arm_fir_f32.c

00001 /* ----------------------------------------------------------------------  
00002 * Copyright (C) 2010 ARM Limited. All rights reserved.  
00003 *  
00004 * $Date:        29. November 2010  
00005 * $Revision:    V1.0.3  
00006 *  
00007 * Project:      CMSIS DSP Library  
00008 * Title:        arm_fir_f32.c  
00009 *  
00010 * Description:  Floating-point FIR filter processing function.  
00011 *  
00012 * Target Processor: Cortex-M4/Cortex-M3
00013 *  
00014 * Version 1.0.3 2010/11/29 
00015 *    Re-organized the CMSIS folders and updated documentation.  
00016 *   
00017 * Version 1.0.2 2010/11/11  
00018 *    Documentation updated.   
00019 *  
00020 * Version 1.0.1 2010/10/05   
00021 *    Production release and review comments incorporated.  
00022 *  
00023 * Version 1.0.0 2010/09/20   
00024 *    Production release and review comments incorporated.  
00025 *  
00026 * Version 0.0.5  2010/04/26   
00027 *    incorporated review comments and updated with latest CMSIS layer  
00028 *  
00029 * Version 0.0.3  2010/03/10   
00030 *    Initial version  
00031 * -------------------------------------------------------------------- */ 
00032  
00033 #include "arm_math.h" 
00034  
00035 /**  
00036  * @ingroup groupFilters  
00037  */ 
00038  
00039 /**  
00040  * @defgroup FIR Finite Impulse Response (FIR) Filters  
00041  *  
00042  * This set of functions implements Finite Impulse Response (FIR) filters  
00043  * for Q7, Q15, Q31, and floating-point data types.  Fast versions of Q15 and Q31 are also provided.  
00044  * The functions operate on blocks of input and output data and each call to the function processes  
00045  * <code>blockSize</code> samples through the filter.  <code>pSrc</code> and  
00046  * <code>pDst</code> points to input and output arrays containing <code>blockSize</code> values.  
00047  *  
00048  * \par Algorithm:  
00049  * The FIR filter algorithm is based upon a sequence of multiply-accumulate (MAC) operations.  
00050  * Each filter coefficient <code>b[n]</code> is multiplied by a state variable which equals a previous input sample <code>x[n]</code>.  
00051  * <pre>  
00052  *    y[n] = b[0] * x[n] + b[1] * x[n-1] + b[2] * x[n-2] + ...+ b[numTaps-1] * x[n-numTaps+1]  
00053  * </pre>  
00054  * \par  
00055  * \image html FIR.gif "Finite Impulse Response filter"  
00056  * \par  
00057  * <code>pCoeffs</code> points to a coefficient array of size <code>numTaps</code>.  
00058  * Coefficients are stored in time reversed order.  
00059  * \par  
00060  * <pre>  
00061  *    {b[numTaps-1], b[numTaps-2], b[N-2], ..., b[1], b[0]}  
00062  * </pre>  
00063  * \par  
00064  * <code>pState</code> points to a state array of size <code>numTaps + blockSize - 1</code>.  
00065  * Samples in the state buffer are stored in the following order.  
00066  * \par  
00067  * <pre>  
00068  *    {x[n-numTaps+1], x[n-numTaps], x[n-numTaps-1], x[n-numTaps-2]....x[0], x[1], ..., x[blockSize-1]}  
00069  * </pre>  
00070  * \par  
00071  * Note that the length of the state buffer exceeds the length of the coefficient array by <code>blockSize-1</code>.  
00072  * The increased state buffer length allows circular addressing, which is traditionally used in the FIR filters,  
00073  * to be avoided and yields a significant speed improvement.  
00074  * The state variables are updated after each block of data is processed; the coefficients are untouched.  
00075  * \par Instance Structure  
00076  * The coefficients and state variables for a filter are stored together in an instance data structure.  
00077  * A separate instance structure must be defined for each filter.  
00078  * Coefficient arrays may be shared among several instances while state variable arrays cannot be shared.  
00079  * There are separate instance structure declarations for each of the 4 supported data types.  
00080  *  
00081  * \par Initialization Functions  
00082  * There is also an associated initialization function for each data type.  
00083  * The initialization function performs the following operations:  
00084  * - Sets the values of the internal structure fields.  
00085  * - Zeros out the values in the state buffer.  
00086  *  
00087  * \par  
00088  * Use of the initialization function is optional.  
00089  * However, if the initialization function is used, then the instance structure cannot be placed into a const data section.  
00090  * To place an instance structure into a const data section, the instance structure must be manually initialized.  
00091  * Set the values in the state buffer to zeros before static initialization.  
00092  * The code below statically initializes each of the 4 different data type filter instance structures  
00093  * <pre>  
00094  *arm_fir_instance_f32 S = {numTaps, pState, pCoeffs};  
00095  *arm_fir_instance_q31 S = {numTaps, pState, pCoeffs};  
00096  *arm_fir_instance_q15 S = {numTaps, pState, pCoeffs};  
00097  *arm_fir_instance_q7 S =  {numTaps, pState, pCoeffs};  
00098  * </pre>  
00099  *  
00100  * where <code>numTaps</code> is the number of filter coefficients in the filter; <code>pState</code> is the address of the state buffer;  
00101  * <code>pCoeffs</code> is the address of the coefficient buffer.  
00102  *  
00103  * \par Fixed-Point Behavior  
00104  * Care must be taken when using the fixed-point versions of the FIR filter functions.  
00105  * In particular, the overflow and saturation behavior of the accumulator used in each function must be considered.  
00106  * Refer to the function specific documentation below for usage guidelines.  
00107  */ 
00108  
00109 /**  
00110  * @addtogroup FIR  
00111  * @{  
00112  */ 
00113  
00114 /**  
00115  *  
00116  * @param[in]  *S points to an instance of the floating-point FIR filter structure.  
00117  * @param[in]  *pSrc points to the block of input data.  
00118  * @param[out] *pDst points to the block of output data.  
00119  * @param[in]  blockSize number of samples to process per call.  
00120  * @return     none.  
00121  *  
00122  */ 
00123  
00124 void arm_fir_f32( 
00125   const arm_fir_instance_f32 * S, 
00126   float32_t * pSrc, 
00127   float32_t * pDst, 
00128   uint32_t blockSize) 
00129 { 
00130   float32_t *pState = S->pState;                 /* State pointer */ 
00131   float32_t *pCoeffs = S->pCoeffs;               /* Coefficient pointer */ 
00132   float32_t *pStateCurnt;                        /* Points to the current sample of the state */ 
00133   float32_t *px, *pb;                            /* Temporary pointers for state and coefficient buffers */ 
00134   float32_t acc0, acc1, acc2, acc3;              /* Accumulators */ 
00135   float32_t x0, x1, x2, x3, c0;                  /* Temporary variables to hold state and coefficient values */ 
00136   uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */ 
00137   uint32_t i, tapCnt, blkCnt;                    /* Loop counters */ 
00138  
00139   /* S->pState points to state array which contains previous frame (numTaps - 1) samples */ 
00140   /* pStateCurnt points to the location where the new input data should be written */ 
00141   pStateCurnt = &(S->pState[(numTaps - 1u)]); 
00142  
00143   /* Apply loop unrolling and compute 4 output values simultaneously.  
00144    * The variables acc0 ... acc3 hold output values that are being computed:  
00145    *  
00146    *    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]  
00147    *    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]  
00148    *    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]  
00149    *    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]  
00150    */ 
00151   blkCnt = blockSize >> 2; 
00152  
00153   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.  
00154    ** a second loop below computes the remaining 1 to 3 samples. */ 
00155   while(blkCnt > 0u) 
00156   { 
00157     /* Copy four new input samples into the state buffer */ 
00158     *pStateCurnt++ = *pSrc++; 
00159     *pStateCurnt++ = *pSrc++; 
00160     *pStateCurnt++ = *pSrc++; 
00161     *pStateCurnt++ = *pSrc++; 
00162  
00163     /* Set all accumulators to zero */ 
00164     acc0 = 0.0f; 
00165     acc1 = 0.0f; 
00166     acc2 = 0.0f; 
00167     acc3 = 0.0f; 
00168  
00169     /* Initialize state pointer */ 
00170     px = pState; 
00171  
00172     /* Initialize coeff pointer */ 
00173     pb = (pCoeffs); 
00174  
00175     /* Read the first three samples from the state buffer:  x[n-numTaps], x[n-numTaps-1], x[n-numTaps-2] */ 
00176     x0 = *px++; 
00177     x1 = *px++; 
00178     x2 = *px++; 
00179  
00180     /* Loop unrolling.  Process 4 taps at a time. */ 
00181     tapCnt = numTaps >> 2u; 
00182  
00183     /* Loop over the number of taps.  Unroll by a factor of 4.  
00184      ** Repeat until we've computed numTaps-4 coefficients. */ 
00185     while(tapCnt > 0u) 
00186     { 
00187       /* Read the b[numTaps-1] coefficient */ 
00188       c0 = *(pb++); 
00189  
00190       /* Read x[n-numTaps-3] sample */ 
00191       x3 = *(px++); 
00192  
00193       /* acc0 +=  b[numTaps-1] * x[n-numTaps] */ 
00194       acc0 += x0 * c0; 
00195  
00196       /* acc1 +=  b[numTaps-1] * x[n-numTaps-1] */ 
00197       acc1 += x1 * c0; 
00198  
00199       /* acc2 +=  b[numTaps-1] * x[n-numTaps-2] */ 
00200       acc2 += x2 * c0; 
00201  
00202       /* acc3 +=  b[numTaps-1] * x[n-numTaps-3] */ 
00203       acc3 += x3 * c0; 
00204  
00205       /* Read the b[numTaps-2] coefficient */ 
00206       c0 = *(pb++); 
00207  
00208       /* Read x[n-numTaps-4] sample */ 
00209       x0 = *(px++); 
00210  
00211       /* Perform the multiply-accumulate */ 
00212       acc0 += x1 * c0; 
00213       acc1 += x2 * c0; 
00214       acc2 += x3 * c0; 
00215       acc3 += x0 * c0; 
00216  
00217       /* Read the b[numTaps-3] coefficient */ 
00218       c0 = *(pb++); 
00219  
00220       /* Read x[n-numTaps-5] sample */ 
00221       x1 = *(px++); 
00222  
00223       /* Perform the multiply-accumulates */ 
00224       acc0 += x2 * c0; 
00225       acc1 += x3 * c0; 
00226       acc2 += x0 * c0; 
00227       acc3 += x1 * c0; 
00228  
00229       /* Read the b[numTaps-4] coefficient */ 
00230       c0 = *(pb++); 
00231  
00232       /* Read x[n-numTaps-6] sample */ 
00233       x2 = *(px++); 
00234  
00235       /* Perform the multiply-accumulates */ 
00236       acc0 += x3 * c0; 
00237       acc1 += x0 * c0; 
00238       acc2 += x1 * c0; 
00239       acc3 += x2 * c0; 
00240  
00241       tapCnt--; 
00242     } 
00243  
00244     /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 
00245     tapCnt = numTaps % 0x4u; 
00246  
00247     while(tapCnt > 0u) 
00248     { 
00249       /* Read coefficients */ 
00250       c0 = *(pb++); 
00251  
00252       /* Fetch 1 state variable */ 
00253       x3 = *(px++); 
00254  
00255       /* Perform the multiply-accumulates */ 
00256       acc0 += x0 * c0; 
00257       acc1 += x1 * c0; 
00258       acc2 += x2 * c0; 
00259       acc3 += x3 * c0; 
00260  
00261       /* Reuse the present sample states for next sample */ 
00262       x0 = x1; 
00263       x1 = x2; 
00264       x2 = x3; 
00265  
00266       /* Decrement the loop counter */ 
00267       tapCnt--; 
00268     } 
00269  
00270     /* Advance the state pointer by 4 to process the next group of 4 samples */ 
00271     pState = pState + 4; 
00272  
00273     /* The results in the 4 accumulators, store in the destination buffer. */ 
00274     *pDst++ = acc0; 
00275     *pDst++ = acc1; 
00276     *pDst++ = acc2; 
00277     *pDst++ = acc3; 
00278  
00279     blkCnt--; 
00280   } 
00281  
00282   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.  
00283    ** No loop unrolling is used. */ 
00284   blkCnt = blockSize % 0x4u; 
00285  
00286   while(blkCnt > 0u) 
00287   { 
00288     /* Copy one sample at a time into state buffer */ 
00289     *pStateCurnt++ = *pSrc++; 
00290  
00291     /* Set the accumulator to zero */ 
00292     acc0 = 0.0f; 
00293  
00294     /* Initialize state pointer */ 
00295     px = pState; 
00296  
00297     /* Initialize Coefficient pointer */ 
00298     pb = (pCoeffs); 
00299  
00300     i = numTaps; 
00301  
00302     /* Perform the multiply-accumulates */ 
00303     do 
00304     { 
00305       acc0 += *px++ * *pb++; 
00306       i--; 
00307  
00308     } while(i > 0u); 
00309  
00310     /* The result is store in the destination buffer. */ 
00311     *pDst++ = acc0; 
00312  
00313     /* Advance state pointer by 1 for the next sample */ 
00314     pState = pState + 1; 
00315  
00316     blkCnt--; 
00317   } 
00318  
00319   /* Processing is complete.  
00320    ** Now copy the last numTaps - 1 samples to the satrt of the state buffer.  
00321    ** This prepares the state buffer for the next function call. */ 
00322  
00323   /* Points to the start of the state buffer */ 
00324   pStateCurnt = S->pState; 
00325  
00326   tapCnt = (numTaps - 1u) >> 2u; 
00327  
00328   /* copy data */ 
00329   while(tapCnt > 0u) 
00330   { 
00331     *pStateCurnt++ = *pState++; 
00332     *pStateCurnt++ = *pState++; 
00333     *pStateCurnt++ = *pState++; 
00334     *pStateCurnt++ = *pState++; 
00335  
00336     /* Decrement the loop counter */ 
00337     tapCnt--; 
00338   } 
00339  
00340   /* Calculate remaining number of copies */ 
00341   tapCnt = (numTaps - 1u) % 0x4u; 
00342  
00343   /* Copy the remaining q31_t data */ 
00344   while(tapCnt > 0u) 
00345   { 
00346     *pStateCurnt++ = *pState++; 
00347  
00348     /* Decrement the loop counter */ 
00349     tapCnt--; 
00350   } 
00351 } 
00352  
00353 /**  
00354  * @} end of FIR group  
00355  */