• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* ----------------------------------------------------------------------
2  * Project:      CMSIS DSP Library
3  * Title:        arm_var_q15.c
4  * Description:  Variance of an array of Q15 type
5  *
6  * $Date:        23 April 2021
7  * $Revision:    V1.9.0
8  *
9  * Target Processor: Cortex-M and Cortex-A cores
10  * -------------------------------------------------------------------- */
11 /*
12  * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
13  *
14  * SPDX-License-Identifier: Apache-2.0
15  *
16  * Licensed under the Apache License, Version 2.0 (the License); you may
17  * not use this file except in compliance with the License.
18  * You may obtain a copy of the License at
19  *
20  * www.apache.org/licenses/LICENSE-2.0
21  *
22  * Unless required by applicable law or agreed to in writing, software
23  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
24  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
25  * See the License for the specific language governing permissions and
26  * limitations under the License.
27  */
28 
29 #include "dsp/statistics_functions.h"
30 
31 /**
32   @ingroup groupStats
33  */
34 
35 /**
36   @addtogroup variance
37   @{
38  */
39 
40 /**
41   @brief         Variance of the elements of a Q15 vector.
42   @param[in]     pSrc       points to the input vector
43   @param[in]     blockSize  number of samples in input vector
44   @param[out]    pResult    variance value returned here
45   @return        none
46 
47   @par           Scaling and Overflow Behavior
48                    The function is implemented using a 64-bit internal accumulator.
49                    The input is represented in 1.15 format.
50                    Intermediate multiplication yields a 2.30 format, and this
51                    result is added without saturation to a 64-bit accumulator in 34.30 format.
52                    With 33 guard bits in the accumulator, there is no risk of overflow, and the
53                    full precision of the intermediate multiplication is preserved.
54                    Finally, the 34.30 result is truncated to 34.15 format by discarding the lower
55                    15 bits, and then saturated to yield a result in 1.15 format.
56  */
57 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_var_q15(const q15_t * pSrc,uint32_t blockSize,q15_t * pResult)58 void arm_var_q15(
59   const q15_t * pSrc,
60         uint32_t blockSize,
61         q15_t * pResult)
62 {
63     uint32_t  blkCnt;     /* loop counters */
64     q15x8_t         vecSrc;
65     q63_t           sumOfSquares = 0LL;
66     q63_t           meanOfSquares, squareOfMean;        /* square of mean and mean of square */
67     q63_t           sum = 0LL;
68     q15_t in;
69 
70     if (blockSize <= 1U) {
71         *pResult = 0;
72         return;
73     }
74 
75 
76     blkCnt = blockSize >> 3;
77     while (blkCnt > 0U)
78     {
79         vecSrc = vldrhq_s16(pSrc);
80         /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
81         /* Compute Sum of squares of the input samples
82          * and then store the result in a temporary variable, sumOfSquares. */
83 
84         sumOfSquares = vmlaldavaq_s16(sumOfSquares, vecSrc, vecSrc);
85         sum = vaddvaq_s16(sum, vecSrc);
86 
87         blkCnt --;
88         pSrc += 8;
89     }
90 
91     /* Tail */
92     blkCnt = blockSize & 7;
93     while (blkCnt > 0U)
94     {
95       /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
96       /* C = A[0] + A[1] + ... + A[blockSize-1] */
97 
98       in = *pSrc++;
99       /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */
100 #if defined (ARM_MATH_DSP)
101       sumOfSquares = __SMLALD(in, in, sumOfSquares);
102 #else
103       sumOfSquares += (in * in);
104 #endif /* #if defined (ARM_MATH_DSP) */
105       /* Compute sum and store result in a temporary variable, sum. */
106       sum += in;
107 
108       /* Decrement loop counter */
109       blkCnt--;
110     }
111 
112     /* Compute Mean of squares of the input samples
113      * and then store the result in a temporary variable, meanOfSquares. */
114     meanOfSquares = arm_div_int64_to_int32(sumOfSquares, (blockSize - 1U));
115 
116     /* Compute square of mean */
117     squareOfMean = arm_div_int64_to_int32((q63_t)sum * sum, (q31_t)(blockSize * (blockSize - 1U)));
118 
119     /* mean of the squares minus the square of the mean. */
120     *pResult = (meanOfSquares - squareOfMean) >> 15;
121 }
122 #else
arm_var_q15(const q15_t * pSrc,uint32_t blockSize,q15_t * pResult)123 void arm_var_q15(
124   const q15_t * pSrc,
125         uint32_t blockSize,
126         q15_t * pResult)
127 {
128         uint32_t blkCnt;                               /* Loop counter */
129         q31_t sum = 0;                                 /* Accumulator */
130         q31_t meanOfSquares, squareOfMean;             /* Square of mean and mean of square */
131         q63_t sumOfSquares = 0;                        /* Sum of squares */
132         q15_t in;                                      /* Temporary variable to store input value */
133 
134 #if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP)
135         q31_t in32;                                    /* Temporary variable to store input value */
136 #endif
137 
138   if (blockSize <= 1U)
139   {
140     *pResult = 0;
141     return;
142   }
143 
144 #if defined (ARM_MATH_LOOPUNROLL)
145 
146   /* Loop unrolling: Compute 4 outputs at a time */
147   blkCnt = blockSize >> 2U;
148 
149   while (blkCnt > 0U)
150   {
151     /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
152     /* C = A[0] + A[1] + ... + A[blockSize-1] */
153 
154     /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */
155     /* Compute sum and store result in a temporary variable, sum. */
156 #if defined (ARM_MATH_DSP)
157     in32 = read_q15x2_ia (&pSrc);
158     sumOfSquares = __SMLALD(in32, in32, sumOfSquares);
159     sum += ((in32 << 16U) >> 16U);
160     sum +=  (in32 >> 16U);
161 
162     in32 = read_q15x2_ia (&pSrc);
163     sumOfSquares = __SMLALD(in32, in32, sumOfSquares);
164     sum += ((in32 << 16U) >> 16U);
165     sum +=  (in32 >> 16U);
166 #else
167     in = *pSrc++;
168     sumOfSquares += (in * in);
169     sum += in;
170 
171     in = *pSrc++;
172     sumOfSquares += (in * in);
173     sum += in;
174 
175     in = *pSrc++;
176     sumOfSquares += (in * in);
177     sum += in;
178 
179     in = *pSrc++;
180     sumOfSquares += (in * in);
181     sum += in;
182 #endif /* #if defined (ARM_MATH_DSP) */
183 
184     /* Decrement loop counter */
185     blkCnt--;
186   }
187 
188   /* Loop unrolling: Compute remaining outputs */
189   blkCnt = blockSize % 0x4U;
190 
191 #else
192 
193   /* Initialize blkCnt with number of samples */
194   blkCnt = blockSize;
195 
196 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
197 
198   while (blkCnt > 0U)
199   {
200     /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
201     /* C = A[0] + A[1] + ... + A[blockSize-1] */
202 
203     in = *pSrc++;
204     /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */
205 #if defined (ARM_MATH_DSP)
206     sumOfSquares = __SMLALD(in, in, sumOfSquares);
207 #else
208     sumOfSquares += (in * in);
209 #endif /* #if defined (ARM_MATH_DSP) */
210     /* Compute sum and store result in a temporary variable, sum. */
211     sum += in;
212 
213     /* Decrement loop counter */
214     blkCnt--;
215   }
216 
217   /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */
218   meanOfSquares = (q31_t) (sumOfSquares / (q63_t)(blockSize - 1U));
219 
220   /* Compute square of mean */
221   squareOfMean = (q31_t) ((q63_t) sum * sum / (q63_t)(blockSize * (blockSize - 1U)));
222 
223   /* mean of squares minus the square of mean. */
224   *pResult = (meanOfSquares - squareOfMean) >> 15U;
225 }
226 #endif /* defined(ARM_MATH_MVEI) */
227 
228 /**
229   @} end of variance group
230  */
231