arm_mult_q31.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_mult_q31.c
00004  * Description:  Q31 vector multiplication
00005  *
00006  * $Date:        27. January 2017
00007  * $Revision:    V.1.5.1
00008  *
00009  * Target Processor: Cortex-M cores
00010  * -------------------------------------------------------------------- */
00011 /*
00012  * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
00013  *
00014  * SPDX-License-Identifier: Apache-2.0
00015  *
00016  * Licensed under the Apache License, Version 2.0 (the License); you may
00017  * not use this file except in compliance with the License.
00018  * You may obtain a copy of the License at
00019  *
00020  * www.apache.org/licenses/LICENSE-2.0
00021  *
00022  * Unless required by applicable law or agreed to in writing, software
00023  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
00024  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
00025  * See the License for the specific language governing permissions and
00026  * limitations under the License.
00027  */
00028 
00029 #include "arm_math.h"
00030 
00031 /**
00032  * @ingroup groupMath
00033  */
00034 
00035 /**
00036  * @addtogroup BasicMult
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Q31 vector multiplication.
00042  * @param[in]       *pSrcA points to the first input vector
00043  * @param[in]       *pSrcB points to the second input vector
00044  * @param[out]      *pDst points to the output vector
00045  * @param[in]       blockSize number of samples in each vector
00046  * @return none.
00047  *
00048  * <b>Scaling and Overflow Behavior:</b>
00049  * \par
00050  * The function uses saturating arithmetic.
00051  * Results outside of the allowable Q31 range[0x80000000 0x7FFFFFFF] will be saturated.
00052  */
00053 
00054 void arm_mult_q31(
00055   q31_t * pSrcA,
00056   q31_t * pSrcB,
00057   q31_t * pDst,
00058   uint32_t blockSize)
00059 {
00060   uint32_t blkCnt;                               /* loop counters */
00061 
00062 #if defined (ARM_MATH_DSP)
00063 
00064 /* Run the below code for Cortex-M4 and Cortex-M3 */
00065   q31_t inA1, inA2, inA3, inA4;                  /* temporary input variables */
00066   q31_t inB1, inB2, inB3, inB4;                  /* temporary input variables */
00067   q31_t out1, out2, out3, out4;                  /* temporary output variables */
00068 
00069   /* loop Unrolling */
00070   blkCnt = blockSize >> 2U;
00071 
00072   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00073    ** a second loop below computes the remaining 1 to 3 samples. */
00074   while (blkCnt > 0U)
00075   {
00076     /* C = A * B */
00077     /* Multiply the inputs and then store the results in the destination buffer. */
00078     inA1 = *pSrcA++;
00079     inA2 = *pSrcA++;
00080     inA3 = *pSrcA++;
00081     inA4 = *pSrcA++;
00082     inB1 = *pSrcB++;
00083     inB2 = *pSrcB++;
00084     inB3 = *pSrcB++;
00085     inB4 = *pSrcB++;
00086 
00087     out1 = ((q63_t) inA1 * inB1) >> 32;
00088     out2 = ((q63_t) inA2 * inB2) >> 32;
00089     out3 = ((q63_t) inA3 * inB3) >> 32;
00090     out4 = ((q63_t) inA4 * inB4) >> 32;
00091 
00092     out1 = __SSAT(out1, 31);
00093     out2 = __SSAT(out2, 31);
00094     out3 = __SSAT(out3, 31);
00095     out4 = __SSAT(out4, 31);
00096 
00097     *pDst++ = out1 << 1U;
00098     *pDst++ = out2 << 1U;
00099     *pDst++ = out3 << 1U;
00100     *pDst++ = out4 << 1U;
00101 
00102     /* Decrement the blockSize loop counter */
00103     blkCnt--;
00104   }
00105 
00106   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00107    ** No loop unrolling is used. */
00108   blkCnt = blockSize % 0x4U;
00109 
00110   while (blkCnt > 0U)
00111   {
00112     /* C = A * B */
00113     /* Multiply the inputs and then store the results in the destination buffer. */
00114     inA1 = *pSrcA++;
00115     inB1 = *pSrcB++;
00116     out1 = ((q63_t) inA1 * inB1) >> 32;
00117     out1 = __SSAT(out1, 31);
00118     *pDst++ = out1 << 1U;
00119 
00120     /* Decrement the blockSize loop counter */
00121     blkCnt--;
00122   }
00123 
00124 #else
00125 
00126   /* Run the below code for Cortex-M0 */
00127 
00128   /* Initialize blkCnt with number of samples */
00129   blkCnt = blockSize;
00130 
00131 
00132   while (blkCnt > 0U)
00133   {
00134     /* C = A * B */
00135     /* Multiply the inputs and then store the results in the destination buffer. */
00136     *pDst++ =
00137       (q31_t) clip_q63_to_q31(((q63_t) (*pSrcA++) * (*pSrcB++)) >> 31);
00138 
00139     /* Decrement the blockSize loop counter */
00140     blkCnt--;
00141   }
00142 
00143 #endif /* #if defined (ARM_MATH_DSP) */
00144 }
00145 
00146 /**
00147  * @} end of BasicMult group
00148  */
00149