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
|
#include "config.h"
#include <assert.h>
#include "alMain.h"
#include "alu.h"
#include "alSource.h"
#include "alAuxEffectSlot.h"
static inline ALfloat point32(const ALfloat *restrict vals, ALuint UNUSED(frac))
{ return vals[0]; }
static inline ALfloat lerp32(const ALfloat *restrict vals, ALuint frac)
{ return lerp(vals[0], vals[1], frac * (1.0f/FRACTIONONE)); }
static inline ALfloat fir4_32(const ALfloat *restrict vals, ALuint frac)
{ return resample_fir4(vals[-1], vals[0], vals[1], vals[2], frac); }
static inline ALfloat fir8_32(const ALfloat *restrict vals, ALuint frac)
{ return resample_fir8(vals[-3], vals[-2], vals[-1], vals[0], vals[1], vals[2], vals[3], vals[4], frac); }
const ALfloat *Resample_copy32_C(const BsincState* UNUSED(state), const ALfloat *restrict src, ALuint UNUSED(frac),
ALuint UNUSED(increment), ALfloat *restrict dst, ALuint numsamples)
{
#if defined(HAVE_SSE) || defined(HAVE_NEON)
/* Avoid copying the source data if it's aligned like the destination. */
if((((intptr_t)src)&15) == (((intptr_t)dst)&15))
return src;
#endif
memcpy(dst, src, numsamples*sizeof(ALfloat));
return dst;
}
#define DECL_TEMPLATE(Sampler) \
const ALfloat *Resample_##Sampler##_C(const BsincState* UNUSED(state), \
const ALfloat *restrict src, ALuint frac, ALuint increment, \
ALfloat *restrict dst, ALuint numsamples) \
{ \
ALuint i; \
for(i = 0;i < numsamples;i++) \
{ \
dst[i] = Sampler(src, frac); \
\
frac += increment; \
src += frac>>FRACTIONBITS; \
frac &= FRACTIONMASK; \
} \
return dst; \
}
DECL_TEMPLATE(point32)
DECL_TEMPLATE(lerp32)
DECL_TEMPLATE(fir4_32)
DECL_TEMPLATE(fir8_32)
#undef DECL_TEMPLATE
const ALfloat *Resample_bsinc32_C(const BsincState *state, const ALfloat *restrict src,
ALuint frac, ALuint increment, ALfloat *restrict dst,
ALuint dstlen)
{
const ALfloat *fil, *scd, *phd, *spd;
const ALfloat sf = state->sf;
const ALuint m = state->m;
const ALint l = state->l;
ALuint j_f, pi, i;
ALfloat pf, r;
ALint j_s;
for(i = 0;i < dstlen;i++)
{
// Calculate the phase index and factor.
#define FRAC_PHASE_BITDIFF (FRACTIONBITS-BSINC_PHASE_BITS)
pi = frac >> FRAC_PHASE_BITDIFF;
pf = (frac & ((1<<FRAC_PHASE_BITDIFF)-1)) * (1.0f/(1<<FRAC_PHASE_BITDIFF));
#undef FRAC_PHASE_BITDIFF
fil = state->coeffs[pi].filter;
scd = state->coeffs[pi].scDelta;
phd = state->coeffs[pi].phDelta;
spd = state->coeffs[pi].spDelta;
// Apply the scale and phase interpolated filter.
r = 0.0f;
for(j_f = 0,j_s = l;j_f < m;j_f++,j_s++)
r += (fil[j_f] + sf*scd[j_f] + pf*(phd[j_f] + sf*spd[j_f])) *
src[j_s];
dst[i] = r;
frac += increment;
src += frac>>FRACTIONBITS;
frac &= FRACTIONMASK;
}
return dst;
}
void ALfilterState_processC(ALfilterState *filter, ALfloat *restrict dst, const ALfloat *restrict src, ALuint numsamples)
{
ALuint i;
if(numsamples > 1)
{
dst[0] = filter->b0 * src[0] +
filter->b1 * filter->x[0] +
filter->b2 * filter->x[1] -
filter->a1 * filter->y[0] -
filter->a2 * filter->y[1];
dst[1] = filter->b0 * src[1] +
filter->b1 * src[0] +
filter->b2 * filter->x[0] -
filter->a1 * dst[0] -
filter->a2 * filter->y[0];
for(i = 2;i < numsamples;i++)
dst[i] = filter->b0 * src[i] +
filter->b1 * src[i-1] +
filter->b2 * src[i-2] -
filter->a1 * dst[i-1] -
filter->a2 * dst[i-2];
filter->x[0] = src[i-1];
filter->x[1] = src[i-2];
filter->y[0] = dst[i-1];
filter->y[1] = dst[i-2];
}
else if(numsamples == 1)
{
dst[0] = filter->b0 * src[0] +
filter->b1 * filter->x[0] +
filter->b2 * filter->x[1] -
filter->a1 * filter->y[0] -
filter->a2 * filter->y[1];
filter->x[1] = filter->x[0];
filter->x[0] = src[0];
filter->y[1] = filter->y[0];
filter->y[0] = dst[0];
}
}
static inline void ApplyCoeffsStep(ALuint Offset, ALfloat (*restrict Values)[2],
const ALuint IrSize,
ALfloat (*restrict Coeffs)[2],
const ALfloat (*restrict CoeffStep)[2],
ALfloat left, ALfloat right)
{
ALuint c;
for(c = 0;c < IrSize;c++)
{
const ALuint off = (Offset+c)&HRIR_MASK;
Values[off][0] += Coeffs[c][0] * left;
Values[off][1] += Coeffs[c][1] * right;
Coeffs[c][0] += CoeffStep[c][0];
Coeffs[c][1] += CoeffStep[c][1];
}
}
static inline void ApplyCoeffs(ALuint Offset, ALfloat (*restrict Values)[2],
const ALuint IrSize,
ALfloat (*restrict Coeffs)[2],
ALfloat left, ALfloat right)
{
ALuint c;
for(c = 0;c < IrSize;c++)
{
const ALuint off = (Offset+c)&HRIR_MASK;
Values[off][0] += Coeffs[c][0] * left;
Values[off][1] += Coeffs[c][1] * right;
}
}
#define MixHrtf MixHrtf_C
#define MixDirectHrtf MixDirectHrtf_C
#include "mixer_inc.c"
#undef MixHrtf
void Mix_C(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)[BUFFERSIZE],
ALfloat *CurrentGains, const ALfloat *TargetGains, ALuint Counter, ALuint OutPos,
ALuint BufferSize)
{
ALfloat gain, delta, step;
ALuint c;
delta = (Counter > 0) ? 1.0f/(ALfloat)Counter : 0.0f;
for(c = 0;c < OutChans;c++)
{
ALuint pos = 0;
gain = CurrentGains[c];
step = (TargetGains[c] - gain) * delta;
if(fabsf(step) > FLT_EPSILON)
{
ALuint minsize = minu(BufferSize, Counter);
for(;pos < minsize;pos++)
{
OutBuffer[c][OutPos+pos] += data[pos]*gain;
gain += step;
}
if(pos == Counter)
gain = TargetGains[c];
CurrentGains[c] = gain;
}
if(!(fabsf(gain) > GAIN_SILENCE_THRESHOLD))
continue;
for(;pos < BufferSize;pos++)
OutBuffer[c][OutPos+pos] += data[pos]*gain;
}
}
/* Basically the inverse of the above. Rather than one input going to multiple
* outputs (each with its own gain), it's multiple inputs (each with its own
* gain) going to one output. This applies one row (vs one column) of a matrix
* transform. And as the matrices are more or less static once set up, no
* stepping is necessary.
*/
void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains, const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint InPos, ALuint BufferSize)
{
ALuint c, i;
for(c = 0;c < InChans;c++)
{
ALfloat gain = Gains[c];
if(!(fabsf(gain) > GAIN_SILENCE_THRESHOLD))
continue;
for(i = 0;i < BufferSize;i++)
OutBuffer[i] += data[c][InPos+i] * gain;
}
}
|
