/*

 * http://www.kurims.kyoto-u.ac.jp/~ooura/fft.html

 * Copyright Takuya OOURA, 1996-2001

 *

 * You may use, copy, modify and distribute this code for any purpose (include

 * commercial use) and without fee. Please refer to this package when you modify

 * this code.

 *

 * Changes by the WebRTC authors:

 *    - Trivial type modifications.

 *    - Minimal code subset to do rdft of length 128.

 *    - Optimizations because of known length.

 *

 *  All changes are covered by the WebRTC license and IP grant:

 *  Use of this source code is governed by a BSD-style license

 *  that can be found in the LICENSE file in the root of the source

 *  tree. An additional intellectual property rights grant can be found

 *  in the file PATENTS.  All contributing project authors may

 *  be found in the AUTHORS file in the root of the source tree.

 */

#include "webrtc/modules/audio_processing/aec/aec_rdft.h"

#include <math.h>

#include "webrtc/system_wrappers/interface/cpu_features_wrapper.h"

#include "webrtc/typedefs.h"

// constants shared by all paths (C, SSE2).

float rdft_w[];

// constants used by the C path.

float rdft_wk3ri_first[];

float rdft_wk3ri_second[];

// constants used by SSE2 but initialized in C path.

ALIGN16_BEG float ALIGN16_END rdft_wk1r[];

ALIGN16_BEG float ALIGN16_END rdft_wk2r[];

ALIGN16_BEG float ALIGN16_END rdft_wk3r[];

ALIGN16_BEG float ALIGN16_END rdft_wk1i[];

ALIGN16_BEG float ALIGN16_END rdft_wk2i[];

ALIGN16_BEG float ALIGN16_END rdft_wk3i[];

ALIGN16_BEG float ALIGN16_END cftmdl_wk1r[];

static int ip[];

static void bitrv2_32(int* ip1, float* a) {

  const int n = ;

  int j, j1, k, k1, m, m2;

  float xr, xi, yr, yi;

  ip1[] = ;

  {

    int l = n;

    m = ;

    while ((m << ) < l) {

      l >>= ;

      for (j = ; j < m; j++) {

        ip1[m + j] = ip1[j] + l;

      }

      m <<= ;

    }

  }

  m2 =  * m;

  for (k = ; k < m; k++) {

    for (j = ; j < k; j++) {

      j1 =  * j + ip1[k];

      k1 =  * k + ip1[j];

      xr = a[j1];

      xi = a[j1 + ];

      yr = a[k1];

      yi = a[k1 + ];

      a[j1] = yr;

      a[j1 + ] = yi;

      a[k1] = xr;

      a[k1 + ] = xi;

      j1 += m2;

      k1 +=  * m2;

      xr = a[j1];

      xi = a[j1 + ];

      yr = a[k1];

      yi = a[k1 + ];

      a[j1] = yr;

      a[j1 + ] = yi;

      a[k1] = xr;

      a[k1 + ] = xi;

      j1 += m2;

      k1 -= m2;

      xr = a[j1];

      xi = a[j1 + ];

      yr = a[k1];

      yi = a[k1 + ];

      a[j1] = yr;

      a[j1 + ] = yi;

      a[k1] = xr;

      a[k1 + ] = xi;

      j1 += m2;

      k1 +=  * m2;

      xr = a[j1];

      xi = a[j1 + ];

      yr = a[k1];

      yi = a[k1 + ];

      a[j1] = yr;

      a[j1 + ] = yi;

      a[k1] = xr;

      a[k1 + ] = xi;

    }

    j1 =  * k + m2 + ip1[k];

    k1 = j1 + m2;

    xr = a[j1];

    xi = a[j1 + ];

    yr = a[k1];

    yi = a[k1 + ];

    a[j1] = yr;

    a[j1 + ] = yi;

    a[k1] = xr;

    a[k1 + ] = xi;

  }

}

static void bitrv2_128(float* a) {

  /*

      Following things have been attempted but are no faster:

      (a) Storing the swap indexes in a LUT (index calculations are done

          for 'free' while waiting on memory/L1).

      (b) Consolidate the load/store of two consecutive floats by a 64 bit

          integer (execution is memory/L1 bound).

      (c) Do a mix of floats and 64 bit integer to maximize register

          utilization (execution is memory/L1 bound).

      (d) Replacing ip[i] by ((k<<31)>>25) + ((k >> 1)<<5).

      (e) Hard-coding of the offsets to completely eliminates index

          calculations.

  */

  unsigned int j, j1, k, k1;

  float xr, xi, yr, yi;

  static const int ip2[] = {, , , };

  for (k = ; k < ; k++) {

    for (j = ; j < k; j++) {

      j1 =  * j + ip2[k];

      k1 =  * k + ip2[j];

      xr = a[j1 + ];

      xi = a[j1 + ];

      yr = a[k1 + ];

      yi = a[k1 + ];

      a[j1 + ] = yr;

      a[j1 + ] = yi;

      a[k1 + ] = xr;

      a[k1 + ] = xi;

      j1 += ;

      k1 += ;

      xr = a[j1 + ];

      xi = a[j1 + ];

      yr = a[k1 + ];

      yi = a[k1 + ];

      a[j1 + ] = yr;

      a[j1 + ] = yi;

      a[k1 + ] = xr;

      a[k1 + ] = xi;

      j1 += ;

      k1 -= ;

      xr = a[j1 + ];

      xi = a[j1 + ];

      yr = a[k1 + ];

      yi = a[k1 + ];

      a[j1 + ] = yr;

      a[j1 + ] = yi;

      a[k1 + ] = xr;

      a[k1 + ] = xi;

      j1 += ;

      k1 += ;

      xr = a[j1 + ];

      xi = a[j1 + ];

      yr = a[k1 + ];

      yi = a[k1 + ];

      a[j1 + ] = yr;

      a[j1 + ] = yi;

      a[k1 + ] = xr;

      a[k1 + ] = xi;

    }

    j1 =  * k +  + ip2[k];

    k1 = j1 + ;

    xr = a[j1 + ];

    xi = a[j1 + ];

    yr = a[k1 + ];

    yi = a[k1 + ];

    a[j1 + ] = yr;

    a[j1 + ] = yi;

    a[k1 + ] = xr;

    a[k1 + ] = xi;

  }

}

static void makewt_32(void) {

  const int nw = ;

  int j, nwh;

  float delta, x, y;

  ip[] = nw;

  ip[] = ;

  nwh = nw >> ;

  delta = atanf(1.0f) / nwh;

  rdft_w[] = ;

  rdft_w[] = ;

  rdft_w[nwh] = cosf(delta * nwh);

  rdft_w[nwh + ] = rdft_w[nwh];

  for (j = ; j < nwh; j += ) {

    x = cosf(delta * j);

    y = sinf(delta * j);

    rdft_w[j] = x;

    rdft_w[j + ] = y;

    rdft_w[nw - j] = y;

    rdft_w[nw - j + ] = x;

  }

  bitrv2_32(ip + , rdft_w);

  // pre-calculate constants used by cft1st_128 and cftmdl_128...

  cftmdl_wk1r[] = rdft_w[];

  cftmdl_wk1r[] = rdft_w[];

  cftmdl_wk1r[] = rdft_w[];

  cftmdl_wk1r[] = -rdft_w[];

  {

    int k1;

    for (k1 = , j = ; j < ; j += , k1 += ) {

      const int k2 =  * k1;

      const float wk2r = rdft_w[k1 + ];

      const float wk2i = rdft_w[k1 + ];

      float wk1r, wk1i;

      // ... scalar version.

      wk1r = rdft_w[k2 + ];

      wk1i = rdft_w[k2 + ];

      rdft_wk3ri_first[k1 + ] = wk1r -  * wk2i * wk1i;

      rdft_wk3ri_first[k1 + ] =  * wk2i * wk1r - wk1i;

      wk1r = rdft_w[k2 + ];

      wk1i = rdft_w[k2 + ];

      rdft_wk3ri_second[k1 + ] = wk1r -  * wk2r * wk1i;

      rdft_wk3ri_second[k1 + ] =  * wk2r * wk1r - wk1i;

      // ... vector version.

      rdft_wk1r[k2 + ] = rdft_w[k2 + ];

      rdft_wk1r[k2 + ] = rdft_w[k2 + ];

      rdft_wk1r[k2 + ] = rdft_w[k2 + ];

      rdft_wk1r[k2 + ] = rdft_w[k2 + ];

      rdft_wk2r[k2 + ] = rdft_w[k1 + ];

      rdft_wk2r[k2 + ] = rdft_w[k1 + ];

      rdft_wk2r[k2 + ] = -rdft_w[k1 + ];

      rdft_wk2r[k2 + ] = -rdft_w[k1 + ];

      rdft_wk3r[k2 + ] = rdft_wk3ri_first[k1 + ];

      rdft_wk3r[k2 + ] = rdft_wk3ri_first[k1 + ];

      rdft_wk3r[k2 + ] = rdft_wk3ri_second[k1 + ];

      rdft_wk3r[k2 + ] = rdft_wk3ri_second[k1 + ];

      rdft_wk1i[k2 + ] = -rdft_w[k2 + ];

      rdft_wk1i[k2 + ] = rdft_w[k2 + ];

      rdft_wk1i[k2 + ] = -rdft_w[k2 + ];

      rdft_wk1i[k2 + ] = rdft_w[k2 + ];

      rdft_wk2i[k2 + ] = -rdft_w[k1 + ];

      rdft_wk2i[k2 + ] = rdft_w[k1 + ];

      rdft_wk2i[k2 + ] = -rdft_w[k1 + ];

      rdft_wk2i[k2 + ] = rdft_w[k1 + ];

      rdft_wk3i[k2 + ] = -rdft_wk3ri_first[k1 + ];

      rdft_wk3i[k2 + ] = rdft_wk3ri_first[k1 + ];

      rdft_wk3i[k2 + ] = -rdft_wk3ri_second[k1 + ];

      rdft_wk3i[k2 + ] = rdft_wk3ri_second[k1 + ];

    }

  }

}

static void makect_32(void) {

  float* c = rdft_w + ;

  const int nc = ;

  int j, nch;

  float delta;

  ip[] = nc;

  nch = nc >> ;

  delta = atanf(1.0f) / nch;

  c[] = cosf(delta * nch);

  c[nch] = 0.5f * c[];

  for (j = ; j < nch; j++) {

    c[j] = 0.5f * cosf(delta * j);

    c[nc - j] = 0.5f * sinf(delta * j);

  }

}

static void cft1st_128_C(float* a) {

  const int n = ;

  int j, k1, k2;

  float wk1r, wk1i, wk2r, wk2i, wk3r, wk3i;

  float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

  x0r = a[] + a[];

  x0i = a[] + a[];

  x1r = a[] - a[];

  x1i = a[] - a[];

  x2r = a[] + a[];

  x2i = a[] + a[];

  x3r = a[] - a[];

  x3i = a[] - a[];

  a[] = x0r + x2r;

  a[] = x0i + x2i;

  a[] = x0r - x2r;

  a[] = x0i - x2i;

  a[] = x1r - x3i;

  a[] = x1i + x3r;

  a[] = x1r + x3i;

  a[] = x1i - x3r;

  wk1r = rdft_w[];

  x0r = a[] + a[];

  x0i = a[] + a[];

  x1r = a[] - a[];

  x1i = a[] - a[];

  x2r = a[] + a[];

  x2i = a[] + a[];

  x3r = a[] - a[];

  x3i = a[] - a[];

  a[] = x0r + x2r;

  a[] = x0i + x2i;

  a[] = x2i - x0i;

  a[] = x0r - x2r;

  x0r = x1r - x3i;

  x0i = x1i + x3r;

  a[] = wk1r * (x0r - x0i);

  a[] = wk1r * (x0r + x0i);

  x0r = x3i + x1r;

  x0i = x3r - x1i;

  a[] = wk1r * (x0i - x0r);

  a[] = wk1r * (x0i + x0r);

  k1 = ;

  for (j = ; j < n; j += ) {

    k1 += ;

    k2 =  * k1;

    wk2r = rdft_w[k1 + ];

    wk2i = rdft_w[k1 + ];

    wk1r = rdft_w[k2 + ];

    wk1i = rdft_w[k2 + ];

    wk3r = rdft_wk3ri_first[k1 + ];

    wk3i = rdft_wk3ri_first[k1 + ];

    x0r = a[j + ] + a[j + ];

    x0i = a[j + ] + a[j + ];

    x1r = a[j + ] - a[j + ];

    x1i = a[j + ] - a[j + ];

    x2r = a[j + ] + a[j + ];

    x2i = a[j + ] + a[j + ];

    x3r = a[j + ] - a[j + ];

    x3i = a[j + ] - a[j + ];

    a[j + ] = x0r + x2r;

    a[j + ] = x0i + x2i;

    x0r -= x2r;

    x0i -= x2i;

    a[j + ] = wk2r * x0r - wk2i * x0i;

    a[j + ] = wk2r * x0i + wk2i * x0r;

    x0r = x1r - x3i;

    x0i = x1i + x3r;

    a[j + ] = wk1r * x0r - wk1i * x0i;

    a[j + ] = wk1r * x0i + wk1i * x0r;

    x0r = x1r + x3i;

    x0i = x1i - x3r;

    a[j + ] = wk3r * x0r - wk3i * x0i;

    a[j + ] = wk3r * x0i + wk3i * x0r;

    wk1r = rdft_w[k2 + ];

    wk1i = rdft_w[k2 + ];

    wk3r = rdft_wk3ri_second[k1 + ];

    wk3i = rdft_wk3ri_second[k1 + ];

    x0r = a[j + ] + a[j + ];

    x0i = a[j + ] + a[j + ];

    x1r = a[j + ] - a[j + ];

    x1i = a[j + ] - a[j + ];

    x2r = a[j + ] + a[j + ];

    x2i = a[j + ] + a[j + ];

    x3r = a[j + ] - a[j + ];

    x3i = a[j + ] - a[j + ];

    a[j + ] = x0r + x2r;

    a[j + ] = x0i + x2i;

    x0r -= x2r;

    x0i -= x2i;

    a[j + ] = -wk2i * x0r - wk2r * x0i;

    a[j + ] = -wk2i * x0i + wk2r * x0r;

    x0r = x1r - x3i;

    x0i = x1i + x3r;

    a[j + ] = wk1r * x0r - wk1i * x0i;

    a[j + ] = wk1r * x0i + wk1i * x0r;

    x0r = x1r + x3i;

    x0i = x1i - x3r;

    a[j + ] = wk3r * x0r - wk3i * x0i;

    a[j + ] = wk3r * x0i + wk3i * x0r;

  }

}

static void cftmdl_128_C(float* a) {

  const int l = ;

  const int n = ;

  const int m = ;

  int j0, j1, j2, j3, k, k1, k2, m2;

  float wk1r, wk1i, wk2r, wk2i, wk3r, wk3i;

  float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

  for (j0 = ; j0 < l; j0 += ) {

    j1 = j0 + ;

    j2 = j0 + ;

    j3 = j0 + ;

    x0r = a[j0 + ] + a[j1 + ];

    x0i = a[j0 + ] + a[j1 + ];

    x1r = a[j0 + ] - a[j1 + ];

    x1i = a[j0 + ] - a[j1 + ];

    x2r = a[j2 + ] + a[j3 + ];

    x2i = a[j2 + ] + a[j3 + ];

    x3r = a[j2 + ] - a[j3 + ];

    x3i = a[j2 + ] - a[j3 + ];

    a[j0 + ] = x0r + x2r;

    a[j0 + ] = x0i + x2i;

    a[j2 + ] = x0r - x2r;

    a[j2 + ] = x0i - x2i;

    a[j1 + ] = x1r - x3i;

    a[j1 + ] = x1i + x3r;

    a[j3 + ] = x1r + x3i;

    a[j3 + ] = x1i - x3r;

  }

  wk1r = rdft_w[];

  for (j0 = m; j0 < l + m; j0 += ) {

    j1 = j0 + ;

    j2 = j0 + ;

    j3 = j0 + ;

    x0r = a[j0 + ] + a[j1 + ];

    x0i = a[j0 + ] + a[j1 + ];

    x1r = a[j0 + ] - a[j1 + ];

    x1i = a[j0 + ] - a[j1 + ];

    x2r = a[j2 + ] + a[j3 + ];

    x2i = a[j2 + ] + a[j3 + ];

    x3r = a[j2 + ] - a[j3 + ];

    x3i = a[j2 + ] - a[j3 + ];

    a[j0 + ] = x0r + x2r;

    a[j0 + ] = x0i + x2i;

    a[j2 + ] = x2i - x0i;

    a[j2 + ] = x0r - x2r;

    x0r = x1r - x3i;

    x0i = x1i + x3r;

    a[j1 + ] = wk1r * (x0r - x0i);

    a[j1 + ] = wk1r * (x0r + x0i);

    x0r = x3i + x1r;

    x0i = x3r - x1i;

    a[j3 + ] = wk1r * (x0i - x0r);

    a[j3 + ] = wk1r * (x0i + x0r);

  }

  k1 = ;

  m2 =  * m;

  for (k = m2; k < n; k += m2) {

    k1 += ;

    k2 =  * k1;

    wk2r = rdft_w[k1 + ];

    wk2i = rdft_w[k1 + ];

    wk1r = rdft_w[k2 + ];

    wk1i = rdft_w[k2 + ];

    wk3r = rdft_wk3ri_first[k1 + ];

    wk3i = rdft_wk3ri_first[k1 + ];

    for (j0 = k; j0 < l + k; j0 += ) {

      j1 = j0 + ;

      j2 = j0 + ;

      j3 = j0 + ;

      x0r = a[j0 + ] + a[j1 + ];

      x0i = a[j0 + ] + a[j1 + ];

      x1r = a[j0 + ] - a[j1 + ];

      x1i = a[j0 + ] - a[j1 + ];

      x2r = a[j2 + ] + a[j3 + ];

      x2i = a[j2 + ] + a[j3 + ];

      x3r = a[j2 + ] - a[j3 + ];

      x3i = a[j2 + ] - a[j3 + ];

      a[j0 + ] = x0r + x2r;

      a[j0 + ] = x0i + x2i;

      x0r -= x2r;

      x0i -= x2i;

      a[j2 + ] = wk2r * x0r - wk2i * x0i;

      a[j2 + ] = wk2r * x0i + wk2i * x0r;

      x0r = x1r - x3i;

      x0i = x1i + x3r;

      a[j1 + ] = wk1r * x0r - wk1i * x0i;

      a[j1 + ] = wk1r * x0i + wk1i * x0r;

      x0r = x1r + x3i;

      x0i = x1i - x3r;

      a[j3 + ] = wk3r * x0r - wk3i * x0i;

      a[j3 + ] = wk3r * x0i + wk3i * x0r;

    }

    wk1r = rdft_w[k2 + ];

    wk1i = rdft_w[k2 + ];

    wk3r = rdft_wk3ri_second[k1 + ];

    wk3i = rdft_wk3ri_second[k1 + ];

    for (j0 = k + m; j0 < l + (k + m); j0 += ) {

      j1 = j0 + ;

      j2 = j0 + ;

      j3 = j0 + ;

      x0r = a[j0 + ] + a[j1 + ];

      x0i = a[j0 + ] + a[j1 + ];

      x1r = a[j0 + ] - a[j1 + ];

      x1i = a[j0 + ] - a[j1 + ];

      x2r = a[j2 + ] + a[j3 + ];

      x2i = a[j2 + ] + a[j3 + ];

      x3r = a[j2 + ] - a[j3 + ];

      x3i = a[j2 + ] - a[j3 + ];

      a[j0 + ] = x0r + x2r;

      a[j0 + ] = x0i + x2i;

      x0r -= x2r;

      x0i -= x2i;

      a[j2 + ] = -wk2i * x0r - wk2r * x0i;

      a[j2 + ] = -wk2i * x0i + wk2r * x0r;

      x0r = x1r - x3i;

      x0i = x1i + x3r;

      a[j1 + ] = wk1r * x0r - wk1i * x0i;

      a[j1 + ] = wk1r * x0i + wk1i * x0r;

      x0r = x1r + x3i;

      x0i = x1i - x3r;

      a[j3 + ] = wk3r * x0r - wk3i * x0i;

      a[j3 + ] = wk3r * x0i + wk3i * x0r;

    }

  }

}

static void cftfsub_128(float* a) {

  int j, j1, j2, j3, l;

  float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

  cft1st_128(a);

  cftmdl_128(a);

  l = ;

  for (j = ; j < l; j += ) {

    j1 = j + l;

    j2 = j1 + l;

    j3 = j2 + l;

    x0r = a[j] + a[j1];

    x0i = a[j + ] + a[j1 + ];

    x1r = a[j] - a[j1];

    x1i = a[j + ] - a[j1 + ];

    x2r = a[j2] + a[j3];

    x2i = a[j2 + ] + a[j3 + ];

    x3r = a[j2] - a[j3];

    x3i = a[j2 + ] - a[j3 + ];

    a[j] = x0r + x2r;

    a[j + ] = x0i + x2i;

    a[j2] = x0r - x2r;

    a[j2 + ] = x0i - x2i;

    a[j1] = x1r - x3i;

    a[j1 + ] = x1i + x3r;

    a[j3] = x1r + x3i;

    a[j3 + ] = x1i - x3r;

  }

}

static void cftbsub_128(float* a) {

  int j, j1, j2, j3, l;

  float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

  cft1st_128(a);

  cftmdl_128(a);

  l = ;

  for (j = ; j < l; j += ) {

    j1 = j + l;

    j2 = j1 + l;

    j3 = j2 + l;

    x0r = a[j] + a[j1];

    x0i = -a[j + ] - a[j1 + ];

    x1r = a[j] - a[j1];

    x1i = -a[j + ] + a[j1 + ];

    x2r = a[j2] + a[j3];

    x2i = a[j2 + ] + a[j3 + ];

    x3r = a[j2] - a[j3];

    x3i = a[j2 + ] - a[j3 + ];

    a[j] = x0r + x2r;

    a[j + ] = x0i - x2i;

    a[j2] = x0r - x2r;

    a[j2 + ] = x0i + x2i;

    a[j1] = x1r - x3i;

    a[j1 + ] = x1i - x3r;

    a[j3] = x1r + x3i;

    a[j3 + ] = x1i + x3r;

  }

}

static void rftfsub_128_C(float* a) {

  const float* c = rdft_w + ;

  int j1, j2, k1, k2;

  float wkr, wki, xr, xi, yr, yi;

  for (j1 = , j2 = ; j2 < ; j1 += , j2 += ) {

    k2 =  - j2;

    k1 =  - j1;

    wkr = 0.5f - c[k1];

    wki = c[j1];

    xr = a[j2 + ] - a[k2 + ];

    xi = a[j2 + ] + a[k2 + ];

    yr = wkr * xr - wki * xi;

    yi = wkr * xi + wki * xr;

    a[j2 + ] -= yr;

    a[j2 + ] -= yi;

    a[k2 + ] += yr;

    a[k2 + ] -= yi;

  }

}

static void rftbsub_128_C(float* a) {

  const float* c = rdft_w + ;

  int j1, j2, k1, k2;

  float wkr, wki, xr, xi, yr, yi;

  a[] = -a[];

  for (j1 = , j2 = ; j2 < ; j1 += , j2 += ) {

    k2 =  - j2;

    k1 =  - j1;

    wkr = 0.5f - c[k1];

    wki = c[j1];

    xr = a[j2 + ] - a[k2 + ];

    xi = a[j2 + ] + a[k2 + ];

    yr = wkr * xr + wki * xi;

    yi = wkr * xi - wki * xr;

    a[j2 + ] = a[j2 + ] - yr;

    a[j2 + ] = yi - a[j2 + ];

    a[k2 + ] = yr + a[k2 + ];

    a[k2 + ] = yi - a[k2 + ];

  }

  a[] = -a[];

}

void aec_rdft_forward_128(float* a) {

  float xi;

  bitrv2_128(a);

  cftfsub_128(a);

  rftfsub_128(a);

  xi = a[] - a[];

  a[] += a[];

  a[] = xi;

}

void aec_rdft_inverse_128(float* a) {

  a[] = 0.5f * (a[] - a[]);

  a[] -= a[];

  rftbsub_128(a);

  bitrv2_128(a);

  cftbsub_128(a);

}

// code path selection

rft_sub_128_t cft1st_128;

rft_sub_128_t cftmdl_128;

rft_sub_128_t rftfsub_128;

rft_sub_128_t rftbsub_128;

void aec_rdft_init(void) {

  cft1st_128 = cft1st_128_C;

  cftmdl_128 = cftmdl_128_C;

  rftfsub_128 = rftfsub_128_C;

  rftbsub_128 = rftbsub_128_C;

#if defined(WEBRTC_ARCH_X86_FAMILY)

  if (WebRtc_GetCPUInfo(kSSE2)) {

    aec_rdft_init_sse2();

  }

#endif

  // init library constants.

  makewt_32();

  makect_32();

}