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_q7.c 00009 * 00010 * Description: Q7 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 00061 void arm_dot_prod_q7( 00062 q7_t * pSrcA, 00063 q7_t * pSrcB, 00064 uint32_t blockSize, 00065 q31_t * result) 00066 { 00067 uint32_t blkCnt; /* loop counter */ 00068 00069 q31_t sum = 0; /* Temporary variables to store output */ 00070 00071 #ifndef ARM_MATH_CM0 00072 00073 /* Run the below code for Cortex-M4 and Cortex-M3 */ 00074 00075 q31_t input1, input2; /* Temporary variables to store input */ 00076 q15_t in1, in2; /* Temporary variables to store input */ 00077 00078 00079 00080 /*loop Unrolling */ 00081 blkCnt = blockSize >> 2u; 00082 00083 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00084 ** a second loop below computes the remaining 1 to 3 samples. */ 00085 while(blkCnt > 0u) 00086 { 00087 /* Reading two inputs of SrcA buffer and packing */ 00088 in1 = (q15_t) * pSrcA++; 00089 in2 = (q15_t) * pSrcA++; 00090 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); 00091 00092 /* Reading two inputs of SrcB buffer and packing */ 00093 in1 = (q15_t) * pSrcB++; 00094 in2 = (q15_t) * pSrcB++; 00095 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); 00096 00097 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00098 /* Perform Dot product of 2 packed inputs using SMLALD and store the result in a temporary variable. */ 00099 sum = __SMLAD(input1, input2, sum); 00100 00101 /* Reading two inputs of SrcA buffer and packing */ 00102 in1 = (q15_t) * pSrcA++; 00103 in2 = (q15_t) * pSrcA++; 00104 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); 00105 00106 /* Reading two inputs of SrcB buffer and packing */ 00107 in1 = (q15_t) * pSrcB++; 00108 in2 = (q15_t) * pSrcB++; 00109 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); 00110 00111 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00112 /* Perform Dot product of 2 packed inputs using SMLALD and store the result in a temporary variable. */ 00113 sum = __SMLAD(input1, input2, sum); 00114 00115 00116 00117 /* Decrement the loop counter */ 00118 blkCnt--; 00119 } 00120 00121 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00122 ** No loop unrolling is used. */ 00123 blkCnt = blockSize % 0x4u; 00124 00125 while(blkCnt > 0u) 00126 { 00127 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00128 /* Dot product and then store the results in a temporary buffer. */ 00129 sum = __SMLAD(*pSrcA++, *pSrcB++, sum); 00130 00131 /* Decrement the loop counter */ 00132 blkCnt--; 00133 } 00134 00135 #else 00136 00137 /* Run the below code for Cortex-M0 */ 00138 00139 00140 00141 /* Initialize blkCnt with number of samples */ 00142 blkCnt = blockSize; 00143 00144 while(blkCnt > 0u) 00145 { 00146 /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */ 00147 /* Dot product and then store the results in a temporary buffer. */ 00148 sum += (q31_t) ((q15_t) * pSrcA++ * *pSrcB++); 00149 00150 /* Decrement the loop counter */ 00151 blkCnt--; 00152 } 00153 00154 #endif /* #ifndef ARM_MATH_CM0 */ 00155 00156 00157 /* Store the result in the destination buffer in 18.14 format */ 00158 *result = sum; 00159 } 00160