arm_var_q15.c

00001 /* ----------------------------------------------------------------------  
00002 * Copyright (C) 2010 ARM Limited. All rights reserved.  
00003 *  
00004 * $Date:        29. November 2010  
00005 * $Revision:    V1.0.3  
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
00013 *  
00014 * Version 1.0.3 2010/11/29 
00015 *    Re-organized the CMSIS folders and updated documentation.  
00016 *   
00017 * Version 1.0.2 2010/11/11  
00018 *    Documentation updated.   
00019 *  
00020 * Version 1.0.1 2010/10/05   
00021 *    Production release and review comments incorporated.  
00022 *  
00023 * Version 1.0.0 2010/09/20   
00024 *    Production release and review comments incorporated.  
00025 * -------------------------------------------------------------------- */ 
00026  
00027 #include "arm_math.h" 
00028  
00029 /**  
00030  * @ingroup groupStats  
00031  */ 
00032  
00033 /**  
00034  * @addtogroup variance  
00035  * @{  
00036  */ 
00037  
00038 /**  
00039  * @brief Variance of the elements of a Q15 vector.  
00040  * @param[in]       *pSrc points to the input vector  
00041  * @param[in]       blockSize length of the input vector  
00042  * @param[out]      *pResult variance value returned here  
00043  * @return none.  
00044  *  
00045  * @details  
00046  * <b>Scaling and Overflow Behavior:</b>  
00047  *  
00048  * \par  
00049  * The function is implemented using a 64-bit internal accumulator.  
00050  * The input is represented in 1.15 format. 
00051  * Intermediate multiplication yields a 2.30 format, and this  
00052  * result is added without saturation to a 64-bit accumulator in 34.30 format.  
00053  * With 33 guard bits in the accumulator, there is no risk of overflow, and the  
00054  * full precision of the intermediate multiplication is preserved.  
00055  * Finally, the 34.30 result is truncated to 34.15 format by discarding the lower   
00056  * 15 bits, and then saturated to yield a result in 1.15 format.  
00057  *  
00058  */ 
00059  
00060  
00061 void arm_var_q15( 
00062   q15_t * pSrc, 
00063   uint32_t blockSize, 
00064   q31_t * pResult) 
00065 { 
00066   q63_t sum = 0;                                 /* Accumulator */ 
00067   q31_t meanOfSquares, squareOfMean;             /* Mean of square and square of mean */ 
00068   q15_t mean;                                    /* mean */ 
00069   q31_t in;                                      /* Input variable */ 
00070   q15_t in1;                                     /* Temporary variable */ 
00071   uint32_t blkCnt;                               /* loop counter */ 
00072   q15_t t;                                       /* Temporary variable */ 
00073   q15_t *pIn;                                    /* Temporary pointer */ 
00074  
00075   pIn = pSrc; 
00076  
00077   /*loop Unrolling */ 
00078   blkCnt = blockSize >> 2u; 
00079  
00080   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.  
00081    ** a second loop below computes the remaining 1 to 3 samples. */ 
00082   while(blkCnt > 0u) 
00083   { 
00084     /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1])  */ 
00085     /* Compute Sum of squares of the input samples  
00086      * and then store the result in a temporary variable, sum. */ 
00087     in = *__SIMD32(pSrc)++; 
00088     sum = __SMLALD(in, in, sum); 
00089     in = *__SIMD32(pSrc)++; 
00090     sum = __SMLALD(in, in, sum); 
00091  
00092     /* Decrement the loop counter */ 
00093     blkCnt--; 
00094   } 
00095  
00096   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.  
00097    ** No loop unrolling is used. */ 
00098   blkCnt = blockSize % 0x4u; 
00099  
00100   while(blkCnt > 0u) 
00101   { 
00102     /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */ 
00103     /* Compute Sum of squares of the input samples  
00104      * and then store the result in a temporary variable, sum. */ 
00105     in1 = *pSrc++; 
00106     sum = __SMLALD(in1, in1, sum); 
00107  
00108     /* Decrement the loop counter */ 
00109     blkCnt--; 
00110   } 
00111  
00112   /* Compute Mean of squares of the input samples  
00113    * and then store the result in a temporary variable, meanOfSquares. */ 
00114   t = (q15_t) ((1.0f / (float32_t) (blockSize - 1u)) * 16384); 
00115   sum = __SSAT((sum >> 15u), 16u); 
00116  
00117   meanOfSquares = (q31_t) ((sum * t) >> 14u); 
00118  
00119   /* Reset the accumulator */ 
00120   sum = 0; 
00121  
00122   /*loop Unrolling */ 
00123   blkCnt = blockSize >> 2u; 
00124  
00125   /* Reset the input working pointer */ 
00126   pSrc = pIn; 
00127  
00128   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.  
00129    ** a second loop below computes the remaining 1 to 3 samples. */ 
00130   while(blkCnt > 0u) 
00131   { 
00132     /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ 
00133     /* Compute sum of all input values and then store the result in a temporary variable, sum. */ 
00134     sum += *pSrc++; 
00135     sum += *pSrc++; 
00136     sum += *pSrc++; 
00137     sum += *pSrc++; 
00138  
00139     /* Decrement the loop counter */ 
00140     blkCnt--; 
00141   } 
00142  
00143   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.  
00144    ** No loop unrolling is used. */ 
00145   blkCnt = blockSize % 0x4u; 
00146  
00147   while(blkCnt > 0u) 
00148   { 
00149     /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ 
00150     /* Compute sum of all input values and then store the result in a temporary variable, sum. */ 
00151     sum += *pSrc++; 
00152  
00153     /* Decrement the loop counter */ 
00154     blkCnt--; 
00155   } 
00156  
00157   /* Compute mean of all input values */ 
00158   t = (q15_t) ((1.0f / (float32_t) (blockSize * (blockSize - 1u))) * 32768); 
00159   mean = __SSAT(sum, 16u); 
00160  
00161   /* Compute square of mean */ 
00162   squareOfMean = ((q31_t) mean * mean) >> 15; 
00163   squareOfMean = (q31_t) (((q63_t) squareOfMean * t) >> 15); 
00164  
00165   /* Compute variance and then store the result to the destination */ 
00166   *pResult = (meanOfSquares - squareOfMean); 
00167  
00168 } 
00169  
00170 /**  
00171  * @} end of variance group  
00172  */