CMSIS DSP library
Dependents: KL25Z_FFT_Demo Hat_Board_v5_1 KL25Z_FFT_Demo_tony KL25Z_FFT_Demo_tony ... more
Fork of mbed-dsp by
arm_std_q15.c
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2010-2013 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 17. January 2013 00005 * $Revision: V1.4.1 00006 * 00007 * Project: CMSIS DSP Library 00008 * Title: arm_std_q15.c 00009 * 00010 * Description: Standard deviation of an array of Q15 type. 00011 * 00012 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 00013 * 00014 * Redistribution and use in source and binary forms, with or without 00015 * modification, are permitted provided that the following conditions 00016 * are met: 00017 * - Redistributions of source code must retain the above copyright 00018 * notice, this list of conditions and the following disclaimer. 00019 * - Redistributions in binary form must reproduce the above copyright 00020 * notice, this list of conditions and the following disclaimer in 00021 * the documentation and/or other materials provided with the 00022 * distribution. 00023 * - Neither the name of ARM LIMITED nor the names of its contributors 00024 * may be used to endorse or promote products derived from this 00025 * software without specific prior written permission. 00026 * 00027 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 00028 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 00029 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 00030 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 00031 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 00032 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 00033 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 00034 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 00035 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 00036 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 00037 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 00038 * POSSIBILITY OF SUCH DAMAGE. 00039 * -------------------------------------------------------------------- */ 00040 00041 #include "arm_math.h" 00042 00043 /** 00044 * @ingroup groupStats 00045 */ 00046 00047 /** 00048 * @addtogroup STD 00049 * @{ 00050 */ 00051 00052 /** 00053 * @brief Standard deviation of the elements of a Q15 vector. 00054 * @param[in] *pSrc points to the input vector 00055 * @param[in] blockSize length of the input vector 00056 * @param[out] *pResult standard deviation value returned here 00057 * @return none. 00058 * 00059 * @details 00060 * <b>Scaling and Overflow Behavior:</b> 00061 * 00062 * \par 00063 * The function is implemented using a 64-bit internal accumulator. 00064 * The input is represented in 1.15 format. 00065 * Intermediate multiplication yields a 2.30 format, and this 00066 * result is added without saturation to a 64-bit accumulator in 34.30 format. 00067 * With 33 guard bits in the accumulator, there is no risk of overflow, and the 00068 * full precision of the intermediate multiplication is preserved. 00069 * Finally, the 34.30 result is truncated to 34.15 format by discarding the lower 00070 * 15 bits, and then saturated to yield a result in 1.15 format. 00071 */ 00072 00073 void arm_std_q15( 00074 q15_t * pSrc, 00075 uint32_t blockSize, 00076 q15_t * pResult) 00077 { 00078 q31_t sum = 0; /* Accumulator */ 00079 q31_t meanOfSquares, squareOfMean; /* square of mean and mean of square */ 00080 q15_t mean; /* mean */ 00081 uint32_t blkCnt; /* loop counter */ 00082 q15_t t; /* Temporary variable */ 00083 q63_t sumOfSquares = 0; /* Accumulator */ 00084 00085 #ifndef ARM_MATH_CM0_FAMILY 00086 00087 /* Run the below code for Cortex-M4 and Cortex-M3 */ 00088 00089 q31_t in; /* input value */ 00090 q15_t in1; /* input value */ 00091 00092 /*loop Unrolling */ 00093 blkCnt = blockSize >> 2u; 00094 00095 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00096 ** a second loop below computes the remaining 1 to 3 samples. */ 00097 while(blkCnt > 0u) 00098 { 00099 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00100 /* Compute Sum of squares of the input samples 00101 * and then store the result in a temporary variable, sum. */ 00102 in = *__SIMD32(pSrc)++; 00103 sum += ((in << 16) >> 16); 00104 sum += (in >> 16); 00105 sumOfSquares = __SMLALD(in, in, sumOfSquares); 00106 in = *__SIMD32(pSrc)++; 00107 sum += ((in << 16) >> 16); 00108 sum += (in >> 16); 00109 sumOfSquares = __SMLALD(in, in, sumOfSquares); 00110 00111 /* Decrement the loop counter */ 00112 blkCnt--; 00113 } 00114 00115 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00116 ** No loop unrolling is used. */ 00117 blkCnt = blockSize % 0x4u; 00118 00119 while(blkCnt > 0u) 00120 { 00121 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00122 /* Compute Sum of squares of the input samples 00123 * and then store the result in a temporary variable, sum. */ 00124 in1 = *pSrc++; 00125 sumOfSquares = __SMLALD(in1, in1, sumOfSquares); 00126 sum += in1; 00127 00128 /* Decrement the loop counter */ 00129 blkCnt--; 00130 } 00131 00132 /* Compute Mean of squares of the input samples 00133 * and then store the result in a temporary variable, meanOfSquares. */ 00134 t = (q15_t) ((1.0 / (blockSize - 1)) * 16384LL); 00135 sumOfSquares = __SSAT((sumOfSquares >> 15u), 16u); 00136 00137 meanOfSquares = (q31_t) ((sumOfSquares * t) >> 14u); 00138 00139 /* Compute mean of all input values */ 00140 t = (q15_t) ((1.0 / (blockSize * (blockSize - 1))) * 32768LL); 00141 mean = (q15_t) __SSAT(sum, 16u); 00142 00143 /* Compute square of mean */ 00144 squareOfMean = ((q31_t) mean * mean) >> 15; 00145 squareOfMean = (q31_t) (((q63_t) squareOfMean * t) >> 15); 00146 00147 /* mean of the squares minus the square of the mean. */ 00148 in1 = (q15_t) (meanOfSquares - squareOfMean); 00149 00150 /* Compute standard deviation and store the result to the destination */ 00151 arm_sqrt_q15(in1, pResult); 00152 00153 #else 00154 00155 /* Run the below code for Cortex-M0 */ 00156 q15_t in; /* input value */ 00157 00158 /* Loop over blockSize number of values */ 00159 blkCnt = blockSize; 00160 00161 while(blkCnt > 0u) 00162 { 00163 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00164 /* Compute Sum of squares of the input samples 00165 * and then store the result in a temporary variable, sumOfSquares. */ 00166 in = *pSrc++; 00167 sumOfSquares += (in * in); 00168 00169 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ 00170 /* Compute sum of all input values and then store the result in a temporary variable, sum. */ 00171 sum += in; 00172 00173 /* Decrement the loop counter */ 00174 blkCnt--; 00175 } 00176 00177 /* Compute Mean of squares of the input samples 00178 * and then store the result in a temporary variable, meanOfSquares. */ 00179 t = (q15_t) ((1.0 / (blockSize - 1)) * 16384LL); 00180 sumOfSquares = __SSAT((sumOfSquares >> 15u), 16u); 00181 meanOfSquares = (q31_t) ((sumOfSquares * t) >> 14u); 00182 00183 /* Compute mean of all input values */ 00184 mean = (q15_t) __SSAT(sum, 16u); 00185 00186 /* Compute square of mean of the input samples 00187 * and then store the result in a temporary variable, squareOfMean.*/ 00188 t = (q15_t) ((1.0 / (blockSize * (blockSize - 1))) * 32768LL); 00189 squareOfMean = ((q31_t) mean * mean) >> 15; 00190 squareOfMean = (q31_t) (((q63_t) squareOfMean * t) >> 15); 00191 00192 /* mean of the squares minus the square of the mean. */ 00193 in = (q15_t) (meanOfSquares - squareOfMean); 00194 00195 /* Compute standard deviation and store the result to the destination */ 00196 arm_sqrt_q15(in, pResult); 00197 00198 #endif /* #ifndef ARM_MATH_CM0_FAMILY */ 00199 00200 00201 } 00202 00203 /** 00204 * @} end of STD group 00205 */
Generated on Tue Jul 12 2022 12:36:57 by 1.7.2