k60-keil

K60-Keil版本（下载安装MDK4.23）

        /* acc1 +=  x[3] * y[2] + x[4] * y[3] */
        acc1 = __SMLAD(x3, c0, acc1);

        /* Read x[4], x[5] */
        x0 = *(q31_t *) (px++);

        /* Read x[5], x[6] */
        x1 = *(q31_t *) (px++);

        /* acc2 +=  x[4] * y[2] + x[5] * y[3] */
        acc2 = __SMLAD(x0, c0, acc2);

        /* acc3 +=  x[5] * y[2] + x[6] * y[3] */
        acc3 = __SMLAD(x1, c0, acc3);

      } while(--k);

      /* For the next MAC operations, SIMD is not used   
       * So, the 16 bit pointer if inputB, py is updated */
      py = (q15_t *) (pb);

      /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.   
       ** No loop unrolling is used. */
      k = srcBLen % 0x4u;

      if(k == 1u)
      {
        /* Read y[4] */
        c0 = *py;
#ifdef  ARM_MATH_BIG_ENDIAN

        c0 = c0 << 16u;

#else

        c0 = c0 & 0x0000FFFF;

#endif /*      #ifdef  ARM_MATH_BIG_ENDIAN     */

        /* Read x[7] */
        x3 = *(q31_t *) px++;

        /* Perform the multiply-accumulates */
        acc0 = __SMLAD(x0, c0, acc0);
        acc1 = __SMLAD(x1, c0, acc1);
        acc2 = __SMLADX(x1, c0, acc2);
        acc3 = __SMLADX(x3, c0, acc3);
      }

      if(k == 2u)
      {
        /* Read y[4], y[5] */
        c0 = *(pb);

        /* Read x[7], x[8] */
        x3 = *(q31_t *) px++;

        /* Read x[9] */
        x2 = *(q31_t *) px++;

        /* Perform the multiply-accumulates */
        acc0 = __SMLAD(x0, c0, acc0);
        acc1 = __SMLAD(x1, c0, acc1);
        acc2 = __SMLAD(x3, c0, acc2);
        acc3 = __SMLAD(x2, c0, acc3);
      }

      if(k == 3u)
      {
        /* Read y[4], y[5] */
        c0 = *pb++;

        /* Read x[7], x[8] */
        x3 = *(q31_t *) px++;

        /* Read x[9] */
        x2 = *(q31_t *) px++;

        /* Perform the multiply-accumulates */
        acc0 = __SMLAD(x0, c0, acc0);
        acc1 = __SMLAD(x1, c0, acc1);
        acc2 = __SMLAD(x3, c0, acc2);
        acc3 = __SMLAD(x2, c0, acc3);

        /* Read y[6] */
#ifdef  ARM_MATH_BIG_ENDIAN
        c0 = (*pb);
        c0 = c0 & 0xFFFF0000;

#else
        c0 = (q15_t) (*pb);
        c0 = c0 & 0x0000FFFF;

#endif /*      #ifdef  ARM_MATH_BIG_ENDIAN     */

        /* Read x[10] */
        x3 = *(q31_t *) px++;

        /* Perform the multiply-accumulates */
        acc0 = __SMLADX(x1, c0, acc0);
        acc1 = __SMLAD(x2, c0, acc1);
        acc2 = __SMLADX(x2, c0, acc2);
        acc3 = __SMLADX(x3, c0, acc3);
      }

      /* Store the result in the accumulator in the destination buffer. */
      *pOut = (q15_t) (acc0 >> 15);
      /* Destination pointer is updated according to the address modifier, inc */
      pOut += inc;

      *pOut = (q15_t) (acc1 >> 15);
      pOut += inc;

      *pOut = (q15_t) (acc2 >> 15);
      pOut += inc;

      *pOut = (q15_t) (acc3 >> 15);
      pOut += inc;

      /* Increment the pointer pIn1 index, count by 1 */
      count += 4u;

      /* Update the inputA and inputB pointers for next MAC calculation */
      px = pIn1 + count;
      py = pIn2;
      pb = (q31_t *) (py);


      /* Decrement the loop counter */
      blkCnt--;
    }

    /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here.   
     ** No loop unrolling is used. */
    blkCnt = blockSize2 % 0x4u;

    while(blkCnt > 0u)
    {
      /* Accumulator is made zero for every iteration */
      sum = 0;

      /* Apply loop unrolling and compute 4 MACs simultaneously. */
      k = srcBLen >> 2u;

      /* First part of the processing with loop unrolling.  Compute 4 MACs at a time.   
       ** a second loop below computes MACs for the remaining 1 to 3 samples. */
      while(k > 0u)
      {
        /* Perform the multiply-accumulates */
        sum += ((q31_t) * px++ * *py++);
        sum += ((q31_t) * px++ * *py++);
        sum += ((q31_t) * px++ * *py++);
        sum += ((q31_t) * px++ * *py++);

        /* Decrement the loop counter */
        k--;
      }

      /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.   
       ** No loop unrolling is used. */
      k = srcBLen % 0x4u;

      while(k > 0u)
      {
        /* Perform the multiply-accumulates */
        sum += ((q31_t) * px++ * *py++);

        /* Decrement the loop counter */
        k--;
      }

      /* Store the result in the accumulator in the destination buffer. */
      *pOut = (q15_t) (sum >> 15);
      /* Destination pointer is updated according to the address modifier, inc */
      pOut += inc;

      /* Increment the pointer pIn1 index, count by 1 */
      count++;

      /* Update the inputA and inputB pointers for next MAC calculation */
      px = pIn1 + count;
      py = pIn2;

      /* Decrement the loop counter */
      blkCnt--;
    }
  }
  else
  {
    /* If the srcBLen is not a multiple of 4,   
     * the blockSize2 loop cannot be unrolled by 4 */
    blkCnt = blockSize2;

    while(blkCnt > 0u)
    {
      /* Accumulator is made zero for every iteration */
      sum = 0;

      /* Loop over srcBLen */
      k = srcBLen;

      while(k > 0u)
      {
        /* Perform the multiply-accumulate */
        sum += ((q31_t) * px++ * *py++);

        /* Decrement the loop counter */
        k--;
      }

      /* Store the result in the accumulator in the destination buffer. */
      *pOut = (q15_t) (sum >> 15);
      /* Destination pointer is updated according to the address modifier, inc */
      pOut += inc;

      /* Increment the MAC count */
      count++;

      /* Update the inputA and inputB pointers for next MAC calculation */
      px = pIn1 + count;
      py = pIn2;

      /* Decrement the loop counter */
      blkCnt--;
    }
  }

  /* --------------------------   
   * Initializations of stage3   
   * -------------------------*/

  /* sum += x[srcALen-srcBLen+1] * y[0] + x[srcALen-srcBLen+2] * y[1] +...+ x[srcALen-1] * y[srcBLen-1]   
   * sum += x[srcALen-srcBLen+2] * y[0] + x[srcALen-srcBLen+3] * y[1] +...+ x[srcALen-1] * y[srcBLen-1]   
   * ....   
   * sum +=  x[srcALen-2] * y[0] + x[srcALen-1] * y[1]   
   * sum +=  x[srcALen-1] * y[0]   
   */

  /* In this stage the MAC operations are decreased by 1 for every iteration.   
     The count variable holds the number of MAC operations performed */
  count = srcBLen - 1u;

  /* Working pointer of inputA */
  pSrc1 = (pIn1 + srcALen) - (srcBLen - 1u);
  px = pSrc1;

  /* Working pointer of inputB */
  py = pIn2;

  /* -------------------   
   * Stage3 process   
   * ------------------*/

  while(blockSize3 > 0u)
  {
    /* Accumulator is made zero for every iteration */
    sum = 0;

    /* Apply loop unrolling and compute 4 MACs simultaneously. */
    k = count >> 2u;

    /* First part of the processing with loop unrolling.  Compute 4 MACs at a time.   
     ** a second loop below computes MACs for the remaining 1 to 3 samples. */
    while(k > 0u)
    {
      /* Perform the multiply-accumulates */
      /* sum += x[srcALen - srcBLen + 4] * y[3] , sum += x[srcALen - srcBLen + 3] * y[2] */
      sum = __SMLAD(*__SIMD32(px)++, *__SIMD32(py)++, sum);
      /* sum += x[srcALen - srcBLen + 2] * y[1] , sum += x[srcALen - srcBLen + 1] * y[0] */
      sum = __SMLAD(*__SIMD32(px)++, *__SIMD32(py)++, sum);

      /* Decrement the loop counter */
      k--;
    }

    /* If the count is not a multiple of 4, compute any remaining MACs here.   
     ** No loop unrolling is used. */
    k = count % 0x4u;

    while(k > 0u)
    {
      /* Perform the multiply-accumulates */
      sum = __SMLAD(*px++, *py++, sum);

      /* Decrement the loop counter */
      k--;
    }

    /* Store the result in the accumulator in the destination buffer. */
    *pOut = (q15_t) (sum >> 15);
    /* Destination pointer is updated according to the address modifier, inc */
    pOut += inc;

    /* Update the inputA and inputB pointers for next MAC calculation */
    px = ++pSrc1;
    py = pIn2;

    /* Decrement the MAC count */
    count--;

    /* Decrement the loop counter */
    blockSize3--;
  }

}

/**   
 * @} end of Corr group   
 */