falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
100582 Commits
32 Branches
779MB
Tree: a454a0c14f
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a454a0c14f'
${ noResults }
FFmpeg/libavcodec/fft_template.c

630 lines
18KB
Raw Blame History 

  1. /*

  2.  * FFT/IFFT transforms

  3.  * Copyright (c) 2008 Loren Merritt

  4.  * Copyright (c) 2002 Fabrice Bellard

  5.  * Partly based on libdjbfft by D. J. Bernstein

  6.  *

  7.  * This file is part of FFmpeg.

  8.  *

  9.  * FFmpeg is free software; you can redistribute it and/or

  10.  * modify it under the terms of the GNU Lesser General Public

  11.  * License as published by the Free Software Foundation; either

  12.  * version 2.1 of the License, or (at your option) any later version.

  13.  *

  14.  * FFmpeg is distributed in the hope that it will be useful,

  15.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  16.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  17.  * Lesser General Public License for more details.

  18.  *

  19.  * You should have received a copy of the GNU Lesser General Public

  20.  * License along with FFmpeg; if not, write to the Free Software

  21.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  22.  */

  23. 

  24. /**

  25.  * @file

  26.  * FFT/IFFT transforms.

  27.  */

  28. 

  29. #include <stdlib.h>

  30. #include <string.h>

  31. #include "libavutil/mathematics.h"

  32. #include "libavutil/thread.h"

  33. #include "fft.h"

  34. #include "fft-internal.h"

  35. 

  36. #if FFT_FIXED_32

  37. #include "fft_table.h"

  38. #else /* FFT_FIXED_32 */

  39. 

  40. /* cos(2*pi*x/n) for 0<=x<=n/4, followed by its reverse */

  41. #if !CONFIG_HARDCODED_TABLES

  42. COSTABLE(16);

  43. COSTABLE(32);

  44. COSTABLE(64);

  45. COSTABLE(128);

  46. COSTABLE(256);

  47. COSTABLE(512);

  48. COSTABLE(1024);

  49. COSTABLE(2048);

  50. COSTABLE(4096);

  51. COSTABLE(8192);

  52. COSTABLE(16384);

  53. COSTABLE(32768);

  54. COSTABLE(65536);

  55. COSTABLE(131072);

  56. 

  57. static av_cold void init_ff_cos_tabs(int index)

  58. {

  59.     int i;

  60.     int m = 1<<index;

  61.     double freq = 2*M_PI/m;

  62.     FFTSample *tab = FFT_NAME(ff_cos_tabs)[index];

  63.     for(i=0; i<=m/4; i++)

  64.         tab[i] = FIX15(cos(i*freq));

  65.     for(i=1; i<m/4; i++)

  66.         tab[m/2-i] = tab[i];

  67. }

  68. 

  69. typedef struct CosTabsInitOnce {

  70.     void (*func)(void);

  71.     AVOnce control;

  72. } CosTabsInitOnce;

  73. 

  74. #define INIT_FF_COS_TABS_FUNC(index, size)          \

  75. static av_cold void init_ff_cos_tabs_ ## size (void)\

  76. {                                                   \

  77.     init_ff_cos_tabs(index);                        \

  78. }

  79. 

  80. INIT_FF_COS_TABS_FUNC(4, 16)

  81. INIT_FF_COS_TABS_FUNC(5, 32)

  82. INIT_FF_COS_TABS_FUNC(6, 64)

  83. INIT_FF_COS_TABS_FUNC(7, 128)

  84. INIT_FF_COS_TABS_FUNC(8, 256)

  85. INIT_FF_COS_TABS_FUNC(9, 512)

  86. INIT_FF_COS_TABS_FUNC(10, 1024)

  87. INIT_FF_COS_TABS_FUNC(11, 2048)

  88. INIT_FF_COS_TABS_FUNC(12, 4096)

  89. INIT_FF_COS_TABS_FUNC(13, 8192)

  90. INIT_FF_COS_TABS_FUNC(14, 16384)

  91. INIT_FF_COS_TABS_FUNC(15, 32768)

  92. INIT_FF_COS_TABS_FUNC(16, 65536)

  93. INIT_FF_COS_TABS_FUNC(17, 131072)

  94. 

  95. static CosTabsInitOnce cos_tabs_init_once[] = {

  96.     { NULL },

  97.     { NULL },

  98.     { NULL },

  99.     { NULL },

  100.     { init_ff_cos_tabs_16, AV_ONCE_INIT },

  101.     { init_ff_cos_tabs_32, AV_ONCE_INIT },

  102.     { init_ff_cos_tabs_64, AV_ONCE_INIT },

  103.     { init_ff_cos_tabs_128, AV_ONCE_INIT },

  104.     { init_ff_cos_tabs_256, AV_ONCE_INIT },

  105.     { init_ff_cos_tabs_512, AV_ONCE_INIT },

  106.     { init_ff_cos_tabs_1024, AV_ONCE_INIT },

  107.     { init_ff_cos_tabs_2048, AV_ONCE_INIT },

  108.     { init_ff_cos_tabs_4096, AV_ONCE_INIT },

  109.     { init_ff_cos_tabs_8192, AV_ONCE_INIT },

  110.     { init_ff_cos_tabs_16384, AV_ONCE_INIT },

  111.     { init_ff_cos_tabs_32768, AV_ONCE_INIT },

  112.     { init_ff_cos_tabs_65536, AV_ONCE_INIT },

  113.     { init_ff_cos_tabs_131072, AV_ONCE_INIT },

  114. };

  115. 

  116. #endif

  117. COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = {

  118.     NULL, NULL, NULL, NULL,

  119.     FFT_NAME(ff_cos_16),

  120.     FFT_NAME(ff_cos_32),

  121.     FFT_NAME(ff_cos_64),

  122.     FFT_NAME(ff_cos_128),

  123.     FFT_NAME(ff_cos_256),

  124.     FFT_NAME(ff_cos_512),

  125.     FFT_NAME(ff_cos_1024),

  126.     FFT_NAME(ff_cos_2048),

  127.     FFT_NAME(ff_cos_4096),

  128.     FFT_NAME(ff_cos_8192),

  129.     FFT_NAME(ff_cos_16384),

  130.     FFT_NAME(ff_cos_32768),

  131.     FFT_NAME(ff_cos_65536),

  132.     FFT_NAME(ff_cos_131072),

  133. };

  134. 

  135. #endif /* FFT_FIXED_32 */

  136. 

  137. static void fft_permute_c(FFTContext *s, FFTComplex *z);

  138. static void fft_calc_c(FFTContext *s, FFTComplex *z);

  139. 

  140. static int split_radix_permutation(int i, int n, int inverse)

  141. {

  142.     int m;

  143.     if(n <= 2) return i&1;

  144.     m = n >> 1;

  145.     if(!(i&m))            return split_radix_permutation(i, m, inverse)*2;

  146.     m >>= 1;

  147.     if(inverse == !(i&m)) return split_radix_permutation(i, m, inverse)*4 + 1;

  148.     else                  return split_radix_permutation(i, m, inverse)*4 - 1;

  149. }

  150. 

  151. av_cold void ff_init_ff_cos_tabs(int index)

  152. {

  153. #if (!CONFIG_HARDCODED_TABLES) && (!FFT_FIXED_32)

  154.     ff_thread_once(&cos_tabs_init_once[index].control, cos_tabs_init_once[index].func);

  155. #endif

  156. }

  157. 

  158. static const int avx_tab[] = {

  159.     0, 4, 1, 5, 8, 12, 9, 13, 2, 6, 3, 7, 10, 14, 11, 15

  160. };

  161. 

  162. static int is_second_half_of_fft32(int i, int n)

  163. {

  164.     if (n <= 32)

  165.         return i >= 16;

  166.     else if (i < n/2)

  167.         return is_second_half_of_fft32(i, n/2);

  168.     else if (i < 3*n/4)

  169.         return is_second_half_of_fft32(i - n/2, n/4);

  170.     else

  171.         return is_second_half_of_fft32(i - 3*n/4, n/4);

  172. }

  173. 

  174. static av_cold void fft_perm_avx(FFTContext *s)

  175. {

  176.     int i;

  177.     int n = 1 << s->nbits;

  178. 

  179.     for (i = 0; i < n; i += 16) {

  180.         int k;

  181.         if (is_second_half_of_fft32(i, n)) {

  182.             for (k = 0; k < 16; k++)

  183.                 s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] =

  184.                     i + avx_tab[k];

  185. 

  186.         } else {

  187.             for (k = 0; k < 16; k++) {

  188.                 int j = i + k;

  189.                 j = (j & ~7) | ((j >> 1) & 3) | ((j << 2) & 4);

  190.                 s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] = j;

  191.             }

  192.         }

  193.     }

  194. }

  195. 

  196. av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)

  197. {

  198.     int i, j, n;

  199. 

  200.     s->revtab = NULL;

  201.     s->revtab32 = NULL;

  202. 

  203.     if (nbits < 2 || nbits > 17)

  204.         goto fail;

  205.     s->nbits = nbits;

  206.     n = 1 << nbits;

  207. 

  208.     if (nbits <= 16) {

  209.         s->revtab = av_malloc(n * sizeof(uint16_t));

  210.         if (!s->revtab)

  211.             goto fail;

  212.     } else {

  213.         s->revtab32 = av_malloc(n * sizeof(uint32_t));

  214.         if (!s->revtab32)

  215.             goto fail;

  216.     }

  217.     s->tmp_buf = av_malloc(n * sizeof(FFTComplex));

  218.     if (!s->tmp_buf)

  219.         goto fail;

  220.     s->inverse = inverse;

  221.     s->fft_permutation = FF_FFT_PERM_DEFAULT;

  222. 

  223.     s->fft_permute = fft_permute_c;

  224.     s->fft_calc    = fft_calc_c;

  225. #if CONFIG_MDCT

  226.     s->imdct_calc  = ff_imdct_calc_c;

  227.     s->imdct_half  = ff_imdct_half_c;

  228.     s->mdct_calc   = ff_mdct_calc_c;

  229. #endif

  230. 

  231. #if FFT_FIXED_32

  232.     ff_fft_lut_init();

  233. #else /* FFT_FIXED_32 */

  234. #if FFT_FLOAT

  235.     if (ARCH_AARCH64) ff_fft_init_aarch64(s);

  236.     if (ARCH_ARM)     ff_fft_init_arm(s);

  237.     if (ARCH_PPC)     ff_fft_init_ppc(s);

  238.     if (ARCH_X86)     ff_fft_init_x86(s);

  239.     if (CONFIG_MDCT)  s->mdct_calcw = s->mdct_calc;

  240.     if (HAVE_MIPSFPU) ff_fft_init_mips(s);

  241. #else

  242.     if (CONFIG_MDCT)  s->mdct_calcw = ff_mdct_calcw_c;

  243.     if (ARCH_ARM)     ff_fft_fixed_init_arm(s);

  244. #endif

  245.     for(j=4; j<=nbits; j++) {

  246.         ff_init_ff_cos_tabs(j);

  247.     }

  248. #endif /* FFT_FIXED_32 */

  249. 

  250. 

  251.     if (ARCH_X86 && FFT_FLOAT && s->fft_permutation == FF_FFT_PERM_AVX) {

  252.         fft_perm_avx(s);

  253.     } else {

  254. #define PROCESS_FFT_PERM_SWAP_LSBS(num) do {\

  255.     for(i = 0; i < n; i++) {\

  256.         int k;\

  257.         j = i;\

  258.         j = (j & ~3) | ((j >> 1) & 1) | ((j << 1) & 2);\

  259.         k = -split_radix_permutation(i, n, s->inverse) & (n - 1);\

  260.         s->revtab##num[k] = j;\

  261.     } \

  262. } while(0);

  263. 

  264. #define PROCESS_FFT_PERM_DEFAULT(num) do {\

  265.     for(i = 0; i < n; i++) {\

  266.         int k;\

  267.         j = i;\

  268.         k = -split_radix_permutation(i, n, s->inverse) & (n - 1);\

  269.         s->revtab##num[k] = j;\

  270.     } \

  271. } while(0);

  272. 

  273. #define SPLIT_RADIX_PERMUTATION(num) do { \

  274.     if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS) {\

  275.         PROCESS_FFT_PERM_SWAP_LSBS(num) \

  276.     } else {\

  277.         PROCESS_FFT_PERM_DEFAULT(num) \

  278.     }\

  279. } while(0);

  280. 

  281.     if (s->revtab)

  282.         SPLIT_RADIX_PERMUTATION()

  283.     if (s->revtab32)

  284.         SPLIT_RADIX_PERMUTATION(32)

  285. 

  286. #undef PROCESS_FFT_PERM_DEFAULT

  287. #undef PROCESS_FFT_PERM_SWAP_LSBS

  288. #undef SPLIT_RADIX_PERMUTATION

  289.     }

  290. 

  291.     return 0;

  292.  fail:

  293.     av_freep(&s->revtab);

  294.     av_freep(&s->revtab32);

  295.     av_freep(&s->tmp_buf);

  296.     return -1;

  297. }

  298. 

  299. static void fft_permute_c(FFTContext *s, FFTComplex *z)

  300. {

  301.     int j, np;

  302.     const uint16_t *revtab = s->revtab;

  303.     const uint32_t *revtab32 = s->revtab32;

  304.     np = 1 << s->nbits;

  305.     /* TODO: handle split-radix permute in a more optimal way, probably in-place */

  306.     if (revtab) {

  307.         for(j=0;j<np;j++) s->tmp_buf[revtab[j]] = z[j];

  308.     } else

  309.         for(j=0;j<np;j++) s->tmp_buf[revtab32[j]] = z[j];

  310. 

  311.     memcpy(z, s->tmp_buf, np * sizeof(FFTComplex));

  312. }

  313. 

  314. av_cold void ff_fft_end(FFTContext *s)

  315. {

  316.     av_freep(&s->revtab);

  317.     av_freep(&s->revtab32);

  318.     av_freep(&s->tmp_buf);

  319. }

  320. 

  321. #if FFT_FIXED_32

  322. 

  323. static void fft_calc_c(FFTContext *s, FFTComplex *z) {

  324. 

  325.     int nbits, i, n, num_transforms, offset, step;

  326.     int n4, n2, n34;

  327.     unsigned tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;

  328.     FFTComplex *tmpz;

  329.     const int fft_size = (1 << s->nbits);

  330.     int64_t accu;

  331. 

  332.     num_transforms = (0x2aab >> (16 - s->nbits)) | 1;

  333. 

  334.     for (n=0; n<num_transforms; n++){

  335.         offset = ff_fft_offsets_lut[n] << 2;

  336.         tmpz = z + offset;

  337. 

  338.         tmp1 = tmpz[0].re + (unsigned)tmpz[1].re;

  339.         tmp5 = tmpz[2].re + (unsigned)tmpz[3].re;

  340.         tmp2 = tmpz[0].im + (unsigned)tmpz[1].im;

  341.         tmp6 = tmpz[2].im + (unsigned)tmpz[3].im;

  342.         tmp3 = tmpz[0].re - (unsigned)tmpz[1].re;

  343.         tmp8 = tmpz[2].im - (unsigned)tmpz[3].im;

  344.         tmp4 = tmpz[0].im - (unsigned)tmpz[1].im;

  345.         tmp7 = tmpz[2].re - (unsigned)tmpz[3].re;

  346. 

  347.         tmpz[0].re = tmp1 + tmp5;

  348.         tmpz[2].re = tmp1 - tmp5;

  349.         tmpz[0].im = tmp2 + tmp6;

  350.         tmpz[2].im = tmp2 - tmp6;

  351.         tmpz[1].re = tmp3 + tmp8;

  352.         tmpz[3].re = tmp3 - tmp8;

  353.         tmpz[1].im = tmp4 - tmp7;

  354.         tmpz[3].im = tmp4 + tmp7;

  355.     }

  356. 

  357.     if (fft_size < 8)

  358.         return;

  359. 

  360.     num_transforms = (num_transforms >> 1) | 1;

  361. 

  362.     for (n=0; n<num_transforms; n++){

  363.         offset = ff_fft_offsets_lut[n] << 3;

  364.         tmpz = z + offset;

  365. 

  366.         tmp1 = tmpz[4].re + (unsigned)tmpz[5].re;

  367.         tmp3 = tmpz[6].re + (unsigned)tmpz[7].re;

  368.         tmp2 = tmpz[4].im + (unsigned)tmpz[5].im;

  369.         tmp4 = tmpz[6].im + (unsigned)tmpz[7].im;

  370.         tmp5 = tmp1 + tmp3;

  371.         tmp7 = tmp1 - tmp3;

  372.         tmp6 = tmp2 + tmp4;

  373.         tmp8 = tmp2 - tmp4;

  374. 

  375.         tmp1 = tmpz[4].re - (unsigned)tmpz[5].re;

  376.         tmp2 = tmpz[4].im - (unsigned)tmpz[5].im;

  377.         tmp3 = tmpz[6].re - (unsigned)tmpz[7].re;

  378.         tmp4 = tmpz[6].im - (unsigned)tmpz[7].im;

  379. 

  380.         tmpz[4].re = tmpz[0].re - tmp5;

  381.         tmpz[0].re = tmpz[0].re + tmp5;

  382.         tmpz[4].im = tmpz[0].im - tmp6;

  383.         tmpz[0].im = tmpz[0].im + tmp6;

  384.         tmpz[6].re = tmpz[2].re - tmp8;

  385.         tmpz[2].re = tmpz[2].re + tmp8;

  386.         tmpz[6].im = tmpz[2].im + tmp7;

  387.         tmpz[2].im = tmpz[2].im - tmp7;

  388. 

  389.         accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp1 + tmp2);

  390.         tmp5 = (int32_t)((accu + 0x40000000) >> 31);

  391.         accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp3 - tmp4);

  392.         tmp7 = (int32_t)((accu + 0x40000000) >> 31);

  393.         accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp2 - tmp1);

  394.         tmp6 = (int32_t)((accu + 0x40000000) >> 31);

  395.         accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp3 + tmp4);

  396.         tmp8 = (int32_t)((accu + 0x40000000) >> 31);

  397.         tmp1 = tmp5 + tmp7;

  398.         tmp3 = tmp5 - tmp7;

  399.         tmp2 = tmp6 + tmp8;

  400.         tmp4 = tmp6 - tmp8;

  401. 

  402.         tmpz[5].re = tmpz[1].re - tmp1;

  403.         tmpz[1].re = tmpz[1].re + tmp1;

  404.         tmpz[5].im = tmpz[1].im - tmp2;

  405.         tmpz[1].im = tmpz[1].im + tmp2;

  406.         tmpz[7].re = tmpz[3].re - tmp4;

  407.         tmpz[3].re = tmpz[3].re + tmp4;

  408.         tmpz[7].im = tmpz[3].im + tmp3;

  409.         tmpz[3].im = tmpz[3].im - tmp3;

  410.     }

  411. 

  412.     step = 1 << ((MAX_LOG2_NFFT-4) - 4);

  413.     n4 = 4;

  414. 

  415.     for (nbits=4; nbits<=s->nbits; nbits++){

  416.         n2  = 2*n4;

  417.         n34 = 3*n4;

  418.         num_transforms = (num_transforms >> 1) | 1;

  419. 

  420.         for (n=0; n<num_transforms; n++){

  421.             const FFTSample *w_re_ptr = ff_w_tab_sr + step;

  422.             const FFTSample *w_im_ptr = ff_w_tab_sr + MAX_FFT_SIZE/(4*16) - step;

  423.             offset = ff_fft_offsets_lut[n] << nbits;

  424.             tmpz = z + offset;

  425. 

  426.             tmp5 = tmpz[ n2].re + (unsigned)tmpz[n34].re;

  427.             tmp1 = tmpz[ n2].re - (unsigned)tmpz[n34].re;

  428.             tmp6 = tmpz[ n2].im + (unsigned)tmpz[n34].im;

  429.             tmp2 = tmpz[ n2].im - (unsigned)tmpz[n34].im;

  430. 

  431.             tmpz[ n2].re = tmpz[ 0].re - tmp5;

  432.             tmpz[  0].re = tmpz[ 0].re + tmp5;

  433.             tmpz[ n2].im = tmpz[ 0].im - tmp6;

  434.             tmpz[  0].im = tmpz[ 0].im + tmp6;

  435.             tmpz[n34].re = tmpz[n4].re - tmp2;

  436.             tmpz[ n4].re = tmpz[n4].re + tmp2;

  437.             tmpz[n34].im = tmpz[n4].im + tmp1;

  438.             tmpz[ n4].im = tmpz[n4].im - tmp1;

  439. 

  440.             for (i=1; i<n4; i++){

  441.                 FFTSample w_re = w_re_ptr[0];

  442.                 FFTSample w_im = w_im_ptr[0];

  443.                 accu  = (int64_t)w_re*tmpz[ n2+i].re;

  444.                 accu += (int64_t)w_im*tmpz[ n2+i].im;

  445.                 tmp1 = (int32_t)((accu + 0x40000000) >> 31);

  446.                 accu  = (int64_t)w_re*tmpz[ n2+i].im;

  447.                 accu -= (int64_t)w_im*tmpz[ n2+i].re;

  448.                 tmp2 = (int32_t)((accu + 0x40000000) >> 31);

  449.                 accu  = (int64_t)w_re*tmpz[n34+i].re;

  450.                 accu -= (int64_t)w_im*tmpz[n34+i].im;

  451.                 tmp3 = (int32_t)((accu + 0x40000000) >> 31);

  452.                 accu  = (int64_t)w_re*tmpz[n34+i].im;

  453.                 accu += (int64_t)w_im*tmpz[n34+i].re;

  454.                 tmp4 = (int32_t)((accu + 0x40000000) >> 31);

  455. 

  456.                 tmp5 = tmp1 + tmp3;

  457.                 tmp1 = tmp1 - tmp3;

  458.                 tmp6 = tmp2 + tmp4;

  459.                 tmp2 = tmp2 - tmp4;

  460. 

  461.                 tmpz[ n2+i].re = tmpz[   i].re - tmp5;

  462.                 tmpz[    i].re = tmpz[   i].re + tmp5;

  463.                 tmpz[ n2+i].im = tmpz[   i].im - tmp6;

  464.                 tmpz[    i].im = tmpz[   i].im + tmp6;

  465.                 tmpz[n34+i].re = tmpz[n4+i].re - tmp2;

  466.                 tmpz[ n4+i].re = tmpz[n4+i].re + tmp2;

  467.                 tmpz[n34+i].im = tmpz[n4+i].im + tmp1;

  468.                 tmpz[ n4+i].im = tmpz[n4+i].im - tmp1;

  469. 

  470.                 w_re_ptr += step;

  471.                 w_im_ptr -= step;

  472.             }

  473.         }

  474.         step >>= 1;

  475.         n4   <<= 1;

  476.     }

  477. }

  478. 

  479. #else /* FFT_FIXED_32 */

  480. 

  481. #define BUTTERFLIES(a0,a1,a2,a3) {\

  482.     BF(t3, t5, t5, t1);\

  483.     BF(a2.re, a0.re, a0.re, t5);\

  484.     BF(a3.im, a1.im, a1.im, t3);\

  485.     BF(t4, t6, t2, t6);\

  486.     BF(a3.re, a1.re, a1.re, t4);\

  487.     BF(a2.im, a0.im, a0.im, t6);\

  488. }

  489. 

  490. // force loading all the inputs before storing any.

  491. // this is slightly slower for small data, but avoids store->load aliasing

  492. // for addresses separated by large powers of 2.

  493. #define BUTTERFLIES_BIG(a0,a1,a2,a3) {\

  494.     FFTSample r0=a0.re, i0=a0.im, r1=a1.re, i1=a1.im;\

  495.     BF(t3, t5, t5, t1);\

  496.     BF(a2.re, a0.re, r0, t5);\

  497.     BF(a3.im, a1.im, i1, t3);\

  498.     BF(t4, t6, t2, t6);\

  499.     BF(a3.re, a1.re, r1, t4);\

  500.     BF(a2.im, a0.im, i0, t6);\

  501. }

  502. 

  503. #define TRANSFORM(a0,a1,a2,a3,wre,wim) {\

  504.     CMUL(t1, t2, a2.re, a2.im, wre, -wim);\

  505.     CMUL(t5, t6, a3.re, a3.im, wre,  wim);\

  506.     BUTTERFLIES(a0,a1,a2,a3)\

  507. }

  508. 

  509. #define TRANSFORM_ZERO(a0,a1,a2,a3) {\

  510.     t1 = a2.re;\

  511.     t2 = a2.im;\

  512.     t5 = a3.re;\

  513.     t6 = a3.im;\

  514.     BUTTERFLIES(a0,a1,a2,a3)\

  515. }

  516. 

  517. /* z[0...8n-1], w[1...2n-1] */

  518. #define PASS(name)\

  519. static void name(FFTComplex *z, const FFTSample *wre, unsigned int n)\

  520. {\

  521.     FFTDouble t1, t2, t3, t4, t5, t6;\

  522.     int o1 = 2*n;\

  523.     int o2 = 4*n;\

  524.     int o3 = 6*n;\

  525.     const FFTSample *wim = wre+o1;\

  526.     n--;\

  527. \

  528.     TRANSFORM_ZERO(z[0],z[o1],z[o2],z[o3]);\

  529.     TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\

  530.     do {\

  531.         z += 2;\

  532.         wre += 2;\

  533.         wim -= 2;\

  534.         TRANSFORM(z[0],z[o1],z[o2],z[o3],wre[0],wim[0]);\

  535.         TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\

  536.     } while(--n);\

  537. }

  538. 

  539. PASS(pass)

  540. #if !CONFIG_SMALL

  541. #undef BUTTERFLIES

  542. #define BUTTERFLIES BUTTERFLIES_BIG

  543. PASS(pass_big)

  544. #endif

  545. 

  546. #define DECL_FFT(n,n2,n4)\

  547. static void fft##n(FFTComplex *z)\

  548. {\

  549.     fft##n2(z);\

  550.     fft##n4(z+n4*2);\

  551.     fft##n4(z+n4*3);\

  552.     pass(z,FFT_NAME(ff_cos_##n),n4/2);\

  553. }

  554. 

  555. static void fft4(FFTComplex *z)

  556. {

  557.     FFTDouble t1, t2, t3, t4, t5, t6, t7, t8;

  558. 

  559.     BF(t3, t1, z[0].re, z[1].re);

  560.     BF(t8, t6, z[3].re, z[2].re);

  561.     BF(z[2].re, z[0].re, t1, t6);

  562.     BF(t4, t2, z[0].im, z[1].im);

  563.     BF(t7, t5, z[2].im, z[3].im);

  564.     BF(z[3].im, z[1].im, t4, t8);

  565.     BF(z[3].re, z[1].re, t3, t7);

  566.     BF(z[2].im, z[0].im, t2, t5);

  567. }

  568. 

  569. static void fft8(FFTComplex *z)

  570. {

  571.     FFTDouble t1, t2, t3, t4, t5, t6;

  572. 

  573.     fft4(z);

  574. 

  575.     BF(t1, z[5].re, z[4].re, -z[5].re);

  576.     BF(t2, z[5].im, z[4].im, -z[5].im);

  577.     BF(t5, z[7].re, z[6].re, -z[7].re);

  578.     BF(t6, z[7].im, z[6].im, -z[7].im);

  579. 

  580.     BUTTERFLIES(z[0],z[2],z[4],z[6]);

  581.     TRANSFORM(z[1],z[3],z[5],z[7],sqrthalf,sqrthalf);

  582. }

  583. 

  584. #if !CONFIG_SMALL

  585. static void fft16(FFTComplex *z)

  586. {

  587.     FFTDouble t1, t2, t3, t4, t5, t6;

  588.     FFTSample cos_16_1 = FFT_NAME(ff_cos_16)[1];

  589.     FFTSample cos_16_3 = FFT_NAME(ff_cos_16)[3];

  590. 

  591.     fft8(z);

  592.     fft4(z+8);

  593.     fft4(z+12);

  594. 

  595.     TRANSFORM_ZERO(z[0],z[4],z[8],z[12]);

  596.     TRANSFORM(z[2],z[6],z[10],z[14],sqrthalf,sqrthalf);

  597.     TRANSFORM(z[1],z[5],z[9],z[13],cos_16_1,cos_16_3);

  598.     TRANSFORM(z[3],z[7],z[11],z[15],cos_16_3,cos_16_1);

  599. }

  600. #else

  601. DECL_FFT(16,8,4)

  602. #endif

  603. DECL_FFT(32,16,8)

  604. DECL_FFT(64,32,16)

  605. DECL_FFT(128,64,32)

  606. DECL_FFT(256,128,64)

  607. DECL_FFT(512,256,128)

  608. #if !CONFIG_SMALL

  609. #define pass pass_big

  610. #endif

  611. DECL_FFT(1024,512,256)

  612. DECL_FFT(2048,1024,512)

  613. DECL_FFT(4096,2048,1024)

  614. DECL_FFT(8192,4096,2048)

  615. DECL_FFT(16384,8192,4096)

  616. DECL_FFT(32768,16384,8192)

  617. DECL_FFT(65536,32768,16384)

  618. DECL_FFT(131072,65536,32768)

  619. 

  620. static void (* const fft_dispatch[])(FFTComplex*) = {

  621.     fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024,

  622.     fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, fft131072

  623. };

  624. 

  625. static void fft_calc_c(FFTContext *s, FFTComplex *z)

  626. {

  627.     fft_dispatch[s->nbits-2](z);

  628. }

  629. #endif /* FFT_FIXED_32 */