1 /*
2 * Copyright (c) 2013 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 // Modified from the Chromium original:
12 // src/media/base/sinc_resampler.cc
13
14 #include <arm_neon.h>
15
16 #include "common_audio/resampler/sinc_resampler.h"
17
18 namespace webrtc {
19
Convolve_NEON(const float * input_ptr,const float * k1,const float * k2,double kernel_interpolation_factor)20 float SincResampler::Convolve_NEON(const float* input_ptr,
21 const float* k1,
22 const float* k2,
23 double kernel_interpolation_factor) {
24 float32x4_t m_input;
25 float32x4_t m_sums1 = vmovq_n_f32(0);
26 float32x4_t m_sums2 = vmovq_n_f32(0);
27
28 const float* upper = input_ptr + kKernelSize;
29 for (; input_ptr < upper;) {
30 m_input = vld1q_f32(input_ptr);
31 input_ptr += 4;
32 m_sums1 = vmlaq_f32(m_sums1, m_input, vld1q_f32(k1));
33 k1 += 4;
34 m_sums2 = vmlaq_f32(m_sums2, m_input, vld1q_f32(k2));
35 k2 += 4;
36 }
37
38 // Linearly interpolate the two "convolutions".
39 m_sums1 = vmlaq_f32(
40 vmulq_f32(m_sums1, vmovq_n_f32(1.0 - kernel_interpolation_factor)),
41 m_sums2, vmovq_n_f32(kernel_interpolation_factor));
42
43 // Sum components together.
44 float32x2_t m_half = vadd_f32(vget_high_f32(m_sums1), vget_low_f32(m_sums1));
45 return vget_lane_f32(vpadd_f32(m_half, m_half), 0);
46 }
47
48 } // namespace webrtc
49