arm_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_mult_q31.c 00009 * 00010 * Description: Q31 vector 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 00050 void arm_mult_q31( 00051 q31_t * pSrcA, 00052 q31_t * pSrcB, 00053 q31_t * pDst, 00054 uint32_t blockSize) 00055 { 00056 uint32_t blkCnt; /* loop counters */ 00057 q31_t inA1, inA2, inA3, inA4; /* temporary input variables */ 00058 q31_t inB1, inB2, inB3, inB4; /* temporary input variables */ 00059 q31_t out1, out2, out3, out4; /* temporary output variables */ 00060 00061 /* loop Unrolling */ 00062 blkCnt = blockSize >> 2u; 00063 00064 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00065 ** a second loop below computes the remaining 1 to 3 samples. */ 00066 while(blkCnt > 0u) 00067 { 00068 /* read sample from sourceA */ 00069 inA1 = *pSrcA; 00070 /* read sample from sourceB */ 00071 inB1 = *pSrcB; 00072 /* read sample from sourceA */ 00073 inA2 = *(pSrcA + 1); 00074 /* read sample from sourceB */ 00075 inB2 = *(pSrcB + 1); 00076 00077 /* sourceA * sourceB */ 00078 inA1 = ((q63_t) inA1 * inB1) >> 32; 00079 00080 /* read sample from sourceA */ 00081 inA3 = *(pSrcA + 2); 00082 00083 /* sourceA * sourceB */ 00084 inA2 = ((q63_t) inA2 * inB2) >> 32; 00085 00086 /* read sample from sourceB */ 00087 inB3 = *(pSrcB + 2); 00088 /* read sample from sourceA */ 00089 inA4 = *(pSrcA + 3); 00090 00091 /* sourceA * sourceB */ 00092 inA3 = ((q63_t) inA3 * inB3) >> 32; 00093 00094 /* read sample from sourceB */ 00095 inB4 = *(pSrcB + 3); 00096 00097 /* sourceA * sourceB */ 00098 inA4 = ((q63_t) inA4 * inB4) >> 32; 00099 00100 00101 /* saturate the result */ 00102 #ifdef CCS 00103 out1 = __SSATA(inA1, 0, 31); 00104 out2 = __SSATA(inA2, 0, 31); 00105 out3 = __SSATA(inA3, 0, 31); 00106 out4 = __SSATA(inA4, 0, 31); 00107 #else 00108 out1 = __SSAT(inA1, 31); 00109 out2 = __SSAT(inA2, 31); 00110 out3 = __SSAT(inA3, 31); 00111 out4 = __SSAT(inA4, 31); 00112 #endif // #ifdef CCS 00113 00114 /* get the output to 1.31 format */ 00115 out1 = out1 << 1u; 00116 out2 = out2 << 1u; 00117 00118 /* store result to destination buffer */ 00119 *pDst = out1; 00120 00121 /* get the output to 1.31 format */ 00122 out3 = out3 << 1u; 00123 00124 /* store result to destination buffer */ 00125 *(pDst + 1) = out2; 00126 00127 /* get the output to 1.31 format */ 00128 out4 = out4 << 1u; 00129 00130 /* store result to destination buffer */ 00131 *(pDst + 2) = out3; 00132 00133 /* incremnet sourceA buffer by 4 to process next samples */ 00134 pSrcA += 4u; 00135 00136 /* store result to destination buffer */ 00137 *(pDst + 3) = out4; 00138 00139 /* incremnet sourceB buffer by 4 to process next samples */ 00140 pSrcB += 4u; 00141 /* incremnet destination buffer by 4 to process next samples */ 00142 pDst += 4u; 00143 00144 /* Decrement the blockSize loop counter */ 00145 blkCnt--; 00146 } 00147 00148 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00149 ** No loop unrolling is used. */ 00150 blkCnt = blockSize % 0x4u; 00151 00152 while(blkCnt > 0u) 00153 { 00154 /* C = A * B */ 00155 /* Multiply the inputs and then store the results in the destination buffer. */ 00156 *pDst++ = (q31_t) clip_q63_to_q31(((q63_t) (*pSrcA++) * (*pSrcB++)) >> 31); 00157 00158 /* Decrement the blockSize loop counter */ 00159 blkCnt--; 00160 } 00161 } 00162
