mips: ac3 downmix updated to the new data layout.
[ffmpeg.git] / libavcodec / mips / ac3dsp_mips.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  * Authors:  Branimir Vasic (bvasic@mips.com)
30  *           Nedeljko Babic (nbabic@mips.com)
31  *
32  * Various AC-3 DSP Utils optimized for MIPS
33  *
34  * This file is part of FFmpeg.
35  *
36  * FFmpeg is free software; you can redistribute it and/or
37  * modify it under the terms of the GNU Lesser General Public
38  * License as published by the Free Software Foundation; either
39  * version 2.1 of the License, or (at your option) any later version.
40  *
41  * FFmpeg is distributed in the hope that it will be useful,
42  * but WITHOUT ANY WARRANTY; without even the implied warranty of
43  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
44  * Lesser General Public License for more details.
45  *
46  * You should have received a copy of the GNU Lesser General Public
47  * License along with FFmpeg; if not, write to the Free Software
48  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
49  */
50
51 /**
52  * @file
53  * Reference: libavcodec/ac3dsp.c
54  */
55
56 #include "config.h"
57 #include "libavcodec/ac3dsp.h"
58 #include "libavcodec/ac3.h"
59
60
61 #if HAVE_INLINE_ASM
62 #if HAVE_MIPSDSPR1
63 static void ac3_bit_alloc_calc_bap_mips(int16_t *mask, int16_t *psd,
64                                         int start, int end,
65                                         int snr_offset, int floor,
66                                         const uint8_t *bap_tab, uint8_t *bap)
67 {
68     int band, band_end, cond;
69     int m, address1, address2;
70     int16_t *psd1, *psd_end;
71     uint8_t *bap1;
72
73     if (snr_offset == -960) {
74         memset(bap, 0, AC3_MAX_COEFS);
75         return;
76     }
77
78     psd1 = &psd[start];
79     bap1 = &bap[start];
80     band = ff_ac3_bin_to_band_tab[start];
81
82     do {
83         m = (FFMAX(mask[band] - snr_offset - floor, 0) & 0x1FE0) + floor;
84         band_end = ff_ac3_band_start_tab[++band];
85         band_end = FFMIN(band_end, end);
86         psd_end = psd + band_end - 1;
87
88         __asm__ volatile (
89             "slt        %[cond],        %[psd1],        %[psd_end]  \n\t"
90             "beqz       %[cond],        1f                          \n\t"
91             "2:                                                     \n\t"
92             "lh         %[address1],    0(%[psd1])                  \n\t"
93             "lh         %[address2],    2(%[psd1])                  \n\t"
94             "addiu      %[psd1],        %[psd1],        4           \n\t"
95             "subu       %[address1],    %[address1],    %[m]        \n\t"
96             "sra        %[address1],    %[address1],    5           \n\t"
97             "addiu      %[address1],    %[address1],    -32         \n\t"
98             "shll_s.w   %[address1],    %[address1],    26          \n\t"
99             "subu       %[address2],    %[address2],    %[m]        \n\t"
100             "sra        %[address2],    %[address2],    5           \n\t"
101             "sra        %[address1],    %[address1],    26          \n\t"
102             "addiu      %[address1],    %[address1],    32          \n\t"
103             "lbux       %[address1],    %[address1](%[bap_tab])     \n\t"
104             "addiu      %[address2],    %[address2],    -32         \n\t"
105             "shll_s.w   %[address2],    %[address2],    26          \n\t"
106             "sb         %[address1],    0(%[bap1])                  \n\t"
107             "slt        %[cond],        %[psd1],        %[psd_end]  \n\t"
108             "sra        %[address2],    %[address2],    26          \n\t"
109             "addiu      %[address2],    %[address2],    32          \n\t"
110             "lbux       %[address2],    %[address2](%[bap_tab])     \n\t"
111             "sb         %[address2],    1(%[bap1])                  \n\t"
112             "addiu      %[bap1],        %[bap1],        2           \n\t"
113             "bnez       %[cond],        2b                          \n\t"
114             "addiu      %[psd_end],     %[psd_end],     2           \n\t"
115             "slt        %[cond],        %[psd1],        %[psd_end]  \n\t"
116             "beqz       %[cond],        3f                          \n\t"
117             "1:                                                     \n\t"
118             "lh         %[address1],    0(%[psd1])                  \n\t"
119             "addiu      %[psd1],        %[psd1],        2           \n\t"
120             "subu       %[address1],    %[address1],    %[m]        \n\t"
121             "sra        %[address1],    %[address1],    5           \n\t"
122             "addiu      %[address1],    %[address1],    -32         \n\t"
123             "shll_s.w   %[address1],    %[address1],    26          \n\t"
124             "sra        %[address1],    %[address1],    26          \n\t"
125             "addiu      %[address1],    %[address1],    32          \n\t"
126             "lbux       %[address1],    %[address1](%[bap_tab])     \n\t"
127             "sb         %[address1],    0(%[bap1])                  \n\t"
128             "addiu      %[bap1],        %[bap1],        1           \n\t"
129             "3:                                                     \n\t"
130
131             : [address1]"=&r"(address1), [address2]"=&r"(address2),
132               [cond]"=&r"(cond), [bap1]"+r"(bap1),
133               [psd1]"+r"(psd1), [psd_end]"+r"(psd_end)
134             : [m]"r"(m), [bap_tab]"r"(bap_tab)
135             : "memory"
136         );
137     } while (end > band_end);
138 }
139
140 static void ac3_update_bap_counts_mips(uint16_t mant_cnt[16], uint8_t *bap,
141                                        int len)
142 {
143     int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
144
145     __asm__ volatile (
146         "andi   %[temp3],   %[len],         3               \n\t"
147         "addu   %[temp2],   %[bap],         %[len]          \n\t"
148         "addu   %[temp4],   %[bap],         %[temp3]        \n\t"
149         "beq    %[temp2],   %[temp4],       4f              \n\t"
150         "1:                                                 \n\t"
151         "lbu    %[temp0],   -1(%[temp2])                    \n\t"
152         "lbu    %[temp5],   -2(%[temp2])                    \n\t"
153         "lbu    %[temp6],   -3(%[temp2])                    \n\t"
154         "sll    %[temp0],   %[temp0],       1               \n\t"
155         "addu   %[temp0],   %[mant_cnt],    %[temp0]        \n\t"
156         "sll    %[temp5],   %[temp5],       1               \n\t"
157         "addu   %[temp5],   %[mant_cnt],    %[temp5]        \n\t"
158         "lhu    %[temp1],   0(%[temp0])                     \n\t"
159         "sll    %[temp6],   %[temp6],       1               \n\t"
160         "addu   %[temp6],   %[mant_cnt],    %[temp6]        \n\t"
161         "addiu  %[temp1],   %[temp1],       1               \n\t"
162         "sh     %[temp1],   0(%[temp0])                     \n\t"
163         "lhu    %[temp1],   0(%[temp5])                     \n\t"
164         "lbu    %[temp7],   -4(%[temp2])                    \n\t"
165         "addiu  %[temp2],   %[temp2],       -4              \n\t"
166         "addiu  %[temp1],   %[temp1],       1               \n\t"
167         "sh     %[temp1],   0(%[temp5])                     \n\t"
168         "lhu    %[temp1],   0(%[temp6])                     \n\t"
169         "sll    %[temp7],   %[temp7],       1               \n\t"
170         "addu   %[temp7],   %[mant_cnt],    %[temp7]        \n\t"
171         "addiu  %[temp1],   %[temp1],1                      \n\t"
172         "sh     %[temp1],   0(%[temp6])                     \n\t"
173         "lhu    %[temp1],   0(%[temp7])                     \n\t"
174         "addiu  %[temp1],   %[temp1],       1               \n\t"
175         "sh     %[temp1],   0(%[temp7])                     \n\t"
176         "bne    %[temp2],   %[temp4],       1b              \n\t"
177         "4:                                                 \n\t"
178         "beqz   %[temp3],   2f                              \n\t"
179         "3:                                                 \n\t"
180         "addiu  %[temp3],   %[temp3],       -1              \n\t"
181         "lbu    %[temp0],   -1(%[temp2])                    \n\t"
182         "addiu  %[temp2],   %[temp2],       -1              \n\t"
183         "sll    %[temp0],   %[temp0],       1               \n\t"
184         "addu   %[temp0],   %[mant_cnt],    %[temp0]        \n\t"
185         "lhu    %[temp1],   0(%[temp0])                     \n\t"
186         "addiu  %[temp1],   %[temp1],       1               \n\t"
187         "sh     %[temp1],   0(%[temp0])                     \n\t"
188         "bgtz   %[temp3],   3b                              \n\t"
189         "2:                                                 \n\t"
190
191         : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1),
192           [temp2] "=&r" (temp2), [temp3] "=&r" (temp3),
193           [temp4] "=&r" (temp4), [temp5] "=&r" (temp5),
194           [temp6] "=&r" (temp6), [temp7] "=&r" (temp7)
195         : [len] "r" (len), [bap] "r" (bap),
196           [mant_cnt] "r" (mant_cnt)
197         : "memory"
198     );
199 }
200 #endif
201
202 #if HAVE_MIPSFPU && HAVE_MIPS32R2
203 static void float_to_fixed24_mips(int32_t *dst, const float *src, unsigned int len)
204 {
205     const float scale = 1 << 24;
206     float src0, src1, src2, src3, src4, src5, src6, src7;
207     int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
208
209     do {
210         __asm__ volatile (
211             "lwc1       %[src0],    0(%[src])               \n\t"
212             "lwc1       %[src1],    4(%[src])               \n\t"
213             "lwc1       %[src2],    8(%[src])               \n\t"
214             "lwc1       %[src3],    12(%[src])              \n\t"
215             "lwc1       %[src4],    16(%[src])              \n\t"
216             "lwc1       %[src5],    20(%[src])              \n\t"
217             "lwc1       %[src6],    24(%[src])              \n\t"
218             "lwc1       %[src7],    28(%[src])              \n\t"
219             "mul.s      %[src0],    %[src0],    %[scale]    \n\t"
220             "mul.s      %[src1],    %[src1],    %[scale]    \n\t"
221             "mul.s      %[src2],    %[src2],    %[scale]    \n\t"
222             "mul.s      %[src3],    %[src3],    %[scale]    \n\t"
223             "mul.s      %[src4],    %[src4],    %[scale]    \n\t"
224             "mul.s      %[src5],    %[src5],    %[scale]    \n\t"
225             "mul.s      %[src6],    %[src6],    %[scale]    \n\t"
226             "mul.s      %[src7],    %[src7],    %[scale]    \n\t"
227             "cvt.w.s    %[src0],    %[src0]                 \n\t"
228             "cvt.w.s    %[src1],    %[src1]                 \n\t"
229             "cvt.w.s    %[src2],    %[src2]                 \n\t"
230             "cvt.w.s    %[src3],    %[src3]                 \n\t"
231             "cvt.w.s    %[src4],    %[src4]                 \n\t"
232             "cvt.w.s    %[src5],    %[src5]                 \n\t"
233             "cvt.w.s    %[src6],    %[src6]                 \n\t"
234             "cvt.w.s    %[src7],    %[src7]                 \n\t"
235             "mfc1       %[temp0],   %[src0]                 \n\t"
236             "mfc1       %[temp1],   %[src1]                 \n\t"
237             "mfc1       %[temp2],   %[src2]                 \n\t"
238             "mfc1       %[temp3],   %[src3]                 \n\t"
239             "mfc1       %[temp4],   %[src4]                 \n\t"
240             "mfc1       %[temp5],   %[src5]                 \n\t"
241             "mfc1       %[temp6],   %[src6]                 \n\t"
242             "mfc1       %[temp7],   %[src7]                 \n\t"
243             "sw         %[temp0],   0(%[dst])               \n\t"
244             "sw         %[temp1],   4(%[dst])               \n\t"
245             "sw         %[temp2],   8(%[dst])               \n\t"
246             "sw         %[temp3],   12(%[dst])              \n\t"
247             "sw         %[temp4],   16(%[dst])              \n\t"
248             "sw         %[temp5],   20(%[dst])              \n\t"
249             "sw         %[temp6],   24(%[dst])              \n\t"
250             "sw         %[temp7],   28(%[dst])              \n\t"
251
252             : [dst] "+r" (dst), [src] "+r" (src),
253               [src0] "=&f" (src0), [src1] "=&f" (src1),
254               [src2] "=&f" (src2), [src3] "=&f" (src3),
255               [src4] "=&f" (src4), [src5] "=&f" (src5),
256               [src6] "=&f" (src6), [src7] "=&f" (src7),
257               [temp0] "=r" (temp0), [temp1] "=r" (temp1),
258               [temp2] "=r" (temp2), [temp3] "=r" (temp3),
259               [temp4] "=r" (temp4), [temp5] "=r" (temp5),
260               [temp6] "=r" (temp6), [temp7] "=r" (temp7)
261             : [scale] "f" (scale)
262             : "memory"
263         );
264         src = src + 8;
265         dst = dst + 8;
266         len -= 8;
267     } while (len > 0);
268 }
269
270 static void ac3_downmix_mips(float **samples, float (*matrix)[2],
271                           int out_ch, int in_ch, int len)
272 {
273     int i, j, i1, i2, i3;
274     float v0, v1, v2, v3;
275     float v4, v5, v6, v7;
276     float samples0, samples1, samples2, samples3, matrix_j, matrix_j2;
277     float *samples_p,*matrix_p, **samples_x, **samples_end, **samples_sw;
278
279     __asm__ volatile(
280         ".set   push                                                \n\t"
281         ".set   noreorder                                           \n\t"
282
283         "li     %[i1],          2                                   \n\t"
284         "sll    %[len],         2                                   \n\t"
285         "move   %[i],           $zero                               \n\t"
286         "sll    %[j],           %[in_ch],               2           \n\t"
287
288         "bne    %[out_ch],      %[i1],                  3f          \n\t"   // if (out_ch == 2)
289         " li    %[i2],          1                                   \n\t"
290
291         "2:                                                         \n\t"   // start of the for loop (for (i = 0; i < len; i+=4))
292         "move   %[matrix_p],    %[matrix]                           \n\t"
293         "move   %[samples_x],   %[samples]                          \n\t"
294         "mtc1   $zero,          %[v0]                               \n\t"
295         "mtc1   $zero,          %[v1]                               \n\t"
296         "mtc1   $zero,          %[v2]                               \n\t"
297         "mtc1   $zero,          %[v3]                               \n\t"
298         "mtc1   $zero,          %[v4]                               \n\t"
299         "mtc1   $zero,          %[v5]                               \n\t"
300         "mtc1   $zero,          %[v6]                               \n\t"
301         "mtc1   $zero,          %[v7]                               \n\t"
302         "addiu  %[i1],          %[i],                  4            \n\t"
303         "addiu  %[i2],          %[i],                  8            \n\t"
304         "lw     %[samples_p],   0(%[samples_x])                     \n\t"
305         "addiu  %[i3],          %[i],                  12           \n\t"
306         "addu   %[samples_end], %[samples_x],          %[j]         \n\t"
307         "move   %[samples_sw],  %[samples_p]                        \n\t"
308
309         "1:                                                         \n\t"   // start of the inner for loop (for (j = 0; j < in_ch; j++))
310         "lwc1   %[matrix_j],    0(%[matrix_p])                      \n\t"
311         "lwc1   %[matrix_j2],   4(%[matrix_p])                      \n\t"
312         "lwxc1  %[samples0],    %[i](%[samples_p])                  \n\t"
313         "lwxc1  %[samples1],    %[i1](%[samples_p])                 \n\t"
314         "lwxc1  %[samples2],    %[i2](%[samples_p])                 \n\t"
315         "lwxc1  %[samples3],    %[i3](%[samples_p])                 \n\t"
316         "addiu  %[matrix_p],    8                                   \n\t"
317         "addiu  %[samples_x],   4                                   \n\t"
318         "madd.s %[v0],          %[v0],  %[samples0],    %[matrix_j] \n\t"
319         "madd.s %[v1],          %[v1],  %[samples1],    %[matrix_j] \n\t"
320         "madd.s %[v2],          %[v2],  %[samples2],    %[matrix_j] \n\t"
321         "madd.s %[v3],          %[v3],  %[samples3],    %[matrix_j] \n\t"
322         "madd.s %[v4],          %[v4],  %[samples0],    %[matrix_j2]\n\t"
323         "madd.s %[v5],          %[v5],  %[samples1],    %[matrix_j2]\n\t"
324         "madd.s %[v6],          %[v6],  %[samples2],    %[matrix_j2]\n\t"
325         "madd.s %[v7],          %[v7],  %[samples3],    %[matrix_j2]\n\t"
326         "bne    %[samples_x],   %[samples_end],         1b          \n\t"
327         " lw    %[samples_p],   0(%[samples_x])                     \n\t"
328
329         "lw     %[samples_p],   4(%[samples])                       \n\t"
330         "swxc1  %[v0],          %[i](%[samples_sw])                 \n\t"
331         "swxc1  %[v1],          %[i1](%[samples_sw])                \n\t"
332         "swxc1  %[v2],          %[i2](%[samples_sw])                \n\t"
333         "swxc1  %[v3],          %[i3](%[samples_sw])                \n\t"
334         "swxc1  %[v4],          %[i](%[samples_p])                  \n\t"
335         "addiu  %[i],           16                                  \n\t"
336         "swxc1  %[v5],          %[i1](%[samples_p])                 \n\t"
337         "swxc1  %[v6],          %[i2](%[samples_p])                 \n\t"
338         "bne    %[i],           %[len],                 2b          \n\t"
339         " swxc1 %[v7],          %[i3](%[samples_p])                 \n\t"
340
341         "3:                                                         \n\t"
342         "bne    %[out_ch],      %[i2],                  6f          \n\t"   // if (out_ch == 1)
343         " nop                                                       \n\t"
344
345         "5:                                                         \n\t"   // start of the outer for loop (for (i = 0; i < len; i+=4))
346         "move   %[matrix_p],    %[matrix]                           \n\t"
347         "move   %[samples_x],   %[samples]                          \n\t"
348         "mtc1   $zero,          %[v0]                               \n\t"
349         "mtc1   $zero,          %[v1]                               \n\t"
350         "mtc1   $zero,          %[v2]                               \n\t"
351         "mtc1   $zero,          %[v3]                               \n\t"
352         "addiu  %[i1],          %[i],                  4            \n\t"
353         "addiu  %[i2],          %[i],                  8            \n\t"
354         "lw     %[samples_p],   0(%[samples_x])                     \n\t"
355         "addiu  %[i3],          %[i],                  12           \n\t"
356         "addu   %[samples_end], %[samples_x],          %[j]         \n\t"
357         "move   %[samples_sw],  %[samples_p]                        \n\t"
358
359         "4:                                                         \n\t"   // start of the inner for loop (for (j = 0; j < in_ch; j++))
360         "lwc1   %[matrix_j],    0(%[matrix_p])                      \n\t"
361         "lwxc1  %[samples0],    %[i](%[samples_p])                  \n\t"
362         "lwxc1  %[samples1],    %[i1](%[samples_p])                 \n\t"
363         "lwxc1  %[samples2],    %[i2](%[samples_p])                 \n\t"
364         "lwxc1  %[samples3],    %[i3](%[samples_p])                 \n\t"
365         "addiu  %[matrix_p],    8                                   \n\t"
366         "addiu  %[samples_x],   4                                   \n\t"
367         "madd.s %[v0],          %[v0],  %[samples0],    %[matrix_j] \n\t"
368         "madd.s %[v1],          %[v1],  %[samples1],    %[matrix_j] \n\t"
369         "madd.s %[v2],          %[v2],  %[samples2],    %[matrix_j] \n\t"
370         "madd.s %[v3],          %[v3],  %[samples3],    %[matrix_j] \n\t"
371         "bne    %[samples_x],   %[samples_end],         4b          \n\t"
372         " lw    %[samples_p],   0(%[samples_x])                     \n\t"
373
374         "swxc1  %[v0],          %[i](%[samples_sw])                 \n\t"
375         "addiu  %[i],           16                                  \n\t"
376         "swxc1  %[v1],          %[i1](%[samples_sw])                \n\t"
377         "swxc1  %[v2],          %[i2](%[samples_sw])                \n\t"
378         "bne    %[i],           %[len],                 5b          \n\t"
379         " swxc1 %[v3],          %[i3](%[samples_sw])                \n\t"
380         "6:                                                         \n\t"
381
382         ".set   pop"
383         :[samples_p]"=&r"(samples_p), [matrix_j]"=&f"(matrix_j), [matrix_j2]"=&f"(matrix_j2),
384          [samples0]"=&f"(samples0), [samples1]"=&f"(samples1),
385          [samples2]"=&f"(samples2), [samples3]"=&f"(samples3),
386          [v0]"=&f"(v0), [v1]"=&f"(v1), [v2]"=&f"(v2), [v3]"=&f"(v3),
387          [v4]"=&f"(v4), [v5]"=&f"(v5), [v6]"=&f"(v6), [v7]"=&f"(v7),
388          [samples_x]"=&r"(samples_x), [matrix_p]"=&r"(matrix_p),
389          [samples_end]"=&r"(samples_end), [samples_sw]"=&r"(samples_sw),
390          [i1]"=&r"(i1), [i2]"=&r"(i2), [i3]"=&r"(i3), [i]"=&r"(i),
391          [j]"=&r"(j), [len]"+r"(len)
392         :[samples]"r"(samples), [matrix]"r"(matrix),
393          [in_ch]"r"(in_ch), [out_ch]"r"(out_ch)
394         :"memory"
395     );
396 }
397 #endif
398 #endif /* HAVE_INLINE_ASM */
399
400 void ff_ac3dsp_init_mips(AC3DSPContext *c, int bit_exact) {
401 #if HAVE_INLINE_ASM
402 #if HAVE_MIPSDSPR1
403     c->bit_alloc_calc_bap = ac3_bit_alloc_calc_bap_mips;
404     c->update_bap_counts  = ac3_update_bap_counts_mips;
405 #endif
406 #if HAVE_MIPSFPU && HAVE_MIPS32R2
407     c->float_to_fixed24 = float_to_fixed24_mips;
408     c->downmix          = ac3_downmix_mips;
409 #endif
410 #endif
411
412 }