andrew@webrtc.org | c6a3755 | 2013-05-08 20:35:43 | [diff] [blame] | 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 "webrtc/common_audio/resampler/sinc_resampler.h" |
| 15 | |
| 16 | #include <arm_neon.h> |
| 17 | |
| 18 | namespace webrtc { |
| 19 | |
| 20 | float SincResampler::Convolve_NEON(const float* input_ptr, const float* k1, |
| 21 | const float* k2, |
| 22 | double kernel_interpolation_factor) { |
| 23 | float32x4_t m_input; |
| 24 | float32x4_t m_sums1 = vmovq_n_f32(0); |
| 25 | float32x4_t m_sums2 = vmovq_n_f32(0); |
| 26 | |
| 27 | const float* upper = input_ptr + kKernelSize; |
| 28 | for (; input_ptr < upper; ) { |
| 29 | m_input = vld1q_f32(input_ptr); |
| 30 | input_ptr += 4; |
| 31 | m_sums1 = vmlaq_f32(m_sums1, m_input, vld1q_f32(k1)); |
| 32 | k1 += 4; |
| 33 | m_sums2 = vmlaq_f32(m_sums2, m_input, vld1q_f32(k2)); |
| 34 | k2 += 4; |
| 35 | } |
| 36 | |
| 37 | // Linearly interpolate the two "convolutions". |
| 38 | m_sums1 = vmlaq_f32( |
| 39 | vmulq_f32(m_sums1, vmovq_n_f32(1.0 - kernel_interpolation_factor)), |
| 40 | m_sums2, vmovq_n_f32(kernel_interpolation_factor)); |
| 41 | |
| 42 | // Sum components together. |
| 43 | float32x2_t m_half = vadd_f32(vget_high_f32(m_sums1), vget_low_f32(m_sums1)); |
| 44 | return vget_lane_f32(vpadd_f32(m_half, m_half), 0); |
| 45 | } |
| 46 | |
| 47 | } // namespace webrtc |