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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
|
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_vlog_q15
* Description: Q15 vector log
*
* $Date: 19 July 2021
* $Revision: V1.10.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/fast_math_functions.h"
#define LOG_Q15_ACCURACY 15
/* Bit to represent the normalization factor
It is Ceiling[Log2[LOG_Q15_ACCURACY]] of the previous value.
The Log2 algorithm is assuming that the value x is
1 <= x < 2.
But input value could be as small a 2^-LOG_Q15_ACCURACY
which would give an integer part of -15.
*/
#define LOG_Q15_INTEGER_PART 4
/* 2.0 in q14 */
#define LOQ_Q15_THRESHOLD (1u << LOG_Q15_ACCURACY)
/* HALF */
#define LOQ_Q15_Q16_HALF LOQ_Q15_THRESHOLD
#define LOQ_Q15_Q14_HALF (LOQ_Q15_Q16_HALF >> 2)
/* 1.0 / Log2[Exp[1]] in q15 */
#define LOG_Q15_INVLOG2EXP 0x58b9u
/* Clay Turner algorithm */
static uint16_t arm_scalar_log_q15(uint16_t src)
{
int i;
int16_t c = __CLZ(src)-16;
int16_t normalization=0;
/* 0.5 in q11 */
uint16_t inc = LOQ_Q15_Q16_HALF >> (LOG_Q15_INTEGER_PART + 1);
/* Will compute y = log2(x) for 1 <= x < 2.0 */
uint16_t x;
/* q11 */
uint16_t y=0;
/* q11 */
int16_t tmp;
/* Normalize and convert to q14 format */
x = src;
if ((c-1) < 0)
{
x = x >> (1-c);
}
else
{
x = x << (c-1);
}
normalization = c;
/* Compute the Log2. Result is in q11 instead of q16
because we know 0 <= y < 1.0 but
we want a result allowing to do a
product on int16 rather than having to go
through int32
*/
for(i = 0; i < LOG_Q15_ACCURACY ; i++)
{
x = (((int32_t)x*x)) >> (LOG_Q15_ACCURACY - 1);
if (x >= LOQ_Q15_THRESHOLD)
{
y += inc ;
x = x >> 1;
}
inc = inc >> 1;
}
/*
Convert the Log2 to Log and apply normalization.
We compute (y - normalisation) * (1 / Log2[e]).
*/
/* q11 */
//tmp = y - ((int32_t)normalization << (LOG_Q15_ACCURACY + 1));
tmp = (int16_t)y - (normalization << (LOG_Q15_ACCURACY - LOG_Q15_INTEGER_PART));
/* q4.11 */
y = ((int32_t)tmp * LOG_Q15_INVLOG2EXP) >> 15;
return(y);
}
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
q15x8_t vlogq_q15(q15x8_t src)
{
int i;
int16x8_t c = vclzq_s16(src);
int16x8_t normalization = c;
/* 0.5 in q11 */
uint16_t inc = LOQ_Q15_Q16_HALF >> (LOG_Q15_INTEGER_PART + 1);
/* Will compute y = log2(x) for 1 <= x < 2.0 */
uint16x8_t x;
/* q11 */
uint16x8_t y = vdupq_n_u16(0);
/* q11 */
int16x8_t vtmp;
mve_pred16_t p;
/* Normalize and convert to q14 format */
vtmp = vsubq_n_s16(c,1);
x = vshlq_u16((uint16x8_t)src,vtmp);
/* Compute the Log2. Result is in q11 instead of q16
because we know 0 <= y < 1.0 but
we want a result allowing to do a
product on int16 rather than having to go
through int32
*/
for(i = 0; i < LOG_Q15_ACCURACY ; i++)
{
x = vmulhq_u16(x,x);
x = vshlq_n_u16(x,2);
p = vcmphiq_u16(x,vdupq_n_u16(LOQ_Q15_THRESHOLD));
y = vaddq_m_n_u16(y, y,inc,p);
x = vshrq_m_n_u16(x,x,1,p);
inc = inc >> 1;
}
/*
Convert the Log2 to Log and apply normalization.
We compute (y - normalisation) * (1 / Log2[e]).
*/
/* q11 */
// tmp = (int16_t)y - (normalization << (LOG_Q15_ACCURACY - LOG_Q15_INTEGER_PART));
vtmp = vshlq_n_s16(normalization,LOG_Q15_ACCURACY - LOG_Q15_INTEGER_PART);
vtmp = vsubq_s16((int16x8_t)y,vtmp);
/* q4.11 */
// y = ((int32_t)tmp * LOG_Q15_INVLOG2EXP) >> 15;
vtmp = vqdmulhq_n_s16(vtmp,LOG_Q15_INVLOG2EXP);
return(vtmp);
}
#endif
/**
@ingroup groupFastMath
*/
/**
@addtogroup vlog
@{
*/
/**
@brief q15 vector of log values.
@param[in] pSrc points to the input vector in q15
@param[out] pDst points to the output vector in q4.11
@param[in] blockSize number of samples in each vector
@return none
*/
void arm_vlog_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
q15x8_t src;
q15x8_t dst;
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
src = vld1q(pSrc);
dst = vlogq_q15(src);
vst1q(pDst, dst);
pSrc += 8;
pDst += 8;
/* Decrement loop counter */
blkCnt--;
}
blkCnt = blockSize & 7;
#else
blkCnt = blockSize;
#endif
while (blkCnt > 0U)
{
*pDst++ = arm_scalar_log_q15(*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
}
/**
@} end of vlog group
*/
|
