k60-keil

K60-Keil版本（下载安装MDK4.23）

          /* acc2 +=  x[4] * y[srcBLen - 3] + x[5] * y[srcBLen - 4] */
          acc2 = __SMLADX(x0, c0, acc2);

          /* acc3 +=  x[5] * y[srcBLen - 3] + x[6] * y[srcBLen - 4] */
          acc3 = __SMLADX(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;
        py = py + 1;

        /* 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[srcBLen - 5] */
          c0 = *(py);
#ifdef  ARM_MATH_BIG_ENDIAN

          c0 = c0 << 16;

#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[srcBLen - 5], y[srcBLen - 6] */
          c0 = *(pb);

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

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

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

        if(k == 3u)
        {
          /* Read y[srcBLen - 5], y[srcBLen - 6] */
          c0 = *pb--;

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

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

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

          /* Read y[srcBLen - 7] */
#ifdef  ARM_MATH_BIG_ENDIAN

          c0 = (*pb);
          c0 = (c0) << 16;

#else

          c0 = (q15_t) (*pb >> 16);

#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 results in the accumulators in the destination buffer. */
#ifndef ARM_MATH_BIG_ENDIAN

        *__SIMD32(pOut)++ = __PKHBT(acc0 >> 15, acc1 >> 15, 16);
        *__SIMD32(pOut)++ = __PKHBT(acc2 >> 15, acc3 >> 15, 16);

#else

        *__SIMD32(pOut)++ = __PKHBT(acc1 >> 15, acc0 >> 15, 16);
        *__SIMD32(pOut)++ = __PKHBT(acc3 >> 15, acc2 >> 15, 16);

#endif /*      #ifndef  ARM_MATH_BIG_ENDIAN    */

        /* Update the inputA and inputB pointers for next MAC calculation */
        px = pIn1 + (count * 4u);
        py = pSrc2;
        pb = (q31_t *) (py - 1);

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

        /* 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 = (uint32_t) 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);

        /* Update the inputA and inputB pointers for next MAC calculation */
        px = pIn1 + count;
        py = pSrc2;

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

        /* Decrement the loop counter */
        blkCnt--;
      }
    }
    else
    {
      /* If the srcBLen is not a multiple of 4,   
       * the blockSize2 loop cannot be unrolled by 4 */
      blkCnt = (uint32_t) blockSize2;

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

        /* srcBLen number of MACS should be performed */
        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);

        /* Update the inputA and inputB pointers for next MAC calculation */
        px = pIn1 + count;
        py = pSrc2;

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

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


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

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

    /* 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 */
    pSrc2 = pIn2 + (srcBLen - 1u);
    pIn2 = pSrc2 - 1u;
    py = pIn2;

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

    /* For loop unrolling by 4, this stage is divided into two. */
    /* First part of this stage computes the MAC operations greater than 4 */
    /* Second part of this stage computes the MAC operations less than or equal to 4 */

    /* The first part of the stage starts here */
    j = count >> 2u;

    while((j > 0u) && (blockSize3 > 0))
    {
      /* 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)
      {
        /* x[srcALen - srcBLen + 1], x[srcALen - srcBLen + 2] are multiplied   
         * with y[srcBLen - 1], y[srcBLen - 2] respectively */
        sum = __SMLADX(*__SIMD32(px)++, *__SIMD32(py)--, sum);
        /* x[srcALen - srcBLen + 3], x[srcALen - srcBLen + 4] are multiplied   
         * with y[srcBLen - 3], y[srcBLen - 4] respectively */
        sum = __SMLADX(*__SIMD32(px)++, *__SIMD32(py)--, sum);

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

      /* For the next MAC operations, the pointer py is used without SIMD   
       * So, py is incremented by 1 */
      py = py + 1u;

      /* 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)
      {
        /* sum += x[srcALen - srcBLen + 5] * y[srcBLen - 5] */
        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);

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

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

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

      j--;
    }

    /* The second part of the stage starts here */
    /* SIMD is not used for the next MAC operations,   
     * so pointer py is updated to read only one sample at a time */
    py = py + 1u;

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

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

      while(k > 0u)
      {
        /* Perform the multiply-accumulates */
        /* sum +=  x[srcALen-1] * y[srcBLen-1] */
        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);

      /* Update the inputA and inputB pointers for next MAC calculation */
      px = ++pSrc1;
      py = pSrc2;

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

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

    /* set status as ARM_MATH_SUCCESS */
    status = ARM_MATH_SUCCESS;
  }

  /* Return to application */
  return (status);

}

/**   
 * @} end of PartialConv group   
 */