k60-keil

K60-Keil版本（下载安装MDK4.23）

        in1 = (q15_t) x2;
        in2 = (q15_t) x3;

        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* y[srcBLen - 3] and y[srcBLen - 4] are packed */
        in1 = (q15_t) c0;
        in2 = (q15_t) c1;

        input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* acc0 += x[2] * y[srcBLen - 3] + x[3] * y[srcBLen - 4]  */
        acc0 = __SMLAD(input1, input2, acc0);

        /* x[3] and x[4] are packed */
        in1 = (q15_t) x3;
        in2 = (q15_t) x0;

        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* acc1 += x[3] * y[srcBLen - 3] + x[4] * y[srcBLen - 4]  */
        acc1 = __SMLAD(input1, input2, acc1);

        /* x[4] and x[5] are packed */
        in1 = (q15_t) x0;
        in2 = (q15_t) x1;

        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* acc2 += x[4] * y[srcBLen - 3] + x[5] * y[srcBLen - 4]  */
        acc2 = __SMLAD(input1, input2, acc2);

        /* Read x[6] sample */
        x2 = *(px++);

        /* x[5] and x[6] are packed */
        in1 = (q15_t) x1;
        in2 = (q15_t) x2;

        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* acc3 += x[5] * y[srcBLen - 3] + x[6] * y[srcBLen - 4]  */
        acc3 = __SMLAD(input1, input2, acc3);

      } while(--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)
      {
        /* Read y[srcBLen - 5] sample */
        c0 = *(py--);

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

        /* Perform the multiply-accumulates */
        /* acc0 +=  x[4] * y[srcBLen - 5] */
        acc0 += ((q15_t) x0 * c0);
        /* acc1 +=  x[5] * y[srcBLen - 5] */
        acc1 += ((q15_t) x1 * c0);
        /* acc2 +=  x[6] * y[srcBLen - 5] */
        acc2 += ((q15_t) x2 * c0);
        /* acc3 +=  x[7] * y[srcBLen - 5] */
        acc3 += ((q15_t) x3 * c0);

        /* Reuse the present samples for the next MAC */
        x0 = x1;
        x1 = x2;
        x2 = x3;

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


      /* Store the result in the accumulator in the destination buffer. */
      *pOut++ = (q7_t) (__SSAT(acc0 >> 7u, 8));
      *pOut++ = (q7_t) (__SSAT(acc1 >> 7u, 8));
      *pOut++ = (q7_t) (__SSAT(acc2 >> 7u, 8));
      *pOut++ = (q7_t) (__SSAT(acc3 >> 7u, 8));

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

      /* 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 = 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)
      {

        /* Reading two inputs of SrcA buffer and packing */
        in1 = (q15_t) * px++;
        in2 = (q15_t) * px++;
        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* Reading two inputs of SrcB buffer and packing */
        in1 = (q15_t) * py--;
        in2 = (q15_t) * py--;
        input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* Perform the multiply-accumulates */
        sum = __SMLAD(input1, input2, sum);

        /* Reading two inputs of SrcA buffer and packing */
        in1 = (q15_t) * px++;
        in2 = (q15_t) * px++;
        input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* Reading two inputs of SrcB buffer and packing */
        in1 = (q15_t) * py--;
        in2 = (q15_t) * py--;
        input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

        /* Perform the multiply-accumulates */
        sum = __SMLAD(input1, input2, sum);

        /* 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 += ((q15_t) * px++ * *py--);

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

      /* Store the result in the accumulator in the destination buffer. */
      *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));

      /* 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 = 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 += ((q15_t) * px++ * *py--);

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

      /* Store the result in the accumulator in the destination buffer. */
      *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));

      /* 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 blockSize3 variable holds the number of MAC operations performed */

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

  /* Working pointer of inputB */
  pSrc2 = pIn2 + (srcBLen - 1u);
  py = pSrc2;

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

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

    /* Apply loop unrolling and compute 4 MACs simultaneously. */
    k = blockSize3 >> 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)
    {
      /* Reading two inputs, x[srcALen - srcBLen + 1] and x[srcALen - srcBLen + 2] of SrcA buffer and packing */
      in1 = (q15_t) * px++;
      in2 = (q15_t) * px++;
      input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

      /* Reading two inputs, y[srcBLen - 1] and y[srcBLen - 2] of SrcB buffer and packing */
      in1 = (q15_t) * py--;
      in2 = (q15_t) * py--;
      input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

      /* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */
      /* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */
      sum = __SMLAD(input1, input2, sum);

      /* Reading two inputs, x[srcALen - srcBLen + 3] and x[srcALen - srcBLen + 4] of SrcA buffer and packing */
      in1 = (q15_t) * px++;
      in2 = (q15_t) * px++;
      input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

      /* Reading two inputs, y[srcBLen - 3] and y[srcBLen - 4] of SrcB buffer and packing */
      in1 = (q15_t) * py--;
      in2 = (q15_t) * py--;
      input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);

      /* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */
      /* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */
      sum = __SMLAD(input1, input2, sum);

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

    /* If the blockSize3 is not a multiple of 4, compute any remaining MACs here.   
     ** No loop unrolling is used. */
    k = blockSize3 % 0x4u;

    while(k > 0u)
    {
      /* Perform the multiply-accumulates */
      sum += ((q15_t) * px++ * *py--);

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

    /* Store the result in the accumulator in the destination buffer. */
    *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));

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

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

#else

  /* Run the below code for Cortex-M0 */

  q7_t *pIn1 = pSrcA;                            /* input pointer */
  q7_t *pIn2 = pSrcB;                            /* coefficient pointer */
  q31_t sum;                                     /* Accumulator */
  uint32_t i, j;                                 /* loop counter */

  /* Loop to calculate output of convolution for output length number of times */
  for (i = 0; i < (srcALen + srcBLen - 1); i++)
  {
    /* Initialize sum with zero to carry on MAC operations */
    sum = 0;

    /* Loop to perform MAC operations according to convolution equation */
    for (j = 0; j <= i; j++)
    {
      /* Check the array limitations */
      if(((i - j) < srcBLen) && (j < srcALen))
      {
        /* z[i] += x[i-j] * y[j] */
        sum += (q15_t) pIn1[j] * (pIn2[i - j]);
      }
    }

    /* Store the output in the destination buffer */
    pDst[i] = (q7_t) __SSAT((sum >> 7u), 8u);
  }

#endif /*   #ifndef ARM_MATH_CM0        */

}

/**   
 * @} end of Conv group   
 */