arm_dot_prod_q7.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        19. March 2015
00005 * $Revision:    V.1.4.5
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 * Redistribution and use in source and binary forms, with or without 
00015 * modification, are permitted provided that the following conditions
00016 * are met:
00017 *   - Redistributions of source code must retain the above copyright
00018 *     notice, this list of conditions and the following disclaimer.
00019 *   - Redistributions in binary form must reproduce the above copyright
00020 *     notice, this list of conditions and the following disclaimer in
00021 *     the documentation and/or other materials provided with the 
00022 *     distribution.
00023 *   - Neither the name of ARM LIMITED nor the names of its contributors
00024 *     may be used to endorse or promote products derived from this
00025 *     software without specific prior written permission.
00026 *
00027 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
00028 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
00029 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
00030 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 
00031 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
00032 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
00033 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
00034 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
00035 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
00036 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
00037 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
00038 * POSSIBILITY OF SUCH DAMAGE.
00039 * -------------------------------------------------------------------- */
00040 
00041 #include "arm_math.h"
00042 
00043 /**    
00044  * @ingroup groupMath    
00045  */
00046 
00047 /**    
00048  * @addtogroup dot_prod    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief Dot product of Q7 vectors.    
00054  * @param[in]       *pSrcA points to the first input vector    
00055  * @param[in]       *pSrcB points to the second input vector    
00056  * @param[in]       blockSize number of samples in each vector    
00057  * @param[out]      *result output result returned here    
00058  * @return none.    
00059  *    
00060  * <b>Scaling and Overflow Behavior:</b>    
00061  * \par    
00062  * The intermediate multiplications are in 1.7 x 1.7 = 2.14 format and these    
00063  * results are added to an accumulator in 18.14 format.    
00064  * Nonsaturating additions are used and there is no danger of wrap around as long as    
00065  * the vectors are less than 2^18 elements long.    
00066  * The return result is in 18.14 format.    
00067  */
00068 
00069 void arm_dot_prod_q7(
00070   q7_t * pSrcA,
00071   q7_t * pSrcB,
00072   uint32_t blockSize,
00073   q31_t * result)
00074 {
00075   uint32_t blkCnt;                               /* loop counter */
00076 
00077   q31_t sum = 0;                                 /* Temporary variables to store output */
00078 
00079 #ifndef ARM_MATH_CM0_FAMILY
00080 
00081 /* Run the below code for Cortex-M4 and Cortex-M3 */
00082 
00083   q31_t input1, input2;                          /* Temporary variables to store input */
00084   q31_t inA1, inA2, inB1, inB2;                  /* Temporary variables to store input */
00085 
00086 
00087 
00088   /*loop Unrolling */
00089   blkCnt = blockSize >> 2u;
00090 
00091   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00092    ** a second loop below computes the remaining 1 to 3 samples. */
00093   while(blkCnt > 0u)
00094   {
00095     /* read 4 samples at a time from sourceA */
00096     input1 = *__SIMD32(pSrcA)++;
00097     /* read 4 samples at a time from sourceB */
00098     input2 = *__SIMD32(pSrcB)++;
00099 
00100     /* extract two q7_t samples to q15_t samples */
00101     inA1 = __SXTB16(__ROR(input1, 8));
00102     /* extract reminaing two samples */
00103     inA2 = __SXTB16(input1);
00104     /* extract two q7_t samples to q15_t samples */
00105     inB1 = __SXTB16(__ROR(input2, 8));
00106     /* extract reminaing two samples */
00107     inB2 = __SXTB16(input2);
00108 
00109     /* multiply and accumulate two samples at a time */
00110     sum = __SMLAD(inA1, inB1, sum);
00111     sum = __SMLAD(inA2, inB2, sum);
00112 
00113     /* Decrement the loop counter */
00114     blkCnt--;
00115   }
00116 
00117   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.    
00118    ** No loop unrolling is used. */
00119   blkCnt = blockSize % 0x4u;
00120 
00121   while(blkCnt > 0u)
00122   {
00123     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
00124     /* Dot product and then store the results in a temporary buffer. */
00125     sum = __SMLAD(*pSrcA++, *pSrcB++, sum);
00126 
00127     /* Decrement the loop counter */
00128     blkCnt--;
00129   }
00130 
00131 #else
00132 
00133   /* Run the below code for Cortex-M0 */
00134 
00135 
00136 
00137   /* Initialize blkCnt with number of samples */
00138   blkCnt = blockSize;
00139 
00140   while(blkCnt > 0u)
00141   {
00142     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
00143     /* Dot product and then store the results in a temporary buffer. */
00144     sum += (q31_t) ((q15_t) * pSrcA++ * *pSrcB++);
00145 
00146     /* Decrement the loop counter */
00147     blkCnt--;
00148   }
00149 
00150 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
00151 
00152 
00153   /* Store the result in the destination buffer in 18.14 format */
00154   *result = sum;
00155 }
00156 
00157 /**    
00158  * @} end of dot_prod group    
00159  */