1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
|
#include "config.h"
#include <cassert>
#include <cmath>
#include <limits>
#include "alnumeric.h"
#include "core/bsinc_tables.h"
#include "defs.h"
#include "hrtfbase.h"
struct CTag;
struct CopyTag;
struct PointTag;
struct LerpTag;
struct CubicTag;
struct BSincTag;
struct FastBSincTag;
namespace {
constexpr uint FracPhaseBitDiff{MixerFracBits - BSincPhaseBits};
constexpr uint FracPhaseDiffOne{1 << FracPhaseBitDiff};
inline float do_point(const InterpState&, const float *RESTRICT vals, const uint)
{ return vals[0]; }
inline float do_lerp(const InterpState&, const float *RESTRICT vals, const uint frac)
{ return lerp(vals[0], vals[1], static_cast<float>(frac)*(1.0f/MixerFracOne)); }
inline float do_cubic(const InterpState&, const float *RESTRICT vals, const uint frac)
{ return cubic(vals[0], vals[1], vals[2], vals[3], static_cast<float>(frac)*(1.0f/MixerFracOne)); }
inline float do_bsinc(const InterpState &istate, const float *RESTRICT vals, const uint frac)
{
const size_t m{istate.bsinc.m};
// Calculate the phase index and factor.
const uint pi{frac >> FracPhaseBitDiff};
const float pf{static_cast<float>(frac & (FracPhaseDiffOne-1)) * (1.0f/FracPhaseDiffOne)};
const float *fil{istate.bsinc.filter + m*pi*4};
const float *phd{fil + m};
const float *scd{phd + m};
const float *spd{scd + m};
// Apply the scale and phase interpolated filter.
float r{0.0f};
for(size_t j_f{0};j_f < m;j_f++)
r += (fil[j_f] + istate.bsinc.sf*scd[j_f] + pf*(phd[j_f] + istate.bsinc.sf*spd[j_f])) * vals[j_f];
return r;
}
inline float do_fastbsinc(const InterpState &istate, const float *RESTRICT vals, const uint frac)
{
const size_t m{istate.bsinc.m};
// Calculate the phase index and factor.
const uint pi{frac >> FracPhaseBitDiff};
const float pf{static_cast<float>(frac & (FracPhaseDiffOne-1)) * (1.0f/FracPhaseDiffOne)};
const float *fil{istate.bsinc.filter + m*pi*4};
const float *phd{fil + m};
// Apply the phase interpolated filter.
float r{0.0f};
for(size_t j_f{0};j_f < m;j_f++)
r += (fil[j_f] + pf*phd[j_f]) * vals[j_f];
return r;
}
using SamplerT = float(&)(const InterpState&, const float*RESTRICT, const uint);
template<SamplerT Sampler>
float *DoResample(const InterpState *state, float *RESTRICT src, uint frac, uint increment,
const al::span<float> dst)
{
const InterpState istate{*state};
for(float &out : dst)
{
out = Sampler(istate, src, frac);
frac += increment;
src += frac>>MixerFracBits;
frac &= MixerFracMask;
}
return dst.data();
}
inline void ApplyCoeffs(float2 *RESTRICT Values, const size_t IrSize, const ConstHrirSpan Coeffs,
const float left, const float right)
{
ASSUME(IrSize >= MinIrLength);
for(size_t c{0};c < IrSize;++c)
{
Values[c][0] += Coeffs[c][0] * left;
Values[c][1] += Coeffs[c][1] * right;
}
}
} // namespace
template<>
float *Resample_<CopyTag,CTag>(const InterpState*, float *RESTRICT src, uint, uint,
const al::span<float> dst)
{
#if defined(HAVE_SSE) || defined(HAVE_NEON)
/* Avoid copying the source data if it's aligned like the destination. */
if((reinterpret_cast<intptr_t>(src)&15) == (reinterpret_cast<intptr_t>(dst.data())&15))
return src;
#endif
std::copy_n(src, dst.size(), dst.begin());
return dst.data();
}
template<>
float *Resample_<PointTag,CTag>(const InterpState *state, float *RESTRICT src, uint frac,
uint increment, const al::span<float> dst)
{ return DoResample<do_point>(state, src, frac, increment, dst); }
template<>
float *Resample_<LerpTag,CTag>(const InterpState *state, float *RESTRICT src, uint frac,
uint increment, const al::span<float> dst)
{ return DoResample<do_lerp>(state, src, frac, increment, dst); }
template<>
float *Resample_<CubicTag,CTag>(const InterpState *state, float *RESTRICT src, uint frac,
uint increment, const al::span<float> dst)
{ return DoResample<do_cubic>(state, src-1, frac, increment, dst); }
template<>
float *Resample_<BSincTag,CTag>(const InterpState *state, float *RESTRICT src, uint frac,
uint increment, const al::span<float> dst)
{ return DoResample<do_bsinc>(state, src-state->bsinc.l, frac, increment, dst); }
template<>
float *Resample_<FastBSincTag,CTag>(const InterpState *state, float *RESTRICT src, uint frac,
uint increment, const al::span<float> dst)
{ return DoResample<do_fastbsinc>(state, src-state->bsinc.l, frac, increment, dst); }
template<>
void MixHrtf_<CTag>(const float *InSamples, float2 *AccumSamples, const uint IrSize,
const MixHrtfFilter *hrtfparams, const size_t BufferSize)
{ MixHrtfBase<ApplyCoeffs>(InSamples, AccumSamples, IrSize, hrtfparams, BufferSize); }
template<>
void MixHrtfBlend_<CTag>(const float *InSamples, float2 *AccumSamples, const uint IrSize,
const HrtfFilter *oldparams, const MixHrtfFilter *newparams, const size_t BufferSize)
{
MixHrtfBlendBase<ApplyCoeffs>(InSamples, AccumSamples, IrSize, oldparams, newparams,
BufferSize);
}
template<>
void MixDirectHrtf_<CTag>(const FloatBufferSpan LeftOut, const FloatBufferSpan RightOut,
const al::span<const FloatBufferLine> InSamples, float2 *AccumSamples,
float *TempBuf, HrtfChannelState *ChanState, const size_t IrSize, const size_t BufferSize)
{
MixDirectHrtfBase<ApplyCoeffs>(LeftOut, RightOut, InSamples, AccumSamples, TempBuf, ChanState,
IrSize, BufferSize);
}
template<>
void Mix_<CTag>(const al::span<const float> InSamples, const al::span<FloatBufferLine> OutBuffer,
float *CurrentGains, const float *TargetGains, const size_t Counter, const size_t OutPos)
{
const float delta{(Counter > 0) ? 1.0f / static_cast<float>(Counter) : 0.0f};
const auto min_len = minz(Counter, InSamples.size());
for(FloatBufferLine &output : OutBuffer)
{
float *RESTRICT dst{al::assume_aligned<16>(output.data()+OutPos)};
float gain{*CurrentGains};
const float step{(*TargetGains-gain) * delta};
size_t pos{0};
if(!(std::abs(step) > std::numeric_limits<float>::epsilon()))
gain = *TargetGains;
else
{
float step_count{0.0f};
for(;pos != min_len;++pos)
{
dst[pos] += InSamples[pos] * (gain + step*step_count);
step_count += 1.0f;
}
if(pos == Counter)
gain = *TargetGains;
else
gain += step*step_count;
}
*CurrentGains = gain;
++CurrentGains;
++TargetGains;
if(!(std::abs(gain) > GainSilenceThreshold))
continue;
for(;pos != InSamples.size();++pos)
dst[pos] += InSamples[pos] * gain;
}
}
|
