arm_cmplx_mag_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_mag_q31.c 00009 * 00010 * Description: Q31 complex magnitude 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 00048 void arm_cmplx_mag_q31( 00049 q31_t * pSrc, 00050 q31_t * pDst, 00051 uint32_t numSamples) 00052 { 00053 q31_t real1, real2, imag1, imag2; /* Temporary variables to hold input values */ 00054 q31_t out1, out2, out3, out4; /* Accumulators */ 00055 q63_t mul1, mul2, mul3, mul4; /* Temporary variables */ 00056 uint32_t blkCnt; /* loop counter */ 00057 00058 00059 /*loop Unrolling */ 00060 blkCnt = numSamples >> 2u; 00061 00062 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00063 ** a second loop below computes the remaining 1 to 3 samples. */ 00064 while(blkCnt > 0u) 00065 { 00066 /* read complex input from source buffer */ 00067 real1 = pSrc[0]; 00068 imag1 = pSrc[1]; 00069 real2 = pSrc[2]; 00070 imag2 = pSrc[3]; 00071 00072 /* calculate power of input values */ 00073 mul1 = (q63_t) real1 * real1; 00074 mul2 = (q63_t) imag1 * imag1; 00075 mul3 = (q63_t) real2 * real2; 00076 mul4 = (q63_t) imag2 * imag2; 00077 00078 /* get the result to 3.29 format */ 00079 out1 = (q31_t)(mul1 >> 33); 00080 out2 = (q31_t)(mul2 >> 33); 00081 out3 = (q31_t)(mul3 >> 33); 00082 out4 = (q31_t)(mul4 >> 33); 00083 00084 /* add real and imaginary accumulators */ 00085 out1 = out1 + out2; 00086 out3 = out3 + out4; 00087 00088 /* read complex input from source buffer */ 00089 real1 = pSrc[4]; 00090 imag1 = pSrc[5]; 00091 real2 = pSrc[6]; 00092 imag2 = pSrc[7]; 00093 00094 /* calculate square root */ 00095 arm_sqrt_q31(out1, &pDst[0]); 00096 00097 /* calculate power of input values */ 00098 mul1 = (q63_t) real1 * real1; 00099 00100 /* calculate square root */ 00101 arm_sqrt_q31(out3, &pDst[1]); 00102 00103 /* calculate power of input values */ 00104 mul2 = (q63_t) imag1 * imag1; 00105 mul3 = (q63_t) real2 * real2; 00106 mul4 = (q63_t) imag2 * imag2; 00107 00108 /* get the result to 3.29 format */ 00109 out1 = (q31_t)(mul1 >> 33); 00110 out2 = (q31_t)(mul2 >> 33); 00111 out3 = (q31_t)(mul3 >> 33); 00112 out4 = (q31_t)(mul4 >> 33); 00113 00114 /* add real and imaginary accumulators */ 00115 out1 = out1 + out2; 00116 out3 = out3 + out4; 00117 00118 /* calculate square root */ 00119 arm_sqrt_q31(out1, &pDst[2]); 00120 00121 /* increment destination by 8 to process next samples */ 00122 pSrc += 8u; 00123 00124 /* calculate square root */ 00125 arm_sqrt_q31(out3, &pDst[3]); 00126 00127 /* increment destination by 4 to process next samples */ 00128 pDst += 4u; 00129 00130 /* Decrement the loop counter */ 00131 blkCnt--; 00132 } 00133 00134 /* If the numSamples is not a multiple of 4, compute any remaining output samples here. 00135 ** No loop unrolling is used. */ 00136 blkCnt = numSamples % 0x4u; 00137 00138 while(blkCnt > 0u) 00139 { 00140 /* C[0] = sqrt(A[0] * A[0] + A[1] * A[1]) */ 00141 real1 = *pSrc++; 00142 imag1 = *pSrc++; 00143 out1 = (q31_t) (((q63_t) real1 * real1) >> 33); 00144 out2 = (q31_t) (((q63_t) imag1 * imag1) >> 33); 00145 /* store the result in 2.30 format in the destination buffer. */ 00146 arm_sqrt_q31(out1 + out2, pDst++); 00147 00148 /* Decrement the loop counter */ 00149 blkCnt--; 00150 } 00151 } 00152
