tor-browser

The Tor Browser
git clone https://git.dasho.dev/tor-browser.git
Log | Files | Refs | README | LICENSE

vector_math_avx2.cc (2221B)

      1 /*
      2 *  Copyright (c) 2020 The WebRTC project authors. All Rights Reserved.
      3 *
      4 *  Use of this source code is governed by a BSD-style license
      5 *  that can be found in the LICENSE file in the root of the source
      6 *  tree. An additional intellectual property rights grant can be found
      7 *  in the file PATENTS.  All contributing project authors may
      8 *  be found in the AUTHORS file in the root of the source tree.
      9 */
     10 
     11 #include <immintrin.h>
     12 
     13 #include <cmath>
     14 
     15 #include "api/array_view.h"
     16 #include "modules/audio_processing/aec3/vector_math.h"
     17 #include "rtc_base/checks.h"
     18 
     19 namespace webrtc {
     20 namespace aec3 {
     21 
     22 // Elementwise square root.
     23 void VectorMath::SqrtAVX2(ArrayView<float> x) {
     24  const int x_size = static_cast<int>(x.size());
     25  const int vector_limit = x_size >> 3;
     26 
     27  int j = 0;
     28  for (; j < vector_limit * 8; j += 8) {
     29    __m256 g = _mm256_loadu_ps(&x[j]);
     30    g = _mm256_sqrt_ps(g);
     31    _mm256_storeu_ps(&x[j], g);
     32  }
     33 
     34  for (; j < x_size; ++j) {
     35    x[j] = sqrtf(x[j]);
     36  }
     37 }
     38 
     39 // Elementwise vector multiplication z = x * y.
     40 void VectorMath::MultiplyAVX2(ArrayView<const float> x,
     41                              ArrayView<const float> y,
     42                              ArrayView<float> z) {
     43  RTC_DCHECK_EQ(z.size(), x.size());
     44  RTC_DCHECK_EQ(z.size(), y.size());
     45  const int x_size = static_cast<int>(x.size());
     46  const int vector_limit = x_size >> 3;
     47 
     48  int j = 0;
     49  for (; j < vector_limit * 8; j += 8) {
     50    const __m256 x_j = _mm256_loadu_ps(&x[j]);
     51    const __m256 y_j = _mm256_loadu_ps(&y[j]);
     52    const __m256 z_j = _mm256_mul_ps(x_j, y_j);
     53    _mm256_storeu_ps(&z[j], z_j);
     54  }
     55 
     56  for (; j < x_size; ++j) {
     57    z[j] = x[j] * y[j];
     58  }
     59 }
     60 
     61 // Elementwise vector accumulation z += x.
     62 void VectorMath::AccumulateAVX2(ArrayView<const float> x, ArrayView<float> z) {
     63  RTC_DCHECK_EQ(z.size(), x.size());
     64  const int x_size = static_cast<int>(x.size());
     65  const int vector_limit = x_size >> 3;
     66 
     67  int j = 0;
     68  for (; j < vector_limit * 8; j += 8) {
     69    const __m256 x_j = _mm256_loadu_ps(&x[j]);
     70    __m256 z_j = _mm256_loadu_ps(&z[j]);
     71    z_j = _mm256_add_ps(x_j, z_j);
     72    _mm256_storeu_ps(&z[j], z_j);
     73  }
     74 
     75  for (; j < x_size; ++j) {
     76    z[j] += x[j];
     77  }
     78 }
     79 
     80 }  // namespace aec3
     81 }  // namespace webrtc