arm_fir_lattice_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_lattice_f32.c  
00009 *  
00010 * Description:  Processing function for the floating-point FIR Lattice filter.  
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 0.0.7  2010/06/10   
00027 *    Misra-C changes done  
00028 * -------------------------------------------------------------------- */ 
00029  
00030 #include "arm_math.h" 
00031  
00032 /**  
00033  * @ingroup groupFilters  
00034  */ 
00035  
00036 /**  
00037  * @defgroup FIR_Lattice Finite Impulse Response (FIR) Lattice Filters  
00038  *  
00039  * This set of functions implements Finite Impulse Response (FIR) lattice filters  
00040  * for Q15, Q31 and floating-point data types.  Lattice filters are used in a   
00041  * variety of adaptive filter applications.  The filter structure is feedforward and  
00042  * the net impulse response is finite length.  
00043  * The functions operate on blocks  
00044  * 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> point to input and output arrays containing <code>blockSize</code> values.  
00047  *  
00048  * \par Algorithm:  
00049  * \image html FIRLattice.gif "Finite Impulse Response Lattice filter"  
00050  * The following difference equation is implemented:  
00051  * <pre>  
00052  *    f0[n] = g0[n] = x[n]  
00053  *    fm[n] = fm-1[n] + km * gm-1[n-1] for m = 1, 2, ...M  
00054  *    gm[n] = km * fm-1[n] + gm-1[n-1] for m = 1, 2, ...M  
00055  *    y[n] = fM[n]  
00056  * </pre>  
00057  * \par  
00058  * <code>pCoeffs</code> points to tha array of reflection coefficients of size <code>numStages</code>.  
00059  * Reflection Coefficients are stored in the following order.  
00060  * \par  
00061  * <pre>  
00062  *    {k1, k2, ..., kM}  
00063  * </pre>  
00064  * where M is number of stages  
00065  * \par  
00066  * <code>pState</code> points to a state array of size <code>numStages</code>.  
00067  * The state variables (g values) hold previous inputs and are stored in the following order.  
00068  * <pre>  
00069  *    {g0[n], g1[n], g2[n] ...gM-1[n]}  
00070  * </pre>  
00071  * The state variables are updated after each block of data is processed; the coefficients are untouched.  
00072  * \par Instance Structure  
00073  * The coefficients and state variables for a filter are stored together in an instance data structure.  
00074  * A separate instance structure must be defined for each filter.  
00075  * Coefficient arrays may be shared among several instances while state variable arrays cannot be shared.  
00076  * There are separate instance structure declarations for each of the 3 supported data types.  
00077  *  
00078  * \par Initialization Functions  
00079  * There is also an associated initialization function for each data type.  
00080  * The initialization function performs the following operations:  
00081  * - Sets the values of the internal structure fields.  
00082  * - Zeros out the values in the state buffer.  
00083  *  
00084  * \par  
00085  * Use of the initialization function is optional.  
00086  * However, if the initialization function is used, then the instance structure cannot be placed into a const data section.  
00087  * To place an instance structure into a const data section, the instance structure must be manually initialized.  
00088  * Set the values in the state buffer to zeros and then manually initialize the instance structure as follows:  
00089  * <pre>  
00090  *arm_fir_lattice_instance_f32 S = {numStages, pState, pCoeffs};  
00091  *arm_fir_lattice_instance_q31 S = {numStages, pState, pCoeffs};  
00092  *arm_fir_lattice_instance_q15 S = {numStages, pState, pCoeffs};  
00093  * </pre>  
00094  * \par  
00095  * where <code>numStages</code> is the number of stages in the filter; <code>pState</code> is the address of the state buffer;  
00096  * <code>pCoeffs</code> is the address of the coefficient buffer.  
00097  * \par Fixed-Point Behavior  
00098  * Care must be taken when using the fixed-point versions of the FIR Lattice filter functions.  
00099  * In particular, the overflow and saturation behavior of the accumulator used in each function must be considered.  
00100  * Refer to the function specific documentation below for usage guidelines.  
00101  */ 
00102  
00103 /**  
00104  * @addtogroup FIR_Lattice  
00105  * @{  
00106  */ 
00107  
00108  
00109   /**  
00110    * @brief Processing function for the floating-point FIR lattice filter.  
00111    * @param[in]  *S        points to an instance of the floating-point FIR lattice structure.  
00112    * @param[in]  *pSrc     points to the block of input data.  
00113    * @param[out] *pDst     points to the block of output data  
00114    * @param[in]  blockSize number of samples to process.  
00115    * @return none.  
00116    */ 
00117  
00118 void arm_fir_lattice_f32( 
00119   const arm_fir_lattice_instance_f32 * S, 
00120   float32_t * pSrc, 
00121   float32_t * pDst, 
00122   uint32_t blockSize) 
00123 { 
00124   float32_t *pState;                             /* State pointer */ 
00125   float32_t *pCoeffs = S->pCoeffs;               /* Coefficient pointer */ 
00126   float32_t *px;                                 /* temporary state pointer */ 
00127   float32_t *pk;                                 /* temporary coefficient pointer */ 
00128   float32_t fcurr1, fnext1, gcurr1, gnext1;      /* temporary variables for first sample in loop unrolling */ 
00129   float32_t fcurr2, fnext2, gnext2;              /* temporary variables for second sample in loop unrolling */ 
00130   float32_t fcurr3, fnext3, gnext3;              /* temporary variables for third sample in loop unrolling */ 
00131   float32_t fcurr4, fnext4, gnext4;              /* temporary variables for fourth sample in loop unrolling */ 
00132   uint32_t numStages = S->numStages;             /* Number of stages in the filter */ 
00133   uint32_t blkCnt, stageCnt;                     /* temporary variables for counts */ 
00134  
00135   gcurr1 = 0.0f; 
00136   pState = &S->pState[0]; 
00137  
00138   blkCnt = blockSize >> 2; 
00139  
00140   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.  
00141      a second loop below computes the remaining 1 to 3 samples. */ 
00142   while(blkCnt > 0u) 
00143   { 
00144  
00145     /* Read two samples from input buffer */ 
00146     /* f0(n) = x(n) */ 
00147     fcurr1 = *pSrc++; 
00148     fcurr2 = *pSrc++; 
00149  
00150     /* Initialize coeff pointer */ 
00151     pk = (pCoeffs); 
00152  
00153     /* Initialize state pointer */ 
00154     px = pState; 
00155  
00156     /* Read g0(n-1) from state */ 
00157     gcurr1 = *px; 
00158  
00159     /* Process first sample for first tap */ 
00160     /* f1(n) = f0(n) +  K1 * g0(n-1) */ 
00161     fnext1 = fcurr1 + ((*pk) * gcurr1); 
00162     /* g1(n) = f0(n) * K1  +  g0(n-1) */ 
00163     gnext1 = (fcurr1 * (*pk)) + gcurr1; 
00164  
00165     /* Process second sample for first tap */ 
00166     /* for sample 2 processing */ 
00167     fnext2 = fcurr2 + ((*pk) * fcurr1); 
00168     gnext2 = (fcurr2 * (*pk)) + fcurr1; 
00169  
00170     /* Read next two samples from input buffer */ 
00171     /* f0(n+2) = x(n+2) */ 
00172     fcurr3 = *pSrc++; 
00173     fcurr4 = *pSrc++; 
00174  
00175     /* Copy only last input samples into the state buffer  
00176        which will be used for next four samples processing */ 
00177     *px++ = fcurr4; 
00178  
00179     /* Process third sample for first tap */ 
00180     fnext3 = fcurr3 + ((*pk) * fcurr2); 
00181     gnext3 = (fcurr3 * (*pk)) + fcurr2; 
00182  
00183     /* Process fourth sample for first tap */ 
00184     fnext4 = fcurr4 + ((*pk) * fcurr3); 
00185     gnext4 = (fcurr4 * (*pk++)) + fcurr3; 
00186  
00187     /* Update of f values for next coefficient set processing */ 
00188     fcurr1 = fnext1; 
00189     fcurr2 = fnext2; 
00190     fcurr3 = fnext3; 
00191     fcurr4 = fnext4; 
00192  
00193     /* Loop unrolling.  Process 4 taps at a time . */ 
00194     stageCnt = (numStages - 1u) >> 2u; 
00195  
00196     /* Loop over the number of taps.  Unroll by a factor of 4.  
00197      ** Repeat until we've computed numStages-3 coefficients. */ 
00198  
00199     /* Process 2nd, 3rd, 4th and 5th taps ... here */ 
00200     while(stageCnt > 0u) 
00201     { 
00202       /* Read g1(n-1), g3(n-1) .... from state */ 
00203       gcurr1 = *px; 
00204  
00205       /* save g1(n) in state buffer */ 
00206       *px++ = gnext4; 
00207  
00208       /* Process first sample for 2nd, 6th .. tap */ 
00209       /* Sample processing for K2, K6.... */ 
00210       /* f2(n) = f1(n) +  K2 * g1(n-1) */ 
00211       fnext1 = fcurr1 + ((*pk) * gcurr1); 
00212       /* Process second sample for 2nd, 6th .. tap */ 
00213       /* for sample 2 processing */ 
00214       fnext2 = fcurr2 + ((*pk) * gnext1); 
00215       /* Process third sample for 2nd, 6th .. tap */ 
00216       fnext3 = fcurr3 + ((*pk) * gnext2); 
00217       /* Process fourth sample for 2nd, 6th .. tap */ 
00218       fnext4 = fcurr4 + ((*pk) * gnext3); 
00219  
00220       /* g2(n) = f1(n) * K2  +  g1(n-1) */ 
00221       /* Calculation of state values for next stage */ 
00222       gnext4 = (fcurr4 * (*pk)) + gnext3; 
00223       gnext3 = (fcurr3 * (*pk)) + gnext2; 
00224       gnext2 = (fcurr2 * (*pk)) + gnext1; 
00225       gnext1 = (fcurr1 * (*pk++)) + gcurr1; 
00226  
00227  
00228       /* Read g2(n-1), g4(n-1) .... from state */ 
00229       gcurr1 = *px; 
00230  
00231       /* save g2(n) in state buffer */ 
00232       *px++ = gnext4; 
00233  
00234       /* Sample processing for K3, K7.... */ 
00235       /* Process first sample for 3rd, 7th .. tap */ 
00236       /* f3(n) = f2(n) +  K3 * g2(n-1) */ 
00237       fcurr1 = fnext1 + ((*pk) * gcurr1); 
00238       /* Process second sample for 3rd, 7th .. tap */ 
00239       fcurr2 = fnext2 + ((*pk) * gnext1); 
00240       /* Process third sample for 3rd, 7th .. tap */ 
00241       fcurr3 = fnext3 + ((*pk) * gnext2); 
00242       /* Process fourth sample for 3rd, 7th .. tap */ 
00243       fcurr4 = fnext4 + ((*pk) * gnext3); 
00244  
00245       /* Calculation of state values for next stage */ 
00246       /* g3(n) = f2(n) * K3  +  g2(n-1) */ 
00247       gnext4 = (fnext4 * (*pk)) + gnext3; 
00248       gnext3 = (fnext3 * (*pk)) + gnext2; 
00249       gnext2 = (fnext2 * (*pk)) + gnext1; 
00250       gnext1 = (fnext1 * (*pk++)) + gcurr1; 
00251  
00252  
00253       /* Read g1(n-1), g3(n-1) .... from state */ 
00254       gcurr1 = *px; 
00255  
00256       /* save g3(n) in state buffer */ 
00257       *px++ = gnext4; 
00258  
00259       /* Sample processing for K4, K8.... */ 
00260       /* Process first sample for 4th, 8th .. tap */ 
00261       /* f4(n) = f3(n) +  K4 * g3(n-1) */ 
00262       fnext1 = fcurr1 + ((*pk) * gcurr1); 
00263       /* Process second sample for 4th, 8th .. tap */ 
00264       /* for sample 2 processing */ 
00265       fnext2 = fcurr2 + ((*pk) * gnext1); 
00266       /* Process third sample for 4th, 8th .. tap */ 
00267       fnext3 = fcurr3 + ((*pk) * gnext2); 
00268       /* Process fourth sample for 4th, 8th .. tap */ 
00269       fnext4 = fcurr4 + ((*pk) * gnext3); 
00270  
00271       /* g4(n) = f3(n) * K4  +  g3(n-1) */ 
00272       /* Calculation of state values for next stage */ 
00273       gnext4 = (fcurr4 * (*pk)) + gnext3; 
00274       gnext3 = (fcurr3 * (*pk)) + gnext2; 
00275       gnext2 = (fcurr2 * (*pk)) + gnext1; 
00276       gnext1 = (fcurr1 * (*pk++)) + gcurr1; 
00277  
00278       /* Read g2(n-1), g4(n-1) .... from state */ 
00279       gcurr1 = *px; 
00280  
00281       /* save g4(n) in state buffer */ 
00282       *px++ = gnext4; 
00283  
00284       /* Sample processing for K5, K9.... */ 
00285       /* Process first sample for 5th, 9th .. tap */ 
00286       /* f5(n) = f4(n) +  K5 * g4(n-1) */ 
00287       fcurr1 = fnext1 + ((*pk) * gcurr1); 
00288       /* Process second sample for 5th, 9th .. tap */ 
00289       fcurr2 = fnext2 + ((*pk) * gnext1); 
00290       /* Process third sample for 5th, 9th .. tap */ 
00291       fcurr3 = fnext3 + ((*pk) * gnext2); 
00292       /* Process fourth sample for 5th, 9th .. tap */ 
00293       fcurr4 = fnext4 + ((*pk) * gnext3); 
00294  
00295       /* Calculation of state values for next stage */ 
00296       /* g5(n) = f4(n) * K5  +  g4(n-1) */ 
00297       gnext4 = (fnext4 * (*pk)) + gnext3; 
00298       gnext3 = (fnext3 * (*pk)) + gnext2; 
00299       gnext2 = (fnext2 * (*pk)) + gnext1; 
00300       gnext1 = (fnext1 * (*pk++)) + gcurr1; 
00301  
00302       stageCnt--; 
00303     } 
00304  
00305     /* If the (filter length -1) is not a multiple of 4, compute the remaining filter taps */ 
00306     stageCnt = (numStages - 1u) % 0x4u; 
00307  
00308     while(stageCnt > 0u) 
00309     { 
00310       gcurr1 = *px; 
00311  
00312       /* save g value in state buffer */ 
00313       *px++ = gnext4; 
00314  
00315       /* Process four samples for last three taps here */ 
00316       fnext1 = fcurr1 + ((*pk) * gcurr1); 
00317       fnext2 = fcurr2 + ((*pk) * gnext1); 
00318       fnext3 = fcurr3 + ((*pk) * gnext2); 
00319       fnext4 = fcurr4 + ((*pk) * gnext3); 
00320  
00321       /* g1(n) = f0(n) * K1  +  g0(n-1) */ 
00322       gnext4 = (fcurr4 * (*pk)) + gnext3; 
00323       gnext3 = (fcurr3 * (*pk)) + gnext2; 
00324       gnext2 = (fcurr2 * (*pk)) + gnext1; 
00325       gnext1 = (fcurr1 * (*pk++)) + gcurr1; 
00326  
00327       /* Update of f values for next coefficient set processing */ 
00328       fcurr1 = fnext1; 
00329       fcurr2 = fnext2; 
00330       fcurr3 = fnext3; 
00331       fcurr4 = fnext4; 
00332  
00333       stageCnt--; 
00334  
00335     } 
00336  
00337     /* The results in the 4 accumulators, store in the destination buffer. */ 
00338     /* y(n) = fN(n) */ 
00339     *pDst++ = fcurr1; 
00340     *pDst++ = fcurr2; 
00341     *pDst++ = fcurr3; 
00342     *pDst++ = fcurr4; 
00343  
00344     blkCnt--; 
00345   } 
00346  
00347   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.  
00348    ** No loop unrolling is used. */ 
00349   blkCnt = blockSize % 0x4u; 
00350  
00351   while(blkCnt > 0u) 
00352   { 
00353     /* f0(n) = x(n) */ 
00354     fcurr1 = *pSrc++; 
00355  
00356     /* Initialize coeff pointer */ 
00357     pk = (pCoeffs); 
00358  
00359     /* Initialize state pointer */ 
00360     px = pState; 
00361  
00362     /* read g2(n) from state buffer */ 
00363     gcurr1 = *px; 
00364  
00365     /* for sample 1 processing */ 
00366     /* f1(n) = f0(n) +  K1 * g0(n-1) */ 
00367     fnext1 = fcurr1 + ((*pk) * gcurr1); 
00368     /* g1(n) = f0(n) * K1  +  g0(n-1) */ 
00369     gnext1 = (fcurr1 * (*pk++)) + gcurr1; 
00370  
00371     /* save g1(n) in state buffer */ 
00372     *px++ = fcurr1; 
00373  
00374     /* f1(n) is saved in fcurr1  
00375        for next stage processing */ 
00376     fcurr1 = fnext1; 
00377  
00378     stageCnt = (numStages - 1u); 
00379  
00380     /* stage loop */ 
00381     while(stageCnt > 0u) 
00382     { 
00383       /* read g2(n) from state buffer */ 
00384       gcurr1 = *px; 
00385  
00386       /* save g1(n) in state buffer */ 
00387       *px++ = gnext1; 
00388  
00389       /* Sample processing for K2, K3.... */ 
00390       /* f2(n) = f1(n) +  K2 * g1(n-1) */ 
00391       fnext1 = fcurr1 + ((*pk) * gcurr1); 
00392       /* g2(n) = f1(n) * K2  +  g1(n-1) */ 
00393       gnext1 = (fcurr1 * (*pk++)) + gcurr1; 
00394  
00395       /* f1(n) is saved in fcurr1  
00396          for next stage processing */ 
00397       fcurr1 = fnext1; 
00398  
00399       stageCnt--; 
00400  
00401     } 
00402  
00403     /* y(n) = fN(n) */ 
00404     *pDst++ = fcurr1; 
00405  
00406     blkCnt--; 
00407  
00408   } 
00409 } 
00410  
00411 /**  
00412  * @} end of FIR_Lattice group  
00413  */