Line data Source code
1 : /* Copyright (C) 2007-2008 Jean-Marc Valin
2 : Copyright (C) 2008 Thorvald Natvig
3 :
4 : File: resample.c
5 : Arbitrary resampling code
6 :
7 : Redistribution and use in source and binary forms, with or without
8 : modification, are permitted provided that the following conditions are
9 : met:
10 :
11 : 1. Redistributions of source code must retain the above copyright notice,
12 : this list of conditions and the following disclaimer.
13 :
14 : 2. Redistributions in binary form must reproduce the above copyright
15 : notice, this list of conditions and the following disclaimer in the
16 : documentation and/or other materials provided with the distribution.
17 :
18 : 3. The name of the author may not be used to endorse or promote products
19 : derived from this software without specific prior written permission.
20 :
21 : THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
22 : IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
23 : OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 : DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
25 : INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
26 : (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
27 : SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 : HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
29 : STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 : ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 : POSSIBILITY OF SUCH DAMAGE.
32 : */
33 :
34 : /*
35 : The design goals of this code are:
36 : - Very fast algorithm
37 : - SIMD-friendly algorithm
38 : - Low memory requirement
39 : - Good *perceptual* quality (and not best SNR)
40 :
41 : Warning: This resampler is relatively new. Although I think I got rid of
42 : all the major bugs and I don't expect the API to change anymore, there
43 : may be something I've missed. So use with caution.
44 :
45 : This algorithm is based on this original resampling algorithm:
46 : Smith, Julius O. Digital Audio Resampling Home Page
47 : Center for Computer Research in Music and Acoustics (CCRMA),
48 : Stanford University, 2007.
49 : Web published at http://ccrma.stanford.edu/~jos/resample/.
50 :
51 : There is one main difference, though. This resampler uses cubic
52 : interpolation instead of linear interpolation in the above paper. This
53 : makes the table much smaller and makes it possible to compute that table
54 : on a per-stream basis. In turn, being able to tweak the table for each
55 : stream makes it possible to both reduce complexity on simple ratios
56 : (e.g. 2/3), and get rid of the rounding operations in the inner loop.
57 : The latter both reduces CPU time and makes the algorithm more SIMD-friendly.
58 : */
59 :
60 : #ifdef HAVE_CONFIG_H
61 : #include "config.h"
62 : #endif
63 :
64 : #define RESAMPLE_HUGEMEM 1
65 :
66 : #ifdef OUTSIDE_SPEEX
67 : #include <stdlib.h>
68 0 : static void *speex_alloc (int size) {return calloc(size,1);}
69 0 : static void *speex_realloc (void *ptr, int size) {return realloc(ptr, size);}
70 0 : static void speex_free (void *ptr) {free(ptr);}
71 : #include "speex_resampler.h"
72 : #include "arch.h"
73 : #else /* OUTSIDE_SPEEX */
74 :
75 : #include "speex/speex_resampler.h"
76 : #include "arch.h"
77 : #include "os_support.h"
78 : #endif /* OUTSIDE_SPEEX */
79 :
80 : #include "stack_alloc.h"
81 : #include <math.h>
82 : #include <limits.h>
83 :
84 : #ifndef M_PI
85 : #define M_PI 3.14159265358979323846
86 : #endif
87 :
88 : #define IMAX(a,b) ((a) > (b) ? (a) : (b))
89 : #define IMIN(a,b) ((a) < (b) ? (a) : (b))
90 :
91 : #ifndef NULL
92 : #define NULL 0
93 : #endif
94 :
95 : #ifndef UINT32_MAX
96 : #define UINT32_MAX 4294967296U
97 : #endif
98 :
99 : #include "simd_detect.h"
100 :
101 : /* Numer of elements to allocate on the stack */
102 : #ifdef VAR_ARRAYS
103 : #define FIXED_STACK_ALLOC 8192
104 : #else
105 : #define FIXED_STACK_ALLOC 1024
106 : #endif
107 :
108 : typedef int (*resampler_basic_func)(SpeexResamplerState *, spx_uint32_t , const spx_word16_t *, spx_uint32_t *, spx_word16_t *, spx_uint32_t *);
109 :
110 : struct SpeexResamplerState_ {
111 : spx_uint32_t in_rate;
112 : spx_uint32_t out_rate;
113 : spx_uint32_t num_rate;
114 : spx_uint32_t den_rate;
115 :
116 : int quality;
117 : spx_uint32_t nb_channels;
118 : spx_uint32_t filt_len;
119 : spx_uint32_t mem_alloc_size;
120 : spx_uint32_t buffer_size;
121 : int int_advance;
122 : int frac_advance;
123 : float cutoff;
124 : spx_uint32_t oversample;
125 : int initialised;
126 : int started;
127 :
128 : /* These are per-channel */
129 : spx_int32_t *last_sample;
130 : spx_uint32_t *samp_frac_num;
131 : spx_uint32_t *magic_samples;
132 :
133 : spx_word16_t *mem;
134 : spx_word16_t *sinc_table;
135 : spx_uint32_t sinc_table_length;
136 : resampler_basic_func resampler_ptr;
137 :
138 : int in_stride;
139 : int out_stride;
140 : } ;
141 :
142 : static const double kaiser12_table[68] = {
143 : 0.99859849, 1.00000000, 0.99859849, 0.99440475, 0.98745105, 0.97779076,
144 : 0.96549770, 0.95066529, 0.93340547, 0.91384741, 0.89213598, 0.86843014,
145 : 0.84290116, 0.81573067, 0.78710866, 0.75723148, 0.72629970, 0.69451601,
146 : 0.66208321, 0.62920216, 0.59606986, 0.56287762, 0.52980938, 0.49704014,
147 : 0.46473455, 0.43304576, 0.40211431, 0.37206735, 0.34301800, 0.31506490,
148 : 0.28829195, 0.26276832, 0.23854851, 0.21567274, 0.19416736, 0.17404546,
149 : 0.15530766, 0.13794294, 0.12192957, 0.10723616, 0.09382272, 0.08164178,
150 : 0.07063950, 0.06075685, 0.05193064, 0.04409466, 0.03718069, 0.03111947,
151 : 0.02584161, 0.02127838, 0.01736250, 0.01402878, 0.01121463, 0.00886058,
152 : 0.00691064, 0.00531256, 0.00401805, 0.00298291, 0.00216702, 0.00153438,
153 : 0.00105297, 0.00069463, 0.00043489, 0.00025272, 0.00013031, 0.0000527734,
154 : 0.00001000, 0.00000000};
155 : /*
156 : static const double kaiser12_table[36] = {
157 : 0.99440475, 1.00000000, 0.99440475, 0.97779076, 0.95066529, 0.91384741,
158 : 0.86843014, 0.81573067, 0.75723148, 0.69451601, 0.62920216, 0.56287762,
159 : 0.49704014, 0.43304576, 0.37206735, 0.31506490, 0.26276832, 0.21567274,
160 : 0.17404546, 0.13794294, 0.10723616, 0.08164178, 0.06075685, 0.04409466,
161 : 0.03111947, 0.02127838, 0.01402878, 0.00886058, 0.00531256, 0.00298291,
162 : 0.00153438, 0.00069463, 0.00025272, 0.0000527734, 0.00000500, 0.00000000};
163 : */
164 : static const double kaiser10_table[36] = {
165 : 0.99537781, 1.00000000, 0.99537781, 0.98162644, 0.95908712, 0.92831446,
166 : 0.89005583, 0.84522401, 0.79486424, 0.74011713, 0.68217934, 0.62226347,
167 : 0.56155915, 0.50119680, 0.44221549, 0.38553619, 0.33194107, 0.28205962,
168 : 0.23636152, 0.19515633, 0.15859932, 0.12670280, 0.09935205, 0.07632451,
169 : 0.05731132, 0.04193980, 0.02979584, 0.02044510, 0.01345224, 0.00839739,
170 : 0.00488951, 0.00257636, 0.00115101, 0.00035515, 0.00000000, 0.00000000};
171 :
172 : static const double kaiser8_table[36] = {
173 : 0.99635258, 1.00000000, 0.99635258, 0.98548012, 0.96759014, 0.94302200,
174 : 0.91223751, 0.87580811, 0.83439927, 0.78875245, 0.73966538, 0.68797126,
175 : 0.63451750, 0.58014482, 0.52566725, 0.47185369, 0.41941150, 0.36897272,
176 : 0.32108304, 0.27619388, 0.23465776, 0.19672670, 0.16255380, 0.13219758,
177 : 0.10562887, 0.08273982, 0.06335451, 0.04724088, 0.03412321, 0.02369490,
178 : 0.01563093, 0.00959968, 0.00527363, 0.00233883, 0.00050000, 0.00000000};
179 :
180 : static const double kaiser6_table[36] = {
181 : 0.99733006, 1.00000000, 0.99733006, 0.98935595, 0.97618418, 0.95799003,
182 : 0.93501423, 0.90755855, 0.87598009, 0.84068475, 0.80211977, 0.76076565,
183 : 0.71712752, 0.67172623, 0.62508937, 0.57774224, 0.53019925, 0.48295561,
184 : 0.43647969, 0.39120616, 0.34752997, 0.30580127, 0.26632152, 0.22934058,
185 : 0.19505503, 0.16360756, 0.13508755, 0.10953262, 0.08693120, 0.06722600,
186 : 0.05031820, 0.03607231, 0.02432151, 0.01487334, 0.00752000, 0.00000000};
187 :
188 : struct FuncDef {
189 : const double *table;
190 : int oversample;
191 : };
192 :
193 : static const struct FuncDef _KAISER12 = {kaiser12_table, 64};
194 : #define KAISER12 (&_KAISER12)
195 : /*static struct FuncDef _KAISER12 = {kaiser12_table, 32};
196 : #define KAISER12 (&_KAISER12)*/
197 : static const struct FuncDef _KAISER10 = {kaiser10_table, 32};
198 : #define KAISER10 (&_KAISER10)
199 : static const struct FuncDef _KAISER8 = {kaiser8_table, 32};
200 : #define KAISER8 (&_KAISER8)
201 : static const struct FuncDef _KAISER6 = {kaiser6_table, 32};
202 : #define KAISER6 (&_KAISER6)
203 :
204 : struct QualityMapping {
205 : int base_length;
206 : int oversample;
207 : float downsample_bandwidth;
208 : float upsample_bandwidth;
209 : const struct FuncDef *window_func;
210 : };
211 :
212 :
213 : /* This table maps conversion quality to internal parameters. There are two
214 : reasons that explain why the up-sampling bandwidth is larger than the
215 : down-sampling bandwidth:
216 : 1) When up-sampling, we can assume that the spectrum is already attenuated
217 : close to the Nyquist rate (from an A/D or a previous resampling filter)
218 : 2) Any aliasing that occurs very close to the Nyquist rate will be masked
219 : by the sinusoids/noise just below the Nyquist rate (guaranteed only for
220 : up-sampling).
221 : */
222 : static const struct QualityMapping quality_map[11] = {
223 : { 8, 4, 0.830f, 0.860f, KAISER6 }, /* Q0 */
224 : { 16, 4, 0.850f, 0.880f, KAISER6 }, /* Q1 */
225 : { 32, 4, 0.882f, 0.910f, KAISER6 }, /* Q2 */ /* 82.3% cutoff ( ~60 dB stop) 6 */
226 : { 48, 8, 0.895f, 0.917f, KAISER8 }, /* Q3 */ /* 84.9% cutoff ( ~80 dB stop) 8 */
227 : { 64, 8, 0.921f, 0.940f, KAISER8 }, /* Q4 */ /* 88.7% cutoff ( ~80 dB stop) 8 */
228 : { 80, 16, 0.922f, 0.940f, KAISER10}, /* Q5 */ /* 89.1% cutoff (~100 dB stop) 10 */
229 : { 96, 16, 0.940f, 0.945f, KAISER10}, /* Q6 */ /* 91.5% cutoff (~100 dB stop) 10 */
230 : {128, 16, 0.950f, 0.950f, KAISER10}, /* Q7 */ /* 93.1% cutoff (~100 dB stop) 10 */
231 : {160, 16, 0.960f, 0.960f, KAISER10}, /* Q8 */ /* 94.5% cutoff (~100 dB stop) 10 */
232 : {192, 32, 0.968f, 0.968f, KAISER12}, /* Q9 */ /* 95.5% cutoff (~100 dB stop) 10 */
233 : {256, 32, 0.975f, 0.975f, KAISER12}, /* Q10 */ /* 96.6% cutoff (~100 dB stop) 10 */
234 : };
235 : /*8,24,40,56,80,104,128,160,200,256,320*/
236 0 : static double compute_func(float x, const struct FuncDef *func)
237 : {
238 : float y, frac;
239 : double interp[4];
240 : int ind;
241 0 : y = x*func->oversample;
242 0 : ind = (int)floor(y);
243 0 : frac = (y-ind);
244 : /* CSE with handle the repeated powers */
245 0 : interp[3] = -0.1666666667*frac + 0.1666666667*(frac*frac*frac);
246 0 : interp[2] = frac + 0.5*(frac*frac) - 0.5*(frac*frac*frac);
247 : /*interp[2] = 1.f - 0.5f*frac - frac*frac + 0.5f*frac*frac*frac;*/
248 0 : interp[0] = -0.3333333333*frac + 0.5*(frac*frac) - 0.1666666667*(frac*frac*frac);
249 : /* Just to make sure we don't have rounding problems */
250 0 : interp[1] = 1.f-interp[3]-interp[2]-interp[0];
251 :
252 : /*sum = frac*accum[1] + (1-frac)*accum[2];*/
253 0 : return interp[0]*func->table[ind] + interp[1]*func->table[ind+1] + interp[2]*func->table[ind+2] + interp[3]*func->table[ind+3];
254 : }
255 :
256 : #if 0
257 : #include <stdio.h>
258 : int main(int argc, char **argv)
259 : {
260 : int i;
261 : for (i=0;i<256;i++)
262 : {
263 : printf ("%f\n", compute_func(i/256., KAISER12));
264 : }
265 : return 0;
266 : }
267 : #endif
268 :
269 : #ifdef FIXED_POINT
270 : /* The slow way of computing a sinc for the table. Should improve that some day */
271 : static spx_word16_t sinc(float cutoff, float x, int N, const struct FuncDef *window_func)
272 : {
273 : /*fprintf (stderr, "%f ", x);*/
274 : float xx = x * cutoff;
275 : if (fabs(x)<1e-6f)
276 : return WORD2INT(32768.*cutoff);
277 : else if (fabs(x) > .5f*N)
278 : return 0;
279 : /*FIXME: Can it really be any slower than this? */
280 : return WORD2INT(32768.*cutoff*sin(M_PI*xx)/(M_PI*xx) * compute_func(fabs(2.*x/N), window_func));
281 : }
282 : #else
283 : /* The slow way of computing a sinc for the table. Should improve that some day */
284 0 : static spx_word16_t sinc(float cutoff, float x, int N, const struct FuncDef *window_func)
285 : {
286 : /*fprintf (stderr, "%f ", x);*/
287 0 : float xx = x * cutoff;
288 0 : if (fabs(x)<1e-6)
289 0 : return cutoff;
290 0 : else if (fabs(x) > .5*N)
291 0 : return 0;
292 : /*FIXME: Can it really be any slower than this? */
293 0 : return cutoff*sin(M_PI*xx)/(M_PI*xx) * compute_func(fabs(2.*x/N), window_func);
294 : }
295 : #endif
296 :
297 : #ifdef FIXED_POINT
298 : static void cubic_coef(spx_word16_t x, spx_word16_t interp[4])
299 : {
300 : /* Compute interpolation coefficients. I'm not sure whether this corresponds to cubic interpolation
301 : but I know it's MMSE-optimal on a sinc */
302 : spx_word16_t x2, x3;
303 : x2 = MULT16_16_P15(x, x);
304 : x3 = MULT16_16_P15(x, x2);
305 : interp[0] = PSHR32(MULT16_16(QCONST16(-0.16667f, 15),x) + MULT16_16(QCONST16(0.16667f, 15),x3),15);
306 : interp[1] = EXTRACT16(EXTEND32(x) + SHR32(SUB32(EXTEND32(x2),EXTEND32(x3)),1));
307 : interp[3] = PSHR32(MULT16_16(QCONST16(-0.33333f, 15),x) + MULT16_16(QCONST16(.5f,15),x2) - MULT16_16(QCONST16(0.16667f, 15),x3),15);
308 : /* Just to make sure we don't have rounding problems */
309 : interp[2] = Q15_ONE-interp[0]-interp[1]-interp[3];
310 : if (interp[2]<32767)
311 : interp[2]+=1;
312 : }
313 : #else
314 0 : static void cubic_coef(spx_word16_t frac, spx_word16_t interp[4])
315 : {
316 : /* Compute interpolation coefficients. I'm not sure whether this corresponds to cubic interpolation
317 : but I know it's MMSE-optimal on a sinc */
318 0 : interp[0] = -0.16667f*frac + 0.16667f*frac*frac*frac;
319 0 : interp[1] = frac + 0.5f*frac*frac - 0.5f*frac*frac*frac;
320 : /*interp[2] = 1.f - 0.5f*frac - frac*frac + 0.5f*frac*frac*frac;*/
321 0 : interp[3] = -0.33333f*frac + 0.5f*frac*frac - 0.16667f*frac*frac*frac;
322 : /* Just to make sure we don't have rounding problems */
323 0 : interp[2] = 1.-interp[0]-interp[1]-interp[3];
324 0 : }
325 : #endif
326 :
327 0 : static int resampler_basic_direct_single(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_word16_t *in, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
328 : {
329 0 : const int N = st->filt_len;
330 0 : int out_sample = 0;
331 0 : int last_sample = st->last_sample[channel_index];
332 0 : spx_uint32_t samp_frac_num = st->samp_frac_num[channel_index];
333 0 : const spx_word16_t *sinc_table = st->sinc_table;
334 0 : const int out_stride = st->out_stride;
335 0 : const int int_advance = st->int_advance;
336 0 : const int frac_advance = st->frac_advance;
337 0 : const spx_uint32_t den_rate = st->den_rate;
338 : spx_word32_t sum;
339 :
340 0 : while (!(last_sample >= (spx_int32_t)*in_len || out_sample >= (spx_int32_t)*out_len))
341 : {
342 0 : const spx_word16_t *sinct = & sinc_table[samp_frac_num*N];
343 0 : const spx_word16_t *iptr = & in[last_sample];
344 :
345 : #ifdef OVERRIDE_INNER_PRODUCT_SINGLE
346 0 : if (!moz_speex_have_single_simd()) {
347 : #endif
348 : int j;
349 0 : sum = 0;
350 0 : for(j=0;j<N;j++) sum += MULT16_16(sinct[j], iptr[j]);
351 :
352 : /* This code is slower on most DSPs which have only 2 accumulators.
353 : Plus this this forces truncation to 32 bits and you lose the HW guard bits.
354 : I think we can trust the compiler and let it vectorize and/or unroll itself.
355 : spx_word32_t accum[4] = {0,0,0,0};
356 : for(j=0;j<N;j+=4) {
357 : accum[0] += MULT16_16(sinct[j], iptr[j]);
358 : accum[1] += MULT16_16(sinct[j+1], iptr[j+1]);
359 : accum[2] += MULT16_16(sinct[j+2], iptr[j+2]);
360 : accum[3] += MULT16_16(sinct[j+3], iptr[j+3]);
361 : }
362 : sum = accum[0] + accum[1] + accum[2] + accum[3];
363 : */
364 0 : sum = SATURATE32PSHR(sum, 15, 32767);
365 : #ifdef OVERRIDE_INNER_PRODUCT_SINGLE
366 : } else {
367 0 : sum = inner_product_single(sinct, iptr, N);
368 : }
369 : #endif
370 :
371 0 : out[out_stride * out_sample++] = sum;
372 0 : last_sample += int_advance;
373 0 : samp_frac_num += frac_advance;
374 0 : if (samp_frac_num >= den_rate)
375 : {
376 0 : samp_frac_num -= den_rate;
377 0 : last_sample++;
378 : }
379 : }
380 :
381 0 : st->last_sample[channel_index] = last_sample;
382 0 : st->samp_frac_num[channel_index] = samp_frac_num;
383 0 : return out_sample;
384 : }
385 :
386 : #ifdef FIXED_POINT
387 : #else
388 : /* This is the same as the previous function, except with a double-precision accumulator */
389 0 : static int resampler_basic_direct_double(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_word16_t *in, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
390 : {
391 0 : const int N = st->filt_len;
392 0 : int out_sample = 0;
393 0 : int last_sample = st->last_sample[channel_index];
394 0 : spx_uint32_t samp_frac_num = st->samp_frac_num[channel_index];
395 0 : const spx_word16_t *sinc_table = st->sinc_table;
396 0 : const int out_stride = st->out_stride;
397 0 : const int int_advance = st->int_advance;
398 0 : const int frac_advance = st->frac_advance;
399 0 : const spx_uint32_t den_rate = st->den_rate;
400 : double sum;
401 :
402 0 : while (!(last_sample >= (spx_int32_t)*in_len || out_sample >= (spx_int32_t)*out_len))
403 : {
404 0 : const spx_word16_t *sinct = & sinc_table[samp_frac_num*N];
405 0 : const spx_word16_t *iptr = & in[last_sample];
406 :
407 : #ifdef OVERRIDE_INNER_PRODUCT_DOUBLE
408 0 : if(moz_speex_have_double_simd()) {
409 : #endif
410 : int j;
411 0 : double accum[4] = {0,0,0,0};
412 :
413 0 : for(j=0;j<N;j+=4) {
414 0 : accum[0] += sinct[j]*iptr[j];
415 0 : accum[1] += sinct[j+1]*iptr[j+1];
416 0 : accum[2] += sinct[j+2]*iptr[j+2];
417 0 : accum[3] += sinct[j+3]*iptr[j+3];
418 : }
419 0 : sum = accum[0] + accum[1] + accum[2] + accum[3];
420 : #ifdef OVERRIDE_INNER_PRODUCT_DOUBLE
421 : } else {
422 0 : sum = inner_product_double(sinct, iptr, N);
423 : }
424 : #endif
425 :
426 0 : out[out_stride * out_sample++] = PSHR32(sum, 15);
427 0 : last_sample += int_advance;
428 0 : samp_frac_num += frac_advance;
429 0 : if (samp_frac_num >= den_rate)
430 : {
431 0 : samp_frac_num -= den_rate;
432 0 : last_sample++;
433 : }
434 : }
435 :
436 0 : st->last_sample[channel_index] = last_sample;
437 0 : st->samp_frac_num[channel_index] = samp_frac_num;
438 0 : return out_sample;
439 : }
440 : #endif
441 :
442 0 : static int resampler_basic_interpolate_single(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_word16_t *in, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
443 : {
444 0 : const int N = st->filt_len;
445 0 : int out_sample = 0;
446 0 : int last_sample = st->last_sample[channel_index];
447 0 : spx_uint32_t samp_frac_num = st->samp_frac_num[channel_index];
448 0 : const int out_stride = st->out_stride;
449 0 : const int int_advance = st->int_advance;
450 0 : const int frac_advance = st->frac_advance;
451 0 : const spx_uint32_t den_rate = st->den_rate;
452 : spx_word32_t sum;
453 :
454 0 : while (!(last_sample >= (spx_int32_t)*in_len || out_sample >= (spx_int32_t)*out_len))
455 : {
456 0 : const spx_word16_t *iptr = & in[last_sample];
457 :
458 0 : const int offset = samp_frac_num*st->oversample/st->den_rate;
459 : #ifdef FIXED_POINT
460 : const spx_word16_t frac = PDIV32(SHL32((samp_frac_num*st->oversample) % st->den_rate,15),st->den_rate);
461 : #else
462 0 : const spx_word16_t frac = ((float)((samp_frac_num*st->oversample) % st->den_rate))/st->den_rate;
463 : #endif
464 : spx_word16_t interp[4];
465 :
466 :
467 : #ifdef OVERRIDE_INTERPOLATE_PRODUCT_SINGLE
468 0 : if (!moz_speex_have_single_simd()) {
469 : #endif
470 : int j;
471 0 : spx_word32_t accum[4] = {0,0,0,0};
472 :
473 0 : for(j=0;j<N;j++) {
474 0 : const spx_word16_t curr_in=iptr[j];
475 0 : accum[0] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset-2]);
476 0 : accum[1] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset-1]);
477 0 : accum[2] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset]);
478 0 : accum[3] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset+1]);
479 : }
480 :
481 0 : cubic_coef(frac, interp);
482 0 : sum = MULT16_32_Q15(interp[0],SHR32(accum[0], 1)) + MULT16_32_Q15(interp[1],SHR32(accum[1], 1)) + MULT16_32_Q15(interp[2],SHR32(accum[2], 1)) + MULT16_32_Q15(interp[3],SHR32(accum[3], 1));
483 0 : sum = SATURATE32PSHR(sum, 15, 32767);
484 : #ifdef OVERRIDE_INTERPOLATE_PRODUCT_SINGLE
485 : } else {
486 0 : cubic_coef(frac, interp);
487 0 : sum = interpolate_product_single(iptr, st->sinc_table + st->oversample + 4 - offset - 2, N, st->oversample, interp);
488 : }
489 : #endif
490 :
491 0 : out[out_stride * out_sample++] = sum;
492 0 : last_sample += int_advance;
493 0 : samp_frac_num += frac_advance;
494 0 : if (samp_frac_num >= den_rate)
495 : {
496 0 : samp_frac_num -= den_rate;
497 0 : last_sample++;
498 : }
499 : }
500 :
501 0 : st->last_sample[channel_index] = last_sample;
502 0 : st->samp_frac_num[channel_index] = samp_frac_num;
503 0 : return out_sample;
504 : }
505 :
506 : #ifdef FIXED_POINT
507 : #else
508 : /* This is the same as the previous function, except with a double-precision accumulator */
509 0 : static int resampler_basic_interpolate_double(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_word16_t *in, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
510 : {
511 0 : const int N = st->filt_len;
512 0 : int out_sample = 0;
513 0 : int last_sample = st->last_sample[channel_index];
514 0 : spx_uint32_t samp_frac_num = st->samp_frac_num[channel_index];
515 0 : const int out_stride = st->out_stride;
516 0 : const int int_advance = st->int_advance;
517 0 : const int frac_advance = st->frac_advance;
518 0 : const spx_uint32_t den_rate = st->den_rate;
519 : spx_word32_t sum;
520 :
521 0 : while (!(last_sample >= (spx_int32_t)*in_len || out_sample >= (spx_int32_t)*out_len))
522 : {
523 0 : const spx_word16_t *iptr = & in[last_sample];
524 :
525 0 : const int offset = samp_frac_num*st->oversample/st->den_rate;
526 : #ifdef FIXED_POINT
527 : const spx_word16_t frac = PDIV32(SHL32((samp_frac_num*st->oversample) % st->den_rate,15),st->den_rate);
528 : #else
529 0 : const spx_word16_t frac = ((float)((samp_frac_num*st->oversample) % st->den_rate))/st->den_rate;
530 : #endif
531 : spx_word16_t interp[4];
532 :
533 :
534 : #ifdef OVERRIDE_INTERPOLATE_PRODUCT_DOUBLE
535 0 : if (!moz_speex_have_double_simd()) {
536 : #endif
537 : int j;
538 0 : double accum[4] = {0,0,0,0};
539 :
540 0 : for(j=0;j<N;j++) {
541 0 : const double curr_in=iptr[j];
542 0 : accum[0] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset-2]);
543 0 : accum[1] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset-1]);
544 0 : accum[2] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset]);
545 0 : accum[3] += MULT16_16(curr_in,st->sinc_table[4+(j+1)*st->oversample-offset+1]);
546 : }
547 :
548 0 : cubic_coef(frac, interp);
549 0 : sum = MULT16_32_Q15(interp[0],accum[0]) + MULT16_32_Q15(interp[1],accum[1]) + MULT16_32_Q15(interp[2],accum[2]) + MULT16_32_Q15(interp[3],accum[3]);
550 : #ifdef OVERRIDE_INTERPOLATE_PRODUCT_DOUBLE
551 : } else {
552 0 : cubic_coef(frac, interp);
553 0 : sum = interpolate_product_double(iptr, st->sinc_table + st->oversample + 4 - offset - 2, N, st->oversample, interp);
554 : }
555 : #endif
556 :
557 0 : out[out_stride * out_sample++] = PSHR32(sum,15);
558 0 : last_sample += int_advance;
559 0 : samp_frac_num += frac_advance;
560 0 : if (samp_frac_num >= den_rate)
561 : {
562 0 : samp_frac_num -= den_rate;
563 0 : last_sample++;
564 : }
565 : }
566 :
567 0 : st->last_sample[channel_index] = last_sample;
568 0 : st->samp_frac_num[channel_index] = samp_frac_num;
569 0 : return out_sample;
570 : }
571 : #endif
572 :
573 : /* This resampler is used to produce zero output in situations where memory
574 : for the filter could not be allocated. The expected numbers of input and
575 : output samples are still processed so that callers failing to check error
576 : codes are not surprised, possibly getting into infinite loops. */
577 0 : static int resampler_basic_zero(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_word16_t *in, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
578 : {
579 0 : int out_sample = 0;
580 0 : int last_sample = st->last_sample[channel_index];
581 0 : spx_uint32_t samp_frac_num = st->samp_frac_num[channel_index];
582 0 : const int out_stride = st->out_stride;
583 0 : const int int_advance = st->int_advance;
584 0 : const int frac_advance = st->frac_advance;
585 0 : const spx_uint32_t den_rate = st->den_rate;
586 :
587 0 : while (!(last_sample >= (spx_int32_t)*in_len || out_sample >= (spx_int32_t)*out_len))
588 : {
589 0 : out[out_stride * out_sample++] = 0;
590 0 : last_sample += int_advance;
591 0 : samp_frac_num += frac_advance;
592 0 : if (samp_frac_num >= den_rate)
593 : {
594 0 : samp_frac_num -= den_rate;
595 0 : last_sample++;
596 : }
597 : }
598 :
599 0 : st->last_sample[channel_index] = last_sample;
600 0 : st->samp_frac_num[channel_index] = samp_frac_num;
601 0 : return out_sample;
602 : }
603 :
604 0 : static int _muldiv(spx_uint32_t *result, spx_uint32_t value, spx_uint32_t mul, spx_uint32_t div)
605 : {
606 : speex_assert(result);
607 0 : spx_uint32_t major = value / div;
608 0 : spx_uint32_t remainder = value % div;
609 : /* TODO: Could use 64 bits operation to check for overflow. But only guaranteed in C99+ */
610 0 : if (remainder > UINT32_MAX / mul || major > UINT32_MAX / mul
611 0 : || major * mul > UINT32_MAX - remainder * mul / div)
612 0 : return RESAMPLER_ERR_OVERFLOW;
613 0 : *result = remainder * mul / div + major * mul;
614 0 : return RESAMPLER_ERR_SUCCESS;
615 : }
616 :
617 0 : static int update_filter(SpeexResamplerState *st)
618 : {
619 0 : spx_uint32_t old_length = st->filt_len;
620 0 : spx_uint32_t old_alloc_size = st->mem_alloc_size;
621 : int use_direct;
622 : spx_uint32_t min_sinc_table_length;
623 : spx_uint32_t min_alloc_size;
624 :
625 0 : st->int_advance = st->num_rate/st->den_rate;
626 0 : st->frac_advance = st->num_rate%st->den_rate;
627 0 : st->oversample = quality_map[st->quality].oversample;
628 0 : st->filt_len = quality_map[st->quality].base_length;
629 :
630 0 : if (st->num_rate > st->den_rate)
631 : {
632 : /* down-sampling */
633 0 : st->cutoff = quality_map[st->quality].downsample_bandwidth * st->den_rate / st->num_rate;
634 0 : if (_muldiv(&st->filt_len,st->filt_len,st->num_rate,st->den_rate) != RESAMPLER_ERR_SUCCESS)
635 0 : goto fail;
636 : /* Round up to make sure we have a multiple of 8 for SSE */
637 0 : st->filt_len = ((st->filt_len-1)&(~0x7))+8;
638 0 : if (2*st->den_rate < st->num_rate)
639 0 : st->oversample >>= 1;
640 0 : if (4*st->den_rate < st->num_rate)
641 0 : st->oversample >>= 1;
642 0 : if (8*st->den_rate < st->num_rate)
643 0 : st->oversample >>= 1;
644 0 : if (16*st->den_rate < st->num_rate)
645 0 : st->oversample >>= 1;
646 0 : if (st->oversample < 1)
647 0 : st->oversample = 1;
648 : } else {
649 : /* up-sampling */
650 0 : st->cutoff = quality_map[st->quality].upsample_bandwidth;
651 : }
652 :
653 0 : use_direct =
654 : #ifdef RESAMPLE_HUGEMEM
655 : /* Choose the direct resampler, even with higher initialization costs,
656 : when resampling any multiple of 100 to 44100. */
657 0 : st->den_rate <= 441
658 : #else
659 : /* Choose the resampling type that requires the least amount of memory */
660 : st->filt_len*st->den_rate <= st->filt_len*st->oversample+8
661 : #endif
662 0 : && INT_MAX/sizeof(spx_word16_t)/st->den_rate >= st->filt_len;
663 0 : if (use_direct)
664 : {
665 0 : min_sinc_table_length = st->filt_len*st->den_rate;
666 : } else {
667 0 : if ((INT_MAX/sizeof(spx_word16_t)-8)/st->oversample < st->filt_len)
668 0 : goto fail;
669 :
670 0 : min_sinc_table_length = st->filt_len*st->oversample+8;
671 : }
672 0 : if (st->sinc_table_length < min_sinc_table_length)
673 : {
674 0 : spx_word16_t *sinc_table = (spx_word16_t *)speex_realloc(st->sinc_table,min_sinc_table_length*sizeof(spx_word16_t));
675 0 : if (!sinc_table)
676 0 : goto fail;
677 :
678 0 : st->sinc_table = sinc_table;
679 0 : st->sinc_table_length = min_sinc_table_length;
680 : }
681 0 : if (use_direct)
682 : {
683 : spx_uint32_t i;
684 0 : for (i=0;i<st->den_rate;i++)
685 : {
686 : spx_int32_t j;
687 0 : for (j=0;j<st->filt_len;j++)
688 : {
689 0 : st->sinc_table[i*st->filt_len+j] = sinc(st->cutoff,((j-(spx_int32_t)st->filt_len/2+1)-((float)i)/st->den_rate), st->filt_len, quality_map[st->quality].window_func);
690 : }
691 : }
692 : #ifdef FIXED_POINT
693 : st->resampler_ptr = resampler_basic_direct_single;
694 : #else
695 0 : if (st->quality>8)
696 0 : st->resampler_ptr = resampler_basic_direct_double;
697 : else
698 0 : st->resampler_ptr = resampler_basic_direct_single;
699 : #endif
700 : /*fprintf (stderr, "resampler uses direct sinc table and normalised cutoff %f\n", cutoff);*/
701 : } else {
702 : spx_int32_t i;
703 0 : for (i=-4;i<(spx_int32_t)(st->oversample*st->filt_len+4);i++)
704 0 : st->sinc_table[i+4] = sinc(st->cutoff,(i/(float)st->oversample - st->filt_len/2), st->filt_len, quality_map[st->quality].window_func);
705 : #ifdef FIXED_POINT
706 : st->resampler_ptr = resampler_basic_interpolate_single;
707 : #else
708 0 : if (st->quality>8)
709 0 : st->resampler_ptr = resampler_basic_interpolate_double;
710 : else
711 0 : st->resampler_ptr = resampler_basic_interpolate_single;
712 : #endif
713 : /*fprintf (stderr, "resampler uses interpolated sinc table and normalised cutoff %f\n", cutoff);*/
714 : }
715 :
716 : /* Here's the place where we update the filter memory to take into account
717 : the change in filter length. It's probably the messiest part of the code
718 : due to handling of lots of corner cases. */
719 :
720 : /* Adding buffer_size to filt_len won't overflow here because filt_len
721 : could be multiplied by sizeof(spx_word16_t) above. */
722 0 : min_alloc_size = st->filt_len-1 + st->buffer_size;
723 0 : if (min_alloc_size > st->mem_alloc_size)
724 : {
725 : spx_word16_t *mem;
726 0 : if (INT_MAX/sizeof(spx_word16_t)/st->nb_channels < min_alloc_size)
727 0 : goto fail;
728 0 : else if (!(mem = (spx_word16_t*)speex_realloc(st->mem, st->nb_channels*min_alloc_size * sizeof(*mem))))
729 0 : goto fail;
730 :
731 0 : st->mem = mem;
732 0 : st->mem_alloc_size = min_alloc_size;
733 : }
734 0 : if (!st->started)
735 : {
736 : spx_uint32_t i;
737 0 : for (i=0;i<st->nb_channels*st->mem_alloc_size;i++)
738 0 : st->mem[i] = 0;
739 : /*speex_warning("reinit filter");*/
740 0 : } else if (st->filt_len > old_length)
741 : {
742 : spx_uint32_t i;
743 : /* Increase the filter length */
744 : /*speex_warning("increase filter size");*/
745 0 : for (i=st->nb_channels;i--;)
746 : {
747 : spx_uint32_t j;
748 0 : spx_uint32_t olen = old_length;
749 : /*if (st->magic_samples[i])*/
750 : {
751 : /* Try and remove the magic samples as if nothing had happened */
752 :
753 : /* FIXME: This is wrong but for now we need it to avoid going over the array bounds */
754 0 : olen = old_length + 2*st->magic_samples[i];
755 0 : for (j=old_length-1+st->magic_samples[i];j--;)
756 0 : st->mem[i*st->mem_alloc_size+j+st->magic_samples[i]] = st->mem[i*old_alloc_size+j];
757 0 : for (j=0;j<st->magic_samples[i];j++)
758 0 : st->mem[i*st->mem_alloc_size+j] = 0;
759 0 : st->magic_samples[i] = 0;
760 : }
761 0 : if (st->filt_len > olen)
762 : {
763 : /* If the new filter length is still bigger than the "augmented" length */
764 : /* Copy data going backward */
765 0 : for (j=0;j<olen-1;j++)
766 0 : st->mem[i*st->mem_alloc_size+(st->filt_len-2-j)] = st->mem[i*st->mem_alloc_size+(olen-2-j)];
767 : /* Then put zeros for lack of anything better */
768 0 : for (;j<st->filt_len-1;j++)
769 0 : st->mem[i*st->mem_alloc_size+(st->filt_len-2-j)] = 0;
770 : /* Adjust last_sample */
771 0 : st->last_sample[i] += (st->filt_len - olen)/2;
772 : } else {
773 : /* Put back some of the magic! */
774 0 : st->magic_samples[i] = (olen - st->filt_len)/2;
775 0 : for (j=0;j<st->filt_len-1+st->magic_samples[i];j++)
776 0 : st->mem[i*st->mem_alloc_size+j] = st->mem[i*st->mem_alloc_size+j+st->magic_samples[i]];
777 : }
778 : }
779 0 : } else if (st->filt_len < old_length)
780 : {
781 : spx_uint32_t i;
782 : /* Reduce filter length, this a bit tricky. We need to store some of the memory as "magic"
783 : samples so they can be used directly as input the next time(s) */
784 0 : for (i=0;i<st->nb_channels;i++)
785 : {
786 : spx_uint32_t j;
787 0 : spx_uint32_t old_magic = st->magic_samples[i];
788 0 : st->magic_samples[i] = (old_length - st->filt_len)/2;
789 : /* We must copy some of the memory that's no longer used */
790 : /* Copy data going backward */
791 0 : for (j=0;j<st->filt_len-1+st->magic_samples[i]+old_magic;j++)
792 0 : st->mem[i*st->mem_alloc_size+j] = st->mem[i*st->mem_alloc_size+j+st->magic_samples[i]];
793 0 : st->magic_samples[i] += old_magic;
794 : }
795 : }
796 0 : return RESAMPLER_ERR_SUCCESS;
797 :
798 : fail:
799 0 : st->resampler_ptr = resampler_basic_zero;
800 : /* st->mem may still contain consumed input samples for the filter.
801 : Restore filt_len so that filt_len - 1 still points to the position after
802 : the last of these samples. */
803 0 : st->filt_len = old_length;
804 0 : return RESAMPLER_ERR_ALLOC_FAILED;
805 : }
806 :
807 0 : EXPORT SpeexResamplerState *speex_resampler_init(spx_uint32_t nb_channels, spx_uint32_t in_rate, spx_uint32_t out_rate, int quality, int *err)
808 : {
809 0 : return speex_resampler_init_frac(nb_channels, in_rate, out_rate, in_rate, out_rate, quality, err);
810 : }
811 :
812 0 : EXPORT SpeexResamplerState *speex_resampler_init_frac(spx_uint32_t nb_channels, spx_uint32_t ratio_num, spx_uint32_t ratio_den, spx_uint32_t in_rate, spx_uint32_t out_rate, int quality, int *err)
813 : {
814 : spx_uint32_t i;
815 : SpeexResamplerState *st;
816 : int filter_err;
817 :
818 0 : if (nb_channels == 0 || ratio_num == 0 || ratio_den == 0 || quality > 10 || quality < 0)
819 : {
820 0 : if (err)
821 0 : *err = RESAMPLER_ERR_INVALID_ARG;
822 0 : return NULL;
823 : }
824 0 : st = (SpeexResamplerState *)speex_alloc(sizeof(SpeexResamplerState));
825 0 : if (!st)
826 : {
827 0 : if (err)
828 0 : *err = RESAMPLER_ERR_ALLOC_FAILED;
829 0 : return NULL;
830 : }
831 0 : st->initialised = 0;
832 0 : st->started = 0;
833 0 : st->in_rate = 0;
834 0 : st->out_rate = 0;
835 0 : st->num_rate = 0;
836 0 : st->den_rate = 0;
837 0 : st->quality = -1;
838 0 : st->sinc_table_length = 0;
839 0 : st->mem_alloc_size = 0;
840 0 : st->filt_len = 0;
841 0 : st->mem = 0;
842 0 : st->resampler_ptr = 0;
843 :
844 0 : st->cutoff = 1.f;
845 0 : st->nb_channels = nb_channels;
846 0 : st->in_stride = 1;
847 0 : st->out_stride = 1;
848 :
849 0 : st->buffer_size = 160;
850 :
851 : /* Per channel data */
852 0 : if (!(st->last_sample = (spx_int32_t*)speex_alloc(nb_channels*sizeof(spx_int32_t))))
853 0 : goto fail;
854 0 : if (!(st->magic_samples = (spx_uint32_t*)speex_alloc(nb_channels*sizeof(spx_uint32_t))))
855 0 : goto fail;
856 0 : if (!(st->samp_frac_num = (spx_uint32_t*)speex_alloc(nb_channels*sizeof(spx_uint32_t))))
857 0 : goto fail;
858 :
859 0 : speex_resampler_set_quality(st, quality);
860 0 : speex_resampler_set_rate_frac(st, ratio_num, ratio_den, in_rate, out_rate);
861 :
862 0 : filter_err = update_filter(st);
863 0 : if (filter_err == RESAMPLER_ERR_SUCCESS)
864 : {
865 0 : st->initialised = 1;
866 : } else {
867 0 : speex_resampler_destroy(st);
868 0 : st = NULL;
869 : }
870 0 : if (err)
871 0 : *err = filter_err;
872 :
873 0 : return st;
874 :
875 : fail:
876 0 : if (err)
877 0 : *err = RESAMPLER_ERR_ALLOC_FAILED;
878 0 : speex_resampler_destroy(st);
879 0 : return NULL;
880 : }
881 :
882 0 : EXPORT void speex_resampler_destroy(SpeexResamplerState *st)
883 : {
884 0 : speex_free(st->mem);
885 0 : speex_free(st->sinc_table);
886 0 : speex_free(st->last_sample);
887 0 : speex_free(st->magic_samples);
888 0 : speex_free(st->samp_frac_num);
889 0 : speex_free(st);
890 0 : }
891 :
892 0 : static int speex_resampler_process_native(SpeexResamplerState *st, spx_uint32_t channel_index, spx_uint32_t *in_len, spx_word16_t *out, spx_uint32_t *out_len)
893 : {
894 0 : int j=0;
895 0 : const int N = st->filt_len;
896 0 : int out_sample = 0;
897 0 : spx_word16_t *mem = st->mem + channel_index * st->mem_alloc_size;
898 : spx_uint32_t ilen;
899 :
900 0 : st->started = 1;
901 :
902 : /* Call the right resampler through the function ptr */
903 0 : out_sample = st->resampler_ptr(st, channel_index, mem, in_len, out, out_len);
904 :
905 0 : if (st->last_sample[channel_index] < (spx_int32_t)*in_len)
906 0 : *in_len = st->last_sample[channel_index];
907 0 : *out_len = out_sample;
908 0 : st->last_sample[channel_index] -= *in_len;
909 :
910 0 : ilen = *in_len;
911 :
912 0 : for(j=0;j<N-1;++j)
913 0 : mem[j] = mem[j+ilen];
914 :
915 0 : return RESAMPLER_ERR_SUCCESS;
916 : }
917 :
918 0 : static int speex_resampler_magic(SpeexResamplerState *st, spx_uint32_t channel_index, spx_word16_t **out, spx_uint32_t out_len) {
919 0 : spx_uint32_t tmp_in_len = st->magic_samples[channel_index];
920 0 : spx_word16_t *mem = st->mem + channel_index * st->mem_alloc_size;
921 0 : const int N = st->filt_len;
922 :
923 0 : speex_resampler_process_native(st, channel_index, &tmp_in_len, *out, &out_len);
924 :
925 0 : st->magic_samples[channel_index] -= tmp_in_len;
926 :
927 : /* If we couldn't process all "magic" input samples, save the rest for next time */
928 0 : if (st->magic_samples[channel_index])
929 : {
930 : spx_uint32_t i;
931 0 : for (i=0;i<st->magic_samples[channel_index];i++)
932 0 : mem[N-1+i]=mem[N-1+i+tmp_in_len];
933 : }
934 0 : *out += out_len*st->out_stride;
935 0 : return out_len;
936 : }
937 :
938 : #ifdef FIXED_POINT
939 : EXPORT int speex_resampler_process_int(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_int16_t *in, spx_uint32_t *in_len, spx_int16_t *out, spx_uint32_t *out_len)
940 : #else
941 0 : EXPORT int speex_resampler_process_float(SpeexResamplerState *st, spx_uint32_t channel_index, const float *in, spx_uint32_t *in_len, float *out, spx_uint32_t *out_len)
942 : #endif
943 : {
944 : int j;
945 0 : spx_uint32_t ilen = *in_len;
946 0 : spx_uint32_t olen = *out_len;
947 0 : spx_word16_t *x = st->mem + channel_index * st->mem_alloc_size;
948 0 : const int filt_offs = st->filt_len - 1;
949 0 : const spx_uint32_t xlen = st->mem_alloc_size - filt_offs;
950 0 : const int istride = st->in_stride;
951 :
952 0 : if (st->magic_samples[channel_index])
953 0 : olen -= speex_resampler_magic(st, channel_index, &out, olen);
954 0 : if (! st->magic_samples[channel_index]) {
955 0 : while (ilen && olen) {
956 0 : spx_uint32_t ichunk = (ilen > xlen) ? xlen : ilen;
957 0 : spx_uint32_t ochunk = olen;
958 :
959 0 : if (in) {
960 0 : for(j=0;j<ichunk;++j)
961 0 : x[j+filt_offs]=in[j*istride];
962 : } else {
963 0 : for(j=0;j<ichunk;++j)
964 0 : x[j+filt_offs]=0;
965 : }
966 0 : speex_resampler_process_native(st, channel_index, &ichunk, out, &ochunk);
967 0 : ilen -= ichunk;
968 0 : olen -= ochunk;
969 0 : out += ochunk * st->out_stride;
970 0 : if (in)
971 0 : in += ichunk * istride;
972 : }
973 : }
974 0 : *in_len -= ilen;
975 0 : *out_len -= olen;
976 0 : return st->resampler_ptr == resampler_basic_zero ? RESAMPLER_ERR_ALLOC_FAILED : RESAMPLER_ERR_SUCCESS;
977 : }
978 :
979 : #ifdef FIXED_POINT
980 : EXPORT int speex_resampler_process_float(SpeexResamplerState *st, spx_uint32_t channel_index, const float *in, spx_uint32_t *in_len, float *out, spx_uint32_t *out_len)
981 : #else
982 0 : EXPORT int speex_resampler_process_int(SpeexResamplerState *st, spx_uint32_t channel_index, const spx_int16_t *in, spx_uint32_t *in_len, spx_int16_t *out, spx_uint32_t *out_len)
983 : #endif
984 : {
985 : int j;
986 0 : const int istride_save = st->in_stride;
987 0 : const int ostride_save = st->out_stride;
988 0 : spx_uint32_t ilen = *in_len;
989 0 : spx_uint32_t olen = *out_len;
990 0 : spx_word16_t *x = st->mem + channel_index * st->mem_alloc_size;
991 0 : const spx_uint32_t xlen = st->mem_alloc_size - (st->filt_len - 1);
992 : #ifdef VAR_ARRAYS
993 : const unsigned int ylen = (olen < FIXED_STACK_ALLOC) ? olen : FIXED_STACK_ALLOC;
994 : VARDECL(spx_word16_t *ystack);
995 : ALLOC(ystack, ylen, spx_word16_t);
996 : #else
997 0 : const unsigned int ylen = FIXED_STACK_ALLOC;
998 : spx_word16_t ystack[FIXED_STACK_ALLOC];
999 : #endif
1000 :
1001 0 : st->out_stride = 1;
1002 :
1003 0 : while (ilen && olen) {
1004 0 : spx_word16_t *y = ystack;
1005 0 : spx_uint32_t ichunk = (ilen > xlen) ? xlen : ilen;
1006 0 : spx_uint32_t ochunk = (olen > ylen) ? ylen : olen;
1007 0 : spx_uint32_t omagic = 0;
1008 :
1009 0 : if (st->magic_samples[channel_index]) {
1010 0 : omagic = speex_resampler_magic(st, channel_index, &y, ochunk);
1011 0 : ochunk -= omagic;
1012 0 : olen -= omagic;
1013 : }
1014 0 : if (! st->magic_samples[channel_index]) {
1015 0 : if (in) {
1016 0 : for(j=0;j<ichunk;++j)
1017 : #ifdef FIXED_POINT
1018 : x[j+st->filt_len-1]=WORD2INT(in[j*istride_save]);
1019 : #else
1020 0 : x[j+st->filt_len-1]=in[j*istride_save];
1021 : #endif
1022 : } else {
1023 0 : for(j=0;j<ichunk;++j)
1024 0 : x[j+st->filt_len-1]=0;
1025 : }
1026 :
1027 0 : speex_resampler_process_native(st, channel_index, &ichunk, y, &ochunk);
1028 : } else {
1029 0 : ichunk = 0;
1030 0 : ochunk = 0;
1031 : }
1032 :
1033 0 : for (j=0;j<ochunk+omagic;++j)
1034 : #ifdef FIXED_POINT
1035 : out[j*ostride_save] = ystack[j];
1036 : #else
1037 0 : out[j*ostride_save] = WORD2INT(ystack[j]);
1038 : #endif
1039 :
1040 0 : ilen -= ichunk;
1041 0 : olen -= ochunk;
1042 0 : out += (ochunk+omagic) * ostride_save;
1043 0 : if (in)
1044 0 : in += ichunk * istride_save;
1045 : }
1046 0 : st->out_stride = ostride_save;
1047 0 : *in_len -= ilen;
1048 0 : *out_len -= olen;
1049 :
1050 0 : return st->resampler_ptr == resampler_basic_zero ? RESAMPLER_ERR_ALLOC_FAILED : RESAMPLER_ERR_SUCCESS;
1051 : }
1052 :
1053 0 : EXPORT int speex_resampler_process_interleaved_float(SpeexResamplerState *st, const float *in, spx_uint32_t *in_len, float *out, spx_uint32_t *out_len)
1054 : {
1055 : spx_uint32_t i;
1056 : int istride_save, ostride_save;
1057 0 : spx_uint32_t bak_out_len = *out_len;
1058 0 : spx_uint32_t bak_in_len = *in_len;
1059 0 : istride_save = st->in_stride;
1060 0 : ostride_save = st->out_stride;
1061 0 : st->in_stride = st->out_stride = st->nb_channels;
1062 0 : for (i=0;i<st->nb_channels;i++)
1063 : {
1064 0 : *out_len = bak_out_len;
1065 0 : *in_len = bak_in_len;
1066 0 : if (in != NULL)
1067 0 : speex_resampler_process_float(st, i, in+i, in_len, out+i, out_len);
1068 : else
1069 0 : speex_resampler_process_float(st, i, NULL, in_len, out+i, out_len);
1070 : }
1071 0 : st->in_stride = istride_save;
1072 0 : st->out_stride = ostride_save;
1073 0 : return st->resampler_ptr == resampler_basic_zero ? RESAMPLER_ERR_ALLOC_FAILED : RESAMPLER_ERR_SUCCESS;
1074 : }
1075 :
1076 0 : EXPORT int speex_resampler_process_interleaved_int(SpeexResamplerState *st, const spx_int16_t *in, spx_uint32_t *in_len, spx_int16_t *out, spx_uint32_t *out_len)
1077 : {
1078 : spx_uint32_t i;
1079 : int istride_save, ostride_save;
1080 0 : spx_uint32_t bak_out_len = *out_len;
1081 0 : spx_uint32_t bak_in_len = *in_len;
1082 0 : istride_save = st->in_stride;
1083 0 : ostride_save = st->out_stride;
1084 0 : st->in_stride = st->out_stride = st->nb_channels;
1085 0 : for (i=0;i<st->nb_channels;i++)
1086 : {
1087 0 : *out_len = bak_out_len;
1088 0 : *in_len = bak_in_len;
1089 0 : if (in != NULL)
1090 0 : speex_resampler_process_int(st, i, in+i, in_len, out+i, out_len);
1091 : else
1092 0 : speex_resampler_process_int(st, i, NULL, in_len, out+i, out_len);
1093 : }
1094 0 : st->in_stride = istride_save;
1095 0 : st->out_stride = ostride_save;
1096 0 : return st->resampler_ptr == resampler_basic_zero ? RESAMPLER_ERR_ALLOC_FAILED : RESAMPLER_ERR_SUCCESS;
1097 : }
1098 :
1099 0 : EXPORT int speex_resampler_set_rate(SpeexResamplerState *st, spx_uint32_t in_rate, spx_uint32_t out_rate)
1100 : {
1101 0 : return speex_resampler_set_rate_frac(st, in_rate, out_rate, in_rate, out_rate);
1102 : }
1103 :
1104 0 : EXPORT void speex_resampler_get_rate(SpeexResamplerState *st, spx_uint32_t *in_rate, spx_uint32_t *out_rate)
1105 : {
1106 0 : *in_rate = st->in_rate;
1107 0 : *out_rate = st->out_rate;
1108 0 : }
1109 :
1110 0 : static inline spx_uint32_t _gcd(spx_uint32_t a, spx_uint32_t b)
1111 : {
1112 0 : while (b != 0)
1113 : {
1114 0 : spx_uint32_t temp = a;
1115 :
1116 0 : a = b;
1117 0 : b = temp % b;
1118 : }
1119 0 : return a;
1120 : }
1121 :
1122 0 : EXPORT int speex_resampler_set_rate_frac(SpeexResamplerState *st, spx_uint32_t ratio_num, spx_uint32_t ratio_den, spx_uint32_t in_rate, spx_uint32_t out_rate)
1123 : {
1124 : spx_uint32_t fact;
1125 : spx_uint32_t old_den;
1126 : spx_uint32_t i;
1127 :
1128 0 : if (ratio_num == 0 || ratio_den == 0)
1129 0 : return RESAMPLER_ERR_INVALID_ARG;
1130 :
1131 0 : if (st->in_rate == in_rate && st->out_rate == out_rate && st->num_rate == ratio_num && st->den_rate == ratio_den)
1132 0 : return RESAMPLER_ERR_SUCCESS;
1133 :
1134 0 : old_den = st->den_rate;
1135 0 : st->in_rate = in_rate;
1136 0 : st->out_rate = out_rate;
1137 0 : st->num_rate = ratio_num;
1138 0 : st->den_rate = ratio_den;
1139 :
1140 0 : fact = _gcd (st->num_rate, st->den_rate);
1141 :
1142 0 : st->num_rate /= fact;
1143 0 : st->den_rate /= fact;
1144 :
1145 0 : if (old_den > 0)
1146 : {
1147 0 : for (i=0;i<st->nb_channels;i++)
1148 : {
1149 0 : if (_muldiv(&st->samp_frac_num[i],st->samp_frac_num[i],st->den_rate,old_den) != RESAMPLER_ERR_SUCCESS) {
1150 0 : st->samp_frac_num[i] = st->den_rate-1;
1151 : }
1152 : /* Safety net */
1153 0 : if (st->samp_frac_num[i] >= st->den_rate)
1154 0 : st->samp_frac_num[i] = st->den_rate-1;
1155 : }
1156 : }
1157 :
1158 0 : if (st->initialised)
1159 0 : return update_filter(st);
1160 0 : return RESAMPLER_ERR_SUCCESS;
1161 : }
1162 :
1163 0 : EXPORT void speex_resampler_get_ratio(SpeexResamplerState *st, spx_uint32_t *ratio_num, spx_uint32_t *ratio_den)
1164 : {
1165 0 : *ratio_num = st->num_rate;
1166 0 : *ratio_den = st->den_rate;
1167 0 : }
1168 :
1169 0 : EXPORT int speex_resampler_set_quality(SpeexResamplerState *st, int quality)
1170 : {
1171 0 : if (quality > 10 || quality < 0)
1172 0 : return RESAMPLER_ERR_INVALID_ARG;
1173 0 : if (st->quality == quality)
1174 0 : return RESAMPLER_ERR_SUCCESS;
1175 0 : st->quality = quality;
1176 0 : if (st->initialised)
1177 0 : return update_filter(st);
1178 0 : return RESAMPLER_ERR_SUCCESS;
1179 : }
1180 :
1181 0 : EXPORT void speex_resampler_get_quality(SpeexResamplerState *st, int *quality)
1182 : {
1183 0 : *quality = st->quality;
1184 0 : }
1185 :
1186 0 : EXPORT void speex_resampler_set_input_stride(SpeexResamplerState *st, spx_uint32_t stride)
1187 : {
1188 0 : st->in_stride = stride;
1189 0 : }
1190 :
1191 0 : EXPORT void speex_resampler_get_input_stride(SpeexResamplerState *st, spx_uint32_t *stride)
1192 : {
1193 0 : *stride = st->in_stride;
1194 0 : }
1195 :
1196 0 : EXPORT void speex_resampler_set_output_stride(SpeexResamplerState *st, spx_uint32_t stride)
1197 : {
1198 0 : st->out_stride = stride;
1199 0 : }
1200 :
1201 0 : EXPORT void speex_resampler_get_output_stride(SpeexResamplerState *st, spx_uint32_t *stride)
1202 : {
1203 0 : *stride = st->out_stride;
1204 0 : }
1205 :
1206 0 : EXPORT int speex_resampler_get_input_latency(SpeexResamplerState *st)
1207 : {
1208 0 : return st->filt_len / 2;
1209 : }
1210 :
1211 0 : EXPORT int speex_resampler_get_output_latency(SpeexResamplerState *st)
1212 : {
1213 0 : return ((st->filt_len / 2) * st->den_rate + (st->num_rate >> 1)) / st->num_rate;
1214 : }
1215 :
1216 0 : EXPORT int speex_resampler_skip_zeros(SpeexResamplerState *st)
1217 : {
1218 : spx_uint32_t i;
1219 0 : for (i=0;i<st->nb_channels;i++)
1220 0 : st->last_sample[i] = st->filt_len/2;
1221 0 : return RESAMPLER_ERR_SUCCESS;
1222 : }
1223 :
1224 0 : EXPORT int speex_resampler_set_skip_frac_num(SpeexResamplerState *st, spx_uint32_t skip_frac_num)
1225 : {
1226 : spx_uint32_t i;
1227 0 : spx_uint32_t last_sample = skip_frac_num / st->den_rate;
1228 0 : spx_uint32_t samp_frac_num = skip_frac_num % st->den_rate;
1229 0 : for (i=0;i<st->nb_channels;i++) {
1230 0 : st->last_sample[i] = last_sample;
1231 0 : st->samp_frac_num[i] = samp_frac_num;
1232 : }
1233 0 : return RESAMPLER_ERR_SUCCESS;
1234 : }
1235 :
1236 0 : EXPORT int speex_resampler_reset_mem(SpeexResamplerState *st)
1237 : {
1238 : spx_uint32_t i;
1239 0 : for (i=0;i<st->nb_channels;i++)
1240 : {
1241 0 : st->last_sample[i] = 0;
1242 0 : st->magic_samples[i] = 0;
1243 0 : st->samp_frac_num[i] = 0;
1244 : }
1245 0 : for (i=0;i<st->nb_channels*(st->filt_len-1);i++)
1246 0 : st->mem[i] = 0;
1247 0 : return RESAMPLER_ERR_SUCCESS;
1248 : }
1249 :
1250 0 : EXPORT const char *speex_resampler_strerror(int err)
1251 : {
1252 0 : switch (err)
1253 : {
1254 : case RESAMPLER_ERR_SUCCESS:
1255 0 : return "Success.";
1256 : case RESAMPLER_ERR_ALLOC_FAILED:
1257 0 : return "Memory allocation failed.";
1258 : case RESAMPLER_ERR_BAD_STATE:
1259 0 : return "Bad resampler state.";
1260 : case RESAMPLER_ERR_INVALID_ARG:
1261 0 : return "Invalid argument.";
1262 : case RESAMPLER_ERR_PTR_OVERLAP:
1263 0 : return "Input and output buffers overlap.";
1264 : default:
1265 0 : return "Unknown error. Bad error code or strange version mismatch.";
1266 : }
1267 : }
|
