Robert Lopez / CMSIS5
Embed: (wiki syntax)

« Back to documentation index

Show/hide line numbers arm_mat_trans_f32.c Source File

arm_mat_trans_f32.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_mat_trans_f32.c
00004  * Description:  Floating-point matrix transpose
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 /**
00030  * @defgroup MatrixTrans Matrix Transpose
00031  *
00032  * Tranposes a matrix.
00033  * Transposing an <code>M x N</code> matrix flips it around the center diagonal and results in an <code>N x M</code> matrix.
00034  * \image html MatrixTranspose.gif "Transpose of a 3 x 3 matrix"
00035  */
00036 
00037 #include "arm_math.h"
00038 
00039 /**
00040  * @ingroup groupMatrix
00041  */
00042 
00043 /**
00044  * @addtogroup MatrixTrans
00045  * @{
00046  */
00047 
00048 /**
00049   * @brief Floating-point matrix transpose.
00050   * @param[in]  *pSrc points to the input matrix
00051   * @param[out] *pDst points to the output matrix
00052   * @return     The function returns either  <code>ARM_MATH_SIZE_MISMATCH</code>
00053   * or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
00054   */
00055 
00056 
00057 arm_status arm_mat_trans_f32(
00058   const arm_matrix_instance_f32 * pSrc,
00059   arm_matrix_instance_f32 * pDst)
00060 {
00061   float32_t *pIn = pSrc->pData;                  /* input data matrix pointer */
00062   float32_t *pOut = pDst->pData;                 /* output data matrix pointer */
00063   float32_t *px;                                 /* Temporary output data matrix pointer */
00064   uint16_t nRows = pSrc->numRows;                /* number of rows */
00065   uint16_t nColumns = pSrc->numCols;             /* number of columns */
00066 
00067 #if defined (ARM_MATH_DSP)
00068 
00069   /* Run the below code for Cortex-M4 and Cortex-M3 */
00070 
00071   uint16_t blkCnt, i = 0U, row = nRows;          /* loop counters */
00072   arm_status status;                             /* status of matrix transpose  */
00073 
00074 
00075 #ifdef ARM_MATH_MATRIX_CHECK
00076 
00077 
00078   /* Check for matrix mismatch condition */
00079   if ((pSrc->numRows != pDst->numCols) || (pSrc->numCols != pDst->numRows))
00080   {
00081     /* Set status as ARM_MATH_SIZE_MISMATCH */
00082     status = ARM_MATH_SIZE_MISMATCH;
00083   }
00084   else
00085 #endif /*    #ifdef ARM_MATH_MATRIX_CHECK    */
00086 
00087   {
00088     /* Matrix transpose by exchanging the rows with columns */
00089     /* row loop     */
00090     do
00091     {
00092       /* Loop Unrolling */
00093       blkCnt = nColumns >> 2;
00094 
00095       /* The pointer px is set to starting address of the column being processed */
00096       px = pOut + i;
00097 
00098       /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00099        ** a second loop below computes the remaining 1 to 3 samples. */
00100       while (blkCnt > 0U)        /* column loop */
00101       {
00102         /* Read and store the input element in the destination */
00103         *px = *pIn++;
00104 
00105         /* Update the pointer px to point to the next row of the transposed matrix */
00106         px += nRows;
00107 
00108         /* Read and store the input element in the destination */
00109         *px = *pIn++;
00110 
00111         /* Update the pointer px to point to the next row of the transposed matrix */
00112         px += nRows;
00113 
00114         /* Read and store the input element in the destination */
00115         *px = *pIn++;
00116 
00117         /* Update the pointer px to point to the next row of the transposed matrix */
00118         px += nRows;
00119 
00120         /* Read and store the input element in the destination */
00121         *px = *pIn++;
00122 
00123         /* Update the pointer px to point to the next row of the transposed matrix */
00124         px += nRows;
00125 
00126         /* Decrement the column loop counter */
00127         blkCnt--;
00128       }
00129 
00130       /* Perform matrix transpose for last 3 samples here. */
00131       blkCnt = nColumns % 0x4U;
00132 
00133       while (blkCnt > 0U)
00134       {
00135         /* Read and store the input element in the destination */
00136         *px = *pIn++;
00137 
00138         /* Update the pointer px to point to the next row of the transposed matrix */
00139         px += nRows;
00140 
00141         /* Decrement the column loop counter */
00142         blkCnt--;
00143       }
00144 
00145 #else
00146 
00147   /* Run the below code for Cortex-M0 */
00148 
00149   uint16_t col, i = 0U, row = nRows;             /* loop counters */
00150   arm_status status;                             /* status of matrix transpose  */
00151 
00152 
00153 #ifdef ARM_MATH_MATRIX_CHECK
00154 
00155   /* Check for matrix mismatch condition */
00156   if ((pSrc->numRows != pDst->numCols) || (pSrc->numCols != pDst->numRows))
00157   {
00158     /* Set status as ARM_MATH_SIZE_MISMATCH */
00159     status = ARM_MATH_SIZE_MISMATCH;
00160   }
00161   else
00162 #endif /*      #ifdef ARM_MATH_MATRIX_CHECK    */
00163 
00164   {
00165     /* Matrix transpose by exchanging the rows with columns */
00166     /* row loop     */
00167     do
00168     {
00169       /* The pointer px is set to starting address of the column being processed */
00170       px = pOut + i;
00171 
00172       /* Initialize column loop counter */
00173       col = nColumns;
00174 
00175       while (col > 0U)
00176       {
00177         /* Read and store the input element in the destination */
00178         *px = *pIn++;
00179 
00180         /* Update the pointer px to point to the next row of the transposed matrix */
00181         px += nRows;
00182 
00183         /* Decrement the column loop counter */
00184         col--;
00185       }
00186 
00187 #endif /* #if defined (ARM_MATH_DSP) */
00188 
00189       i++;
00190 
00191       /* Decrement the row loop counter */
00192       row--;
00193 
00194     } while (row > 0U);          /* row loop end  */
00195 
00196     /* Set status as ARM_MATH_SUCCESS */
00197     status = ARM_MATH_SUCCESS;
00198   }
00199 
00200   /* Return to application */
00201   return (status);
00202 }
00203 
00204 /**
00205  * @} end of MatrixTrans group
00206  */
00207