FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
aacpsdsp_mips.c
Go to the documentation of this file.
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  * Authors: Darko Laus (darko@mips.com)
30  * Djordje Pesut (djordje@mips.com)
31  * Mirjana Vulin (mvulin@mips.com)
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/aacpsdsp.c
53  */
54 
55 #include "config.h"
56 #include "libavcodec/aacpsdsp.h"
57 #include "libavutil/mips/asmdefs.h"
58 
59 #if HAVE_INLINE_ASM
60 static void ps_hybrid_analysis_ileave_mips(float (*out)[32][2], float L[2][38][64],
61  int i, int len)
62 {
63  int temp0, temp1, temp2, temp3;
64  int temp4, temp5, temp6, temp7;
65  float *out1=&out[i][0][0];
66  float *L1=&L[0][0][i];
67  float *j=out1+ len*2;
68 
69  for (; i < 64; i++) {
70 
71  /* loop unrolled 8 times */
72  __asm__ volatile (
73  "1: \n\t"
74  "lw %[temp0], 0(%[L1]) \n\t"
75  "lw %[temp1], 9728(%[L1]) \n\t"
76  "lw %[temp2], 256(%[L1]) \n\t"
77  "lw %[temp3], 9984(%[L1]) \n\t"
78  "lw %[temp4], 512(%[L1]) \n\t"
79  "lw %[temp5], 10240(%[L1]) \n\t"
80  "lw %[temp6], 768(%[L1]) \n\t"
81  "lw %[temp7], 10496(%[L1]) \n\t"
82  "sw %[temp0], 0(%[out1]) \n\t"
83  "sw %[temp1], 4(%[out1]) \n\t"
84  "sw %[temp2], 8(%[out1]) \n\t"
85  "sw %[temp3], 12(%[out1]) \n\t"
86  "sw %[temp4], 16(%[out1]) \n\t"
87  "sw %[temp5], 20(%[out1]) \n\t"
88  "sw %[temp6], 24(%[out1]) \n\t"
89  "sw %[temp7], 28(%[out1]) \n\t"
90  PTR_ADDIU "%[out1], %[out1], 32 \n\t"
91  PTR_ADDIU "%[L1], %[L1], 1024 \n\t"
92  "bne %[out1], %[j], 1b \n\t"
93 
94  : [out1]"+r"(out1), [L1]"+r"(L1), [j]"+r"(j),
95  [temp0]"=&r"(temp0), [temp1]"=&r"(temp1),
96  [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),
97  [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
98  [temp6]"=&r"(temp6), [temp7]"=&r"(temp7)
99  : [len]"r"(len)
100  : "memory"
101  );
102  out1-=(len<<1)-64;
103  L1-=(len<<6)-1;
104  j+=len*2;
105  }
106 }
107 
108 static void ps_hybrid_synthesis_deint_mips(float out[2][38][64],
109  float (*in)[32][2],
110  int i, int len)
111 {
112  int n;
113  int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
114  float *out1 = (float*)out + i;
115  float *out2 = (float*)out + 2432 + i;
116  float *in1 = (float*)in + 64 * i;
117  float *in2 = (float*)in + 64 * i + 1;
118 
119  for (; i < 64; i++) {
120  for (n = 0; n < 7; n++) {
121 
122  /* loop unrolled 8 times */
123  __asm__ volatile (
124  "lw %[temp0], 0(%[in1]) \n\t"
125  "lw %[temp1], 0(%[in2]) \n\t"
126  "lw %[temp2], 8(%[in1]) \n\t"
127  "lw %[temp3], 8(%[in2]) \n\t"
128  "lw %[temp4], 16(%[in1]) \n\t"
129  "lw %[temp5], 16(%[in2]) \n\t"
130  "lw %[temp6], 24(%[in1]) \n\t"
131  "lw %[temp7], 24(%[in2]) \n\t"
132  PTR_ADDIU "%[out1], %[out1], 1024 \n\t"
133  PTR_ADDIU "%[out2], %[out2], 1024 \n\t"
134  PTR_ADDIU "%[in1], %[in1], 32 \n\t"
135  PTR_ADDIU "%[in2], %[in2], 32 \n\t"
136  "sw %[temp0], -1024(%[out1]) \n\t"
137  "sw %[temp1], -1024(%[out2]) \n\t"
138  "sw %[temp2], -768(%[out1]) \n\t"
139  "sw %[temp3], -768(%[out2]) \n\t"
140  "sw %[temp4], -512(%[out1]) \n\t"
141  "sw %[temp5], -512(%[out2]) \n\t"
142  "sw %[temp6], -256(%[out1]) \n\t"
143  "sw %[temp7], -256(%[out2]) \n\t"
144 
145  : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1),
146  [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),
147  [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
148  [temp6]"=&r"(temp6), [temp7]"=&r"(temp7),
149  [out1]"+r"(out1), [out2]"+r"(out2),
150  [in1]"+r"(in1), [in2]"+r"(in2)
151  :
152  : "memory"
153  );
154  }
155  /* loop unrolled 8 times */
156  __asm__ volatile (
157  "lw %[temp0], 0(%[in1]) \n\t"
158  "lw %[temp1], 0(%[in2]) \n\t"
159  "lw %[temp2], 8(%[in1]) \n\t"
160  "lw %[temp3], 8(%[in2]) \n\t"
161  "lw %[temp4], 16(%[in1]) \n\t"
162  "lw %[temp5], 16(%[in2]) \n\t"
163  "lw %[temp6], 24(%[in1]) \n\t"
164  "lw %[temp7], 24(%[in2]) \n\t"
165  PTR_ADDIU "%[out1], %[out1], -7164 \n\t"
166  PTR_ADDIU "%[out2], %[out2], -7164 \n\t"
167  PTR_ADDIU "%[in1], %[in1], 32 \n\t"
168  PTR_ADDIU "%[in2], %[in2], 32 \n\t"
169  "sw %[temp0], 7164(%[out1]) \n\t"
170  "sw %[temp1], 7164(%[out2]) \n\t"
171  "sw %[temp2], 7420(%[out1]) \n\t"
172  "sw %[temp3], 7420(%[out2]) \n\t"
173  "sw %[temp4], 7676(%[out1]) \n\t"
174  "sw %[temp5], 7676(%[out2]) \n\t"
175  "sw %[temp6], 7932(%[out1]) \n\t"
176  "sw %[temp7], 7932(%[out2]) \n\t"
177 
178  : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1),
179  [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),
180  [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
181  [temp6]"=&r"(temp6), [temp7]"=&r"(temp7),
182  [out1]"+r"(out1), [out2]"+r"(out2),
183  [in1]"+r"(in1), [in2]"+r"(in2)
184  :
185  : "memory"
186  );
187  }
188 }
189 
190 #if HAVE_MIPSFPU
191 static void ps_add_squares_mips(float *dst, const float (*src)[2], int n)
192 {
193  int i;
194  float temp0, temp1, temp2, temp3, temp4, temp5;
195  float temp6, temp7, temp8, temp9, temp10, temp11;
196  float *src0 = (float*)&src[0][0];
197  float *dst0 = &dst[0];
198 
199  for (i = 0; i < 8; i++) {
200  /* loop unrolled 4 times */
201  __asm__ volatile (
202  "lwc1 %[temp0], 0(%[src0]) \n\t"
203  "lwc1 %[temp1], 4(%[src0]) \n\t"
204  "lwc1 %[temp2], 8(%[src0]) \n\t"
205  "lwc1 %[temp3], 12(%[src0]) \n\t"
206  "lwc1 %[temp4], 16(%[src0]) \n\t"
207  "lwc1 %[temp5], 20(%[src0]) \n\t"
208  "lwc1 %[temp6], 24(%[src0]) \n\t"
209  "lwc1 %[temp7], 28(%[src0]) \n\t"
210  "lwc1 %[temp8], 0(%[dst0]) \n\t"
211  "lwc1 %[temp9], 4(%[dst0]) \n\t"
212  "lwc1 %[temp10], 8(%[dst0]) \n\t"
213  "lwc1 %[temp11], 12(%[dst0]) \n\t"
214  "mul.s %[temp1], %[temp1], %[temp1] \n\t"
215  "mul.s %[temp3], %[temp3], %[temp3] \n\t"
216  "mul.s %[temp5], %[temp5], %[temp5] \n\t"
217  "mul.s %[temp7], %[temp7], %[temp7] \n\t"
218  "madd.s %[temp0], %[temp1], %[temp0], %[temp0] \n\t"
219  "madd.s %[temp2], %[temp3], %[temp2], %[temp2] \n\t"
220  "madd.s %[temp4], %[temp5], %[temp4], %[temp4] \n\t"
221  "madd.s %[temp6], %[temp7], %[temp6], %[temp6] \n\t"
222  "add.s %[temp0], %[temp8], %[temp0] \n\t"
223  "add.s %[temp2], %[temp9], %[temp2] \n\t"
224  "add.s %[temp4], %[temp10], %[temp4] \n\t"
225  "add.s %[temp6], %[temp11], %[temp6] \n\t"
226  "swc1 %[temp0], 0(%[dst0]) \n\t"
227  "swc1 %[temp2], 4(%[dst0]) \n\t"
228  "swc1 %[temp4], 8(%[dst0]) \n\t"
229  "swc1 %[temp6], 12(%[dst0]) \n\t"
230  PTR_ADDIU "%[dst0], %[dst0], 16 \n\t"
231  PTR_ADDIU "%[src0], %[src0], 32 \n\t"
232 
233  : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2),
234  [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5),
235  [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8),
236  [temp9]"=&f"(temp9), [dst0]"+r"(dst0), [src0]"+r"(src0),
237  [temp10]"=&f"(temp10), [temp11]"=&f"(temp11)
238  :
239  : "memory"
240  );
241  }
242 }
243 
244 static void ps_mul_pair_single_mips(float (*dst)[2], float (*src0)[2], float *src1,
245  int n)
246 {
247  float temp0, temp1, temp2;
248  float *p_d, *p_s0, *p_s1, *end;
249  p_d = &dst[0][0];
250  p_s0 = &src0[0][0];
251  p_s1 = &src1[0];
252  end = p_s1 + n;
253 
254  __asm__ volatile(
255  ".set push \n\t"
256  ".set noreorder \n\t"
257  "1: \n\t"
258  "lwc1 %[temp2], 0(%[p_s1]) \n\t"
259  "lwc1 %[temp0], 0(%[p_s0]) \n\t"
260  "lwc1 %[temp1], 4(%[p_s0]) \n\t"
261  PTR_ADDIU "%[p_d], %[p_d], 8 \n\t"
262  "mul.s %[temp0], %[temp0], %[temp2] \n\t"
263  "mul.s %[temp1], %[temp1], %[temp2] \n\t"
264  PTR_ADDIU "%[p_s0], %[p_s0], 8 \n\t"
265  "swc1 %[temp0], -8(%[p_d]) \n\t"
266  "swc1 %[temp1], -4(%[p_d]) \n\t"
267  "bne %[p_s1], %[end], 1b \n\t"
268  PTR_ADDIU "%[p_s1], %[p_s1], 4 \n\t"
269  ".set pop \n\t"
270 
271  : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1),
272  [temp2]"=&f"(temp2), [p_d]"+r"(p_d),
273  [p_s0]"+r"(p_s0), [p_s1]"+r"(p_s1)
274  : [end]"r"(end)
275  : "memory"
276  );
277 }
278 
279 static void ps_decorrelate_mips(float (*out)[2], float (*delay)[2],
280  float (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2],
281  const float phi_fract[2], const float (*Q_fract)[2],
282  const float *transient_gain,
283  float g_decay_slope,
284  int len)
285 {
286  float *p_delay = &delay[0][0];
287  float *p_out = &out[0][0];
288  float *p_ap_delay = &ap_delay[0][0][0];
289  const float *p_t_gain = transient_gain;
290  const float *p_Q_fract = &Q_fract[0][0];
291  float ag0, ag1, ag2;
292  float phi_fract0 = phi_fract[0];
293  float phi_fract1 = phi_fract[1];
294  float temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9;
295 
296  float *p_delay_end = (p_delay + (len << 1));
297 
298  /* merged 2 loops */
299  __asm__ volatile(
300  ".set push \n\t"
301  ".set noreorder \n\t"
302  "li.s %[ag0], 0.65143905753106 \n\t"
303  "li.s %[ag1], 0.56471812200776 \n\t"
304  "li.s %[ag2], 0.48954165955695 \n\t"
305  "mul.s %[ag0], %[ag0], %[g_decay_slope] \n\t"
306  "mul.s %[ag1], %[ag1], %[g_decay_slope] \n\t"
307  "mul.s %[ag2], %[ag2], %[g_decay_slope] \n\t"
308  "1: \n\t"
309  "lwc1 %[temp0], 0(%[p_delay]) \n\t"
310  "lwc1 %[temp1], 4(%[p_delay]) \n\t"
311  "lwc1 %[temp4], 16(%[p_ap_delay]) \n\t"
312  "lwc1 %[temp5], 20(%[p_ap_delay]) \n\t"
313  "mul.s %[temp3], %[temp0], %[phi_fract1] \n\t"
314  "lwc1 %[temp6], 0(%[p_Q_fract]) \n\t"
315  "mul.s %[temp2], %[temp1], %[phi_fract1] \n\t"
316  "lwc1 %[temp7], 4(%[p_Q_fract]) \n\t"
317  "madd.s %[temp3], %[temp3], %[temp1], %[phi_fract0] \n\t"
318  "msub.s %[temp2], %[temp2], %[temp0], %[phi_fract0] \n\t"
319  "mul.s %[temp8], %[temp5], %[temp7] \n\t"
320  "mul.s %[temp9], %[temp4], %[temp7] \n\t"
321  "lwc1 %[temp7], 12(%[p_Q_fract]) \n\t"
322  "mul.s %[temp0], %[ag0], %[temp2] \n\t"
323  "mul.s %[temp1], %[ag0], %[temp3] \n\t"
324  "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t"
325  "lwc1 %[temp4], 304(%[p_ap_delay]) \n\t"
326  "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t"
327  "lwc1 %[temp5], 308(%[p_ap_delay]) \n\t"
328  "sub.s %[temp0], %[temp8], %[temp0] \n\t"
329  "sub.s %[temp1], %[temp9], %[temp1] \n\t"
330  "madd.s %[temp2], %[temp2], %[ag0], %[temp0] \n\t"
331  "lwc1 %[temp6], 8(%[p_Q_fract]) \n\t"
332  "madd.s %[temp3], %[temp3], %[ag0], %[temp1] \n\t"
333  "mul.s %[temp8], %[temp5], %[temp7] \n\t"
334  "mul.s %[temp9], %[temp4], %[temp7] \n\t"
335  "lwc1 %[temp7], 20(%[p_Q_fract]) \n\t"
336  "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t"
337  "swc1 %[temp2], 40(%[p_ap_delay]) \n\t"
338  "mul.s %[temp2], %[ag1], %[temp0] \n\t"
339  "swc1 %[temp3], 44(%[p_ap_delay]) \n\t"
340  "mul.s %[temp3], %[ag1], %[temp1] \n\t"
341  "lwc1 %[temp4], 592(%[p_ap_delay]) \n\t"
342  "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t"
343  "lwc1 %[temp5], 596(%[p_ap_delay]) \n\t"
344  "sub.s %[temp2], %[temp8], %[temp2] \n\t"
345  "sub.s %[temp3], %[temp9], %[temp3] \n\t"
346  "lwc1 %[temp6], 16(%[p_Q_fract]) \n\t"
347  "madd.s %[temp0], %[temp0], %[ag1], %[temp2] \n\t"
348  "madd.s %[temp1], %[temp1], %[ag1], %[temp3] \n\t"
349  "mul.s %[temp8], %[temp5], %[temp7] \n\t"
350  "mul.s %[temp9], %[temp4], %[temp7] \n\t"
351  "msub.s %[temp8], %[temp8], %[temp4], %[temp6] \n\t"
352  "madd.s %[temp9], %[temp9], %[temp5], %[temp6] \n\t"
353  "swc1 %[temp0], 336(%[p_ap_delay]) \n\t"
354  "mul.s %[temp0], %[ag2], %[temp2] \n\t"
355  "swc1 %[temp1], 340(%[p_ap_delay]) \n\t"
356  "mul.s %[temp1], %[ag2], %[temp3] \n\t"
357  "lwc1 %[temp4], 0(%[p_t_gain]) \n\t"
358  "sub.s %[temp0], %[temp8], %[temp0] \n\t"
359  PTR_ADDIU "%[p_ap_delay], %[p_ap_delay], 8 \n\t"
360  "sub.s %[temp1], %[temp9], %[temp1] \n\t"
361  PTR_ADDIU "%[p_t_gain], %[p_t_gain], 4 \n\t"
362  "madd.s %[temp2], %[temp2], %[ag2], %[temp0] \n\t"
363  PTR_ADDIU "%[p_delay], %[p_delay], 8 \n\t"
364  "madd.s %[temp3], %[temp3], %[ag2], %[temp1] \n\t"
365  PTR_ADDIU "%[p_out], %[p_out], 8 \n\t"
366  "mul.s %[temp5], %[temp4], %[temp0] \n\t"
367  "mul.s %[temp6], %[temp4], %[temp1] \n\t"
368  "swc1 %[temp2], 624(%[p_ap_delay]) \n\t"
369  "swc1 %[temp3], 628(%[p_ap_delay]) \n\t"
370  "swc1 %[temp5], -8(%[p_out]) \n\t"
371  "swc1 %[temp6], -4(%[p_out]) \n\t"
372  "bne %[p_delay], %[p_delay_end],1b \n\t"
373  " swc1 %[temp6], -4(%[p_out]) \n\t"
374  ".set pop \n\t"
375 
376  : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2),
377  [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5),
378  [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), [temp8]"=&f"(temp8),
379  [temp9]"=&f"(temp9), [p_delay]"+r"(p_delay), [p_ap_delay]"+r"(p_ap_delay),
380  [p_Q_fract]"+r"(p_Q_fract), [p_t_gain]"+r"(p_t_gain), [p_out]"+r"(p_out),
381  [ag0]"=&f"(ag0), [ag1]"=&f"(ag1), [ag2]"=&f"(ag2)
382  : [phi_fract0]"f"(phi_fract0), [phi_fract1]"f"(phi_fract1),
383  [p_delay_end]"r"(p_delay_end), [g_decay_slope]"f"(g_decay_slope)
384  : "memory"
385  );
386 }
387 
388 static void ps_stereo_interpolate_mips(float (*l)[2], float (*r)[2],
389  float h[2][4], float h_step[2][4],
390  int len)
391 {
392  float h0 = h[0][0];
393  float h1 = h[0][1];
394  float h2 = h[0][2];
395  float h3 = h[0][3];
396  float hs0 = h_step[0][0];
397  float hs1 = h_step[0][1];
398  float hs2 = h_step[0][2];
399  float hs3 = h_step[0][3];
400  float temp0, temp1, temp2, temp3;
401  float l_re, l_im, r_re, r_im;
402 
403  float *l_end = ((float *)l + (len << 1));
404 
405  __asm__ volatile(
406  ".set push \n\t"
407  ".set noreorder \n\t"
408  "1: \n\t"
409  "add.s %[h0], %[h0], %[hs0] \n\t"
410  "lwc1 %[l_re], 0(%[l]) \n\t"
411  "add.s %[h1], %[h1], %[hs1] \n\t"
412  "lwc1 %[r_re], 0(%[r]) \n\t"
413  "add.s %[h2], %[h2], %[hs2] \n\t"
414  "lwc1 %[l_im], 4(%[l]) \n\t"
415  "add.s %[h3], %[h3], %[hs3] \n\t"
416  "lwc1 %[r_im], 4(%[r]) \n\t"
417  "mul.s %[temp0], %[h0], %[l_re] \n\t"
418  PTR_ADDIU "%[l], %[l], 8 \n\t"
419  "mul.s %[temp2], %[h1], %[l_re] \n\t"
420  PTR_ADDIU "%[r], %[r], 8 \n\t"
421  "madd.s %[temp0], %[temp0], %[h2], %[r_re] \n\t"
422  "madd.s %[temp2], %[temp2], %[h3], %[r_re] \n\t"
423  "mul.s %[temp1], %[h0], %[l_im] \n\t"
424  "mul.s %[temp3], %[h1], %[l_im] \n\t"
425  "madd.s %[temp1], %[temp1], %[h2], %[r_im] \n\t"
426  "madd.s %[temp3], %[temp3], %[h3], %[r_im] \n\t"
427  "swc1 %[temp0], -8(%[l]) \n\t"
428  "swc1 %[temp2], -8(%[r]) \n\t"
429  "swc1 %[temp1], -4(%[l]) \n\t"
430  "bne %[l], %[l_end], 1b \n\t"
431  " swc1 %[temp3], -4(%[r]) \n\t"
432  ".set pop \n\t"
433 
434  : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1),
435  [temp2]"=&f"(temp2), [temp3]"=&f"(temp3),
436  [h0]"+f"(h0), [h1]"+f"(h1), [h2]"+f"(h2),
437  [h3]"+f"(h3), [l]"+r"(l), [r]"+r"(r),
438  [l_re]"=&f"(l_re), [l_im]"=&f"(l_im),
439  [r_re]"=&f"(r_re), [r_im]"=&f"(r_im)
440  : [hs0]"f"(hs0), [hs1]"f"(hs1), [hs2]"f"(hs2),
441  [hs3]"f"(hs3), [l_end]"r"(l_end)
442  : "memory"
443  );
444 }
445 #endif /* HAVE_MIPSFPU */
446 #endif /* HAVE_INLINE_ASM */
447 
449 {
450 #if HAVE_INLINE_ASM
451  s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_mips;
452  s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_mips;
453 #if HAVE_MIPSFPU
454  s->add_squares = ps_add_squares_mips;
455  s->mul_pair_single = ps_mul_pair_single_mips;
456  s->decorrelate = ps_decorrelate_mips;
457  s->stereo_interpolate[0] = ps_stereo_interpolate_mips;
458 #endif /* HAVE_MIPSFPU */
459 #endif /* HAVE_INLINE_ASM */
460 }