1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_mult_q7.c
  4   * Description:  Q7 vector multiplication
  5   *
  6   * $Date:        23 April 2021
  7   * $Revision:    V1.9.0
  8   *
  9   * Target Processor: Cortex-M and Cortex-A cores
 10   * -------------------------------------------------------------------- */
 11  /*
 12   * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
 13   *
 14   * SPDX-License-Identifier: Apache-2.0
 15   *
 16   * Licensed under the Apache License, Version 2.0 (the License); you may
 17   * not use this file except in compliance with the License.
 18   * You may obtain a copy of the License at
 19   *
 20   * www.apache.org/licenses/LICENSE-2.0
 21   *
 22   * Unless required by applicable law or agreed to in writing, software
 23   * distributed under the License is distributed on an AS IS BASIS, WITHOUT
 24   * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 25   * See the License for the specific language governing permissions and
 26   * limitations under the License.
 27   */
 28  
 29  #include "dsp/basic_math_functions.h"
 30  
 31  /**
 32    @ingroup groupMath
 33   */
 34  
 35  /**
 36    @addtogroup BasicMult
 37    @{
 38   */
 39  
 40  /**
 41    @brief         Q7 vector multiplication
 42    @param[in]     pSrcA      points to the first input vector
 43    @param[in]     pSrcB      points to the second input vector
 44    @param[out]    pDst       points to the output vector
 45    @param[in]     blockSize  number of samples in each vector
 46    @return        none
 47  
 48    @par           Scaling and Overflow Behavior
 49                     The function uses saturating arithmetic.
 50                     Results outside of the allowable Q7 range [0x80 0x7F] are saturated.
 51   */
 52  #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
 53  
 54  #include "arm_helium_utils.h"
 55  
 56  void arm_mult_q7(
 57      const q7_t * pSrcA,
 58      const q7_t * pSrcB,
 59      q7_t * pDst,
 60      uint32_t blockSize)
 61  {
 62      uint32_t  blkCnt;           /* loop counters */
 63      q7x16_t vecA, vecB;
 64  
 65      /* Compute 16 outputs at a time */
 66      blkCnt = blockSize >> 4;
 67      while (blkCnt > 0U)
 68      {
 69          /*
 70           * C = A * B
 71           * Multiply the inputs and then store the results in the destination buffer.
 72           */
 73          vecA = vld1q(pSrcA);
 74          vecB = vld1q(pSrcB);
 75          vst1q(pDst, vqdmulhq(vecA, vecB));
 76          /*
 77           * Decrement the blockSize loop counter
 78           */
 79          blkCnt--;
 80          /*
 81           * advance vector source and destination pointers
 82           */
 83          pSrcA  += 16;
 84          pSrcB  += 16;
 85          pDst   += 16;
 86      }
 87      /*
 88       * tail
 89       */
 90      blkCnt = blockSize & 0xF;
 91      if (blkCnt > 0U)
 92      {
 93          mve_pred16_t p0 = vctp8q(blkCnt);
 94          vecA = vld1q(pSrcA);
 95          vecB = vld1q(pSrcB);
 96          vstrbq_p(pDst, vqdmulhq(vecA, vecB), p0);
 97      }
 98  }
 99  
100  #else
101  void arm_mult_q7(
102    const q7_t * pSrcA,
103    const q7_t * pSrcB,
104          q7_t * pDst,
105          uint32_t blockSize)
106  {
107          uint32_t blkCnt;                               /* Loop counter */
108  
109  #if defined (ARM_MATH_LOOPUNROLL)
110  
111  #if defined (ARM_MATH_DSP)
112    q7_t out1, out2, out3, out4;                   /* Temporary output variables */
113  #endif
114  
115    /* Loop unrolling: Compute 4 outputs at a time */
116    blkCnt = blockSize >> 2U;
117  
118    while (blkCnt > 0U)
119    {
120      /* C = A * B */
121  
122  #if defined (ARM_MATH_DSP)
123      /* Multiply inputs and store results in temporary variables */
124      out1 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
125      out2 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
126      out3 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
127      out4 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
128  
129      /* Pack and store result in destination buffer (in single write) */
130      write_q7x4_ia (&pDst, __PACKq7(out1, out2, out3, out4));
131  #else
132      *pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
133      *pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
134      *pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
135      *pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
136  #endif
137  
138      /* Decrement loop counter */
139      blkCnt--;
140    }
141  
142    /* Loop unrolling: Compute remaining outputs */
143    blkCnt = blockSize % 0x4U;
144  
145  #else
146  
147    /* Initialize blkCnt with number of samples */
148    blkCnt = blockSize;
149  
150  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
151  
152    while (blkCnt > 0U)
153    {
154      /* C = A * B */
155  
156      /* Multiply input and store result in destination buffer. */
157      *pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
158  
159      /* Decrement loop counter */
160      blkCnt--;
161    }
162  
163  }
164  #endif /* defined(ARM_MATH_MVEI) */
165  
166  /**
167    @} end of BasicMult group
168   */