arm_fir_lattice_q31.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        17. January 2013
00005 * $Revision:    V1.4.1
00006 *    
00007 * Project:      CMSIS DSP Library    
00008 * Title:        arm_fir_lattice_q31.c    
00009 *    
00010 * Description:  Q31 FIR lattice filter processing function.    
00011 *    
00012 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
00013 *  
00014 * Redistribution and use in source and binary forms, with or without 
00015 * modification, are permitted provided that the following conditions
00016 * are met:
00017 *   - Redistributions of source code must retain the above copyright
00018 *     notice, this list of conditions and the following disclaimer.
00019 *   - Redistributions in binary form must reproduce the above copyright
00020 *     notice, this list of conditions and the following disclaimer in
00021 *     the documentation and/or other materials provided with the 
00022 *     distribution.
00023 *   - Neither the name of ARM LIMITED nor the names of its contributors
00024 *     may be used to endorse or promote products derived from this
00025 *     software without specific prior written permission.
00026 *
00027 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
00028 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
00029 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
00030 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 
00031 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
00032 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
00033 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
00034 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
00035 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
00036 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
00037 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
00038 * POSSIBILITY OF SUCH DAMAGE.  
00039 * -------------------------------------------------------------------- */
00040 
00041 #include "arm_math.h"
00042 
00043 /**    
00044  * @ingroup groupFilters    
00045  */
00046 
00047 /**    
00048  * @addtogroup FIR_Lattice    
00049  * @{    
00050  */
00051 
00052 
00053 /**    
00054  * @brief Processing function for the Q31 FIR lattice filter.    
00055  * @param[in]  *S        points to an instance of the Q31 FIR lattice structure.    
00056  * @param[in]  *pSrc     points to the block of input data.    
00057  * @param[out] *pDst     points to the block of output data    
00058  * @param[in]  blockSize number of samples to process.    
00059  * @return none.    
00060  *    
00061  * @details    
00062  * <b>Scaling and Overflow Behavior:</b>    
00063  * In order to avoid overflows the input signal must be scaled down by 2*log2(numStages) bits.    
00064  */
00065 
00066 #ifndef ARM_MATH_CM0_FAMILY
00067 
00068   /* Run the below code for Cortex-M4 and Cortex-M3 */
00069 
00070 void arm_fir_lattice_q31(
00071   const arm_fir_lattice_instance_q31 * S,
00072   q31_t * pSrc,
00073   q31_t * pDst,
00074   uint32_t blockSize)
00075 {
00076   q31_t *pState;                                 /* State pointer */
00077   q31_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */
00078   q31_t *px;                                     /* temporary state pointer */
00079   q31_t *pk;                                     /* temporary coefficient pointer */
00080   q31_t fcurr1, fnext1, gcurr1 = 0, gnext1;      /* temporary variables for first sample in loop unrolling */
00081   q31_t fcurr2, fnext2, gnext2;                  /* temporary variables for second sample in loop unrolling */
00082   uint32_t numStages = S->numStages;             /* Length of the filter */
00083   uint32_t blkCnt, stageCnt;                     /* temporary variables for counts */
00084   q31_t k;
00085 
00086   pState = &S->pState[0];
00087 
00088   blkCnt = blockSize >> 1u;
00089 
00090   /* First part of the processing with loop unrolling.  Compute 2 outputs at a time.        
00091      a second loop below computes the remaining 1 sample. */
00092   while(blkCnt > 0u)
00093   {
00094     /* f0(n) = x(n) */
00095     fcurr1 = *pSrc++;
00096 
00097     /* f0(n) = x(n) */
00098     fcurr2 = *pSrc++;
00099 
00100     /* Initialize coeff pointer */
00101     pk = (pCoeffs);
00102 
00103     /* Initialize state pointer */
00104     px = pState;
00105 
00106     /* read g0(n - 1) from state buffer */
00107     gcurr1 = *px;
00108 
00109     /* Read the reflection coefficient */
00110     k = *pk++;
00111 
00112     /* for sample 1 processing */
00113     /* f1(n) = f0(n) +  K1 * g0(n-1) */
00114     fnext1 = (q31_t) (((q63_t) gcurr1 * k) >> 32);
00115 
00116     /* g1(n) = f0(n) * K1  +  g0(n-1) */
00117     gnext1 = (q31_t) (((q63_t) fcurr1 * (k)) >> 32);
00118     fnext1 = fcurr1 + (fnext1 << 1u);
00119     gnext1 = gcurr1 + (gnext1 << 1u);
00120 
00121     /* for sample 1 processing */
00122     /* f1(n) = f0(n) +  K1 * g0(n-1) */
00123     fnext2 = (q31_t) (((q63_t) fcurr1 * k) >> 32);
00124 
00125     /* g1(n) = f0(n) * K1  +  g0(n-1) */
00126     gnext2 = (q31_t) (((q63_t) fcurr2 * (k)) >> 32);
00127     fnext2 = fcurr2 + (fnext2 << 1u);
00128     gnext2 = fcurr1 + (gnext2 << 1u);
00129 
00130     /* save g1(n) in state buffer */
00131     *px++ = fcurr2;
00132 
00133     /* f1(n) is saved in fcurr1        
00134        for next stage processing */
00135     fcurr1 = fnext1;
00136     fcurr2 = fnext2;
00137 
00138     stageCnt = (numStages - 1u);
00139 
00140     /* stage loop */
00141     while(stageCnt > 0u)
00142     {
00143 
00144       /* Read the reflection coefficient */
00145       k = *pk++;
00146 
00147       /* read g2(n) from state buffer */
00148       gcurr1 = *px;
00149 
00150       /* save g1(n) in state buffer */
00151       *px++ = gnext2;
00152 
00153       /* Sample processing for K2, K3.... */
00154       /* f2(n) = f1(n) +  K2 * g1(n-1) */
00155       fnext1 = (q31_t) (((q63_t) gcurr1 * k) >> 32);
00156       fnext2 = (q31_t) (((q63_t) gnext1 * k) >> 32);
00157 
00158       fnext1 = fcurr1 + (fnext1 << 1u);
00159       fnext2 = fcurr2 + (fnext2 << 1u);
00160 
00161       /* g2(n) = f1(n) * K2  +  g1(n-1) */
00162       gnext2 = (q31_t) (((q63_t) fcurr2 * (k)) >> 32);
00163       gnext2 = gnext1 + (gnext2 << 1u);
00164 
00165       /* g2(n) = f1(n) * K2  +  g1(n-1) */
00166       gnext1 = (q31_t) (((q63_t) fcurr1 * (k)) >> 32);
00167       gnext1 = gcurr1 + (gnext1 << 1u);
00168 
00169       /* f1(n) is saved in fcurr1        
00170          for next stage processing */
00171       fcurr1 = fnext1;
00172       fcurr2 = fnext2;
00173 
00174       stageCnt--;
00175 
00176     }
00177 
00178     /* y(n) = fN(n) */
00179     *pDst++ = fcurr1;
00180     *pDst++ = fcurr2;
00181 
00182     blkCnt--;
00183 
00184   }
00185 
00186   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.        
00187    ** No loop unrolling is used. */
00188   blkCnt = blockSize % 0x2u;
00189 
00190   while(blkCnt > 0u)
00191   {
00192     /* f0(n) = x(n) */
00193     fcurr1 = *pSrc++;
00194 
00195     /* Initialize coeff pointer */
00196     pk = (pCoeffs);
00197 
00198     /* Initialize state pointer */
00199     px = pState;
00200 
00201     /* read g0(n - 1) from state buffer */
00202     gcurr1 = *px;
00203 
00204     /* Read the reflection coefficient */
00205     k = *pk++;
00206 
00207     /* for sample 1 processing */
00208     /* f1(n) = f0(n) +  K1 * g0(n-1) */
00209     fnext1 = (q31_t) (((q63_t) gcurr1 * k) >> 32);
00210     fnext1 = fcurr1 + (fnext1 << 1u);
00211 
00212     /* g1(n) = f0(n) * K1  +  g0(n-1) */
00213     gnext1 = (q31_t) (((q63_t) fcurr1 * (k)) >> 32);
00214     gnext1 = gcurr1 + (gnext1 << 1u);
00215 
00216     /* save g1(n) in state buffer */
00217     *px++ = fcurr1;
00218 
00219     /* f1(n) is saved in fcurr1        
00220        for next stage processing */
00221     fcurr1 = fnext1;
00222 
00223     stageCnt = (numStages - 1u);
00224 
00225     /* stage loop */
00226     while(stageCnt > 0u)
00227     {
00228       /* Read the reflection coefficient */
00229       k = *pk++;
00230 
00231       /* read g2(n) from state buffer */
00232       gcurr1 = *px;
00233 
00234       /* save g1(n) in state buffer */
00235       *px++ = gnext1;
00236 
00237       /* Sample processing for K2, K3.... */
00238       /* f2(n) = f1(n) +  K2 * g1(n-1) */
00239       fnext1 = (q31_t) (((q63_t) gcurr1 * k) >> 32);
00240       fnext1 = fcurr1 + (fnext1 << 1u);
00241 
00242       /* g2(n) = f1(n) * K2  +  g1(n-1) */
00243       gnext1 = (q31_t) (((q63_t) fcurr1 * (k)) >> 32);
00244       gnext1 = gcurr1 + (gnext1 << 1u);
00245 
00246       /* f1(n) is saved in fcurr1        
00247          for next stage processing */
00248       fcurr1 = fnext1;
00249 
00250       stageCnt--;
00251 
00252     }
00253 
00254 
00255     /* y(n) = fN(n) */
00256     *pDst++ = fcurr1;
00257 
00258     blkCnt--;
00259 
00260   }
00261 
00262 
00263 }
00264 
00265 
00266 #else
00267 
00268 /* Run the below code for Cortex-M0 */
00269 
00270 void arm_fir_lattice_q31(
00271   const arm_fir_lattice_instance_q31 * S,
00272   q31_t * pSrc,
00273   q31_t * pDst,
00274   uint32_t blockSize)
00275 {
00276   q31_t *pState;                                 /* State pointer */
00277   q31_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */
00278   q31_t *px;                                     /* temporary state pointer */
00279   q31_t *pk;                                     /* temporary coefficient pointer */
00280   q31_t fcurr, fnext, gcurr, gnext;              /* temporary variables */
00281   uint32_t numStages = S->numStages;             /* Length of the filter */
00282   uint32_t blkCnt, stageCnt;                     /* temporary variables for counts */
00283 
00284   pState = &S->pState[0];
00285 
00286   blkCnt = blockSize;
00287 
00288   while(blkCnt > 0u)
00289   {
00290     /* f0(n) = x(n) */
00291     fcurr = *pSrc++;
00292 
00293     /* Initialize coeff pointer */
00294     pk = (pCoeffs);
00295 
00296     /* Initialize state pointer */
00297     px = pState;
00298 
00299     /* read g0(n-1) from state buffer */
00300     gcurr = *px;
00301 
00302     /* for sample 1 processing */
00303     /* f1(n) = f0(n) +  K1 * g0(n-1) */
00304     fnext = (q31_t) (((q63_t) gcurr * (*pk)) >> 31) + fcurr;
00305     /* g1(n) = f0(n) * K1  +  g0(n-1) */
00306     gnext = (q31_t) (((q63_t) fcurr * (*pk++)) >> 31) + gcurr;
00307     /* save g1(n) in state buffer */
00308     *px++ = fcurr;
00309 
00310     /* f1(n) is saved in fcurr1            
00311        for next stage processing */
00312     fcurr = fnext;
00313 
00314     stageCnt = (numStages - 1u);
00315 
00316     /* stage loop */
00317     while(stageCnt > 0u)
00318     {
00319       /* read g2(n) from state buffer */
00320       gcurr = *px;
00321 
00322       /* save g1(n) in state buffer */
00323       *px++ = gnext;
00324 
00325       /* Sample processing for K2, K3.... */
00326       /* f2(n) = f1(n) +  K2 * g1(n-1) */
00327       fnext = (q31_t) (((q63_t) gcurr * (*pk)) >> 31) + fcurr;
00328       /* g2(n) = f1(n) * K2  +  g1(n-1) */
00329       gnext = (q31_t) (((q63_t) fcurr * (*pk++)) >> 31) + gcurr;
00330 
00331       /* f1(n) is saved in fcurr1            
00332          for next stage processing */
00333       fcurr = fnext;
00334 
00335       stageCnt--;
00336 
00337     }
00338 
00339     /* y(n) = fN(n) */
00340     *pDst++ = fcurr;
00341 
00342     blkCnt--;
00343 
00344   }
00345 
00346 }
00347 
00348 #endif /*   #ifndef ARM_MATH_CM0_FAMILY */
00349 
00350 
00351 /**    
00352  * @} end of FIR_Lattice group    
00353  */