update code.
This commit is contained in:
548
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft.cc
vendored
Normal file
548
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft.cc
vendored
Normal file
@ -0,0 +1,548 @@
|
||||
/*
|
||||
* 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.
|
||||
* - Removed the global variables by moving the code in to a class in order
|
||||
* to make it thread safe.
|
||||
*
|
||||
* 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 "common_audio/third_party/ooura/fft_size_128/ooura_fft.h"
|
||||
|
||||
#include "common_audio/third_party/ooura/fft_size_128/ooura_fft_tables_common.h"
|
||||
#include "rtc_base/system/arch.h"
|
||||
#include "system_wrappers/include/cpu_features_wrapper.h"
|
||||
|
||||
namespace webrtc {
|
||||
|
||||
namespace {
|
||||
|
||||
#if !(defined(MIPS_FPU_LE) || defined(WEBRTC_HAS_NEON))
|
||||
static void cft1st_128_C(float* a) {
|
||||
const int n = 128;
|
||||
int j, k1, k2;
|
||||
float wk1r, wk1i, wk2r, wk2i, wk3r, wk3i;
|
||||
float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;
|
||||
|
||||
// The processing of the first set of elements was simplified in C to avoid
|
||||
// some operations (multiplication by zero or one, addition of two elements
|
||||
// multiplied by the same weight, ...).
|
||||
x0r = a[0] + a[2];
|
||||
x0i = a[1] + a[3];
|
||||
x1r = a[0] - a[2];
|
||||
x1i = a[1] - a[3];
|
||||
x2r = a[4] + a[6];
|
||||
x2i = a[5] + a[7];
|
||||
x3r = a[4] - a[6];
|
||||
x3i = a[5] - a[7];
|
||||
a[0] = x0r + x2r;
|
||||
a[1] = x0i + x2i;
|
||||
a[4] = x0r - x2r;
|
||||
a[5] = x0i - x2i;
|
||||
a[2] = x1r - x3i;
|
||||
a[3] = x1i + x3r;
|
||||
a[6] = x1r + x3i;
|
||||
a[7] = x1i - x3r;
|
||||
wk1r = rdft_w[2];
|
||||
x0r = a[8] + a[10];
|
||||
x0i = a[9] + a[11];
|
||||
x1r = a[8] - a[10];
|
||||
x1i = a[9] - a[11];
|
||||
x2r = a[12] + a[14];
|
||||
x2i = a[13] + a[15];
|
||||
x3r = a[12] - a[14];
|
||||
x3i = a[13] - a[15];
|
||||
a[8] = x0r + x2r;
|
||||
a[9] = x0i + x2i;
|
||||
a[12] = x2i - x0i;
|
||||
a[13] = x0r - x2r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[10] = wk1r * (x0r - x0i);
|
||||
a[11] = wk1r * (x0r + x0i);
|
||||
x0r = x3i + x1r;
|
||||
x0i = x3r - x1i;
|
||||
a[14] = wk1r * (x0i - x0r);
|
||||
a[15] = wk1r * (x0i + x0r);
|
||||
k1 = 0;
|
||||
for (j = 16; j < n; j += 16) {
|
||||
k1 += 2;
|
||||
k2 = 2 * k1;
|
||||
wk2r = rdft_w[k1 + 0];
|
||||
wk2i = rdft_w[k1 + 1];
|
||||
wk1r = rdft_w[k2 + 0];
|
||||
wk1i = rdft_w[k2 + 1];
|
||||
wk3r = rdft_wk3ri_first[k1 + 0];
|
||||
wk3i = rdft_wk3ri_first[k1 + 1];
|
||||
x0r = a[j + 0] + a[j + 2];
|
||||
x0i = a[j + 1] + a[j + 3];
|
||||
x1r = a[j + 0] - a[j + 2];
|
||||
x1i = a[j + 1] - a[j + 3];
|
||||
x2r = a[j + 4] + a[j + 6];
|
||||
x2i = a[j + 5] + a[j + 7];
|
||||
x3r = a[j + 4] - a[j + 6];
|
||||
x3i = a[j + 5] - a[j + 7];
|
||||
a[j + 0] = x0r + x2r;
|
||||
a[j + 1] = x0i + x2i;
|
||||
x0r -= x2r;
|
||||
x0i -= x2i;
|
||||
a[j + 4] = wk2r * x0r - wk2i * x0i;
|
||||
a[j + 5] = wk2r * x0i + wk2i * x0r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[j + 2] = wk1r * x0r - wk1i * x0i;
|
||||
a[j + 3] = wk1r * x0i + wk1i * x0r;
|
||||
x0r = x1r + x3i;
|
||||
x0i = x1i - x3r;
|
||||
a[j + 6] = wk3r * x0r - wk3i * x0i;
|
||||
a[j + 7] = wk3r * x0i + wk3i * x0r;
|
||||
wk1r = rdft_w[k2 + 2];
|
||||
wk1i = rdft_w[k2 + 3];
|
||||
wk3r = rdft_wk3ri_second[k1 + 0];
|
||||
wk3i = rdft_wk3ri_second[k1 + 1];
|
||||
x0r = a[j + 8] + a[j + 10];
|
||||
x0i = a[j + 9] + a[j + 11];
|
||||
x1r = a[j + 8] - a[j + 10];
|
||||
x1i = a[j + 9] - a[j + 11];
|
||||
x2r = a[j + 12] + a[j + 14];
|
||||
x2i = a[j + 13] + a[j + 15];
|
||||
x3r = a[j + 12] - a[j + 14];
|
||||
x3i = a[j + 13] - a[j + 15];
|
||||
a[j + 8] = x0r + x2r;
|
||||
a[j + 9] = x0i + x2i;
|
||||
x0r -= x2r;
|
||||
x0i -= x2i;
|
||||
a[j + 12] = -wk2i * x0r - wk2r * x0i;
|
||||
a[j + 13] = -wk2i * x0i + wk2r * x0r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[j + 10] = wk1r * x0r - wk1i * x0i;
|
||||
a[j + 11] = wk1r * x0i + wk1i * x0r;
|
||||
x0r = x1r + x3i;
|
||||
x0i = x1i - x3r;
|
||||
a[j + 14] = wk3r * x0r - wk3i * x0i;
|
||||
a[j + 15] = wk3r * x0i + wk3i * x0r;
|
||||
}
|
||||
}
|
||||
|
||||
static void cftmdl_128_C(float* a) {
|
||||
const int l = 8;
|
||||
const int n = 128;
|
||||
const int m = 32;
|
||||
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 = 0; j0 < l; j0 += 2) {
|
||||
j1 = j0 + 8;
|
||||
j2 = j0 + 16;
|
||||
j3 = j0 + 24;
|
||||
x0r = a[j0 + 0] + a[j1 + 0];
|
||||
x0i = a[j0 + 1] + a[j1 + 1];
|
||||
x1r = a[j0 + 0] - a[j1 + 0];
|
||||
x1i = a[j0 + 1] - a[j1 + 1];
|
||||
x2r = a[j2 + 0] + a[j3 + 0];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2 + 0] - a[j3 + 0];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j0 + 0] = x0r + x2r;
|
||||
a[j0 + 1] = x0i + x2i;
|
||||
a[j2 + 0] = x0r - x2r;
|
||||
a[j2 + 1] = x0i - x2i;
|
||||
a[j1 + 0] = x1r - x3i;
|
||||
a[j1 + 1] = x1i + x3r;
|
||||
a[j3 + 0] = x1r + x3i;
|
||||
a[j3 + 1] = x1i - x3r;
|
||||
}
|
||||
wk1r = rdft_w[2];
|
||||
for (j0 = m; j0 < l + m; j0 += 2) {
|
||||
j1 = j0 + 8;
|
||||
j2 = j0 + 16;
|
||||
j3 = j0 + 24;
|
||||
x0r = a[j0 + 0] + a[j1 + 0];
|
||||
x0i = a[j0 + 1] + a[j1 + 1];
|
||||
x1r = a[j0 + 0] - a[j1 + 0];
|
||||
x1i = a[j0 + 1] - a[j1 + 1];
|
||||
x2r = a[j2 + 0] + a[j3 + 0];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2 + 0] - a[j3 + 0];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j0 + 0] = x0r + x2r;
|
||||
a[j0 + 1] = x0i + x2i;
|
||||
a[j2 + 0] = x2i - x0i;
|
||||
a[j2 + 1] = x0r - x2r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[j1 + 0] = wk1r * (x0r - x0i);
|
||||
a[j1 + 1] = wk1r * (x0r + x0i);
|
||||
x0r = x3i + x1r;
|
||||
x0i = x3r - x1i;
|
||||
a[j3 + 0] = wk1r * (x0i - x0r);
|
||||
a[j3 + 1] = wk1r * (x0i + x0r);
|
||||
}
|
||||
k1 = 0;
|
||||
m2 = 2 * m;
|
||||
for (k = m2; k < n; k += m2) {
|
||||
k1 += 2;
|
||||
k2 = 2 * k1;
|
||||
wk2r = rdft_w[k1 + 0];
|
||||
wk2i = rdft_w[k1 + 1];
|
||||
wk1r = rdft_w[k2 + 0];
|
||||
wk1i = rdft_w[k2 + 1];
|
||||
wk3r = rdft_wk3ri_first[k1 + 0];
|
||||
wk3i = rdft_wk3ri_first[k1 + 1];
|
||||
for (j0 = k; j0 < l + k; j0 += 2) {
|
||||
j1 = j0 + 8;
|
||||
j2 = j0 + 16;
|
||||
j3 = j0 + 24;
|
||||
x0r = a[j0 + 0] + a[j1 + 0];
|
||||
x0i = a[j0 + 1] + a[j1 + 1];
|
||||
x1r = a[j0 + 0] - a[j1 + 0];
|
||||
x1i = a[j0 + 1] - a[j1 + 1];
|
||||
x2r = a[j2 + 0] + a[j3 + 0];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2 + 0] - a[j3 + 0];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j0 + 0] = x0r + x2r;
|
||||
a[j0 + 1] = x0i + x2i;
|
||||
x0r -= x2r;
|
||||
x0i -= x2i;
|
||||
a[j2 + 0] = wk2r * x0r - wk2i * x0i;
|
||||
a[j2 + 1] = wk2r * x0i + wk2i * x0r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[j1 + 0] = wk1r * x0r - wk1i * x0i;
|
||||
a[j1 + 1] = wk1r * x0i + wk1i * x0r;
|
||||
x0r = x1r + x3i;
|
||||
x0i = x1i - x3r;
|
||||
a[j3 + 0] = wk3r * x0r - wk3i * x0i;
|
||||
a[j3 + 1] = wk3r * x0i + wk3i * x0r;
|
||||
}
|
||||
wk1r = rdft_w[k2 + 2];
|
||||
wk1i = rdft_w[k2 + 3];
|
||||
wk3r = rdft_wk3ri_second[k1 + 0];
|
||||
wk3i = rdft_wk3ri_second[k1 + 1];
|
||||
for (j0 = k + m; j0 < l + (k + m); j0 += 2) {
|
||||
j1 = j0 + 8;
|
||||
j2 = j0 + 16;
|
||||
j3 = j0 + 24;
|
||||
x0r = a[j0 + 0] + a[j1 + 0];
|
||||
x0i = a[j0 + 1] + a[j1 + 1];
|
||||
x1r = a[j0 + 0] - a[j1 + 0];
|
||||
x1i = a[j0 + 1] - a[j1 + 1];
|
||||
x2r = a[j2 + 0] + a[j3 + 0];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2 + 0] - a[j3 + 0];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j0 + 0] = x0r + x2r;
|
||||
a[j0 + 1] = x0i + x2i;
|
||||
x0r -= x2r;
|
||||
x0i -= x2i;
|
||||
a[j2 + 0] = -wk2i * x0r - wk2r * x0i;
|
||||
a[j2 + 1] = -wk2i * x0i + wk2r * x0r;
|
||||
x0r = x1r - x3i;
|
||||
x0i = x1i + x3r;
|
||||
a[j1 + 0] = wk1r * x0r - wk1i * x0i;
|
||||
a[j1 + 1] = wk1r * x0i + wk1i * x0r;
|
||||
x0r = x1r + x3i;
|
||||
x0i = x1i - x3r;
|
||||
a[j3 + 0] = wk3r * x0r - wk3i * x0i;
|
||||
a[j3 + 1] = wk3r * x0i + wk3i * x0r;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
static void rftfsub_128_C(float* a) {
|
||||
const float* c = rdft_w + 32;
|
||||
int j1, j2, k1, k2;
|
||||
float wkr, wki, xr, xi, yr, yi;
|
||||
|
||||
for (j1 = 1, j2 = 2; j2 < 64; j1 += 1, j2 += 2) {
|
||||
k2 = 128 - j2;
|
||||
k1 = 32 - j1;
|
||||
wkr = 0.5f - c[k1];
|
||||
wki = c[j1];
|
||||
xr = a[j2 + 0] - a[k2 + 0];
|
||||
xi = a[j2 + 1] + a[k2 + 1];
|
||||
yr = wkr * xr - wki * xi;
|
||||
yi = wkr * xi + wki * xr;
|
||||
a[j2 + 0] -= yr;
|
||||
a[j2 + 1] -= yi;
|
||||
a[k2 + 0] += yr;
|
||||
a[k2 + 1] -= yi;
|
||||
}
|
||||
}
|
||||
|
||||
static void rftbsub_128_C(float* a) {
|
||||
const float* c = rdft_w + 32;
|
||||
int j1, j2, k1, k2;
|
||||
float wkr, wki, xr, xi, yr, yi;
|
||||
|
||||
a[1] = -a[1];
|
||||
for (j1 = 1, j2 = 2; j2 < 64; j1 += 1, j2 += 2) {
|
||||
k2 = 128 - j2;
|
||||
k1 = 32 - j1;
|
||||
wkr = 0.5f - c[k1];
|
||||
wki = c[j1];
|
||||
xr = a[j2 + 0] - a[k2 + 0];
|
||||
xi = a[j2 + 1] + a[k2 + 1];
|
||||
yr = wkr * xr + wki * xi;
|
||||
yi = wkr * xi - wki * xr;
|
||||
a[j2 + 0] = a[j2 + 0] - yr;
|
||||
a[j2 + 1] = yi - a[j2 + 1];
|
||||
a[k2 + 0] = yr + a[k2 + 0];
|
||||
a[k2 + 1] = yi - a[k2 + 1];
|
||||
}
|
||||
a[65] = -a[65];
|
||||
}
|
||||
#endif
|
||||
|
||||
} // namespace
|
||||
|
||||
OouraFft::OouraFft(bool sse2_available) {
|
||||
#if defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
use_sse2_ = sse2_available;
|
||||
#else
|
||||
use_sse2_ = false;
|
||||
#endif
|
||||
}
|
||||
|
||||
OouraFft::OouraFft() {
|
||||
#if defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
use_sse2_ = (GetCPUInfo(kSSE2) != 0);
|
||||
#else
|
||||
use_sse2_ = false;
|
||||
#endif
|
||||
}
|
||||
|
||||
OouraFft::~OouraFft() = default;
|
||||
|
||||
void OouraFft::Fft(float* a) const {
|
||||
float xi;
|
||||
bitrv2_128(a);
|
||||
cftfsub_128(a);
|
||||
rftfsub_128(a);
|
||||
xi = a[0] - a[1];
|
||||
a[0] += a[1];
|
||||
a[1] = xi;
|
||||
}
|
||||
void OouraFft::InverseFft(float* a) const {
|
||||
a[1] = 0.5f * (a[0] - a[1]);
|
||||
a[0] -= a[1];
|
||||
rftbsub_128(a);
|
||||
bitrv2_128(a);
|
||||
cftbsub_128(a);
|
||||
}
|
||||
|
||||
void OouraFft::cft1st_128(float* a) const {
|
||||
#if defined(MIPS_FPU_LE)
|
||||
cft1st_128_mips(a);
|
||||
#elif defined(WEBRTC_HAS_NEON)
|
||||
cft1st_128_neon(a);
|
||||
#elif defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
if (use_sse2_) {
|
||||
cft1st_128_SSE2(a);
|
||||
} else {
|
||||
cft1st_128_C(a);
|
||||
}
|
||||
#else
|
||||
cft1st_128_C(a);
|
||||
#endif
|
||||
}
|
||||
void OouraFft::cftmdl_128(float* a) const {
|
||||
#if defined(MIPS_FPU_LE)
|
||||
cftmdl_128_mips(a);
|
||||
#elif defined(WEBRTC_HAS_NEON)
|
||||
cftmdl_128_neon(a);
|
||||
#elif defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
if (use_sse2_) {
|
||||
cftmdl_128_SSE2(a);
|
||||
} else {
|
||||
cftmdl_128_C(a);
|
||||
}
|
||||
#else
|
||||
cftmdl_128_C(a);
|
||||
#endif
|
||||
}
|
||||
void OouraFft::rftfsub_128(float* a) const {
|
||||
#if defined(MIPS_FPU_LE)
|
||||
rftfsub_128_mips(a);
|
||||
#elif defined(WEBRTC_HAS_NEON)
|
||||
rftfsub_128_neon(a);
|
||||
#elif defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
if (use_sse2_) {
|
||||
rftfsub_128_SSE2(a);
|
||||
} else {
|
||||
rftfsub_128_C(a);
|
||||
}
|
||||
#else
|
||||
rftfsub_128_C(a);
|
||||
#endif
|
||||
}
|
||||
|
||||
void OouraFft::rftbsub_128(float* a) const {
|
||||
#if defined(MIPS_FPU_LE)
|
||||
rftbsub_128_mips(a);
|
||||
#elif defined(WEBRTC_HAS_NEON)
|
||||
rftbsub_128_neon(a);
|
||||
#elif defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
if (use_sse2_) {
|
||||
rftbsub_128_SSE2(a);
|
||||
} else {
|
||||
rftbsub_128_C(a);
|
||||
}
|
||||
#else
|
||||
rftbsub_128_C(a);
|
||||
#endif
|
||||
}
|
||||
|
||||
void OouraFft::cftbsub_128(float* a) const {
|
||||
int j, j1, j2, j3, l;
|
||||
float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;
|
||||
|
||||
cft1st_128(a);
|
||||
cftmdl_128(a);
|
||||
l = 32;
|
||||
|
||||
for (j = 0; j < l; j += 2) {
|
||||
j1 = j + l;
|
||||
j2 = j1 + l;
|
||||
j3 = j2 + l;
|
||||
x0r = a[j] + a[j1];
|
||||
x0i = -a[j + 1] - a[j1 + 1];
|
||||
x1r = a[j] - a[j1];
|
||||
x1i = -a[j + 1] + a[j1 + 1];
|
||||
x2r = a[j2] + a[j3];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2] - a[j3];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j] = x0r + x2r;
|
||||
a[j + 1] = x0i - x2i;
|
||||
a[j2] = x0r - x2r;
|
||||
a[j2 + 1] = x0i + x2i;
|
||||
a[j1] = x1r - x3i;
|
||||
a[j1 + 1] = x1i - x3r;
|
||||
a[j3] = x1r + x3i;
|
||||
a[j3 + 1] = x1i + x3r;
|
||||
}
|
||||
}
|
||||
|
||||
void OouraFft::cftfsub_128(float* a) const {
|
||||
int j, j1, j2, j3, l;
|
||||
float x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;
|
||||
|
||||
cft1st_128(a);
|
||||
cftmdl_128(a);
|
||||
l = 32;
|
||||
for (j = 0; j < l; j += 2) {
|
||||
j1 = j + l;
|
||||
j2 = j1 + l;
|
||||
j3 = j2 + l;
|
||||
x0r = a[j] + a[j1];
|
||||
x0i = a[j + 1] + a[j1 + 1];
|
||||
x1r = a[j] - a[j1];
|
||||
x1i = a[j + 1] - a[j1 + 1];
|
||||
x2r = a[j2] + a[j3];
|
||||
x2i = a[j2 + 1] + a[j3 + 1];
|
||||
x3r = a[j2] - a[j3];
|
||||
x3i = a[j2 + 1] - a[j3 + 1];
|
||||
a[j] = x0r + x2r;
|
||||
a[j + 1] = x0i + x2i;
|
||||
a[j2] = x0r - x2r;
|
||||
a[j2 + 1] = x0i - x2i;
|
||||
a[j1] = x1r - x3i;
|
||||
a[j1 + 1] = x1i + x3r;
|
||||
a[j3] = x1r + x3i;
|
||||
a[j3 + 1] = x1i - x3r;
|
||||
}
|
||||
}
|
||||
|
||||
void OouraFft::bitrv2_128(float* a) const {
|
||||
/*
|
||||
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;
|
||||
|
||||
const int ip[4] = {0, 64, 32, 96};
|
||||
for (k = 0; k < 4; k++) {
|
||||
for (j = 0; j < k; j++) {
|
||||
j1 = 2 * j + ip[k];
|
||||
k1 = 2 * k + ip[j];
|
||||
xr = a[j1 + 0];
|
||||
xi = a[j1 + 1];
|
||||
yr = a[k1 + 0];
|
||||
yi = a[k1 + 1];
|
||||
a[j1 + 0] = yr;
|
||||
a[j1 + 1] = yi;
|
||||
a[k1 + 0] = xr;
|
||||
a[k1 + 1] = xi;
|
||||
j1 += 8;
|
||||
k1 += 16;
|
||||
xr = a[j1 + 0];
|
||||
xi = a[j1 + 1];
|
||||
yr = a[k1 + 0];
|
||||
yi = a[k1 + 1];
|
||||
a[j1 + 0] = yr;
|
||||
a[j1 + 1] = yi;
|
||||
a[k1 + 0] = xr;
|
||||
a[k1 + 1] = xi;
|
||||
j1 += 8;
|
||||
k1 -= 8;
|
||||
xr = a[j1 + 0];
|
||||
xi = a[j1 + 1];
|
||||
yr = a[k1 + 0];
|
||||
yi = a[k1 + 1];
|
||||
a[j1 + 0] = yr;
|
||||
a[j1 + 1] = yi;
|
||||
a[k1 + 0] = xr;
|
||||
a[k1 + 1] = xi;
|
||||
j1 += 8;
|
||||
k1 += 16;
|
||||
xr = a[j1 + 0];
|
||||
xi = a[j1 + 1];
|
||||
yr = a[k1 + 0];
|
||||
yi = a[k1 + 1];
|
||||
a[j1 + 0] = yr;
|
||||
a[j1 + 1] = yi;
|
||||
a[k1 + 0] = xr;
|
||||
a[k1 + 1] = xi;
|
||||
}
|
||||
j1 = 2 * k + 8 + ip[k];
|
||||
k1 = j1 + 8;
|
||||
xr = a[j1 + 0];
|
||||
xi = a[j1 + 1];
|
||||
yr = a[k1 + 0];
|
||||
yi = a[k1 + 1];
|
||||
a[j1 + 0] = yr;
|
||||
a[j1 + 1] = yi;
|
||||
a[k1 + 0] = xr;
|
||||
a[k1 + 1] = xi;
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace webrtc
|
64
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft.h
vendored
Normal file
64
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft.h
vendored
Normal file
@ -0,0 +1,64 @@
|
||||
/*
|
||||
* Copyright (c) 2016 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* 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.
|
||||
*/
|
||||
|
||||
#ifndef MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_H_
|
||||
#define MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_H_
|
||||
|
||||
#include "rtc_base/system/arch.h"
|
||||
|
||||
namespace webrtc {
|
||||
|
||||
#if defined(WEBRTC_ARCH_X86_FAMILY)
|
||||
void cft1st_128_SSE2(float* a);
|
||||
void cftmdl_128_SSE2(float* a);
|
||||
void rftfsub_128_SSE2(float* a);
|
||||
void rftbsub_128_SSE2(float* a);
|
||||
#endif
|
||||
|
||||
#if defined(MIPS_FPU_LE)
|
||||
void cft1st_128_mips(float* a);
|
||||
void cftmdl_128_mips(float* a);
|
||||
void rftfsub_128_mips(float* a);
|
||||
void rftbsub_128_mips(float* a);
|
||||
#endif
|
||||
|
||||
#if defined(WEBRTC_HAS_NEON)
|
||||
void cft1st_128_neon(float* a);
|
||||
void cftmdl_128_neon(float* a);
|
||||
void rftfsub_128_neon(float* a);
|
||||
void rftbsub_128_neon(float* a);
|
||||
#endif
|
||||
|
||||
class OouraFft {
|
||||
public:
|
||||
// Ctor allowing the availability of SSE2 support to be specified.
|
||||
explicit OouraFft(bool sse2_available);
|
||||
|
||||
// Deprecated: This Ctor will soon be removed.
|
||||
OouraFft();
|
||||
~OouraFft();
|
||||
void Fft(float* a) const;
|
||||
void InverseFft(float* a) const;
|
||||
|
||||
private:
|
||||
void cft1st_128(float* a) const;
|
||||
void cftmdl_128(float* a) const;
|
||||
void rftfsub_128(float* a) const;
|
||||
void rftbsub_128(float* a) const;
|
||||
|
||||
void cftfsub_128(float* a) const;
|
||||
void cftbsub_128(float* a) const;
|
||||
void bitrv2_128(float* a) const;
|
||||
bool use_sse2_;
|
||||
};
|
||||
|
||||
} // namespace webrtc
|
||||
|
||||
#endif // MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_H_
|
54
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft_tables_common.h
vendored
Normal file
54
VocieProcess/common_audio/third_party/ooura/fft_size_128/ooura_fft_tables_common.h
vendored
Normal file
@ -0,0 +1,54 @@
|
||||
/*
|
||||
* Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* 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.
|
||||
*/
|
||||
|
||||
#ifndef MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_TABLES_COMMON_H_
|
||||
#define MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_TABLES_COMMON_H_
|
||||
|
||||
#include "common_audio/third_party/ooura/fft_size_128/ooura_fft.h"
|
||||
|
||||
namespace webrtc {
|
||||
|
||||
// This tables used to be computed at run-time. For example, refer to:
|
||||
// https://code.google.com/p/webrtc/source/browse/trunk/webrtc/modules/audio_processing/utility/apm_rdft.c?r=6564
|
||||
// to see the initialization code.
|
||||
// Constants shared by all paths (C, SSE2, NEON).
|
||||
const float rdft_w[64] = {
|
||||
1.0000000000f, 0.0000000000f, 0.7071067691f, 0.7071067691f, 0.9238795638f,
|
||||
0.3826834559f, 0.3826834559f, 0.9238795638f, 0.9807852507f, 0.1950903237f,
|
||||
0.5555702448f, 0.8314695954f, 0.8314695954f, 0.5555702448f, 0.1950903237f,
|
||||
0.9807852507f, 0.9951847196f, 0.0980171412f, 0.6343933344f, 0.7730104327f,
|
||||
0.8819212914f, 0.4713967443f, 0.2902846634f, 0.9569403529f, 0.9569403529f,
|
||||
0.2902846634f, 0.4713967443f, 0.8819212914f, 0.7730104327f, 0.6343933344f,
|
||||
0.0980171412f, 0.9951847196f, 0.7071067691f, 0.4993977249f, 0.4975923598f,
|
||||
0.4945882559f, 0.4903926253f, 0.4850156307f, 0.4784701765f, 0.4707720280f,
|
||||
0.4619397819f, 0.4519946277f, 0.4409606457f, 0.4288643003f, 0.4157347977f,
|
||||
0.4016037583f, 0.3865052164f, 0.3704755902f, 0.3535533845f, 0.3357794881f,
|
||||
0.3171966672f, 0.2978496552f, 0.2777851224f, 0.2570513785f, 0.2356983721f,
|
||||
0.2137775421f, 0.1913417280f, 0.1684449315f, 0.1451423317f, 0.1214900985f,
|
||||
0.0975451618f, 0.0733652338f, 0.0490085706f, 0.0245338380f,
|
||||
};
|
||||
|
||||
// Constants used by the C and MIPS paths.
|
||||
const float rdft_wk3ri_first[16] = {
|
||||
1.000000000f, 0.000000000f, 0.382683456f, 0.923879564f,
|
||||
0.831469536f, 0.555570245f, -0.195090353f, 0.980785251f,
|
||||
0.956940353f, 0.290284693f, 0.098017156f, 0.995184720f,
|
||||
0.634393334f, 0.773010492f, -0.471396863f, 0.881921172f,
|
||||
};
|
||||
const float rdft_wk3ri_second[16] = {
|
||||
-0.707106769f, 0.707106769f, -0.923879564f, -0.382683456f,
|
||||
-0.980785251f, 0.195090353f, -0.555570245f, -0.831469536f,
|
||||
-0.881921172f, 0.471396863f, -0.773010492f, -0.634393334f,
|
||||
-0.995184720f, -0.098017156f, -0.290284693f, -0.956940353f,
|
||||
};
|
||||
|
||||
} // namespace webrtc
|
||||
|
||||
#endif // MODULES_AUDIO_PROCESSING_UTILITY_OOURA_FFT_TABLES_COMMON_H_
|
77
VocieProcess/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.c
vendored
Normal file
77
VocieProcess/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.c
vendored
Normal file
@ -0,0 +1,77 @@
|
||||
/*
|
||||
* Written by Wilco Dijkstra, 1996. The following email exchange establishes the
|
||||
* license.
|
||||
*
|
||||
* From: Wilco Dijkstra <Wilco.Dijkstra@ntlworld.com>
|
||||
* Date: Fri, Jun 24, 2011 at 3:20 AM
|
||||
* Subject: Re: sqrt routine
|
||||
* To: Kevin Ma <kma@google.com>
|
||||
* Hi Kevin,
|
||||
* Thanks for asking. Those routines are public domain (originally posted to
|
||||
* comp.sys.arm a long time ago), so you can use them freely for any purpose.
|
||||
* Cheers,
|
||||
* Wilco
|
||||
*
|
||||
* ----- Original Message -----
|
||||
* From: "Kevin Ma" <kma@google.com>
|
||||
* To: <Wilco.Dijkstra@ntlworld.com>
|
||||
* Sent: Thursday, June 23, 2011 11:44 PM
|
||||
* Subject: Fwd: sqrt routine
|
||||
* Hi Wilco,
|
||||
* I saw your sqrt routine from several web sites, including
|
||||
* http://www.finesse.demon.co.uk/steven/sqrt.html.
|
||||
* Just wonder if there's any copyright information with your Successive
|
||||
* approximation routines, or if I can freely use it for any purpose.
|
||||
* Thanks.
|
||||
* Kevin
|
||||
*/
|
||||
|
||||
// Minor modifications in code style for WebRTC, 2012.
|
||||
|
||||
#include "common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.h"
|
||||
|
||||
/*
|
||||
* Algorithm:
|
||||
* Successive approximation of the equation (root + delta) ^ 2 = N
|
||||
* until delta < 1. If delta < 1 we have the integer part of SQRT (N).
|
||||
* Use delta = 2^i for i = 15 .. 0.
|
||||
*
|
||||
* Output precision is 16 bits. Note for large input values (close to
|
||||
* 0x7FFFFFFF), bit 15 (the highest bit of the low 16-bit half word)
|
||||
* contains the MSB information (a non-sign value). Do with caution
|
||||
* if you need to cast the output to int16_t type.
|
||||
*
|
||||
* If the input value is negative, it returns 0.
|
||||
*/
|
||||
|
||||
#define WEBRTC_SPL_SQRT_ITER(N) \
|
||||
try1 = root + (1 << (N)); \
|
||||
if (value >= try1 << (N)) \
|
||||
{ \
|
||||
value -= try1 << (N); \
|
||||
root |= 2 << (N); \
|
||||
}
|
||||
|
||||
int32_t WebRtcSpl_SqrtFloor(int32_t value)
|
||||
{
|
||||
int32_t root = 0, try1;
|
||||
|
||||
WEBRTC_SPL_SQRT_ITER (15);
|
||||
WEBRTC_SPL_SQRT_ITER (14);
|
||||
WEBRTC_SPL_SQRT_ITER (13);
|
||||
WEBRTC_SPL_SQRT_ITER (12);
|
||||
WEBRTC_SPL_SQRT_ITER (11);
|
||||
WEBRTC_SPL_SQRT_ITER (10);
|
||||
WEBRTC_SPL_SQRT_ITER ( 9);
|
||||
WEBRTC_SPL_SQRT_ITER ( 8);
|
||||
WEBRTC_SPL_SQRT_ITER ( 7);
|
||||
WEBRTC_SPL_SQRT_ITER ( 6);
|
||||
WEBRTC_SPL_SQRT_ITER ( 5);
|
||||
WEBRTC_SPL_SQRT_ITER ( 4);
|
||||
WEBRTC_SPL_SQRT_ITER ( 3);
|
||||
WEBRTC_SPL_SQRT_ITER ( 2);
|
||||
WEBRTC_SPL_SQRT_ITER ( 1);
|
||||
WEBRTC_SPL_SQRT_ITER ( 0);
|
||||
|
||||
return root >> 1;
|
||||
}
|
29
VocieProcess/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.h
vendored
Normal file
29
VocieProcess/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.h
vendored
Normal file
@ -0,0 +1,29 @@
|
||||
/*
|
||||
* Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
|
||||
*
|
||||
* 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 <stdint.h>
|
||||
|
||||
//
|
||||
// WebRtcSpl_SqrtFloor(...)
|
||||
//
|
||||
// Returns the square root of the input value `value`. The precision of this
|
||||
// function is rounding down integer precision, i.e., sqrt(8) gives 2 as answer.
|
||||
// If `value` is a negative number then 0 is returned.
|
||||
//
|
||||
// Algorithm:
|
||||
//
|
||||
// An iterative 4 cylce/bit routine
|
||||
//
|
||||
// Input:
|
||||
// - value : Value to calculate sqrt of
|
||||
//
|
||||
// Return value : Result of the sqrt calculation
|
||||
//
|
||||
int32_t WebRtcSpl_SqrtFloor(int32_t value);
|
Reference in New Issue
Block a user