arm_mat_add_f32.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_mat_add_f32.c
00004  * Description:  Floating-point 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  * @defgroup MatrixAdd Matrix Addition
00037  *
00038  * Adds two matrices.
00039  * \image html MatrixAddition.gif "Addition of two 3 x 3 matrices"
00040  *
00041  * The functions check to make sure that
00042  * <code>pSrcA</code>, <code>pSrcB</code>, and <code>pDst</code> have the same
00043  * number of rows and columns.
00044  */
00045 
00046 /**
00047  * @addtogroup MatrixAdd
00048  * @{
00049  */
00050 
00051 
00052 /**
00053  * @brief Floating-point matrix addition.
00054  * @param[in]       *pSrcA points to the first input matrix structure
00055  * @param[in]       *pSrcB points to the second input matrix structure
00056  * @param[out]      *pDst points to output matrix structure
00057  * @return          The function returns either
00058  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
00059  */
00060 
00061 arm_status arm_mat_add_f32(
00062   const arm_matrix_instance_f32 * pSrcA,
00063   const arm_matrix_instance_f32 * pSrcB,
00064   arm_matrix_instance_f32 * pDst)
00065 {
00066   float32_t *pIn1 = pSrcA->pData;                /* input data matrix pointer A  */
00067   float32_t *pIn2 = pSrcB->pData;                /* input data matrix pointer B  */
00068   float32_t *pOut = pDst->pData;                 /* output data matrix pointer   */
00069 
00070 #if defined (ARM_MATH_DSP)
00071 
00072   float32_t inA1, inA2, inB1, inB2, out1, out2;  /* temporary variables */
00073 
00074 #endif //      #if defined (ARM_MATH_DSP)
00075 
00076   uint32_t numSamples;                           /* total number of elements in the matrix  */
00077   uint32_t blkCnt;                               /* loop counters */
00078   arm_status status;                             /* status of matrix addition */
00079 
00080 #ifdef ARM_MATH_MATRIX_CHECK
00081   /* Check for matrix mismatch condition */
00082   if ((pSrcA->numRows != pSrcB->numRows) ||
00083      (pSrcA->numCols != pSrcB->numCols) ||
00084      (pSrcA->numRows != pDst->numRows) || (pSrcA->numCols != pDst->numCols))
00085   {
00086     /* Set status as ARM_MATH_SIZE_MISMATCH */
00087     status = ARM_MATH_SIZE_MISMATCH;
00088   }
00089   else
00090 #endif
00091   {
00092 
00093     /* Total number of samples in the input matrix */
00094     numSamples = (uint32_t) pSrcA->numRows * pSrcA->numCols;
00095 
00096 #if defined (ARM_MATH_DSP)
00097 
00098     /* Loop unrolling */
00099     blkCnt = numSamples >> 2U;
00100 
00101     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00102      ** a second loop below computes the remaining 1 to 3 samples. */
00103     while (blkCnt > 0U)
00104     {
00105       /* C(m,n) = A(m,n) + B(m,n) */
00106       /* Add and then store the results in the destination buffer. */
00107       /* Read values from source A */
00108       inA1 = pIn1[0];
00109 
00110       /* Read values from source B */
00111       inB1 = pIn2[0];
00112 
00113       /* Read values from source A */
00114       inA2 = pIn1[1];
00115 
00116       /* out = sourceA + sourceB */
00117       out1 = inA1 + inB1;
00118 
00119       /* Read values from source B */
00120       inB2 = pIn2[1];
00121 
00122       /* Read values from source A */
00123       inA1 = pIn1[2];
00124 
00125       /* out = sourceA + sourceB */
00126       out2 = inA2 + inB2;
00127 
00128       /* Read values from source B */
00129       inB1 = pIn2[2];
00130 
00131       /* Store result in destination */
00132       pOut[0] = out1;
00133       pOut[1] = out2;
00134 
00135       /* Read values from source A */
00136       inA2 = pIn1[3];
00137 
00138       /* Read values from source B */
00139       inB2 = pIn2[3];
00140 
00141       /* out = sourceA + sourceB */
00142       out1 = inA1 + inB1;
00143 
00144       /* out = sourceA + sourceB */
00145       out2 = inA2 + inB2;
00146 
00147       /* Store result in destination */
00148       pOut[2] = out1;
00149 
00150       /* Store result in destination */
00151       pOut[3] = out2;
00152 
00153 
00154       /* update pointers to process next sampels */
00155       pIn1 += 4U;
00156       pIn2 += 4U;
00157       pOut += 4U;
00158       /* Decrement the loop counter */
00159       blkCnt--;
00160     }
00161 
00162     /* If the numSamples is not a multiple of 4, compute any remaining output samples here.
00163      ** No loop unrolling is used. */
00164     blkCnt = numSamples % 0x4U;
00165 
00166 #else
00167 
00168     /* Run the below code for Cortex-M0 */
00169 
00170     /* Initialize blkCnt with number of samples */
00171     blkCnt = numSamples;
00172 
00173 #endif /* #if defined (ARM_MATH_DSP) */
00174 
00175     while (blkCnt > 0U)
00176     {
00177       /* C(m,n) = A(m,n) + B(m,n) */
00178       /* Add and then store the results in the destination buffer. */
00179       *pOut++ = (*pIn1++) + (*pIn2++);
00180 
00181       /* Decrement the loop counter */
00182       blkCnt--;
00183     }
00184 
00185     /* set status as ARM_MATH_SUCCESS */
00186     status = ARM_MATH_SUCCESS;
00187 
00188   }
00189 
00190   /* Return to application */
00191   return (status);
00192 }
00193 
00194 /**
00195  * @} end of MatrixAdd group
00196  */
00197