arm_mat_cmplx_mult_q31.c
Go to the documentation of this file.
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2011 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. December 2011 00005 * $Revision: V2.0.0 00006 * 00007 * Project: Cortex-R DSP Library 00008 * Title: arm_mat_cmplx_mult_q31.c 00009 * 00010 * Description: Floating-point matrix multiplication. 00011 * 00012 * Target Processor: Cortex-R4/R5 00013 * 00014 * Version 1.0.0 2011/03/08 00015 * Alpha release. 00016 * 00017 * Version 1.0.1 2011/09/30 00018 * Beta release. 00019 * 00020 * Version 2.0.0 2011/12/15 00021 * Final release. 00022 * 00023 * 00024 * -------------------------------------------------------------------- */ 00025 #include "arm_math.h" 00026 00060 arm_status arm_mat_cmplx_mult_q31( 00061 const arm_matrix_instance_q31 * pSrcA, 00062 const arm_matrix_instance_q31 * pSrcB, 00063 arm_matrix_instance_q31 * pDst) 00064 { 00065 q31_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */ 00066 q31_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */ 00067 q31_t *pInA = pSrcA->pData; /* input data matrix pointer A */ 00068 q31_t *pOut = pDst->pData; /* output data matrix pointer */ 00069 q31_t *px; /* Temporary output data matrix pointer */ 00070 uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */ 00071 uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */ 00072 uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */ 00073 q63_t sumReal1, sumImag1; /* accumulator */ 00074 q31_t a0, b0, c0, d0; 00075 q31_t a1, b1, c1, d1; 00076 00077 00078 /* Run the below code for Cortex-M4 and Cortex-M3 */ 00079 00080 uint16_t col, i = 0u, j, row = numRowsA, colCnt; /* loop counters */ 00081 arm_status status; /* status of matrix multiplication */ 00082 00083 #ifdef ARM_MATH_MATRIX_CHECK 00084 00085 00086 /* Check for matrix mismatch condition */ 00087 if((pSrcA->numCols != pSrcB->numRows) || 00088 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols)) 00089 { 00090 00091 /* Set status as ARM_MATH_SIZE_MISMATCH */ 00092 status = ARM_MATH_SIZE_MISMATCH; 00093 } 00094 else 00095 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */ 00096 00097 { 00098 /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */ 00099 /* row loop */ 00100 do 00101 { 00102 /* Output pointer is set to starting address of the row being processed */ 00103 px = pOut + 2 * i; 00104 00105 /* For every row wise process, the column loop counter is to be initiated */ 00106 col = numColsB; 00107 00108 /* For every row wise process, the pIn2 pointer is set 00109 ** to the starting address of the pSrcB data */ 00110 pIn2 = pSrcB->pData; 00111 00112 j = 0u; 00113 00114 /* column loop */ 00115 do 00116 { 00117 /* Set the variable sum, that acts as accumulator, to zero */ 00118 sumReal1 = 0.0; 00119 sumImag1 = 0.0; 00120 00121 /* Initiate the pointer pIn1 to point to the starting address of the column being processed */ 00122 pIn1 = pInA; 00123 00124 /* Apply loop unrolling and compute 4 MACs simultaneously. */ 00125 colCnt = numColsA >> 2; 00126 00127 /* matrix multiplication */ 00128 while(colCnt > 0u) 00129 { 00130 00131 /* Reading real part of complex matrix A */ 00132 a0 = *pIn1; 00133 00134 /* Reading real part of complex matrix B */ 00135 c0 = *pIn2; 00136 00137 /* Reading imaginary part of complex matrix A */ 00138 b0 = *(pIn1 + 1u); 00139 00140 /* Reading imaginary part of complex matrix B */ 00141 d0 = *(pIn2 + 1u); 00142 00143 /* Multiply and Accumlates */ 00144 sumReal1 += (q63_t)a0 * c0; 00145 sumImag1 += (q63_t)b0 * c0; 00146 00147 /* update pointers */ 00148 pIn1 += 2u; 00149 pIn2 += 2* numColsB; 00150 00151 /* Multiply and Accumlates */ 00152 sumReal1 -= (q63_t)b0 * d0; 00153 sumImag1 += (q63_t)a0 * d0; 00154 00155 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */ 00156 00157 /* read real and imag values from pSrcA and pSrcB buffer */ 00158 a1 = *pIn1; 00159 c1 = *pIn2; 00160 b1 = *(pIn1 + 1u); 00161 d1 = *(pIn2 + 1u); 00162 00163 /* Multiply and Accumlates */ 00164 sumReal1 += (q63_t)a1 * c1; 00165 sumImag1 += (q63_t)b1 * c1; 00166 00167 /* update pointers */ 00168 pIn1 += 2u; 00169 pIn2 += 2* numColsB; 00170 00171 /* Multiply and Accumlates */ 00172 sumReal1 -= (q63_t)b1 * d1; 00173 sumImag1 += (q63_t) a1 * d1; 00174 00175 a0 = *pIn1; 00176 c0 = *pIn2; 00177 00178 b0 = *(pIn1 + 1u); 00179 d0 = *(pIn2 + 1u); 00180 00181 /* Multiply and Accumlates */ 00182 sumReal1 += (q63_t)a0 * c0; 00183 sumImag1 += (q63_t)b0 * c0; 00184 00185 /* update pointers */ 00186 pIn1 += 2u; 00187 pIn2 += 2* numColsB; 00188 00189 /* Multiply and Accumlates */ 00190 sumReal1 -= (q63_t)b0 * d0; 00191 sumImag1 += (q63_t)a0 * d0; 00192 00193 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */ 00194 00195 a1 = *pIn1; 00196 c1 = *pIn2; 00197 00198 b1 = *(pIn1 + 1u); 00199 d1 = *(pIn2 + 1u); 00200 00201 /* Multiply and Accumlates */ 00202 sumReal1 += (q63_t)a1 * c1; 00203 sumImag1 += (q63_t)b1 * c1; 00204 00205 /* update pointers */ 00206 pIn1 += 2u; 00207 pIn2 += 2* numColsB; 00208 00209 /* Multiply and Accumlates */ 00210 sumReal1 -= (q63_t)b1 * d1; 00211 sumImag1 += (q63_t)a1 * d1; 00212 00213 /* Decrement the loop count */ 00214 colCnt--; 00215 } 00216 00217 /* If the columns of pSrcA is not a multiple of 4, compute any remaining MACs here. 00218 ** No loop unrolling is used. */ 00219 colCnt = numColsA % 0x4u; 00220 00221 while(colCnt > 0u) 00222 { 00223 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */ 00224 a1 = *pIn1; 00225 c1 = *pIn2; 00226 00227 b1 = *(pIn1 + 1u); 00228 d1 = *(pIn2 + 1u); 00229 00230 /* Multiply and Accumlates */ 00231 sumReal1 += (q63_t)a1 * c1; 00232 sumImag1 += (q63_t)b1 * c1; 00233 00234 /* update pointers */ 00235 pIn1 += 2u; 00236 pIn2 += 2* numColsB; 00237 00238 /* Multiply and Accumlates */ 00239 sumReal1 -= (q63_t)b1 * d1; 00240 sumImag1 += (q63_t)a1 * d1; 00241 00242 /* Decrement the loop counter */ 00243 colCnt--; 00244 } 00245 00246 /* Store the result in the destination buffer */ 00247 #ifdef CCS 00248 *px++ = (q31_t)__SSATA((sumReal1 >> 31), 0, 31); 00249 *px++ = (q31_t)__SSATA((sumImag1 >> 31), 0, 31); 00250 #else 00251 *px++ = (q31_t)__SSAT((sumReal1 >> 31), 31); 00252 *px++ = (q31_t)__SSAT((sumImag1 >> 31), 31); 00253 #endif 00254 /* Update the pointer pIn2 to point to the starting address of the next column */ 00255 j++; 00256 pIn2 = pSrcB->pData + 2u * j; 00257 00258 /* Decrement the column loop counter */ 00259 col--; 00260 00261 } while(col > 0u); 00262 00263 /* Update the pointer pInA to point to the starting address of the next row */ 00264 i = i + numColsB; 00265 pInA = pInA + 2 * numColsA; 00266 00267 /* Decrement the row loop counter */ 00268 row--; 00269 00270 } while(row > 0u); 00271 00272 /* Set status as ARM_MATH_SUCCESS */ 00273 status = ARM_MATH_SUCCESS; 00274 } 00275 00276 /* Return to application */ 00277 return (status); 00278 } 00279
