arm_cmplx_mult_cmplx_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_cmplx_mult_cmplx_q31.c 00009 * 00010 * Description: Q31 complex-by-complex 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 #include "arm_math.h" 00025 00050 void arm_cmplx_mult_cmplx_q31( 00051 q31_t * pSrcA, 00052 q31_t * pSrcB, 00053 q31_t * pDst, 00054 uint32_t numSamples) 00055 { 00056 q31_t a1, b1, c1, d1; /* Temporary variables to store real and imaginary values */ 00057 q31_t a2, b2, c2, d2; /* Temporary variables to store real and imaginary values */ 00058 q31_t out1, out2; 00059 q31_t mul1, mul2, mul3, mul4; 00060 uint32_t blkCnt; /* loop counters */ 00061 00062 /* loop Unrolling */ 00063 blkCnt = numSamples >> 2u; 00064 00065 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00066 ** a second loop below computes the remaining 1 to 3 samples. */ 00067 while(blkCnt > 0u) 00068 { 00069 /* C[2 * i] = A[2 * i] * B[2 * i] - A[2 * i + 1] * B[2 * i + 1]. */ 00070 /* C[2 * i + 1] = A[2 * i] * B[2 * i + 1] + A[2 * i + 1] * B[2 * i]. */ 00071 a1 = pSrcA[0]; 00072 b1 = pSrcA[1]; 00073 c1 = pSrcB[0]; 00074 d1 = pSrcB[1]; 00075 00076 mul1 = (q31_t)(((q63_t)a1 * c1) >> 32); 00077 mul2 = (q31_t)(((q63_t)b1 * d1) >> 32); 00078 mul3 = (q31_t)(((q63_t)a1 * d1) >> 32); 00079 mul4 = (q31_t)(((q63_t)b1 * c1) >> 32); 00080 00081 a2 = pSrcA[2]; 00082 b2 = pSrcA[3]; 00083 c2 = pSrcB[2]; 00084 a1 = (mul1 >> 1); 00085 d2 = pSrcB[3]; 00086 00087 b1 = (mul2 >> 1); 00088 c1 = (mul3 >> 1); 00089 d1 = (mul4 >> 1); 00090 00091 mul1 = (q31_t)(((q63_t)a2 * c2) >> 32); 00092 mul2 = (q31_t)(((q63_t)b2 * d2) >> 32); 00093 mul3 = (q31_t)(((q63_t)a2 * d2) >> 32); 00094 mul4 = (q31_t)(((q63_t)b2 * c2) >> 32); 00095 00096 out1 = a1 - b1; 00097 a1 = pSrcA[4]; 00098 out2 = c1 + d1; 00099 b1 = pSrcA[5]; 00100 c1 = pSrcB[4]; 00101 d1 = pSrcB[5]; 00102 00103 pDst[0] = out1; 00104 00105 a2 = (mul1 >> 1); 00106 pDst[1] = out2; 00107 b2 = (mul2 >> 1); 00108 c2 = (mul3 >> 1); 00109 d2 = (mul4 >> 1); 00110 00111 out1 = a2 - b2; 00112 a2 = pSrcA[6]; 00113 out2 = c2 + d2; 00114 00115 mul1 = (q31_t)(((q63_t)a1 * c1) >> 32); 00116 mul2 = (q31_t)(((q63_t)b1 * d1) >> 32); 00117 mul3 = (q31_t)(((q63_t)a1 * d1) >> 32); 00118 mul4 = (q31_t)(((q63_t)b1 * c1) >> 32); 00119 00120 b2 = pSrcA[7]; 00121 c2 = pSrcB[6]; 00122 d2 = pSrcB[7]; 00123 00124 pDst[2] = out1; 00125 a1 = (mul1 >> 1); 00126 pDst[3] = out2; 00127 b1 = (mul2 >> 1); 00128 c1 = (mul3 >> 1); 00129 d1 = (mul4 >> 1); 00130 00131 out1 = a1 - b1; 00132 out2 = c1 + d1; 00133 00134 mul1 = (q31_t)(((q63_t)a2 * c2) >> 32); 00135 mul2 = (q31_t)(((q63_t)b2 * d2) >> 32); 00136 mul3 = (q31_t)(((q63_t)a2 * d2) >> 32); 00137 mul4 = (q31_t)(((q63_t)b2 * c2) >> 32); 00138 00139 pDst[4] = out1; 00140 pSrcA += 8u; 00141 pDst[5] = out2; 00142 pSrcB += 8u; 00143 00144 a2 = (mul1 >> 1); 00145 b2 = (mul2 >> 1); 00146 c2 = (mul3 >> 1); 00147 d2 = (mul4 >> 1); 00148 00149 out1 = a2 - b2; 00150 out2 = c2 + d2; 00151 00152 pDst[6] = out1; 00153 pDst[7] = out2; 00154 00155 pDst += 8u; 00156 00157 /* Decrement the blockSize loop counter */ 00158 blkCnt--; 00159 } 00160 00161 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00162 ** No loop unrolling is used. */ 00163 blkCnt = numSamples % 0x4u; 00164 00165 while(blkCnt > 0u) 00166 { 00167 /* C[2 * i] = A[2 * i] * B[2 * i] - A[2 * i + 1] * B[2 * i + 1]. */ 00168 /* C[2 * i + 1] = A[2 * i] * B[2 * i + 1] + A[2 * i + 1] * B[2 * i]. */ 00169 a1 = *pSrcA++; 00170 b1 = *pSrcA++; 00171 c1 = *pSrcB++; 00172 d1 = *pSrcB++; 00173 00174 /* store the result in 3.29 format in the destination buffer. */ 00175 *pDst++ = (q31_t) ((((q63_t) a1 * c1) >> 33) - (((q63_t) b1 * d1) >> 33)); 00176 /* store the result in 3.29 format in the destination buffer. */ 00177 *pDst++ = (q31_t) ((((q63_t) a1 * d1) >> 33) + (((q63_t) b1 * c1) >> 33)); 00178 00179 /* Decrement the blockSize loop counter */ 00180 blkCnt--; 00181 } 00182 } 00183
