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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
|
#include "config.h"
#include <cassert>
#include <cmath>
#include <limits>
#include "alnumeric.h"
#include "core/bsinc_defs.h"
#include "core/cubic_defs.h"
#include "defs.h"
#include "hrtfbase.h"
struct CTag;
struct PointTag;
struct LerpTag;
struct CubicTag;
struct BSincTag;
struct FastBSincTag;
namespace {
constexpr uint BsincPhaseDiffBits{MixerFracBits - BSincPhaseBits};
constexpr uint BsincPhaseDiffOne{1 << BsincPhaseDiffBits};
constexpr uint BsincPhaseDiffMask{BsincPhaseDiffOne - 1u};
constexpr uint CubicPhaseDiffBits{MixerFracBits - CubicPhaseBits};
constexpr uint CubicPhaseDiffOne{1 << CubicPhaseDiffBits};
constexpr uint CubicPhaseDiffMask{CubicPhaseDiffOne - 1u};
inline float do_point(const float *RESTRICT vals, const uint)
{ return vals[0]; }
inline float do_lerp(const float *RESTRICT vals, const uint frac)
{ return lerpf(vals[0], vals[1], static_cast<float>(frac)*(1.0f/MixerFracOne)); }
inline float do_cubic(const CubicState &istate, const float *RESTRICT vals, const uint frac)
{
/* Calculate the phase index and factor. */
const uint pi{frac >> CubicPhaseDiffBits};
const float pf{static_cast<float>(frac&CubicPhaseDiffMask) * (1.0f/CubicPhaseDiffOne)};
const float *RESTRICT fil{al::assume_aligned<16>(istate.filter[pi].mCoeffs.data())};
const float *RESTRICT phd{al::assume_aligned<16>(istate.filter[pi].mDeltas.data())};
/* Apply the phase interpolated filter. */
return (fil[0] + pf*phd[0])*vals[0] + (fil[1] + pf*phd[1])*vals[1]
+ (fil[2] + pf*phd[2])*vals[2] + (fil[3] + pf*phd[3])*vals[3];
}
inline float do_bsinc(const BsincState &istate, const float *RESTRICT vals, const uint frac)
{
const size_t m{istate.m};
ASSUME(m > 0);
/* Calculate the phase index and factor. */
const uint pi{frac >> BsincPhaseDiffBits};
const float pf{static_cast<float>(frac&BsincPhaseDiffMask) * (1.0f/BsincPhaseDiffOne)};
const float *RESTRICT fil{istate.filter + m*pi*2_uz};
const float *RESTRICT phd{fil + m};
const float *RESTRICT scd{fil + BSincPhaseCount*2_uz*m};
const float *RESTRICT 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.sf*scd[j_f] + pf*(phd[j_f] + istate.sf*spd[j_f])) * vals[j_f];
return r;
}
inline float do_fastbsinc(const BsincState &istate, const float *RESTRICT vals, const uint frac)
{
const size_t m{istate.m};
ASSUME(m > 0);
/* Calculate the phase index and factor. */
const uint pi{frac >> BsincPhaseDiffBits};
const float pf{static_cast<float>(frac&BsincPhaseDiffMask) * (1.0f/BsincPhaseDiffOne)};
const float *RESTRICT fil{istate.filter + m*pi*2_uz};
const float *RESTRICT 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 float*RESTRICT, const uint);
template<SamplerT Sampler>
void DoResample(const float *RESTRICT src, uint frac, const uint increment,
const al::span<float> dst)
{
ASSUME(frac < MixerFracOne);
for(float &out : dst)
{
out = Sampler(src, frac);
frac += increment;
src += frac>>MixerFracBits;
frac &= MixerFracMask;
}
}
template<typename T, typename U>
void DoResample(T sampler, const U istate, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{
ASSUME(frac < MixerFracOne);
for(float &out : dst)
{
out = sampler(istate, src, frac);
frac += increment;
src += frac>>MixerFracBits;
frac &= MixerFracMask;
}
}
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;
}
}
force_inline void MixLine(const al::span<const float> InSamples, float *RESTRICT dst,
float &CurrentGain, const float TargetGain, const float delta, const size_t min_len,
size_t Counter)
{
float gain{CurrentGain};
const float step{(TargetGain-gain) * delta};
size_t pos{0};
if(!(std::abs(step) > std::numeric_limits<float>::epsilon()))
gain = TargetGain;
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 = TargetGain;
else
gain += step*step_count;
}
CurrentGain = gain;
if(!(std::abs(gain) > GainSilenceThreshold))
return;
for(;pos != InSamples.size();++pos)
dst[pos] += InSamples[pos] * gain;
}
} // namespace
template<>
void Resample_<PointTag,CTag>(const InterpState*, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{ DoResample<do_point>(src, frac, increment, dst); }
template<>
void Resample_<LerpTag,CTag>(const InterpState*, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{ DoResample<do_lerp>(src, frac, increment, dst); }
template<>
void Resample_<CubicTag,CTag>(const InterpState *state, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{ DoResample(do_cubic, std::get<CubicState>(*state), src-1, frac, increment, dst); }
template<>
void Resample_<BSincTag,CTag>(const InterpState *state, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{
const auto istate = std::get<BsincState>(*state);
DoResample(do_bsinc, istate, src-istate.l, frac, increment, dst);
}
template<>
void Resample_<FastBSincTag,CTag>(const InterpState *state, const float *RESTRICT src, uint frac,
const uint increment, const al::span<float> dst)
{
const auto istate = std::get<BsincState>(*state);
DoResample(do_fastbsinc, istate, src-istate.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)
MixLine(InSamples, al::assume_aligned<16>(output.data()+OutPos), *CurrentGains++,
*TargetGains++, delta, min_len, Counter);
}
template<>
void Mix_<CTag>(const al::span<const float> InSamples, float *OutBuffer, float &CurrentGain,
const float TargetGain, const size_t Counter)
{
const float delta{(Counter > 0) ? 1.0f / static_cast<float>(Counter) : 0.0f};
const auto min_len = minz(Counter, InSamples.size());
MixLine(InSamples, al::assume_aligned<16>(OutBuffer), CurrentGain,
TargetGain, delta, min_len, Counter);
}
|