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
265
266
267
268
269
270
271
272
273
274
275
276
|
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_absmin_f16.c
* Description: Minimum value of absolute values of a floating-point vector
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* 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/statistics_functions_f16.h"
#if defined(ARM_FLOAT16_SUPPORTED)
#if (defined(ARM_MATH_NEON) || defined(ARM_MATH_MVEF)) && !defined(ARM_MATH_AUTOVECTORIZE)
#include <limits.h>
#endif
/**
@ingroup groupStats
*/
/**
@addtogroup AbsMin
@{
*/
/**
@brief Minimum value of absolute values of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
@param[out] pIndex index of minimum value returned here
@return none
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
void arm_absmin_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex)
{
uint16_t blkCnt; /* loop counters */
f16x8_t vecSrc;
float16_t const *pSrcVec;
f16x8_t curExtremValVec = vdupq_n_f16(F16_ABSMAX);
float16_t minValue = F16_ABSMAX;
uint16_t idx = blockSize;
uint16x8_t indexVec;
uint16x8_t curExtremIdxVec;
mve_pred16_t p0;
indexVec = vidupq_u16((uint32_t)0, 1);
curExtremIdxVec = vdupq_n_u16(0);
pSrcVec = (float16_t const *) pSrc;
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
vecSrc = vldrhq_f16(pSrcVec);
pSrcVec += 8;
vecSrc = vabsq(vecSrc);
/*
* Get current max per lane and current index per lane
* when a max is selected
*/
p0 = vcmpleq(vecSrc, curExtremValVec);
curExtremValVec = vpselq(vecSrc, curExtremValVec, p0);
curExtremIdxVec = vpselq(indexVec, curExtremIdxVec, p0);
indexVec = indexVec + 8;
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/*
* tail
* (will be merged thru tail predication)
*/
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
p0 = vctp16q(blkCnt);
vecSrc = vldrhq_f16(pSrcVec);
pSrcVec += 8;
vecSrc = vabsq(vecSrc);
/*
* Get current max per lane and current index per lane
* when a max is selected
*/
p0 = vcmpleq_m(vecSrc, curExtremValVec, p0);
curExtremValVec = vpselq(vecSrc, curExtremValVec, p0);
curExtremIdxVec = vpselq(indexVec, curExtremIdxVec, p0);
}
/*
* Get min value across the vector
*/
minValue = vminnmvq(minValue, curExtremValVec);
/*
* set index for lower values to max possible index
*/
p0 = vcmpleq(curExtremValVec, minValue);
indexVec = vpselq(curExtremIdxVec, vdupq_n_u16(blockSize), p0);
/*
* Get min index which is thus for a max value
*/
idx = vminvq(idx, indexVec);
/*
* Save result
*/
*pIndex = idx;
*pResult = minValue;
}
#else
#if defined(ARM_MATH_LOOPUNROLL)
void arm_absmin_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex)
{
float16_t cur_absmin, out; /* Temporary variables to store the output value. */\
uint32_t blkCnt, outIndex; /* Loop counter */ \
uint32_t index; /* index of maximum value */ \
\
/* Initialize index value to zero. */ \
outIndex = 0U; \
/* Load first input value that act as reference value for comparision */ \
out = *pSrc++; \
out = ((_Float16)out > 0.0f16) ? out : -(_Float16)out; \
/* Initialize index of extrema value. */ \
index = 0U; \
\
/* Loop unrolling: Compute 4 outputs at a time */ \
blkCnt = (blockSize - 1U) >> 2U; \
\
while (blkCnt > 0U) \
{ \
/* Initialize cur_absmin to next consecutive values one by one */ \
cur_absmin = *pSrc++; \
cur_absmin = ((_Float16)cur_absmin > 0.0f16) ? cur_absmin : -(_Float16)cur_absmin; \
/* compare for the extrema value */ \
if ((_Float16)cur_absmin < (_Float16)out) \
{ \
/* Update the extrema value and it's index */ \
out = cur_absmin; \
outIndex = index + 1U; \
} \
\
cur_absmin = *pSrc++; \
cur_absmin = ((_Float16)cur_absmin > 0.0f16) ? cur_absmin : -(_Float16)cur_absmin; \
if ((_Float16)cur_absmin < (_Float16)out) \
{ \
out = cur_absmin; \
outIndex = index + 2U; \
} \
\
cur_absmin = *pSrc++; \
cur_absmin = ((_Float16)cur_absmin > 0.0f16) ? cur_absmin : -(_Float16)cur_absmin; \
if ((_Float16)cur_absmin < (_Float16)out) \
{ \
out = cur_absmin; \
outIndex = index + 3U; \
} \
\
cur_absmin = *pSrc++; \
cur_absmin = ((_Float16)cur_absmin > 0.0f16) ? cur_absmin : -(_Float16)cur_absmin; \
if ((_Float16)cur_absmin < (_Float16)out) \
{ \
out = cur_absmin; \
outIndex = index + 4U; \
} \
\
index += 4U; \
\
/* Decrement loop counter */ \
blkCnt--; \
} \
\
/* Loop unrolling: Compute remaining outputs */ \
blkCnt = (blockSize - 1U) % 4U; \
\
\
while (blkCnt > 0U) \
{ \
cur_absmin = *pSrc++; \
cur_absmin = ((_Float16)cur_absmin > 0.0f16) ? cur_absmin : -(_Float16)cur_absmin; \
if ((_Float16)cur_absmin < (_Float16)out) \
{ \
out = cur_absmin; \
outIndex = blockSize - blkCnt; \
} \
\
/* Decrement loop counter */ \
blkCnt--; \
} \
\
/* Store the extrema value and it's index into destination pointers */ \
*pResult = out; \
*pIndex = outIndex;
}
#else
void arm_absmin_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex)
{
float16_t minVal, out; /* Temporary variables to store the output value. */
uint32_t blkCnt, outIndex; /* Loop counter */
/* Initialise index value to zero. */
outIndex = 0U;
/* Load first input value that act as reference value for comparision */
out = (_Float16)fabsf((float32_t)*pSrc++);
/* Initialize blkCnt with number of samples */
blkCnt = (blockSize - 1U);
while (blkCnt > 0U)
{
/* Initialize minVal to the next consecutive values one by one */
minVal = (_Float16)fabsf((float32_t)*pSrc++);
/* compare for the minimum value */
if ((_Float16)out > (_Float16)minVal)
{
/* Update the minimum value and it's index */
out = minVal;
outIndex = blockSize - blkCnt;
}
/* Decrement loop counter */
blkCnt--;
}
/* Store the minimum value and it's index into destination pointers */
*pResult = out;
*pIndex = outIndex;
}
#endif /* defined(ARM_MATH_LOOPUNROLL) */
#endif /* defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of AbsMin group
*/
#endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
|