arm_mat_cmplx_mult_f32.c

00001 /* ----------------------------------------------------------------------      
00002 * Copyright (C) 2010-2014 ARM Limited. All rights reserved. 
00003 *      
00004 * $Date:        19. March 2015
00005 * $Revision:    V.1.4.5
00006 *      
00007 * Project:      CMSIS DSP Library 
00008 * Title:        arm_mat_cmplx_mult_f32.c      
00009 *      
00010 * Description:  Floating-point matrix multiplication.      
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 #include "arm_math.h"
00041 
00042 /**      
00043  * @ingroup groupMatrix      
00044  */
00045 
00046 /**      
00047  * @defgroup CmplxMatrixMult  Complex Matrix Multiplication     
00048  *     
00049  * Complex Matrix multiplication is only defined if the number of columns of the      
00050  * first matrix equals the number of rows of the second matrix.      
00051  * Multiplying an <code>M x N</code> matrix with an <code>N x P</code> matrix results      
00052  * in an <code>M x P</code> matrix.      
00053  * When matrix size checking is enabled, the functions check: (1) that the inner dimensions of      
00054  * <code>pSrcA</code> and <code>pSrcB</code> are equal; and (2) that the size of the output      
00055  * matrix equals the outer dimensions of <code>pSrcA</code> and <code>pSrcB</code>.      
00056  */
00057 
00058 
00059 /**      
00060  * @addtogroup CmplxMatrixMult      
00061  * @{      
00062  */
00063 
00064 /**      
00065  * @brief Floating-point Complex matrix multiplication.      
00066  * @param[in]       *pSrcA points to the first input complex matrix structure      
00067  * @param[in]       *pSrcB points to the second input complex matrix structure      
00068  * @param[out]      *pDst points to output complex matrix structure      
00069  * @return          The function returns either      
00070  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.      
00071  */
00072 
00073 arm_status arm_mat_cmplx_mult_f32(
00074   const arm_matrix_instance_f32 * pSrcA,
00075   const arm_matrix_instance_f32 * pSrcB,
00076   arm_matrix_instance_f32 * pDst)
00077 {
00078   float32_t *pIn1 = pSrcA->pData;                /* input data matrix pointer A */
00079   float32_t *pIn2 = pSrcB->pData;                /* input data matrix pointer B */
00080   float32_t *pInA = pSrcA->pData;                /* input data matrix pointer A  */
00081   float32_t *pOut = pDst->pData;                 /* output data matrix pointer */
00082   float32_t *px;                                 /* Temporary output data matrix pointer */
00083   uint16_t numRowsA = pSrcA->numRows;            /* number of rows of input matrix A */
00084   uint16_t numColsB = pSrcB->numCols;            /* number of columns of input matrix B */
00085   uint16_t numColsA = pSrcA->numCols;            /* number of columns of input matrix A */
00086   float32_t sumReal1, sumImag1;                  /* accumulator */
00087   float32_t a0, b0, c0, d0;
00088   float32_t a1, b1, c1, d1;
00089   float32_t sumReal2, sumImag2;                  /* accumulator */
00090 
00091 
00092   /* Run the below code for Cortex-M4 and Cortex-M3 */
00093 
00094   uint16_t col, i = 0u, j, row = numRowsA, colCnt;      /* loop counters */
00095   arm_status status;                             /* status of matrix multiplication */
00096 
00097 #ifdef ARM_MATH_MATRIX_CHECK
00098 
00099 
00100   /* Check for matrix mismatch condition */
00101   if((pSrcA->numCols != pSrcB->numRows) ||
00102      (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
00103   {
00104 
00105     /* Set status as ARM_MATH_SIZE_MISMATCH */
00106     status = ARM_MATH_SIZE_MISMATCH;
00107   }
00108   else
00109 #endif /*      #ifdef ARM_MATH_MATRIX_CHECK    */
00110 
00111   {
00112     /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
00113     /* row loop */
00114     do
00115     {
00116       /* Output pointer is set to starting address of the row being processed */
00117       px = pOut + 2 * i;
00118 
00119       /* For every row wise process, the column loop counter is to be initiated */
00120       col = numColsB;
00121 
00122       /* For every row wise process, the pIn2 pointer is set      
00123        ** to the starting address of the pSrcB data */
00124       pIn2 = pSrcB->pData;
00125 
00126       j = 0u;
00127 
00128       /* column loop */
00129       do
00130       {
00131         /* Set the variable sum, that acts as accumulator, to zero */
00132         sumReal1 = 0.0f;
00133         sumImag1 = 0.0f;
00134 
00135         sumReal2 = 0.0f;
00136         sumImag2 = 0.0f;
00137 
00138         /* Initiate the pointer pIn1 to point to the starting address of the column being processed */
00139         pIn1 = pInA;
00140 
00141         /* Apply loop unrolling and compute 4 MACs simultaneously. */
00142         colCnt = numColsA >> 2;
00143 
00144         /* matrix multiplication        */
00145         while(colCnt > 0u)
00146         {
00147 
00148           /* Reading real part of complex matrix A */
00149           a0 = *pIn1;
00150 
00151           /* Reading real part of complex matrix B */
00152           c0 = *pIn2;
00153 
00154           /* Reading imaginary part of complex matrix A */
00155           b0 = *(pIn1 + 1u);
00156 
00157           /* Reading imaginary part of complex matrix B */
00158           d0 = *(pIn2 + 1u);
00159 
00160           sumReal1 += a0 * c0;
00161           sumImag1 += b0 * c0;
00162 
00163           pIn1 += 2u;
00164           pIn2 += 2 * numColsB;
00165 
00166           sumReal2 -= b0 * d0;
00167           sumImag2 += a0 * d0;
00168 
00169           /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
00170 
00171           a1 = *pIn1;
00172           c1 = *pIn2;
00173 
00174           b1 = *(pIn1 + 1u);
00175           d1 = *(pIn2 + 1u);
00176 
00177           sumReal1 += a1 * c1;
00178           sumImag1 += b1 * c1;
00179 
00180           pIn1 += 2u;
00181           pIn2 += 2 * numColsB;
00182 
00183           sumReal2 -= b1 * d1;
00184           sumImag2 += a1 * d1;
00185 
00186           a0 = *pIn1;
00187           c0 = *pIn2;
00188 
00189           b0 = *(pIn1 + 1u);
00190           d0 = *(pIn2 + 1u);
00191 
00192           sumReal1 += a0 * c0;
00193           sumImag1 += b0 * c0;
00194 
00195           pIn1 += 2u;
00196           pIn2 += 2 * numColsB;
00197 
00198           sumReal2 -= b0 * d0;
00199           sumImag2 += a0 * d0;
00200 
00201           /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
00202 
00203           a1 = *pIn1;
00204           c1 = *pIn2;
00205 
00206           b1 = *(pIn1 + 1u);
00207           d1 = *(pIn2 + 1u);
00208 
00209           sumReal1 += a1 * c1;
00210           sumImag1 += b1 * c1;
00211 
00212           pIn1 += 2u;
00213           pIn2 += 2 * numColsB;
00214 
00215           sumReal2 -= b1 * d1;
00216           sumImag2 += a1 * d1;
00217 
00218           /* Decrement the loop count */
00219           colCnt--;
00220         }
00221 
00222         /* If the columns of pSrcA is not a multiple of 4, compute any remaining MACs here.      
00223          ** No loop unrolling is used. */
00224         colCnt = numColsA % 0x4u;
00225 
00226         while(colCnt > 0u)
00227         {
00228           /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
00229           a1 = *pIn1;
00230           c1 = *pIn2;
00231 
00232           b1 = *(pIn1 + 1u);
00233           d1 = *(pIn2 + 1u);
00234 
00235           sumReal1 += a1 * c1;
00236           sumImag1 += b1 * c1;
00237 
00238           pIn1 += 2u;
00239           pIn2 += 2 * numColsB;
00240 
00241           sumReal2 -= b1 * d1;
00242           sumImag2 += a1 * d1;
00243 
00244           /* Decrement the loop counter */
00245           colCnt--;
00246         }
00247 
00248         sumReal1 += sumReal2;
00249         sumImag1 += sumImag2;
00250 
00251         /* Store the result in the destination buffer */
00252         *px++ = sumReal1;
00253         *px++ = sumImag1;
00254 
00255         /* Update the pointer pIn2 to point to the  starting address of the next column */
00256         j++;
00257         pIn2 = pSrcB->pData + 2u * j;
00258 
00259         /* Decrement the column loop counter */
00260         col--;
00261 
00262       } while(col > 0u);
00263 
00264       /* Update the pointer pInA to point to the  starting address of the next row */
00265       i = i + numColsB;
00266       pInA = pInA + 2 * numColsA;
00267 
00268       /* Decrement the row loop counter */
00269       row--;
00270 
00271     } while(row > 0u);
00272 
00273     /* Set status as ARM_MATH_SUCCESS */
00274     status = ARM_MATH_SUCCESS;
00275   }
00276 
00277   /* Return to application */
00278   return (status);
00279 }
00280 
00281 /**      
00282  * @} end of MatrixMult group      
00283  */