arm_mult_q31.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_mult_q31.c    
00009 *    
00010 * Description:  Q31 vector multiplication.    
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 BasicMult    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief Q31 vector multiplication.    
00054  * @param[in]       *pSrcA points to the first input vector    
00055  * @param[in]       *pSrcB points to the second input vector    
00056  * @param[out]      *pDst points to the output vector    
00057  * @param[in]       blockSize number of samples in each vector    
00058  * @return none.    
00059  *    
00060  * <b>Scaling and Overflow Behavior:</b>    
00061  * \par    
00062  * The function uses saturating arithmetic.    
00063  * Results outside of the allowable Q31 range[0x80000000 0x7FFFFFFF] will be saturated.    
00064  */
00065 
00066 void arm_mult_q31(
00067   q31_t * pSrcA,
00068   q31_t * pSrcB,
00069   q31_t * pDst,
00070   uint32_t blockSize)
00071 {
00072   uint32_t blkCnt;                               /* loop counters */
00073 
00074 #ifndef ARM_MATH_CM0_FAMILY
00075 
00076 /* Run the below code for Cortex-M4 and Cortex-M3 */
00077   q31_t inA1, inA2, inA3, inA4;                  /* temporary input variables */
00078   q31_t inB1, inB2, inB3, inB4;                  /* temporary input variables */
00079   q31_t out1, out2, out3, out4;                  /* temporary output variables */
00080 
00081   /* loop Unrolling */
00082   blkCnt = blockSize >> 2u;
00083 
00084   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00085    ** a second loop below computes the remaining 1 to 3 samples. */
00086   while(blkCnt > 0u)
00087   {
00088     /* C = A * B */
00089     /* Multiply the inputs and then store the results in the destination buffer. */
00090     inA1 = *pSrcA++;
00091     inA2 = *pSrcA++;
00092     inA3 = *pSrcA++;
00093     inA4 = *pSrcA++;
00094     inB1 = *pSrcB++;
00095     inB2 = *pSrcB++;
00096     inB3 = *pSrcB++;
00097     inB4 = *pSrcB++;
00098 
00099     out1 = ((q63_t) inA1 * inB1) >> 32;
00100     out2 = ((q63_t) inA2 * inB2) >> 32;
00101     out3 = ((q63_t) inA3 * inB3) >> 32;
00102     out4 = ((q63_t) inA4 * inB4) >> 32;
00103 
00104     out1 = __SSAT(out1, 31);
00105     out2 = __SSAT(out2, 31);
00106     out3 = __SSAT(out3, 31);
00107     out4 = __SSAT(out4, 31);
00108 
00109     *pDst++ = out1 << 1u;
00110     *pDst++ = out2 << 1u;
00111     *pDst++ = out3 << 1u;
00112     *pDst++ = out4 << 1u;
00113 
00114     /* Decrement the blockSize loop counter */
00115     blkCnt--;
00116   }
00117 
00118   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.    
00119    ** No loop unrolling is used. */
00120   blkCnt = blockSize % 0x4u;
00121   
00122   while(blkCnt > 0u)
00123   {
00124     /* C = A * B */
00125     /* Multiply the inputs and then store the results in the destination buffer. */
00126     inA1 = *pSrcA++;
00127     inB1 = *pSrcB++;
00128     out1 = ((q63_t) inA1 * inB1) >> 32;
00129     out1 = __SSAT(out1, 31);
00130     *pDst++ = out1 << 1u;
00131 
00132     /* Decrement the blockSize loop counter */
00133     blkCnt--;
00134   }
00135 
00136 #else
00137 
00138   /* Run the below code for Cortex-M0 */
00139 
00140   /* Initialize blkCnt with number of samples */
00141   blkCnt = blockSize;
00142 
00143 
00144   while(blkCnt > 0u)
00145   {
00146     /* C = A * B */
00147     /* Multiply the inputs and then store the results in the destination buffer. */
00148     *pDst++ =
00149       (q31_t) clip_q63_to_q31(((q63_t) (*pSrcA++) * (*pSrcB++)) >> 31);
00150 
00151     /* Decrement the blockSize loop counter */
00152     blkCnt--;
00153   }
00154   
00155 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
00156 }
00157 
00158 /**    
00159  * @} end of BasicMult group    
00160  */