])                                  \n\t"
599         "lw     %[t1],        2*4(%[tmp])                                  \n\t"
600         "addu   %[temp_reg1], %[t3],        %[t2]                          \n\t"
601         "li     %[temp_reg2], 0x807D2B1E                                   \n\t"
602         "move   %[s1],        $0                                           \n\t"
603         "multu  %[temp_reg2], %[temp_reg1]                                 \n\t"
604         "sra    %[temp_reg1], %[temp_reg1], 31                             \n\t"
605         "movn   %[s1],        %[temp_reg2], %[temp_reg1]                   \n\t"
606         "sub    %[temp_reg3], %[t3],        %[t2]                          \n\t"
607         "li     %[temp_reg4], 0x2de5151                                    \n\t"
608         "mfhi   %[temp_reg2]                                               \n\t"
609         "addu   %[s0],        %[t1],        %[t0]                          \n\t"
610         "lw     %[temp_reg5], 9*4(%[win])                                  \n\t"
611         "mult   $ac1,         %[temp_reg4], %[temp_reg3]                   \n\t"
612         "lw     %[temp_reg6], 4*9*4(%[buf])                                \n\t"
613         "sub    %[s2],        %[t1],        %[t0]                          \n\t"
614         "lw     %[temp_reg3], 29*4(%[win])                                 \n\t"
615         "subu   %[s1],        %[temp_reg2], %[s1]                          \n\t"
616         "lw     %[temp_reg4], 28*4(%[win])                                 \n\t"
617         "add    %[t0],        %[s0],        %[s1]                          \n\t"
618         "extr.w %[s3],        $ac1,23                                      \n\t"
619         "mult   $ac2,         %[t0],        %[temp_reg3]                   \n\t"
620         "sub    %[t1],        %[s0],        %[s1]                          \n\t"
621         "lw     %[temp_reg1], 4*8*4(%[buf])                                \n\t"
622         "mult   %[t1],        %[temp_reg5]                                 \n\t"
623         "lw     %[temp_reg2], 8*4(%[win])                                  \n\t"
624         "mfhi   %[temp_reg3], $ac2                                         \n\t"
625         "mult   $ac3,         %[t0],        %[temp_reg4]                   \n\t"
626         "add    %[t0],        %[s2],        %[s3]                          \n\t"
627         "mfhi   %[temp_reg5]                                               \n\t"
628         "mult   $ac1,         %[t1],        %[temp_reg2]                   \n\t"
629         "sub    %[t1],        %[s2],        %[s3]                          \n\t"
630         "sw     %[temp_reg3], 4*9*4(%[buf])                                \n\t"
631         "mfhi   %[temp_reg4], $ac3                                         \n\t"
632         "lw     %[temp_reg3], 37*4(%[win])                                 \n\t"
633         "mfhi   %[temp_reg2], $ac1                                         \n\t"
634         "add    %[temp_reg5], %[temp_reg5], %[temp_reg6]                   \n\t"
635         "lw     %[temp_reg6], 17*4(%[win])                                 \n\t"
636         "sw     %[temp_reg5], 32*9*4(%[out])                               \n\t"
637         "sw     %[temp_reg4], 4*8*4(%[buf])                                \n\t"
638         "mult   %[t1],        %[temp_reg6]                                 \n\t"
639         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
640         "lw     %[temp_reg2], 0*4(%[win])                                  \n\t"
641         "lw     %[temp_reg5], 4*17*4(%[buf])                               \n\t"
642         "sw     %[temp_reg1], 8*32*4(%[out])                               \n\t"
643         "mfhi   %[temp_reg6]                                               \n\t"
644         "mult   $ac1,         %[t1],        %[temp_reg2]                   \n\t"
645         "lw     %[temp_reg4], 20*4(%[win])                                 \n\t"
646         "lw     %[temp_reg1], 0(%[buf])                                    \n\t"
647         "mult   $ac2,         %[t0],        %[temp_reg3]                   \n\t"
648         "mult   %[t0],        %[temp_reg4]                                 \n\t"
649         "mfhi   %[temp_reg2], $ac1                                         \n\t"
650         "lw     %[t0],        4*4(%[tmp])                                  \n\t"
651         "add    %[temp_reg5], %[temp_reg5], %[temp_reg6]                   \n\t"
652         "mfhi   %[temp_reg3], $ac2                                         \n\t"
653         "mfhi   %[temp_reg4]                                               \n\t"
654         "sw     %[temp_reg5], 17*32*4(%[out])                              \n\t"
655         "lw     %[t1],        6*4(%[tmp])                                  \n\t"
656         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
657         "lw     %[t2],        5*4(%[tmp])                                  \n\t"
658         "sw     %[temp_reg1], 0*32*4(%[out])                               \n\t"
659         "addu   %[s0],        %[t1],        %[t0]                          \n\t"
660         "sw     %[temp_reg3], 4*17*4(%[buf])                               \n\t"
661         "lw     %[t3],        7*4(%[tmp])                                  \n\t"
662         "sub    %[s2],        %[t1],        %[t0]                          \n\t"
663         "sw     %[temp_reg4], 0(%[buf])                                    \n\t"
664         "addu   %[temp_reg5], %[t3],        %[t2]                          \n\t"
665         "li     %[temp_reg6], 0x8483EE0C                                   \n\t"
666         "move   %[s1],        $0                                           \n\t"
667         "multu  %[temp_reg6], %[temp_reg5]                                 \n\t"
668         "sub    %[temp_reg1], %[t3],        %[t2]                          \n\t"
669         "li     %[temp_reg2], 0xf746ea                                     \n\t"
670         "sra    %[temp_reg5], %[temp_reg5], 31                             \n\t"
671         "mult   $ac1,         %[temp_reg2], %[temp_reg1]                   \n\t"
672         "movn   %[s1],        %[temp_reg6], %[temp_reg5]                   \n\t"
673         "mfhi   %[temp_reg5]                                               \n\t"
674         "lw     %[temp_reg3], 10*4(%[win])                                 \n\t"
675         "lw     %[temp_reg4], 4*10*4(%[buf])                               \n\t"
676         "extr.w %[s3],        $ac1,         23                             \n\t"
677         "lw     %[temp_reg1], 4*7*4(%[buf])                                \n\t"
678         "lw     %[temp_reg2], 7*4(%[win])                                  \n\t"
679         "lw     %[temp_reg6], 30*4(%[win])                                 \n\t"
680         "subu   %[s1],        %[temp_reg5], %[s1]                          \n\t"
681         "sub    %[t1],        %[s0],        %[s1]                          \n\t"
682         "add    %[t0],        %[s0],        %[s1]                          \n\t"
683         "mult   $ac2,         %[t1],        %[temp_reg3]                   \n\t"
684         "mult   $ac3,         %[t1],        %[temp_reg2]                   \n\t"
685         "mult   %[t0],        %[temp_reg6]                                 \n\t"
686         "lw     %[temp_reg5], 27*4(%[win])                                 \n\t"
687         "mult   $ac1,         %[t0],        %[temp_reg5]                   \n\t"
688         "mfhi   %[temp_reg3], $ac2                                         \n\t"
689         "mfhi   %[temp_reg2], $ac3                                         \n\t"
690         "mfhi   %[temp_reg6]                                               \n\t"
691         "add    %[t0],        %[s2],        %[s3]                          \n\t"
692         "sub    %[t1],        %[s2],        %[s3]                          \n\t"
693         "add    %[temp_reg3], %[temp_reg3], %[temp_reg4]                   \n\t"
694         "lw     %[temp_reg4], 16*4(%[win])                                 \n\t"
695         "mfhi   %[temp_reg5], $ac1                                         \n\t"
696         "sw     %[temp_reg3], 32*10*4(%[out])                              \n\t"
697         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
698         "lw     %[temp_reg3], 4*16*4(%[buf])                               \n\t"
699         "sw     %[temp_reg6], 4*10*4(%[buf])                               \n\t"
700         "sw     %[temp_reg1], 7*32*4(%[out])                               \n\t"
701         "mult   $ac2,         %[t1],        %[temp_reg4]                   \n\t"
702         "sw     %[temp_reg5], 4*7*4(%[buf])                                \n\t"
703         "lw     %[temp_reg6], 1*4(%[win])                                  \n\t"
704         "lw     %[temp_reg5], 4*1*4(%[buf])                                \n\t"
705         "lw     %[temp_reg1], 36*4(%[win])                                 \n\t"
706         "mult   $ac3,         %[t1],        %[temp_reg6]                   \n\t"
707         "lw     %[temp_reg2], 21*4(%[win])                                 \n\t"
708         "mfhi   %[temp_reg4], $ac2                                         \n\t"
709         "mult   %[t0],        %[temp_reg1]                                 \n\t"
710         "mult   $ac1,         %[t0],%[temp_reg2]                           \n\t"
711         "lw     %[t0],        8*4(%[tmp])                                  \n\t"
712         "mfhi   %[temp_reg6], $ac3                                         \n\t"
713         "lw     %[t1],        10*4(%[tmp])                                 \n\t"
714         "lw     %[t3],        11*4(%[tmp])                                 \n\t"
715         "mfhi   %[temp_reg1]                                               \n\t"
716         "add    %[temp_reg3], %[temp_reg3], %[temp_reg4]                   \n\t"
717         "lw     %[t2],        9*4(%[tmp])                                  \n\t"
718         "mfhi   %[temp_reg2], $ac1                                         \n\t"
719         "add    %[temp_reg5], %[temp_reg5], %[temp_reg6]                   \n\t"
720         "sw     %[temp_reg3], 16*32*4(%[out])                              \n\t"
721         "sw     %[temp_reg5], 1*32*4(%[out])                               \n\t"
722         "sw     %[temp_reg1], 4*16*4(%[buf])                               \n\t"
723         "addu   %[temp_reg3], %[t3],        %[t2]                          \n\t"
724         "li     %[temp_reg4], 0x8D3B7CD6                                   \n\t"
725         "sw     %[temp_reg2], 4*1*4(%[buf])                                \n\t"
726         "multu  %[temp_reg4],%[temp_reg3]                                  \n\t"
727         "sra    %[temp_reg3], %[temp_reg3], 31                             \n\t"
728         "move   %[s1],        $0                                           \n\t"
729         "movn   %[s1],        %[temp_reg4], %[temp_reg3]                   \n\t"
730         "addu   %[s0],        %[t1],        %[t0]                          \n\t"
731         "mfhi   %[temp_reg3]                                               \n\t"
732         "sub    %[s2],        %[t1],        %[t0]                          \n\t"
733         "sub    %[temp_reg5], %[t3],        %[t2]                          \n\t"
734         "li     %[temp_reg6], 0x976fd9                                     \n\t"
735         "lw     %[temp_reg2], 11*4(%[win])                                 \n\t"
736         "lw     %[temp_reg1], 4*11*4(%[buf])                               \n\t"
737         "mult   $ac1,         %[temp_reg6], %[temp_reg5]                   \n\t"
738         "subu   %[s1],        %[temp_reg3], %[s1]                          \n\t"
739         "lw     %[temp_reg5], 31*4(%[win])                                 \n\t"
740         "sub    %[t1],        %[s0],        %[s1]                          \n\t"
741         "add    %[t0],        %[s0],        %[s1]                          \n\t"
742         "mult   $ac2,         %[t1],        %[temp_reg2]                   \n\t"
743         "mult   %[t0],        %[temp_reg5]                                 \n\t"
744         "lw     %[temp_reg4], 6*4(%[win])                                  \n\t"
745         "extr.w %[s3],        $ac1,         23                             \n\t"
746         "lw     %[temp_reg3], 4*6*4(%[buf])                                \n\t"
747         "mfhi   %[temp_reg2], $ac2                                         \n\t"
748         "lw     %[temp_reg6], 26*4(%[win])                                 \n\t"
749         "mfhi   %[temp_reg5]                                               \n\t"
750         "mult   $ac3,         %[t1],        %[temp_reg4]                   \n\t"
751         "mult   $ac1,         %[t0],        %[temp_reg6]                   \n\t"
752         "add    %[t0],        %[s2],        %[s3]                          \n\t"
753         "sub    %[t1],        %[s2],        %[s3]                          \n\t"
754         "add    %[temp_reg2], %[temp_reg2], %[temp_reg1]                   \n\t"
755         "mfhi   %[temp_reg4], $ac3                                         \n\t"
756         "mfhi   %[temp_reg6], $ac1                                         \n\t"
757         "sw     %[temp_reg5], 4*11*4(%[buf])                               \n\t"
758         "sw     %[temp_reg2], 32*11*4(%[out])                              \n\t"
759         "lw     %[temp_reg1], 4*15*4(%[buf])                               \n\t"
760         "add    %[temp_reg3], %[temp_reg3], %[temp_reg4]                   \n\t"
761         "lw     %[temp_reg2], 15*4(%[win])                                 \n\t"
762         "sw     %[temp_reg3], 6*32*4(%[out])                               \n\t"
763         "sw     %[temp_reg6], 4*6*4(%[buf])                                \n\t"
764         "mult   %[t1],        %[temp_reg2]                                 \n\t"
765         "lw     %[temp_reg3], 2*4(%[win])                                  \n\t"
766         "lw     %[temp_reg4], 4*2*4(%[buf])                                \n\t"
767         "lw     %[temp_reg5], 35*4(%[win])                                 \n\t"
768         "mult   $ac1,         %[t1],        %[temp_reg3]                   \n\t"
769         "mfhi   %[temp_reg2]                                               \n\t"
770         "lw     %[temp_reg6], 22*4(%[win])                                 \n\t"
771         "mult   $ac2,         %[t0],        %[temp_reg5]                   \n\t"
772         "lw     %[t1],        14*4(%[tmp])                                 \n\t"
773         "mult   $ac3,         %[t0],        %[temp_reg6]                   \n\t"
774         "lw     %[t0],        12*4(%[tmp])                                 \n\t"
775         "mfhi   %[temp_reg3], $ac1                                         \n\t"
776         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
777         "mfhi   %[temp_reg5], $ac2                                         \n\t"
778         "sw     %[temp_reg1], 15*32*4(%[out])                              \n\t"
779         "mfhi   %[temp_reg6], $ac3                                         \n\t"
780         "lw     %[t2],        13*4(%[tmp])                                 \n\t"
781         "lw     %[t3],        15*4(%[tmp])                                 \n\t"
782         "add    %[temp_reg4], %[temp_reg4], %[temp_reg3]                   \n\t"
783         "sw     %[temp_reg5], 4*15*4(%[buf])                               \n\t"
784         "addu   %[temp_reg1], %[t3],        %[t2]                          \n\t"
785         "li     %[temp_reg2], 0x9C42577C                                   \n\t"
786         "move   %[s1],        $0                                           \n\t"
787         "multu  %[temp_reg2], %[temp_reg1]                                 \n\t"
788         "sw     %[temp_reg4], 2*32*4(%[out])                               \n\t"
789         "sra    %[temp_reg1], %[temp_reg1], 31                             \n\t"
790         "movn   %[s1],        %[temp_reg2], %[temp_reg1]                   \n\t"
791         "sub    %[temp_reg3], %[t3],        %[t2]                          \n\t"
792         "li     %[temp_reg4], 0x6f94a2                                     \n\t"
793         "mfhi   %[temp_reg1]                                               \n\t"
794         "addu   %[s0],        %[t1],        %[t0]                          \n\t"
795         "sw     %[temp_reg6], 4*2*4(%[buf])                                \n\t"
796         "mult   $ac1,         %[temp_reg4], %[temp_reg3]                   \n\t"
797         "sub    %[s2],        %[t1],        %[t0]                          \n\t"
798         "lw     %[temp_reg5], 12*4(%[win])                                 \n\t"
799         "lw     %[temp_reg6], 4*12*4(%[buf])                               \n\t"
800         "subu   %[s1],        %[temp_reg1], %[s1]                          \n\t"
801         "sub    %[t1],        %[s0],        %[s1]                          \n\t"
802         "lw     %[temp_reg3], 32*4(%[win])                                 \n\t"
803         "mult   $ac2,         %[t1],        %[temp_reg5]                   \n\t"
804         "add    %[t0],        %[s0],        %[s1]                          \n\t"
805         "extr.w %[s3],        $ac1,         23                             \n\t"
806         "lw     %[temp_reg2], 5*4(%[win])                                  \n\t"
807         "mult   %[t0],        %[temp_reg3]                                 \n\t"
808         "mfhi   %[temp_reg5], $ac2                                         \n\t"
809         "lw     %[temp_reg4], 25*4(%[win])                                 \n\t"
810         "lw     %[temp_reg1], 4*5*4(%[buf])                                \n\t"
811         "mult   $ac3,         %[t1],        %[temp_reg2]                   \n\t"
812         "mult   $ac1,         %[t0],        %[temp_reg4]                   \n\t"
813         "mfhi   %[temp_reg3]                                               \n\t"
814         "add    %[t0],        %[s2],        %[s3]                          \n\t"
815         "add    %[temp_reg5], %[temp_reg5], %[temp_reg6]                   \n\t"
816         "mfhi   %[temp_reg2], $ac3                                         \n\t"
817         "mfhi   %[temp_reg4], $ac1                                         \n\t"
818         "sub    %[t1],        %[s2],        %[s3]                          \n\t"
819         "sw     %[temp_reg5], 32*12*4(%[out])                              \n\t"
820         "sw     %[temp_reg3], 4*12*4(%[buf])                               \n\t"
821         "lw     %[temp_reg6], 14*4(%[win])                                 \n\t"
822         "lw     %[temp_reg5], 4*14*4(%[buf])                               \n\t"
823         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
824         "sw     %[temp_reg4], 4*5*4(%[buf])                                \n\t"
825         "sw     %[temp_reg1], 5*32*4(%[out])                               \n\t"
826         "mult   %[t1],        %[temp_reg6]                                 \n\t"
827         "lw     %[temp_reg4], 34*4(%[win])                                 \n\t"
828         "lw     %[temp_reg2], 3*4(%[win])                                  \n\t"
829         "lw     %[temp_reg1], 4*3*4(%[buf])                                \n\t"
830         "mult   $ac2,         %[t0],        %[temp_reg4]                   \n\t"
831         "mfhi   %[temp_reg6]                                               \n\t"
832         "mult   $ac1,         %[t1],        %[temp_reg2]                   \n\t"
833         "lw     %[temp_reg3], 23*4(%[win])                                 \n\t"
834         "lw     %[s0],        16*4(%[tmp])                                 \n\t"
835         "mfhi   %[temp_reg4], $ac2                                         \n\t"
836         "lw     %[t1],        17*4(%[tmp])                                 \n\t"
837         "mult   $ac3,         %[t0],        %[temp_reg3]                   \n\t"
838         "move   %[s1],        $0                                           \n\t"
839         "add    %[temp_reg5], %[temp_reg5], %[temp_reg6]                   \n\t"
840         "mfhi   %[temp_reg2], $ac1                                         \n\t"
841         "sw     %[temp_reg5], 14*32*4(%[out])                              \n\t"
842         "sw     %[temp_reg4], 4*14*4(%[buf])                               \n\t"
843         "mfhi   %[temp_reg3], $ac3                                         \n\t"
844         "li     %[temp_reg5], 0xB504F334                                   \n\t"
845         "add    %[temp_reg1], %[temp_reg1], %[temp_reg2]                   \n\t"
846         "multu  %[temp_reg5], %[t1]                                        \n\t"
847         "lw     %[temp_reg2], 4*13*4(%[buf])                               \n\t"
848         "sw     %[temp_reg1], 3*32*4(%[out])                               \n\t"
849         "sra    %[t1],        %[t1],        31                             \n\t"
850         "mfhi   %[temp_reg6]                                               \n\t"
851         "movn   %[s1],        %[temp_reg5], %[t1]                          \n\t"
852         "sw     %[temp_reg3], 4*3*4(%[buf])                                \n\t"
853         "lw     %[temp_reg1], 13*4(%[win])                                 \n\t"
854         "lw     %[temp_reg4], 4*4*4(%[buf])                                \n\t"
855         "lw     %[temp_reg3], 4*4(%[win])                                  \n\t"
856         "lw     %[temp_reg5], 33*4(%[win])                                 \n\t"
857         "subu   %[s1],        %[temp_reg6], %[s1]                          \n\t"
858         "lw     %[temp_reg6], 24*4(%[win])                                 \n\t"
859         "sub    %[t1],        %[s0],        %[s1]                          \n\t"
860         "add    %[t0],        %[s0],        %[s1]                          \n\t"
861         "mult   $ac1,         %[t1],        %[temp_reg1]                   \n\t"
862         "mult   $ac2,         %[t1],        %[temp_reg3]                   \n\t"
863         "mult   $ac3,         %[t0],        %[temp_reg5]                   \n\t"
864         "mult   %[t0],        %[temp_reg6]                                 \n\t"
865         "mfhi   %[temp_reg1], $ac1                                         \n\t"
866         "mfhi   %[temp_reg3], $ac2                                         \n\t"
867         "mfhi   %[temp_reg5], $ac3                                         \n\t"
868         "mfhi   %[temp_reg6]                                               \n\t"
869         "add    %[temp_reg2], %[temp_reg2], %[temp_reg1]                   \n\t"
870         "add    %[temp_reg4], %[temp_reg4], %[temp_reg3]                   \n\t"
871         "sw     %[temp_reg2], 13*32*4(%[out])                              \n\t"
872         "sw     %[temp_reg4], 4*32*4(%[out])                               \n\t"
873         "sw     %[temp_reg5], 4*13*4(%[buf])                               \n\t"
874         "sw     %[temp_reg6], 4*4*4(%[buf])                                \n\t"
875
876         : [t0] "=&r" (t0), [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3),
877           [s0] "=&r" (s0), [s2] "=&r" (s2), [temp_reg1] "=&r" (temp_reg1),
878           [temp_reg2] "=&r" (temp_reg2), [s1] "=&r" (s1), [s3] "=&r" (s3),
879           [temp_reg3] "=&r" (temp_reg3), [temp_reg4] "=&r" (temp_reg4),
880           [temp_reg5] "=&r" (temp_reg5), [temp_reg6] "=&r" (temp_reg6),
881           [out] "+r" (out)
882         : [tmp] "r" (tmp), [win] "r" (win), [buf] "r" (buf)
883         : "memory", "hi", "lo", "$ac1hi", "$ac1lo", "$ac2hi", "$ac2lo",
884           "$ac3hi", "$ac3lo"
885     );
886 }
887
888 static void ff_imdct36_blocks_mips_fixed(int *out, int *buf, int *in,
889                                int count, int switch_point, int block_type)
890 {
891     int j;
892     for (j=0 ; j < count; j++) {
893         /* apply window & overlap with previous buffer */
894
895         /* select window */
896         int win_idx = (switch_point && j < 2) ? 0 : block_type;
897         int *win = ff_mdct_win_fixed[win_idx + (4 & -(j & 1))];
898
899         imdct36_mips_fixed(out, buf, in, win);
900
901         in  += 18;
902         buf += ((j&3) != 3 ? 1 : (72-3));
903         out++;
904     }
905 }
906
907 #endif /* !HAVE_MIPS32R6 && !HAVE_MIPS64R6 */
908 #endif /* HAVE_INLINE_ASM */
909
910 void ff_mpadsp_init_mipsdsp(MPADSPContext *s)
911 {
912 #if HAVE_INLINE_ASM
913 #if !HAVE_MIPS32R6 && !HAVE_MIPS64R6
914     s->apply_window_fixed   = ff_mpadsp_apply_window_mips_fixed;
915     s->imdct36_blocks_fixed = ff_imdct36_blocks_mips_fixed;
916 #endif
917 #endif
918 }