1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/content/media/webaudio/blink/DirectConvolver.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,347 @@
1.4 +/*
1.5 + * Copyright (C) 2012 Intel Inc. All rights reserved.
1.6 + *
1.7 + * Redistribution and use in source and binary forms, with or without
1.8 + * modification, are permitted provided that the following conditions
1.9 + * are met:
1.10 + *
1.11 + * 1. Redistributions of source code must retain the above copyright
1.12 + * notice, this list of conditions and the following disclaimer.
1.13 + * 2. Redistributions in binary form must reproduce the above copyright
1.14 + * notice, this list of conditions and the following disclaimer in the
1.15 + * documentation and/or other materials provided with the distribution.
1.16 + * 3. Neither the name of Apple Computer, Inc. ("Apple") nor the names of
1.17 + * its contributors may be used to endorse or promote products derived
1.18 + * from this software without specific prior written permission.
1.19 + *
1.20 + * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND ANY
1.21 + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1.22 + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1.23 + * DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR ANY
1.24 + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1.25 + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
1.26 + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
1.27 + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
1.28 + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
1.29 + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
1.30 + */
1.31 +
1.32 +#include "DirectConvolver.h"
1.33 +#include "mozilla/PodOperations.h"
1.34 +
1.35 +using namespace mozilla;
1.36 +
1.37 +namespace WebCore {
1.38 +
1.39 +DirectConvolver::DirectConvolver(size_t inputBlockSize)
1.40 + : m_inputBlockSize(inputBlockSize)
1.41 +{
1.42 + m_buffer.SetLength(inputBlockSize * 2);
1.43 + PodZero(m_buffer.Elements(), inputBlockSize * 2);
1.44 +}
1.45 +
1.46 +void DirectConvolver::process(const nsTArray<float>* convolutionKernel, const float* sourceP, float* destP, size_t framesToProcess)
1.47 +{
1.48 + MOZ_ASSERT(framesToProcess == m_inputBlockSize);
1.49 + if (framesToProcess != m_inputBlockSize)
1.50 + return;
1.51 +
1.52 + // Only support kernelSize <= m_inputBlockSize
1.53 + size_t kernelSize = convolutionKernel->Length();
1.54 + MOZ_ASSERT(kernelSize <= m_inputBlockSize);
1.55 + if (kernelSize > m_inputBlockSize)
1.56 + return;
1.57 +
1.58 + const float* kernelP = convolutionKernel->Elements();
1.59 +
1.60 + // Sanity check
1.61 + bool isCopyGood = kernelP && sourceP && destP && m_buffer.Elements();
1.62 + MOZ_ASSERT(isCopyGood);
1.63 + if (!isCopyGood)
1.64 + return;
1.65 +
1.66 + float* inputP = m_buffer.Elements() + m_inputBlockSize;
1.67 +
1.68 + // Copy samples to 2nd half of input buffer.
1.69 + memcpy(inputP, sourceP, sizeof(float) * framesToProcess);
1.70 +
1.71 + // FIXME: The macro can be further optimized to avoid pipeline stalls. One possibility is to maintain 4 separate sums and change the macro to CONVOLVE_FOUR_SAMPLES.
1.72 +#define CONVOLVE_ONE_SAMPLE \
1.73 + sum += inputP[i - j] * kernelP[j]; \
1.74 + j++;
1.75 +
1.76 + size_t i = 0;
1.77 + while (i < framesToProcess) {
1.78 + size_t j = 0;
1.79 + float sum = 0;
1.80 +
1.81 + // FIXME: SSE optimization may be applied here.
1.82 + if (kernelSize == 32) {
1.83 + CONVOLVE_ONE_SAMPLE // 1
1.84 + CONVOLVE_ONE_SAMPLE // 2
1.85 + CONVOLVE_ONE_SAMPLE // 3
1.86 + CONVOLVE_ONE_SAMPLE // 4
1.87 + CONVOLVE_ONE_SAMPLE // 5
1.88 + CONVOLVE_ONE_SAMPLE // 6
1.89 + CONVOLVE_ONE_SAMPLE // 7
1.90 + CONVOLVE_ONE_SAMPLE // 8
1.91 + CONVOLVE_ONE_SAMPLE // 9
1.92 + CONVOLVE_ONE_SAMPLE // 10
1.93 +
1.94 + CONVOLVE_ONE_SAMPLE // 11
1.95 + CONVOLVE_ONE_SAMPLE // 12
1.96 + CONVOLVE_ONE_SAMPLE // 13
1.97 + CONVOLVE_ONE_SAMPLE // 14
1.98 + CONVOLVE_ONE_SAMPLE // 15
1.99 + CONVOLVE_ONE_SAMPLE // 16
1.100 + CONVOLVE_ONE_SAMPLE // 17
1.101 + CONVOLVE_ONE_SAMPLE // 18
1.102 + CONVOLVE_ONE_SAMPLE // 19
1.103 + CONVOLVE_ONE_SAMPLE // 20
1.104 +
1.105 + CONVOLVE_ONE_SAMPLE // 21
1.106 + CONVOLVE_ONE_SAMPLE // 22
1.107 + CONVOLVE_ONE_SAMPLE // 23
1.108 + CONVOLVE_ONE_SAMPLE // 24
1.109 + CONVOLVE_ONE_SAMPLE // 25
1.110 + CONVOLVE_ONE_SAMPLE // 26
1.111 + CONVOLVE_ONE_SAMPLE // 27
1.112 + CONVOLVE_ONE_SAMPLE // 28
1.113 + CONVOLVE_ONE_SAMPLE // 29
1.114 + CONVOLVE_ONE_SAMPLE // 30
1.115 +
1.116 + CONVOLVE_ONE_SAMPLE // 31
1.117 + CONVOLVE_ONE_SAMPLE // 32
1.118 +
1.119 + } else if (kernelSize == 64) {
1.120 + CONVOLVE_ONE_SAMPLE // 1
1.121 + CONVOLVE_ONE_SAMPLE // 2
1.122 + CONVOLVE_ONE_SAMPLE // 3
1.123 + CONVOLVE_ONE_SAMPLE // 4
1.124 + CONVOLVE_ONE_SAMPLE // 5
1.125 + CONVOLVE_ONE_SAMPLE // 6
1.126 + CONVOLVE_ONE_SAMPLE // 7
1.127 + CONVOLVE_ONE_SAMPLE // 8
1.128 + CONVOLVE_ONE_SAMPLE // 9
1.129 + CONVOLVE_ONE_SAMPLE // 10
1.130 +
1.131 + CONVOLVE_ONE_SAMPLE // 11
1.132 + CONVOLVE_ONE_SAMPLE // 12
1.133 + CONVOLVE_ONE_SAMPLE // 13
1.134 + CONVOLVE_ONE_SAMPLE // 14
1.135 + CONVOLVE_ONE_SAMPLE // 15
1.136 + CONVOLVE_ONE_SAMPLE // 16
1.137 + CONVOLVE_ONE_SAMPLE // 17
1.138 + CONVOLVE_ONE_SAMPLE // 18
1.139 + CONVOLVE_ONE_SAMPLE // 19
1.140 + CONVOLVE_ONE_SAMPLE // 20
1.141 +
1.142 + CONVOLVE_ONE_SAMPLE // 21
1.143 + CONVOLVE_ONE_SAMPLE // 22
1.144 + CONVOLVE_ONE_SAMPLE // 23
1.145 + CONVOLVE_ONE_SAMPLE // 24
1.146 + CONVOLVE_ONE_SAMPLE // 25
1.147 + CONVOLVE_ONE_SAMPLE // 26
1.148 + CONVOLVE_ONE_SAMPLE // 27
1.149 + CONVOLVE_ONE_SAMPLE // 28
1.150 + CONVOLVE_ONE_SAMPLE // 29
1.151 + CONVOLVE_ONE_SAMPLE // 30
1.152 +
1.153 + CONVOLVE_ONE_SAMPLE // 31
1.154 + CONVOLVE_ONE_SAMPLE // 32
1.155 + CONVOLVE_ONE_SAMPLE // 33
1.156 + CONVOLVE_ONE_SAMPLE // 34
1.157 + CONVOLVE_ONE_SAMPLE // 35
1.158 + CONVOLVE_ONE_SAMPLE // 36
1.159 + CONVOLVE_ONE_SAMPLE // 37
1.160 + CONVOLVE_ONE_SAMPLE // 38
1.161 + CONVOLVE_ONE_SAMPLE // 39
1.162 + CONVOLVE_ONE_SAMPLE // 40
1.163 +
1.164 + CONVOLVE_ONE_SAMPLE // 41
1.165 + CONVOLVE_ONE_SAMPLE // 42
1.166 + CONVOLVE_ONE_SAMPLE // 43
1.167 + CONVOLVE_ONE_SAMPLE // 44
1.168 + CONVOLVE_ONE_SAMPLE // 45
1.169 + CONVOLVE_ONE_SAMPLE // 46
1.170 + CONVOLVE_ONE_SAMPLE // 47
1.171 + CONVOLVE_ONE_SAMPLE // 48
1.172 + CONVOLVE_ONE_SAMPLE // 49
1.173 + CONVOLVE_ONE_SAMPLE // 50
1.174 +
1.175 + CONVOLVE_ONE_SAMPLE // 51
1.176 + CONVOLVE_ONE_SAMPLE // 52
1.177 + CONVOLVE_ONE_SAMPLE // 53
1.178 + CONVOLVE_ONE_SAMPLE // 54
1.179 + CONVOLVE_ONE_SAMPLE // 55
1.180 + CONVOLVE_ONE_SAMPLE // 56
1.181 + CONVOLVE_ONE_SAMPLE // 57
1.182 + CONVOLVE_ONE_SAMPLE // 58
1.183 + CONVOLVE_ONE_SAMPLE // 59
1.184 + CONVOLVE_ONE_SAMPLE // 60
1.185 +
1.186 + CONVOLVE_ONE_SAMPLE // 61
1.187 + CONVOLVE_ONE_SAMPLE // 62
1.188 + CONVOLVE_ONE_SAMPLE // 63
1.189 + CONVOLVE_ONE_SAMPLE // 64
1.190 +
1.191 + } else if (kernelSize == 128) {
1.192 + CONVOLVE_ONE_SAMPLE // 1
1.193 + CONVOLVE_ONE_SAMPLE // 2
1.194 + CONVOLVE_ONE_SAMPLE // 3
1.195 + CONVOLVE_ONE_SAMPLE // 4
1.196 + CONVOLVE_ONE_SAMPLE // 5
1.197 + CONVOLVE_ONE_SAMPLE // 6
1.198 + CONVOLVE_ONE_SAMPLE // 7
1.199 + CONVOLVE_ONE_SAMPLE // 8
1.200 + CONVOLVE_ONE_SAMPLE // 9
1.201 + CONVOLVE_ONE_SAMPLE // 10
1.202 +
1.203 + CONVOLVE_ONE_SAMPLE // 11
1.204 + CONVOLVE_ONE_SAMPLE // 12
1.205 + CONVOLVE_ONE_SAMPLE // 13
1.206 + CONVOLVE_ONE_SAMPLE // 14
1.207 + CONVOLVE_ONE_SAMPLE // 15
1.208 + CONVOLVE_ONE_SAMPLE // 16
1.209 + CONVOLVE_ONE_SAMPLE // 17
1.210 + CONVOLVE_ONE_SAMPLE // 18
1.211 + CONVOLVE_ONE_SAMPLE // 19
1.212 + CONVOLVE_ONE_SAMPLE // 20
1.213 +
1.214 + CONVOLVE_ONE_SAMPLE // 21
1.215 + CONVOLVE_ONE_SAMPLE // 22
1.216 + CONVOLVE_ONE_SAMPLE // 23
1.217 + CONVOLVE_ONE_SAMPLE // 24
1.218 + CONVOLVE_ONE_SAMPLE // 25
1.219 + CONVOLVE_ONE_SAMPLE // 26
1.220 + CONVOLVE_ONE_SAMPLE // 27
1.221 + CONVOLVE_ONE_SAMPLE // 28
1.222 + CONVOLVE_ONE_SAMPLE // 29
1.223 + CONVOLVE_ONE_SAMPLE // 30
1.224 +
1.225 + CONVOLVE_ONE_SAMPLE // 31
1.226 + CONVOLVE_ONE_SAMPLE // 32
1.227 + CONVOLVE_ONE_SAMPLE // 33
1.228 + CONVOLVE_ONE_SAMPLE // 34
1.229 + CONVOLVE_ONE_SAMPLE // 35
1.230 + CONVOLVE_ONE_SAMPLE // 36
1.231 + CONVOLVE_ONE_SAMPLE // 37
1.232 + CONVOLVE_ONE_SAMPLE // 38
1.233 + CONVOLVE_ONE_SAMPLE // 39
1.234 + CONVOLVE_ONE_SAMPLE // 40
1.235 +
1.236 + CONVOLVE_ONE_SAMPLE // 41
1.237 + CONVOLVE_ONE_SAMPLE // 42
1.238 + CONVOLVE_ONE_SAMPLE // 43
1.239 + CONVOLVE_ONE_SAMPLE // 44
1.240 + CONVOLVE_ONE_SAMPLE // 45
1.241 + CONVOLVE_ONE_SAMPLE // 46
1.242 + CONVOLVE_ONE_SAMPLE // 47
1.243 + CONVOLVE_ONE_SAMPLE // 48
1.244 + CONVOLVE_ONE_SAMPLE // 49
1.245 + CONVOLVE_ONE_SAMPLE // 50
1.246 +
1.247 + CONVOLVE_ONE_SAMPLE // 51
1.248 + CONVOLVE_ONE_SAMPLE // 52
1.249 + CONVOLVE_ONE_SAMPLE // 53
1.250 + CONVOLVE_ONE_SAMPLE // 54
1.251 + CONVOLVE_ONE_SAMPLE // 55
1.252 + CONVOLVE_ONE_SAMPLE // 56
1.253 + CONVOLVE_ONE_SAMPLE // 57
1.254 + CONVOLVE_ONE_SAMPLE // 58
1.255 + CONVOLVE_ONE_SAMPLE // 59
1.256 + CONVOLVE_ONE_SAMPLE // 60
1.257 +
1.258 + CONVOLVE_ONE_SAMPLE // 61
1.259 + CONVOLVE_ONE_SAMPLE // 62
1.260 + CONVOLVE_ONE_SAMPLE // 63
1.261 + CONVOLVE_ONE_SAMPLE // 64
1.262 + CONVOLVE_ONE_SAMPLE // 65
1.263 + CONVOLVE_ONE_SAMPLE // 66
1.264 + CONVOLVE_ONE_SAMPLE // 67
1.265 + CONVOLVE_ONE_SAMPLE // 68
1.266 + CONVOLVE_ONE_SAMPLE // 69
1.267 + CONVOLVE_ONE_SAMPLE // 70
1.268 +
1.269 + CONVOLVE_ONE_SAMPLE // 71
1.270 + CONVOLVE_ONE_SAMPLE // 72
1.271 + CONVOLVE_ONE_SAMPLE // 73
1.272 + CONVOLVE_ONE_SAMPLE // 74
1.273 + CONVOLVE_ONE_SAMPLE // 75
1.274 + CONVOLVE_ONE_SAMPLE // 76
1.275 + CONVOLVE_ONE_SAMPLE // 77
1.276 + CONVOLVE_ONE_SAMPLE // 78
1.277 + CONVOLVE_ONE_SAMPLE // 79
1.278 + CONVOLVE_ONE_SAMPLE // 80
1.279 +
1.280 + CONVOLVE_ONE_SAMPLE // 81
1.281 + CONVOLVE_ONE_SAMPLE // 82
1.282 + CONVOLVE_ONE_SAMPLE // 83
1.283 + CONVOLVE_ONE_SAMPLE // 84
1.284 + CONVOLVE_ONE_SAMPLE // 85
1.285 + CONVOLVE_ONE_SAMPLE // 86
1.286 + CONVOLVE_ONE_SAMPLE // 87
1.287 + CONVOLVE_ONE_SAMPLE // 88
1.288 + CONVOLVE_ONE_SAMPLE // 89
1.289 + CONVOLVE_ONE_SAMPLE // 90
1.290 +
1.291 + CONVOLVE_ONE_SAMPLE // 91
1.292 + CONVOLVE_ONE_SAMPLE // 92
1.293 + CONVOLVE_ONE_SAMPLE // 93
1.294 + CONVOLVE_ONE_SAMPLE // 94
1.295 + CONVOLVE_ONE_SAMPLE // 95
1.296 + CONVOLVE_ONE_SAMPLE // 96
1.297 + CONVOLVE_ONE_SAMPLE // 97
1.298 + CONVOLVE_ONE_SAMPLE // 98
1.299 + CONVOLVE_ONE_SAMPLE // 99
1.300 + CONVOLVE_ONE_SAMPLE // 100
1.301 +
1.302 + CONVOLVE_ONE_SAMPLE // 101
1.303 + CONVOLVE_ONE_SAMPLE // 102
1.304 + CONVOLVE_ONE_SAMPLE // 103
1.305 + CONVOLVE_ONE_SAMPLE // 104
1.306 + CONVOLVE_ONE_SAMPLE // 105
1.307 + CONVOLVE_ONE_SAMPLE // 106
1.308 + CONVOLVE_ONE_SAMPLE // 107
1.309 + CONVOLVE_ONE_SAMPLE // 108
1.310 + CONVOLVE_ONE_SAMPLE // 109
1.311 + CONVOLVE_ONE_SAMPLE // 110
1.312 +
1.313 + CONVOLVE_ONE_SAMPLE // 111
1.314 + CONVOLVE_ONE_SAMPLE // 112
1.315 + CONVOLVE_ONE_SAMPLE // 113
1.316 + CONVOLVE_ONE_SAMPLE // 114
1.317 + CONVOLVE_ONE_SAMPLE // 115
1.318 + CONVOLVE_ONE_SAMPLE // 116
1.319 + CONVOLVE_ONE_SAMPLE // 117
1.320 + CONVOLVE_ONE_SAMPLE // 118
1.321 + CONVOLVE_ONE_SAMPLE // 119
1.322 + CONVOLVE_ONE_SAMPLE // 120
1.323 +
1.324 + CONVOLVE_ONE_SAMPLE // 121
1.325 + CONVOLVE_ONE_SAMPLE // 122
1.326 + CONVOLVE_ONE_SAMPLE // 123
1.327 + CONVOLVE_ONE_SAMPLE // 124
1.328 + CONVOLVE_ONE_SAMPLE // 125
1.329 + CONVOLVE_ONE_SAMPLE // 126
1.330 + CONVOLVE_ONE_SAMPLE // 127
1.331 + CONVOLVE_ONE_SAMPLE // 128
1.332 + } else {
1.333 + while (j < kernelSize) {
1.334 + // Non-optimized using actual while loop.
1.335 + CONVOLVE_ONE_SAMPLE
1.336 + }
1.337 + }
1.338 + destP[i++] = sum;
1.339 + }
1.340 +
1.341 + // Copy 2nd half of input buffer to 1st half.
1.342 + memcpy(m_buffer.Elements(), inputP, sizeof(float) * framesToProcess);
1.343 +}
1.344 +
1.345 +void DirectConvolver::reset()
1.346 +{
1.347 + PodZero(m_buffer.Elements(), m_buffer.Length());
1.348 +}
1.349 +
1.350 +} // namespace WebCore
