arm_power_q7.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_power_q7.c
00004  * Description:  Sum of the squares of the elements of a Q7 vector
00005  *
00006  * $Date:        27. January 2017
00007  * $Revision:    V.1.5.1
00008  *
00009  * Target Processor: Cortex-M cores
00010  * -------------------------------------------------------------------- */
00011 /*
00012  * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
00013  *
00014  * SPDX-License-Identifier: Apache-2.0
00015  *
00016  * Licensed under the Apache License, Version 2.0 (the License); you may
00017  * not use this file except in compliance with the License.
00018  * You may obtain a copy of the License at
00019  *
00020  * www.apache.org/licenses/LICENSE-2.0
00021  *
00022  * Unless required by applicable law or agreed to in writing, software
00023  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
00024  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
00025  * See the License for the specific language governing permissions and
00026  * limitations under the License.
00027  */
00028 
00029 #include "arm_math.h"
00030 
00031 /**
00032  * @ingroup groupStats
00033  */
00034 
00035 /**
00036  * @addtogroup power
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Sum of the squares of the elements of a Q7 vector.
00042  * @param[in]       *pSrc points to the input vector
00043  * @param[in]       blockSize length of the input vector
00044  * @param[out]      *pResult sum of the squares value returned here
00045  * @return none.
00046  *
00047  * @details
00048  * <b>Scaling and Overflow Behavior:</b>
00049  *
00050  * \par
00051  * The function is implemented using a 32-bit internal accumulator.
00052  * The input is represented in 1.7 format.
00053  * Intermediate multiplication yields a 2.14 format, and this
00054  * result is added without saturation to an accumulator in 18.14 format.
00055  * With 17 guard bits in the accumulator, there is no risk of overflow, and the
00056  * full precision of the intermediate multiplication is preserved.
00057  * Finally, the return result is in 18.14 format.
00058  *
00059  */
00060 
00061 void arm_power_q7(
00062   q7_t * pSrc,
00063   uint32_t blockSize,
00064   q31_t * pResult)
00065 {
00066   q31_t sum = 0;                                 /* Temporary result storage */
00067   q7_t in;                                       /* Temporary variable to store input */
00068   uint32_t blkCnt;                               /* loop counter */
00069 
00070 #if defined (ARM_MATH_DSP)
00071   /* Run the below code for Cortex-M4 and Cortex-M3 */
00072 
00073   q31_t input1;                                  /* Temporary variable to store packed input */
00074   q31_t in1, in2;                                /* Temporary variables to store input */
00075 
00076   /*loop Unrolling */
00077   blkCnt = blockSize >> 2U;
00078 
00079   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00080    ** a second loop below computes the remaining 1 to 3 samples. */
00081   while (blkCnt > 0U)
00082   {
00083     /* Reading two inputs of pSrc vector and packing */
00084     input1 = *__SIMD32(pSrc)++;
00085 
00086     in1 = __SXTB16(__ROR(input1, 8));
00087     in2 = __SXTB16(input1);
00088 
00089     /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */
00090     /* calculate power and accumulate to accumulator */
00091     sum = __SMLAD(in1, in1, sum);
00092     sum = __SMLAD(in2, in2, sum);
00093 
00094     /* Decrement the loop counter */
00095     blkCnt--;
00096   }
00097 
00098   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00099    ** No loop unrolling is used. */
00100   blkCnt = blockSize % 0x4U;
00101 
00102 #else
00103   /* Run the below code for Cortex-M0 */
00104 
00105   /* Loop over blockSize number of values */
00106   blkCnt = blockSize;
00107 
00108 #endif /* #if defined (ARM_MATH_DSP) */
00109 
00110   while (blkCnt > 0U)
00111   {
00112     /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */
00113     /* Compute Power and then store the result in a temporary variable, sum. */
00114     in = *pSrc++;
00115     sum += ((q15_t) in * in);
00116 
00117     /* Decrement the loop counter */
00118     blkCnt--;
00119   }
00120 
00121   /* Store the result in 18.14 format  */
00122   *pResult = sum;
00123 }
00124 
00125 /**
00126  * @} end of power group
00127  */
00128