The CMSIS DSP 5 library

Dependents:   Nucleo-Heart-Rate ejercicioVrms2 PROYECTOFINAL ejercicioVrms ... more

Committer:
xorjoep
Date:
Thu Jun 21 11:56:27 2018 +0000
Revision:
3:4098b9d3d571
Parent:
1:24714b45cd1b
headers is a folder not a library

Who changed what in which revision?

UserRevisionLine numberNew contents of line
xorjoep 1:24714b45cd1b 1 /* ----------------------------------------------------------------------
xorjoep 1:24714b45cd1b 2 * Project: CMSIS DSP Library
xorjoep 1:24714b45cd1b 3 * Title: arm_std_q15.c
xorjoep 1:24714b45cd1b 4 * Description: Standard deviation of an array of Q15 vector
xorjoep 1:24714b45cd1b 5 *
xorjoep 1:24714b45cd1b 6 * $Date: 27. January 2017
xorjoep 1:24714b45cd1b 7 * $Revision: V.1.5.1
xorjoep 1:24714b45cd1b 8 *
xorjoep 1:24714b45cd1b 9 * Target Processor: Cortex-M cores
xorjoep 1:24714b45cd1b 10 * -------------------------------------------------------------------- */
xorjoep 1:24714b45cd1b 11 /*
xorjoep 1:24714b45cd1b 12 * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
xorjoep 1:24714b45cd1b 13 *
xorjoep 1:24714b45cd1b 14 * SPDX-License-Identifier: Apache-2.0
xorjoep 1:24714b45cd1b 15 *
xorjoep 1:24714b45cd1b 16 * Licensed under the Apache License, Version 2.0 (the License); you may
xorjoep 1:24714b45cd1b 17 * not use this file except in compliance with the License.
xorjoep 1:24714b45cd1b 18 * You may obtain a copy of the License at
xorjoep 1:24714b45cd1b 19 *
xorjoep 1:24714b45cd1b 20 * www.apache.org/licenses/LICENSE-2.0
xorjoep 1:24714b45cd1b 21 *
xorjoep 1:24714b45cd1b 22 * Unless required by applicable law or agreed to in writing, software
xorjoep 1:24714b45cd1b 23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
xorjoep 1:24714b45cd1b 24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
xorjoep 1:24714b45cd1b 25 * See the License for the specific language governing permissions and
xorjoep 1:24714b45cd1b 26 * limitations under the License.
xorjoep 1:24714b45cd1b 27 */
xorjoep 1:24714b45cd1b 28
xorjoep 1:24714b45cd1b 29 #include "arm_math.h"
xorjoep 1:24714b45cd1b 30
xorjoep 1:24714b45cd1b 31 /**
xorjoep 1:24714b45cd1b 32 * @ingroup groupStats
xorjoep 1:24714b45cd1b 33 */
xorjoep 1:24714b45cd1b 34
xorjoep 1:24714b45cd1b 35 /**
xorjoep 1:24714b45cd1b 36 * @addtogroup STD
xorjoep 1:24714b45cd1b 37 * @{
xorjoep 1:24714b45cd1b 38 */
xorjoep 1:24714b45cd1b 39
xorjoep 1:24714b45cd1b 40 /**
xorjoep 1:24714b45cd1b 41 * @brief Standard deviation of the elements of a Q15 vector.
xorjoep 1:24714b45cd1b 42 * @param[in] *pSrc points to the input vector
xorjoep 1:24714b45cd1b 43 * @param[in] blockSize length of the input vector
xorjoep 1:24714b45cd1b 44 * @param[out] *pResult standard deviation value returned here
xorjoep 1:24714b45cd1b 45 * @return none.
xorjoep 1:24714b45cd1b 46 * @details
xorjoep 1:24714b45cd1b 47 * <b>Scaling and Overflow Behavior:</b>
xorjoep 1:24714b45cd1b 48 *
xorjoep 1:24714b45cd1b 49 * \par
xorjoep 1:24714b45cd1b 50 * The function is implemented using a 64-bit internal accumulator.
xorjoep 1:24714b45cd1b 51 * The input is represented in 1.15 format.
xorjoep 1:24714b45cd1b 52 * Intermediate multiplication yields a 2.30 format, and this
xorjoep 1:24714b45cd1b 53 * result is added without saturation to a 64-bit accumulator in 34.30 format.
xorjoep 1:24714b45cd1b 54 * With 33 guard bits in the accumulator, there is no risk of overflow, and the
xorjoep 1:24714b45cd1b 55 * full precision of the intermediate multiplication is preserved.
xorjoep 1:24714b45cd1b 56 * Finally, the 34.30 result is truncated to 34.15 format by discarding the lower
xorjoep 1:24714b45cd1b 57 * 15 bits, and then saturated to yield a result in 1.15 format.
xorjoep 1:24714b45cd1b 58 */
xorjoep 1:24714b45cd1b 59
xorjoep 1:24714b45cd1b 60 void arm_std_q15(
xorjoep 1:24714b45cd1b 61 q15_t * pSrc,
xorjoep 1:24714b45cd1b 62 uint32_t blockSize,
xorjoep 1:24714b45cd1b 63 q15_t * pResult)
xorjoep 1:24714b45cd1b 64 {
xorjoep 1:24714b45cd1b 65 q31_t sum = 0; /* Accumulator */
xorjoep 1:24714b45cd1b 66 q31_t meanOfSquares, squareOfMean; /* square of mean and mean of square */
xorjoep 1:24714b45cd1b 67 uint32_t blkCnt; /* loop counter */
xorjoep 1:24714b45cd1b 68 q63_t sumOfSquares = 0; /* Accumulator */
xorjoep 1:24714b45cd1b 69 #if defined (ARM_MATH_DSP)
xorjoep 1:24714b45cd1b 70 q31_t in; /* input value */
xorjoep 1:24714b45cd1b 71 q15_t in1; /* input value */
xorjoep 1:24714b45cd1b 72 #else
xorjoep 1:24714b45cd1b 73 q15_t in; /* input value */
xorjoep 1:24714b45cd1b 74 #endif
xorjoep 1:24714b45cd1b 75
xorjoep 1:24714b45cd1b 76 if (blockSize == 1U)
xorjoep 1:24714b45cd1b 77 {
xorjoep 1:24714b45cd1b 78 *pResult = 0;
xorjoep 1:24714b45cd1b 79 return;
xorjoep 1:24714b45cd1b 80 }
xorjoep 1:24714b45cd1b 81
xorjoep 1:24714b45cd1b 82 #if defined (ARM_MATH_DSP)
xorjoep 1:24714b45cd1b 83 /* Run the below code for Cortex-M4 and Cortex-M3 */
xorjoep 1:24714b45cd1b 84
xorjoep 1:24714b45cd1b 85 /*loop Unrolling */
xorjoep 1:24714b45cd1b 86 blkCnt = blockSize >> 2U;
xorjoep 1:24714b45cd1b 87
xorjoep 1:24714b45cd1b 88 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
xorjoep 1:24714b45cd1b 89 ** a second loop below computes the remaining 1 to 3 samples. */
xorjoep 1:24714b45cd1b 90 while (blkCnt > 0U)
xorjoep 1:24714b45cd1b 91 {
xorjoep 1:24714b45cd1b 92 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
xorjoep 1:24714b45cd1b 93 /* Compute Sum of squares of the input samples
xorjoep 1:24714b45cd1b 94 * and then store the result in a temporary variable, sum. */
xorjoep 1:24714b45cd1b 95 in = *__SIMD32(pSrc)++;
xorjoep 1:24714b45cd1b 96 sum += ((in << 16U) >> 16U);
xorjoep 1:24714b45cd1b 97 sum += (in >> 16U);
xorjoep 1:24714b45cd1b 98 sumOfSquares = __SMLALD(in, in, sumOfSquares);
xorjoep 1:24714b45cd1b 99 in = *__SIMD32(pSrc)++;
xorjoep 1:24714b45cd1b 100 sum += ((in << 16U) >> 16U);
xorjoep 1:24714b45cd1b 101 sum += (in >> 16U);
xorjoep 1:24714b45cd1b 102 sumOfSquares = __SMLALD(in, in, sumOfSquares);
xorjoep 1:24714b45cd1b 103
xorjoep 1:24714b45cd1b 104 /* Decrement the loop counter */
xorjoep 1:24714b45cd1b 105 blkCnt--;
xorjoep 1:24714b45cd1b 106 }
xorjoep 1:24714b45cd1b 107
xorjoep 1:24714b45cd1b 108 /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
xorjoep 1:24714b45cd1b 109 ** No loop unrolling is used. */
xorjoep 1:24714b45cd1b 110 blkCnt = blockSize % 0x4U;
xorjoep 1:24714b45cd1b 111
xorjoep 1:24714b45cd1b 112 while (blkCnt > 0U)
xorjoep 1:24714b45cd1b 113 {
xorjoep 1:24714b45cd1b 114 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
xorjoep 1:24714b45cd1b 115 /* Compute Sum of squares of the input samples
xorjoep 1:24714b45cd1b 116 * and then store the result in a temporary variable, sum. */
xorjoep 1:24714b45cd1b 117 in1 = *pSrc++;
xorjoep 1:24714b45cd1b 118 sumOfSquares = __SMLALD(in1, in1, sumOfSquares);
xorjoep 1:24714b45cd1b 119 sum += in1;
xorjoep 1:24714b45cd1b 120
xorjoep 1:24714b45cd1b 121 /* Decrement the loop counter */
xorjoep 1:24714b45cd1b 122 blkCnt--;
xorjoep 1:24714b45cd1b 123 }
xorjoep 1:24714b45cd1b 124
xorjoep 1:24714b45cd1b 125 /* Compute Mean of squares of the input samples
xorjoep 1:24714b45cd1b 126 * and then store the result in a temporary variable, meanOfSquares. */
xorjoep 1:24714b45cd1b 127 meanOfSquares = (q31_t)(sumOfSquares / (q63_t)(blockSize - 1U));
xorjoep 1:24714b45cd1b 128
xorjoep 1:24714b45cd1b 129 /* Compute square of mean */
xorjoep 1:24714b45cd1b 130 squareOfMean = (q31_t)((q63_t)sum * sum / (q63_t)(blockSize * (blockSize - 1U)));
xorjoep 1:24714b45cd1b 131
xorjoep 1:24714b45cd1b 132 /* mean of the squares minus the square of the mean. */
xorjoep 1:24714b45cd1b 133 /* Compute standard deviation and store the result to the destination */
xorjoep 1:24714b45cd1b 134 arm_sqrt_q15(__SSAT((meanOfSquares - squareOfMean) >> 15U, 16U), pResult);
xorjoep 1:24714b45cd1b 135
xorjoep 1:24714b45cd1b 136 #else
xorjoep 1:24714b45cd1b 137 /* Run the below code for Cortex-M0 */
xorjoep 1:24714b45cd1b 138
xorjoep 1:24714b45cd1b 139 /* Loop over blockSize number of values */
xorjoep 1:24714b45cd1b 140 blkCnt = blockSize;
xorjoep 1:24714b45cd1b 141
xorjoep 1:24714b45cd1b 142 while (blkCnt > 0U)
xorjoep 1:24714b45cd1b 143 {
xorjoep 1:24714b45cd1b 144 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
xorjoep 1:24714b45cd1b 145 /* Compute Sum of squares of the input samples
xorjoep 1:24714b45cd1b 146 * and then store the result in a temporary variable, sumOfSquares. */
xorjoep 1:24714b45cd1b 147 in = *pSrc++;
xorjoep 1:24714b45cd1b 148 sumOfSquares += (in * in);
xorjoep 1:24714b45cd1b 149
xorjoep 1:24714b45cd1b 150 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */
xorjoep 1:24714b45cd1b 151 /* Compute sum of all input values and then store the result in a temporary variable, sum. */
xorjoep 1:24714b45cd1b 152 sum += in;
xorjoep 1:24714b45cd1b 153
xorjoep 1:24714b45cd1b 154 /* Decrement the loop counter */
xorjoep 1:24714b45cd1b 155 blkCnt--;
xorjoep 1:24714b45cd1b 156 }
xorjoep 1:24714b45cd1b 157
xorjoep 1:24714b45cd1b 158 /* Compute Mean of squares of the input samples
xorjoep 1:24714b45cd1b 159 * and then store the result in a temporary variable, meanOfSquares. */
xorjoep 1:24714b45cd1b 160 meanOfSquares = (q31_t)(sumOfSquares / (q63_t)(blockSize - 1U));
xorjoep 1:24714b45cd1b 161
xorjoep 1:24714b45cd1b 162 /* Compute square of mean */
xorjoep 1:24714b45cd1b 163 squareOfMean = (q31_t)((q63_t)sum * sum / (q63_t)(blockSize * (blockSize - 1U)));
xorjoep 1:24714b45cd1b 164
xorjoep 1:24714b45cd1b 165 /* mean of the squares minus the square of the mean. */
xorjoep 1:24714b45cd1b 166 /* Compute standard deviation and store the result to the destination */
xorjoep 1:24714b45cd1b 167 arm_sqrt_q15(__SSAT((meanOfSquares - squareOfMean) >> 15U, 16U), pResult);
xorjoep 1:24714b45cd1b 168
xorjoep 1:24714b45cd1b 169 #endif /* #if defined (ARM_MATH_DSP) */
xorjoep 1:24714b45cd1b 170 }
xorjoep 1:24714b45cd1b 171
xorjoep 1:24714b45cd1b 172 /**
xorjoep 1:24714b45cd1b 173 * @} end of STD group
xorjoep 1:24714b45cd1b 174 */