arm_rms_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_rms_q15.c    
00009 *    
00010 * Description:  Root Mean Square of the elements of a Q15 vector.  
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  * @addtogroup RMS    
00045  * @{    
00046  */
00047 
00048 /**    
00049  * @brief Root Mean Square of the elements of a Q15 vector.    
00050  * @param[in]       *pSrc points to the input vector    
00051  * @param[in]       blockSize length of the input vector    
00052  * @param[out]      *pResult rms value returned here    
00053  * @return none.    
00054  *    
00055  * @details    
00056  * <b>Scaling and Overflow Behavior:</b>    
00057  *    
00058  * \par    
00059  * The function is implemented using a 64-bit internal accumulator.    
00060  * The input is represented in 1.15 format.    
00061  * Intermediate multiplication yields a 2.30 format, and this    
00062  * result is added without saturation to a 64-bit accumulator in 34.30 format.    
00063  * With 33 guard bits in the accumulator, there is no risk of overflow, and the    
00064  * full precision of the intermediate multiplication is preserved.    
00065  * Finally, the 34.30 result is truncated to 34.15 format by discarding the lower     
00066  * 15 bits, and then saturated to yield a result in 1.15 format.    
00067  *    
00068  */
00069 
00070 void arm_rms_q15(
00071   q15_t * pSrc,
00072   uint32_t blockSize,
00073   q15_t * pResult)
00074 {
00075   q63_t sum = 0;                                 /* accumulator */
00076 
00077 #ifndef ARM_MATH_CM0_FAMILY
00078 
00079   /* Run the below code for Cortex-M4 and Cortex-M3 */
00080 
00081   q31_t in;                                      /* temporary variable to store the input value */
00082   q15_t in1;                                     /* temporary variable to store the input value */
00083   uint32_t blkCnt;                               /* loop counter */
00084 
00085   /* loop Unrolling */
00086   blkCnt = blockSize >> 2u;
00087 
00088   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00089    ** a second loop below computes the remaining 1 to 3 samples. */
00090   while(blkCnt > 0u)
00091   {
00092     /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
00093     /* Compute sum of the squares and then store the results in a temporary variable, sum */
00094     in = *__SIMD32(pSrc)++;
00095     sum = __SMLALD(in, in, sum);
00096     in = *__SIMD32(pSrc)++;
00097     sum = __SMLALD(in, in, sum);
00098 
00099     /* Decrement the loop counter */
00100     blkCnt--;
00101   }
00102 
00103   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.    
00104    ** No loop unrolling is used. */
00105   blkCnt = blockSize % 0x4u;
00106 
00107   while(blkCnt > 0u)
00108   {
00109     /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
00110     /* Compute sum of the squares and then store the results in a temporary variable, sum */
00111     in1 = *pSrc++;
00112     sum = __SMLALD(in1, in1, sum);
00113 
00114     /* Decrement the loop counter */
00115     blkCnt--;
00116   }
00117 
00118   /* Truncating and saturating the accumulator to 1.15 format */
00119   in = (q31_t)(sum >> 15);
00120 
00121   in1 = __SSAT(in / blockSize, 16);
00122 
00123   /* Store the result in the destination */
00124   arm_sqrt_q15(in1, pResult);
00125 
00126 #else
00127 
00128   /* Run the below code for Cortex-M0 */
00129 
00130   q15_t in;                                      /* temporary variable to store the input value */
00131   q31_t tmp;                                     /* temporary variable to store the input value */
00132   uint32_t blkCnt;                               /* loop counter */
00133 
00134   /* Loop over blockSize number of values */
00135   blkCnt = blockSize;
00136 
00137   while(blkCnt > 0u)
00138   {
00139     /* C = (A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1]) */
00140     /* Compute sum of the squares and then store the results in a temporary variable, sum */
00141     in = *pSrc++;
00142     sum += ((q31_t) in * in);
00143 
00144     /* Decrement the loop counter */
00145     blkCnt--;
00146   }
00147 
00148   /* Truncating and saturating the accumulator to 1.15 format */
00149   tmp = (q31_t)(sum >> 15);
00150 
00151   in = __SSAT(tmp / blockSize, 16);
00152 
00153   /* Store the result in the destination */
00154   arm_sqrt_q15(in, pResult);
00155 
00156 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
00157 
00158 }
00159 
00160 /**    
00161  * @} end of RMS group    
00162  */