00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2010 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. July 2011 00005 * $Revision: V1.0.10 00006 * 00007 * Project: CMSIS DSP Library 00008 * Title: arm_dot_prod_q31.c 00009 * 00010 * Description: Q31 dot product. 00011 * 00012 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 00013 * 00014 * Version 1.0.10 2011/7/15 00015 * Big Endian support added and Merged M0 and M3/M4 Source code. 00016 * 00017 * Version 1.0.3 2010/11/29 00018 * Re-organized the CMSIS folders and updated documentation. 00019 * 00020 * Version 1.0.2 2010/11/11 00021 * Documentation updated. 00022 * 00023 * Version 1.0.1 2010/10/05 00024 * Production release and review comments incorporated. 00025 * 00026 * Version 1.0.0 2010/09/20 00027 * Production release and review comments incorporated. 00028 * 00029 * Version 0.0.7 2010/06/10 00030 * Misra-C changes done 00031 * -------------------------------------------------------------------- */ 00032 00033 #include "arm_math.h" 00034 00062 void arm_dot_prod_q31( 00063 q31_t * pSrcA, 00064 q31_t * pSrcB, 00065 uint32_t blockSize, 00066 q63_t * result) 00067 { 00068 q63_t sum = 0; /* Temporary result storage */ 00069 uint32_t blkCnt; /* loop counter */ 00070 00071 00072 #ifndef ARM_MATH_CM0 00073 00074 /* Run the below code for Cortex-M4 and Cortex-M3 */ 00075 00076 /*loop Unrolling */ 00077 blkCnt = blockSize >> 2u; 00078 00079 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00080 ** a second loop below computes the remaining 1 to 3 samples. */ 00081 while(blkCnt > 0u) 00082 { 00083 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00084 /* Calculate dot product and then store the result in a temporary buffer. */ 00085 sum += ((q63_t) * pSrcA++ * *pSrcB++) >> 14u; 00086 sum += ((q63_t) * pSrcA++ * *pSrcB++) >> 14u; 00087 sum += ((q63_t) * pSrcA++ * *pSrcB++) >> 14u; 00088 sum += ((q63_t) * pSrcA++ * *pSrcB++) >> 14u; 00089 00090 /* Decrement the loop counter */ 00091 blkCnt--; 00092 } 00093 00094 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00095 ** No loop unrolling is used. */ 00096 blkCnt = blockSize % 0x4u; 00097 00098 #else 00099 00100 /* Run the below code for Cortex-M0 */ 00101 00102 /* Initialize blkCnt with number of samples */ 00103 blkCnt = blockSize; 00104 00105 #endif /* #ifndef ARM_MATH_CM0 */ 00106 00107 00108 while(blkCnt > 0u) 00109 { 00110 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00111 /* Calculate dot product and then store the result in a temporary buffer. */ 00112 sum += ((q63_t) * pSrcA++ * *pSrcB++) >> 14u; 00113 00114 /* Decrement the loop counter */ 00115 blkCnt--; 00116 } 00117 00118 /* Store the result in the destination buffer in 16.48 format */ 00119 *result = sum; 00120 } 00121