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_var_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_var_q15.c 00009 * 00010 * Description: Variance 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 variance 00049 * @{ 00050 */ 00051 00052 /** 00053 * @brief Variance 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 variance 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 00074 00075 void arm_var_q15( 00076 q15_t * pSrc, 00077 uint32_t blockSize, 00078 q31_t * pResult) 00079 { 00080 q31_t sum = 0; /* Accumulator */ 00081 q31_t meanOfSquares, squareOfMean; /* Mean of square and square of mean */ 00082 q15_t mean; /* mean */ 00083 uint32_t blkCnt; /* loop counter */ 00084 q15_t t; /* Temporary variable */ 00085 q63_t sumOfSquares = 0; /* Accumulator */ 00086 00087 #ifndef ARM_MATH_CM0_FAMILY 00088 00089 /* Run the below code for Cortex-M4 and Cortex-M3 */ 00090 00091 q31_t in; /* Input variable */ 00092 q15_t in1; /* Temporary variable */ 00093 00094 /*loop Unrolling */ 00095 blkCnt = blockSize >> 2u; 00096 00097 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00098 ** a second loop below computes the remaining 1 to 3 samples. */ 00099 while(blkCnt > 0u) 00100 { 00101 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00102 /* Compute Sum of squares of the input samples 00103 * and then store the result in a temporary variable, sum. */ 00104 in = *__SIMD32(pSrc)++; 00105 sum += ((in << 16) >> 16); 00106 sum += (in >> 16); 00107 sumOfSquares = __SMLALD(in, in, sumOfSquares); 00108 in = *__SIMD32(pSrc)++; 00109 sum += ((in << 16) >> 16); 00110 sum += (in >> 16); 00111 sumOfSquares = __SMLALD(in, in, sumOfSquares); 00112 00113 /* Decrement the loop counter */ 00114 blkCnt--; 00115 } 00116 00117 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00118 ** No loop unrolling is used. */ 00119 blkCnt = blockSize % 0x4u; 00120 00121 while(blkCnt > 0u) 00122 { 00123 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00124 /* Compute Sum of squares of the input samples 00125 * and then store the result in a temporary variable, sum. */ 00126 in1 = *pSrc++; 00127 sum += in1; 00128 sumOfSquares = __SMLALD(in1, in1, sumOfSquares); 00129 00130 /* Decrement the loop counter */ 00131 blkCnt--; 00132 } 00133 00134 /* Compute Mean of squares of the input samples 00135 * and then store the result in a temporary variable, meanOfSquares. */ 00136 t = (q15_t) ((1.0f / (float32_t) (blockSize - 1u)) * 16384); 00137 sumOfSquares = __SSAT((sumOfSquares >> 15u), 16u); 00138 00139 meanOfSquares = (q31_t) ((sumOfSquares * t) >> 14u); 00140 00141 #else 00142 00143 /* Run the below code for Cortex-M0 */ 00144 00145 q15_t in; /* Temporary variable */ 00146 /* Loop over blockSize number of values */ 00147 blkCnt = blockSize; 00148 00149 while(blkCnt > 0u) 00150 { 00151 /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 00152 /* Compute Sum of squares of the input samples 00153 * and then store the result in a temporary variable, sumOfSquares. */ 00154 in = *pSrc++; 00155 sumOfSquares += (in * in); 00156 00157 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ 00158 /* Compute sum of all input values and then store the result in a temporary variable, sum. */ 00159 sum += in; 00160 00161 /* Decrement the loop counter */ 00162 blkCnt--; 00163 } 00164 00165 /* Compute Mean of squares of the input samples 00166 * and then store the result in a temporary variable, meanOfSquares. */ 00167 t = (q15_t) ((1.0f / (float32_t) (blockSize - 1u)) * 16384); 00168 sumOfSquares = __SSAT((sumOfSquares >> 15u), 16u); 00169 meanOfSquares = (q31_t) ((sumOfSquares * t) >> 14u); 00170 00171 #endif /* #ifndef ARM_MATH_CM0_FAMILY */ 00172 00173 /* Compute mean of all input values */ 00174 t = (q15_t) ((1.0f / (float32_t) (blockSize * (blockSize - 1u))) * 32768); 00175 mean = __SSAT(sum, 16u); 00176 00177 /* Compute square of mean */ 00178 squareOfMean = ((q31_t) mean * mean) >> 15; 00179 squareOfMean = (q31_t) (((q63_t) squareOfMean * t) >> 15); 00180 00181 /* Compute variance and then store the result to the destination */ 00182 *pResult = (meanOfSquares - squareOfMean); 00183 00184 } 00185 00186 /** 00187 * @} end of variance group 00188 */
Generated on Tue Jul 12 2022 12:36:57 by 1.7.2