michael@0: /*
michael@0:  * Copyright (C) 2012 Intel Inc. All rights reserved.
michael@0:  *
michael@0:  * Redistribution and use in source and binary forms, with or without
michael@0:  * modification, are permitted provided that the following conditions
michael@0:  * are met:
michael@0:  *
michael@0:  * 1.  Redistributions of source code must retain the above copyright
michael@0:  *     notice, this list of conditions and the following disclaimer.
michael@0:  * 2.  Redistributions in binary form must reproduce the above copyright
michael@0:  *     notice, this list of conditions and the following disclaimer in the
michael@0:  *     documentation and/or other materials provided with the distribution.
michael@0:  * 3.  Neither the name of Apple Computer, Inc. ("Apple") nor the names of
michael@0:  *     its contributors may be used to endorse or promote products derived
michael@0:  *     from this software without specific prior written permission.
michael@0:  *
michael@0:  * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND ANY
michael@0:  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
michael@0:  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
michael@0:  * DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR ANY
michael@0:  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
michael@0:  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
michael@0:  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
michael@0:  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
michael@0:  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
michael@0:  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
michael@0:  */
michael@0: 
michael@0: #include "DirectConvolver.h"
michael@0: #include "mozilla/PodOperations.h"
michael@0: 
michael@0: using namespace mozilla;
michael@0: 
michael@0: namespace WebCore {
michael@0: 
michael@0: DirectConvolver::DirectConvolver(size_t inputBlockSize)
michael@0:     : m_inputBlockSize(inputBlockSize)
michael@0: {
michael@0:   m_buffer.SetLength(inputBlockSize * 2);
michael@0:   PodZero(m_buffer.Elements(), inputBlockSize * 2);
michael@0: }
michael@0: 
michael@0: void DirectConvolver::process(const nsTArray<float>* convolutionKernel, const float* sourceP, float* destP, size_t framesToProcess)
michael@0: {
michael@0:     MOZ_ASSERT(framesToProcess == m_inputBlockSize);
michael@0:     if (framesToProcess != m_inputBlockSize)
michael@0:         return;
michael@0: 
michael@0:     // Only support kernelSize <= m_inputBlockSize
michael@0:     size_t kernelSize = convolutionKernel->Length();
michael@0:     MOZ_ASSERT(kernelSize <= m_inputBlockSize);
michael@0:     if (kernelSize > m_inputBlockSize)
michael@0:         return;
michael@0: 
michael@0:     const float* kernelP = convolutionKernel->Elements();
michael@0: 
michael@0:     // Sanity check
michael@0:     bool isCopyGood = kernelP && sourceP && destP && m_buffer.Elements();
michael@0:     MOZ_ASSERT(isCopyGood);
michael@0:     if (!isCopyGood)
michael@0:         return;
michael@0: 
michael@0:     float* inputP = m_buffer.Elements() + m_inputBlockSize;
michael@0: 
michael@0:     // Copy samples to 2nd half of input buffer.
michael@0:     memcpy(inputP, sourceP, sizeof(float) * framesToProcess);
michael@0: 
michael@0:     // 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.
michael@0: #define CONVOLVE_ONE_SAMPLE             \
michael@0:     sum += inputP[i - j] * kernelP[j];  \
michael@0:     j++;
michael@0: 
michael@0:     size_t i = 0;
michael@0:     while (i < framesToProcess) {
michael@0:         size_t j = 0;
michael@0:         float sum = 0;
michael@0:         
michael@0:         // FIXME: SSE optimization may be applied here.
michael@0:         if (kernelSize == 32) {
michael@0:             CONVOLVE_ONE_SAMPLE // 1
michael@0:             CONVOLVE_ONE_SAMPLE // 2
michael@0:             CONVOLVE_ONE_SAMPLE // 3
michael@0:             CONVOLVE_ONE_SAMPLE // 4
michael@0:             CONVOLVE_ONE_SAMPLE // 5
michael@0:             CONVOLVE_ONE_SAMPLE // 6
michael@0:             CONVOLVE_ONE_SAMPLE // 7
michael@0:             CONVOLVE_ONE_SAMPLE // 8
michael@0:             CONVOLVE_ONE_SAMPLE // 9
michael@0:             CONVOLVE_ONE_SAMPLE // 10
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 11
michael@0:             CONVOLVE_ONE_SAMPLE // 12
michael@0:             CONVOLVE_ONE_SAMPLE // 13
michael@0:             CONVOLVE_ONE_SAMPLE // 14
michael@0:             CONVOLVE_ONE_SAMPLE // 15
michael@0:             CONVOLVE_ONE_SAMPLE // 16
michael@0:             CONVOLVE_ONE_SAMPLE // 17
michael@0:             CONVOLVE_ONE_SAMPLE // 18
michael@0:             CONVOLVE_ONE_SAMPLE // 19
michael@0:             CONVOLVE_ONE_SAMPLE // 20
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 21
michael@0:             CONVOLVE_ONE_SAMPLE // 22
michael@0:             CONVOLVE_ONE_SAMPLE // 23
michael@0:             CONVOLVE_ONE_SAMPLE // 24
michael@0:             CONVOLVE_ONE_SAMPLE // 25
michael@0:             CONVOLVE_ONE_SAMPLE // 26
michael@0:             CONVOLVE_ONE_SAMPLE // 27
michael@0:             CONVOLVE_ONE_SAMPLE // 28
michael@0:             CONVOLVE_ONE_SAMPLE // 29
michael@0:             CONVOLVE_ONE_SAMPLE // 30
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 31
michael@0:             CONVOLVE_ONE_SAMPLE // 32
michael@0: 
michael@0:         } else if (kernelSize == 64) {
michael@0:             CONVOLVE_ONE_SAMPLE // 1
michael@0:             CONVOLVE_ONE_SAMPLE // 2
michael@0:             CONVOLVE_ONE_SAMPLE // 3
michael@0:             CONVOLVE_ONE_SAMPLE // 4
michael@0:             CONVOLVE_ONE_SAMPLE // 5
michael@0:             CONVOLVE_ONE_SAMPLE // 6
michael@0:             CONVOLVE_ONE_SAMPLE // 7
michael@0:             CONVOLVE_ONE_SAMPLE // 8
michael@0:             CONVOLVE_ONE_SAMPLE // 9
michael@0:             CONVOLVE_ONE_SAMPLE // 10
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 11
michael@0:             CONVOLVE_ONE_SAMPLE // 12
michael@0:             CONVOLVE_ONE_SAMPLE // 13
michael@0:             CONVOLVE_ONE_SAMPLE // 14
michael@0:             CONVOLVE_ONE_SAMPLE // 15
michael@0:             CONVOLVE_ONE_SAMPLE // 16
michael@0:             CONVOLVE_ONE_SAMPLE // 17
michael@0:             CONVOLVE_ONE_SAMPLE // 18
michael@0:             CONVOLVE_ONE_SAMPLE // 19
michael@0:             CONVOLVE_ONE_SAMPLE // 20
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 21
michael@0:             CONVOLVE_ONE_SAMPLE // 22
michael@0:             CONVOLVE_ONE_SAMPLE // 23
michael@0:             CONVOLVE_ONE_SAMPLE // 24
michael@0:             CONVOLVE_ONE_SAMPLE // 25
michael@0:             CONVOLVE_ONE_SAMPLE // 26
michael@0:             CONVOLVE_ONE_SAMPLE // 27
michael@0:             CONVOLVE_ONE_SAMPLE // 28
michael@0:             CONVOLVE_ONE_SAMPLE // 29
michael@0:             CONVOLVE_ONE_SAMPLE // 30
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 31
michael@0:             CONVOLVE_ONE_SAMPLE // 32
michael@0:             CONVOLVE_ONE_SAMPLE // 33
michael@0:             CONVOLVE_ONE_SAMPLE // 34
michael@0:             CONVOLVE_ONE_SAMPLE // 35
michael@0:             CONVOLVE_ONE_SAMPLE // 36
michael@0:             CONVOLVE_ONE_SAMPLE // 37
michael@0:             CONVOLVE_ONE_SAMPLE // 38
michael@0:             CONVOLVE_ONE_SAMPLE // 39
michael@0:             CONVOLVE_ONE_SAMPLE // 40
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 41
michael@0:             CONVOLVE_ONE_SAMPLE // 42
michael@0:             CONVOLVE_ONE_SAMPLE // 43
michael@0:             CONVOLVE_ONE_SAMPLE // 44
michael@0:             CONVOLVE_ONE_SAMPLE // 45
michael@0:             CONVOLVE_ONE_SAMPLE // 46
michael@0:             CONVOLVE_ONE_SAMPLE // 47
michael@0:             CONVOLVE_ONE_SAMPLE // 48
michael@0:             CONVOLVE_ONE_SAMPLE // 49
michael@0:             CONVOLVE_ONE_SAMPLE // 50
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 51
michael@0:             CONVOLVE_ONE_SAMPLE // 52
michael@0:             CONVOLVE_ONE_SAMPLE // 53
michael@0:             CONVOLVE_ONE_SAMPLE // 54
michael@0:             CONVOLVE_ONE_SAMPLE // 55
michael@0:             CONVOLVE_ONE_SAMPLE // 56
michael@0:             CONVOLVE_ONE_SAMPLE // 57
michael@0:             CONVOLVE_ONE_SAMPLE // 58
michael@0:             CONVOLVE_ONE_SAMPLE // 59
michael@0:             CONVOLVE_ONE_SAMPLE // 60
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 61
michael@0:             CONVOLVE_ONE_SAMPLE // 62
michael@0:             CONVOLVE_ONE_SAMPLE // 63
michael@0:             CONVOLVE_ONE_SAMPLE // 64
michael@0: 
michael@0:         } else if (kernelSize == 128) {
michael@0:             CONVOLVE_ONE_SAMPLE // 1
michael@0:             CONVOLVE_ONE_SAMPLE // 2
michael@0:             CONVOLVE_ONE_SAMPLE // 3
michael@0:             CONVOLVE_ONE_SAMPLE // 4
michael@0:             CONVOLVE_ONE_SAMPLE // 5
michael@0:             CONVOLVE_ONE_SAMPLE // 6
michael@0:             CONVOLVE_ONE_SAMPLE // 7
michael@0:             CONVOLVE_ONE_SAMPLE // 8
michael@0:             CONVOLVE_ONE_SAMPLE // 9
michael@0:             CONVOLVE_ONE_SAMPLE // 10
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 11
michael@0:             CONVOLVE_ONE_SAMPLE // 12
michael@0:             CONVOLVE_ONE_SAMPLE // 13
michael@0:             CONVOLVE_ONE_SAMPLE // 14
michael@0:             CONVOLVE_ONE_SAMPLE // 15
michael@0:             CONVOLVE_ONE_SAMPLE // 16
michael@0:             CONVOLVE_ONE_SAMPLE // 17
michael@0:             CONVOLVE_ONE_SAMPLE // 18
michael@0:             CONVOLVE_ONE_SAMPLE // 19
michael@0:             CONVOLVE_ONE_SAMPLE // 20
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 21
michael@0:             CONVOLVE_ONE_SAMPLE // 22
michael@0:             CONVOLVE_ONE_SAMPLE // 23
michael@0:             CONVOLVE_ONE_SAMPLE // 24
michael@0:             CONVOLVE_ONE_SAMPLE // 25
michael@0:             CONVOLVE_ONE_SAMPLE // 26
michael@0:             CONVOLVE_ONE_SAMPLE // 27
michael@0:             CONVOLVE_ONE_SAMPLE // 28
michael@0:             CONVOLVE_ONE_SAMPLE // 29
michael@0:             CONVOLVE_ONE_SAMPLE // 30
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 31
michael@0:             CONVOLVE_ONE_SAMPLE // 32
michael@0:             CONVOLVE_ONE_SAMPLE // 33
michael@0:             CONVOLVE_ONE_SAMPLE // 34
michael@0:             CONVOLVE_ONE_SAMPLE // 35
michael@0:             CONVOLVE_ONE_SAMPLE // 36
michael@0:             CONVOLVE_ONE_SAMPLE // 37
michael@0:             CONVOLVE_ONE_SAMPLE // 38
michael@0:             CONVOLVE_ONE_SAMPLE // 39
michael@0:             CONVOLVE_ONE_SAMPLE // 40
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 41
michael@0:             CONVOLVE_ONE_SAMPLE // 42
michael@0:             CONVOLVE_ONE_SAMPLE // 43
michael@0:             CONVOLVE_ONE_SAMPLE // 44
michael@0:             CONVOLVE_ONE_SAMPLE // 45
michael@0:             CONVOLVE_ONE_SAMPLE // 46
michael@0:             CONVOLVE_ONE_SAMPLE // 47
michael@0:             CONVOLVE_ONE_SAMPLE // 48
michael@0:             CONVOLVE_ONE_SAMPLE // 49
michael@0:             CONVOLVE_ONE_SAMPLE // 50
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 51
michael@0:             CONVOLVE_ONE_SAMPLE // 52
michael@0:             CONVOLVE_ONE_SAMPLE // 53
michael@0:             CONVOLVE_ONE_SAMPLE // 54
michael@0:             CONVOLVE_ONE_SAMPLE // 55
michael@0:             CONVOLVE_ONE_SAMPLE // 56
michael@0:             CONVOLVE_ONE_SAMPLE // 57
michael@0:             CONVOLVE_ONE_SAMPLE // 58
michael@0:             CONVOLVE_ONE_SAMPLE // 59
michael@0:             CONVOLVE_ONE_SAMPLE // 60
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 61
michael@0:             CONVOLVE_ONE_SAMPLE // 62
michael@0:             CONVOLVE_ONE_SAMPLE // 63
michael@0:             CONVOLVE_ONE_SAMPLE // 64
michael@0:             CONVOLVE_ONE_SAMPLE // 65
michael@0:             CONVOLVE_ONE_SAMPLE // 66
michael@0:             CONVOLVE_ONE_SAMPLE // 67
michael@0:             CONVOLVE_ONE_SAMPLE // 68
michael@0:             CONVOLVE_ONE_SAMPLE // 69
michael@0:             CONVOLVE_ONE_SAMPLE // 70
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 71
michael@0:             CONVOLVE_ONE_SAMPLE // 72
michael@0:             CONVOLVE_ONE_SAMPLE // 73
michael@0:             CONVOLVE_ONE_SAMPLE // 74
michael@0:             CONVOLVE_ONE_SAMPLE // 75
michael@0:             CONVOLVE_ONE_SAMPLE // 76
michael@0:             CONVOLVE_ONE_SAMPLE // 77
michael@0:             CONVOLVE_ONE_SAMPLE // 78
michael@0:             CONVOLVE_ONE_SAMPLE // 79
michael@0:             CONVOLVE_ONE_SAMPLE // 80
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 81
michael@0:             CONVOLVE_ONE_SAMPLE // 82
michael@0:             CONVOLVE_ONE_SAMPLE // 83
michael@0:             CONVOLVE_ONE_SAMPLE // 84
michael@0:             CONVOLVE_ONE_SAMPLE // 85
michael@0:             CONVOLVE_ONE_SAMPLE // 86
michael@0:             CONVOLVE_ONE_SAMPLE // 87
michael@0:             CONVOLVE_ONE_SAMPLE // 88
michael@0:             CONVOLVE_ONE_SAMPLE // 89
michael@0:             CONVOLVE_ONE_SAMPLE // 90
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 91
michael@0:             CONVOLVE_ONE_SAMPLE // 92
michael@0:             CONVOLVE_ONE_SAMPLE // 93
michael@0:             CONVOLVE_ONE_SAMPLE // 94
michael@0:             CONVOLVE_ONE_SAMPLE // 95
michael@0:             CONVOLVE_ONE_SAMPLE // 96
michael@0:             CONVOLVE_ONE_SAMPLE // 97
michael@0:             CONVOLVE_ONE_SAMPLE // 98
michael@0:             CONVOLVE_ONE_SAMPLE // 99
michael@0:             CONVOLVE_ONE_SAMPLE // 100
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 101
michael@0:             CONVOLVE_ONE_SAMPLE // 102
michael@0:             CONVOLVE_ONE_SAMPLE // 103
michael@0:             CONVOLVE_ONE_SAMPLE // 104
michael@0:             CONVOLVE_ONE_SAMPLE // 105
michael@0:             CONVOLVE_ONE_SAMPLE // 106
michael@0:             CONVOLVE_ONE_SAMPLE // 107
michael@0:             CONVOLVE_ONE_SAMPLE // 108
michael@0:             CONVOLVE_ONE_SAMPLE // 109
michael@0:             CONVOLVE_ONE_SAMPLE // 110
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 111
michael@0:             CONVOLVE_ONE_SAMPLE // 112
michael@0:             CONVOLVE_ONE_SAMPLE // 113
michael@0:             CONVOLVE_ONE_SAMPLE // 114
michael@0:             CONVOLVE_ONE_SAMPLE // 115
michael@0:             CONVOLVE_ONE_SAMPLE // 116
michael@0:             CONVOLVE_ONE_SAMPLE // 117
michael@0:             CONVOLVE_ONE_SAMPLE // 118
michael@0:             CONVOLVE_ONE_SAMPLE // 119
michael@0:             CONVOLVE_ONE_SAMPLE // 120
michael@0: 
michael@0:             CONVOLVE_ONE_SAMPLE // 121
michael@0:             CONVOLVE_ONE_SAMPLE // 122
michael@0:             CONVOLVE_ONE_SAMPLE // 123
michael@0:             CONVOLVE_ONE_SAMPLE // 124
michael@0:             CONVOLVE_ONE_SAMPLE // 125
michael@0:             CONVOLVE_ONE_SAMPLE // 126
michael@0:             CONVOLVE_ONE_SAMPLE // 127
michael@0:             CONVOLVE_ONE_SAMPLE // 128
michael@0:         } else {
michael@0:             while (j < kernelSize) {
michael@0:                 // Non-optimized using actual while loop.
michael@0:                 CONVOLVE_ONE_SAMPLE
michael@0:             }
michael@0:         }
michael@0:         destP[i++] = sum;
michael@0:     }
michael@0: 
michael@0:     // Copy 2nd half of input buffer to 1st half.
michael@0:     memcpy(m_buffer.Elements(), inputP, sizeof(float) * framesToProcess);
michael@0: }
michael@0: 
michael@0: void DirectConvolver::reset()
michael@0: {
michael@0:     PodZero(m_buffer.Elements(), m_buffer.Length());
michael@0: }
michael@0: 
michael@0: } // namespace WebCore