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_f32.c 00009 * 00010 * Description: Floating-point 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_f32( 00063 float32_t * pSrcA, 00064 float32_t * pSrcB, 00065 uint32_t blockSize, 00066 float32_t * result) 00067 { 00068 float32_t sum = 0.0f; /* 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 /*loop Unrolling */ 00076 blkCnt = blockSize >> 2u; 00077 00078 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00079 ** a second loop below computes the remaining 1 to 3 samples. */ 00080 while(blkCnt > 0u) 00081 { 00082 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00083 /* Calculate dot product and then store the result in a temporary buffer */ 00084 sum += (*pSrcA++) * (*pSrcB++); 00085 sum += (*pSrcA++) * (*pSrcB++); 00086 sum += (*pSrcA++) * (*pSrcB++); 00087 sum += (*pSrcA++) * (*pSrcB++); 00088 00089 /* Decrement the loop counter */ 00090 blkCnt--; 00091 } 00092 00093 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00094 ** No loop unrolling is used. */ 00095 blkCnt = blockSize % 0x4u; 00096 00097 #else 00098 00099 /* Run the below code for Cortex-M0 */ 00100 00101 /* Initialize blkCnt with number of samples */ 00102 blkCnt = blockSize; 00103 00104 #endif /* #ifndef ARM_MATH_CM0 */ 00105 00106 00107 while(blkCnt > 0u) 00108 { 00109 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00110 /* Calculate dot product and then store the result in a temporary buffer. */ 00111 sum += (*pSrcA++) * (*pSrcB++); 00112 00113 /* Decrement the loop counter */ 00114 blkCnt--; 00115 } 00116 /* Store the result back in the destination buffer */ 00117 *result = sum; 00118 } 00119