arm_dot_prod_q15.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_dot_prod_q15.c 00009 * 00010 * Description: Q15 dot product. 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 00056 void arm_dot_prod_q15( 00057 q15_t * pSrcA, 00058 q15_t * pSrcB, 00059 uint32_t blockSize, 00060 q63_t * result) 00061 { 00062 q63_t sum = 0; /* Temporary result storage */ 00063 uint32_t blkCnt; /* loop counter */ 00064 q31_t inA1, inA2, inB1, inB2; /* Temporary variables to store input data */ 00065 q31_t inA3, inA4, inB3, inB4; 00066 00067 /*loop Unrolling */ 00068 blkCnt = blockSize >> 3u; 00069 00070 /* First part of the processing with loop unrolling. Compute 8 outputs at a time. 00071 ** a second loop below computes the remaining 1 to 7 samples. */ 00072 while(blkCnt > 0u) 00073 { 00074 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00075 /* Calculate dot product and then store the result in a temporary buffer. */ 00076 /* read two samples at a time from soruceA buffer */ 00077 inA1 = _SIMD32_OFFSET(pSrcA); 00078 /* read two samples at a time from soruceB buffer */ 00079 inB1 = _SIMD32_OFFSET(pSrcB); 00080 /* read two samples at a time from soruceA buffer */ 00081 inA2 = _SIMD32_OFFSET(pSrcA+2); 00082 00083 /* multiply and accumulate two samples at a time */ 00084 sum = __SMLALD(inA1, inB1, sum); 00085 00086 /* read two samples at a time from soruceB buffer */ 00087 inB2 = _SIMD32_OFFSET(pSrcB+2); 00088 /* read two samples at a time from soruceA buffer */ 00089 inA3 = _SIMD32_OFFSET(pSrcA+4); 00090 /* read two samples at a time from soruceB buffer */ 00091 inB3 = _SIMD32_OFFSET(pSrcB+4); 00092 00093 /* multiply and accumulate two samples at a time */ 00094 sum = __SMLALD(inA2, inB2, sum); 00095 00096 /* read two samples at a time from soruceA buffer */ 00097 inA4 = _SIMD32_OFFSET(pSrcA+6); 00098 /* read two samples at a time from soruceB buffer */ 00099 inB4 = _SIMD32_OFFSET(pSrcB+6); 00100 00101 /* increment source A buffer by 8 */ 00102 pSrcA += 8u; 00103 /* increment sourceB buffer by 8 */ 00104 pSrcB += 8u; 00105 00106 /* multiply and accumulate two samples at a time */ 00107 sum = __SMLALD(inA3, inB3, sum); 00108 sum = __SMLALD(inA4, inB4, sum); 00109 00110 /* Decrement the loop counter */ 00111 blkCnt--; 00112 } 00113 /* If the blockSize is not a multiple of 8, compute any remaining output samples here. 00114 ** No loop unrolling is used. */ 00115 blkCnt = blockSize % 0x8u; 00116 00117 while(blkCnt > 0u) 00118 { 00119 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00120 /* Calculate dot product and then store the results in a temporary buffer. */ 00121 sum = __SMLALD(*pSrcA++, *pSrcB++, sum); 00122 00123 /* Decrement the loop counter */ 00124 blkCnt--; 00125 } 00126 00127 /* Store the result in the destination buffer in 34.30 format */ 00128 *result = sum; 00129 } 00130
