arm_mat_add_q15.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_mat_add_q15.c
00004  * Description:  Q15 matrix addition
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 groupMatrix
00033  */
00034 
00035 /**
00036  * @addtogroup MatrixAdd
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Q15 matrix addition.
00042  * @param[in]       *pSrcA points to the first input matrix structure
00043  * @param[in]       *pSrcB points to the second input matrix structure
00044  * @param[out]      *pDst points to output matrix structure
00045  * @return          The function returns either
00046  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
00047  *
00048  * <b>Scaling and Overflow Behavior:</b>
00049  * \par
00050  * The function uses saturating arithmetic.
00051  * Results outside of the allowable Q15 range [0x8000 0x7FFF] will be saturated.
00052  */
00053 
00054 arm_status arm_mat_add_q15(
00055   const arm_matrix_instance_q15 * pSrcA,
00056   const arm_matrix_instance_q15 * pSrcB,
00057   arm_matrix_instance_q15 * pDst)
00058 {
00059   q15_t *pInA = pSrcA->pData;                    /* input data matrix pointer A  */
00060   q15_t *pInB = pSrcB->pData;                    /* input data matrix pointer B */
00061   q15_t *pOut = pDst->pData;                     /* output data matrix pointer */
00062   uint16_t numSamples;                           /* total number of elements in the matrix  */
00063   uint32_t blkCnt;                               /* loop counters  */
00064   arm_status status;                             /* status of matrix addition  */
00065 
00066 #ifdef ARM_MATH_MATRIX_CHECK
00067 
00068 
00069   /* Check for matrix mismatch condition */
00070   if ((pSrcA->numRows != pSrcB->numRows) ||
00071      (pSrcA->numCols != pSrcB->numCols) ||
00072      (pSrcA->numRows != pDst->numRows) || (pSrcA->numCols != pDst->numCols))
00073   {
00074     /* Set status as ARM_MATH_SIZE_MISMATCH */
00075     status = ARM_MATH_SIZE_MISMATCH;
00076   }
00077   else
00078 #endif /*    #ifdef ARM_MATH_MATRIX_CHECK    */
00079 
00080   {
00081     /* Total number of samples in the input matrix */
00082     numSamples = (uint16_t) (pSrcA->numRows * pSrcA->numCols);
00083 
00084 #if defined (ARM_MATH_DSP)
00085 
00086     /* Run the below code for Cortex-M4 and Cortex-M3 */
00087 
00088     /* Loop unrolling */
00089     blkCnt = (uint32_t) numSamples >> 2U;
00090 
00091     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00092      ** a second loop below computes the remaining 1 to 3 samples. */
00093     while (blkCnt > 0U)
00094     {
00095       /* C(m,n) = A(m,n) + B(m,n) */
00096       /* Add, Saturate and then store the results in the destination buffer. */
00097       *__SIMD32(pOut)++ = __QADD16(*__SIMD32(pInA)++, *__SIMD32(pInB)++);
00098       *__SIMD32(pOut)++ = __QADD16(*__SIMD32(pInA)++, *__SIMD32(pInB)++);
00099 
00100       /* Decrement the loop counter */
00101       blkCnt--;
00102     }
00103 
00104     /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00105      ** No loop unrolling is used. */
00106     blkCnt = (uint32_t) numSamples % 0x4U;
00107 
00108     /* q15 pointers of input and output are initialized */
00109 
00110     while (blkCnt > 0U)
00111     {
00112       /* C(m,n) = A(m,n) + B(m,n) */
00113       /* Add, Saturate and then store the results in the destination buffer. */
00114       *pOut++ = (q15_t) __QADD16(*pInA++, *pInB++);
00115 
00116       /* Decrement the loop counter */
00117       blkCnt--;
00118     }
00119 
00120 #else
00121 
00122     /* Run the below code for Cortex-M0 */
00123 
00124     /* Initialize blkCnt with number of samples */
00125     blkCnt = (uint32_t) numSamples;
00126 
00127 
00128     /* q15 pointers of input and output are initialized */
00129     while (blkCnt > 0U)
00130     {
00131       /* C(m,n) = A(m,n) + B(m,n) */
00132       /* Add, Saturate and then store the results in the destination buffer. */
00133       *pOut++ = (q15_t) __SSAT(((q31_t) * pInA++ + *pInB++), 16);
00134 
00135       /* Decrement the loop counter */
00136       blkCnt--;
00137     }
00138 
00139 #endif /* #if defined (ARM_MATH_DSP) */
00140 
00141     /* set status as ARM_MATH_SUCCESS */
00142     status = ARM_MATH_SUCCESS;
00143   }
00144 
00145   /* Return to application */
00146   return (status);
00147 }
00148 
00149 /**
00150  * @} end of MatrixAdd group
00151  */
00152