arm_fir_q31.c

Go to the documentation of this file.

/* ----------------------------------------------------------------------      
* Copyright (C) 2011 ARM Limited. All rights reserved. 
*      
* $Date:        15. December 2011   
* $Revision:    V2.0.0  
*      
* Project:      Cortex-R DSP Library 
* Title:        arm_fir_q31.c      
*      
* Description:  Q31 FIR filter processing function.      
*      
* Target Processor:          Cortex-R4/R5
*
* Version 1.0.0 2011/03/08
*     Alpha release.
*
* Version 1.0.1 2011/09/30
*     Beta release.
*
* Version 2.0.0 2011/12/15
*     Final release. 
* 
* -------------------------------------------------------------------- */     
     
#include "arm_math.h"     
     
void arm_fir_q31(  
  const arm_fir_instance_q31 * S,  
  q31_t * pSrc,  
  q31_t * pDst,  
  uint32_t blockSize)  
{  
  q31_t *pState = S->pState;                     /* State pointer */  
  q31_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */  
  q31_t *pStateCurnt;                            /* Points to the current sample of the state */  
  q31_t x0, x1, x2;                              /* Temporary variables to hold state */  
  q31_t c0;                                      /* Temporary variable to hold coefficient value */  
  q31_t *px;                                     /* Temporary pointer for state */  
  q31_t *pb;                                     /* Temporary pointer for coefficient buffer */  
  q63_t acc0, acc1, acc2;                        /* Accumulators */  
  uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */  
  uint32_t i, tapCnt, blkCnt, tapCntN3;          /* Loop counters */  
  
  /* S->pState points to state array which contains previous frame (numTaps - 1) samples */  
  /* pStateCurnt points to the location where the new input data should be written */  
  pStateCurnt = &(S->pState[(numTaps - 1u)]);  
  
  /* Apply loop unrolling and compute 4 output values simultaneously.   
   * The variables acc0 ... acc3 hold output values that are being computed:   
   *   
   *    acc0 =  b[numTaps-1] * x[n-numTaps-1] + b[numTaps-2] * x[n-numTaps-2] + b[numTaps-3] * x[n-numTaps-3] +...+ b[0] * x[0]   
   *    acc1 =  b[numTaps-1] * x[n-numTaps] +   b[numTaps-2] * x[n-numTaps-1] + b[numTaps-3] * x[n-numTaps-2] +...+ b[0] * x[1]   
   *    acc2 =  b[numTaps-1] * x[n-numTaps+1] + b[numTaps-2] * x[n-numTaps] +   b[numTaps-3] * x[n-numTaps-1] +...+ b[0] * x[2]   
   *    acc3 =  b[numTaps-1] * x[n-numTaps+2] + b[numTaps-2] * x[n-numTaps+1] + b[numTaps-3] * x[n-numTaps]   +...+ b[0] * x[3]   
   */  
  blkCnt = blockSize / 3;  
  blockSize = blockSize - (3*blkCnt);  
  
  tapCnt = numTaps / 3;  
  tapCntN3 = numTaps - (3*tapCnt);  
    
  /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.   
   ** a second loop below computes the remaining 1 to 3 samples. */  
  while(blkCnt > 0u)  
  {  
    /* Copy four new input samples into the state buffer */  
    *pStateCurnt++ = *pSrc++;  
    *pStateCurnt++ = *pSrc++;  
    *pStateCurnt++ = *pSrc++;  
      
    /* Set all accumulators to zero */  
    acc0 = 0;  
    acc1 = 0;  
    acc2 = 0;  
      
    /* Initialize state pointer */  
    px = pState;  
  
    /* Initialize coefficient pointer */  
    pb = pCoeffs;  
  
    /* Read the first three samples from the state buffer:   
     *  x[n-numTaps], x[n-numTaps-1] */  
    x0 = *(px++);  
    x1 = *(px++);  
      
    /* Loop unrolling.  Process 4 taps at a time. */  
    i = tapCnt;  
  
    while(i > 0u)  
    {  
      /* Read the b[numTaps] coefficient */     
      c0 = *pb;  
     
      /* Read x[n-numTaps-2] sample */     
      x2 = *(px++);  
  
      /* Perform the multiply-accumulates */     
      acc0 += ((q63_t) x0 * c0);  
      acc1 += ((q63_t) x1 * c0);  
      acc2 += ((q63_t) x2 * c0);  
        
      c0 = *(pb + 1u);  
      x0 = *(px++);  
  
      /* Perform the multiply-accumulates */     
      acc0 += ((q63_t) x1 * c0);  
      acc1 += ((q63_t) x2 * c0);  
      acc2 += ((q63_t) x0 * c0);  
        
      c0 = *(pb + 2u);  
      x1 = *(px++);  
      pb += 3u;  
        
      acc0 += ((q63_t) x2 * c0);  
      acc1 += ((q63_t) x0 * c0);  
      acc2 += ((q63_t) x1 * c0);  
  
      i--;  
    }  
  
    /* If the filter length is not a multiple of 2, compute the remaining filter taps */  
  
    i=tapCntN3;  
      
    while(i > 0u)  
    {  
      /* Read coefficients */  
      c0 = *(pb++);  
  
      /* Fetch 1 state variable */  
      x2 = *(px++);  
  
      /* Perform the multiply-accumulates */  
      acc0 += ((q63_t) x0 * c0);  
      acc1 += ((q63_t) x1 * c0);  
      acc2 += ((q63_t) x2 * c0);  
        
      /* Reuse the present sample states for next sample */  
      x0 = x1;  
      x1 = x2;  
        
      /* Decrement the loop counter */  
      i--;  
    }  
  
    /* Advance the state pointer by 4 to process the next group of 4 samples */  
    pState = pState + 3;  
  
    /* The results in the 4 accumulators are in 2.30 format.  Convert to 1.31   
     ** Then store the 4 outputs in the destination buffer. */  
    *pDst++ = (q31_t) (acc0 >> 31u);  
    *pDst++ = (q31_t) (acc1 >> 31u);  
    *pDst++ = (q31_t) (acc2 >> 31u);  
      
    /* Decrement the samples loop counter */  
    blkCnt--;  
  }  
  
  /* If the blockSize is not a multiple of 2, compute any remaining output samples here.   
   ** No loop unrolling is used. */  
  
  while(blockSize > 0u)  
  {  
    /* Copy one sample at a time into state buffer */  
    *pStateCurnt++ = *pSrc++;  
  
    /* Set the accumulator to zero */  
    acc0 = 0;  
  
    /* Initialize state pointer */  
    px = pState;  
  
    /* Initialize Coefficient pointer */  
    pb = (pCoeffs);  
  
    i = numTaps;  
  
    /* Perform the multiply-accumulates */  
    do  
    {  
      acc0 += (q63_t) * (px++) * (*(pb++));  
      i--;  
    } while(i > 0u);  
  
    /* The result is in 2.62 format.  Convert to 1.31   
     ** Then store the output in the destination buffer. */  
    *pDst++ = (q31_t) (acc0 >> 31u);  
  
    /* Advance state pointer by 1 for the next sample */  
    pState = pState + 1;  
  
    /* Decrement the samples loop counter */  
    blockSize--;  
  }  
  
  /* Processing is complete.   
   ** Now copy the last numTaps - 1 samples to the satrt of the state buffer.   
   ** This prepares the state buffer for the next function call. */  
  
  /* Points to the start of the state buffer */  
  pStateCurnt = S->pState;  
  
  tapCnt = (numTaps - 1u) >> 2u;  
  
  /* copy data */  
  while(tapCnt > 0u)  
  {  
    *pStateCurnt++ = *pState++;  
    *pStateCurnt++ = *pState++;  
    *pStateCurnt++ = *pState++;  
    *pStateCurnt++ = *pState++;  
  
    /* Decrement the loop counter */  
    tapCnt--;  
  }  
  
  /* Calculate remaining number of copies */  
  tapCnt = (numTaps - 1u) % 0x4u;  
  
  /* Copy the remaining q31_t data */  
  while(tapCnt > 0u)  
  {  
    *pStateCurnt++ = *pState++;  
  
    /* Decrement the loop counter */  
    tapCnt--;  
  }  
  
}