arm_iir_lattice_q31.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        19. March 2015
00005 * $Revision:    V.1.4.5
00006 *    
00007 * Project:      CMSIS DSP Library    
00008 * Title:        arm_iir_lattice_q31.c    
00009 *    
00010 * Description:  Q31 IIR 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 IIR_Lattice    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief Processing function for the Q31 IIR lattice filter.    
00054  * @param[in] *S points to an instance of the Q31 IIR lattice structure.    
00055  * @param[in] *pSrc points to the block of input data.    
00056  * @param[out] *pDst points to the block of output data.    
00057  * @param[in] blockSize number of samples to process.    
00058  * @return none.    
00059  *    
00060  * @details    
00061  * <b>Scaling and Overflow Behavior:</b>    
00062  * \par    
00063  * The function is implemented using an internal 64-bit accumulator.    
00064  * The accumulator has a 2.62 format and maintains full precision of the intermediate multiplication results but provides only a single guard bit.    
00065  * Thus, if the accumulator result overflows it wraps around rather than clip.    
00066  * In order to avoid overflows completely the input signal must be scaled down by 2*log2(numStages) bits.    
00067  * After all multiply-accumulates are performed, the 2.62 accumulator is saturated to 1.32 format and then truncated to 1.31 format.    
00068  */
00069 
00070 void arm_iir_lattice_q31(
00071   const arm_iir_lattice_instance_q31 * S,
00072   q31_t * pSrc,
00073   q31_t * pDst,
00074   uint32_t blockSize)
00075 {
00076   q31_t fcurr, fnext = 0, gcurr = 0, gnext;      /* Temporary variables for lattice stages */
00077   q63_t acc;                                     /* Accumlator */
00078   uint32_t blkCnt, tapCnt;                       /* Temporary variables for counts */
00079   q31_t *px1, *px2, *pk, *pv;                    /* Temporary pointers for state and coef */
00080   uint32_t numStages = S->numStages;             /* number of stages */
00081   q31_t *pState;                                 /* State pointer */
00082   q31_t *pStateCurnt;                            /* State current pointer */
00083 
00084   blkCnt = blockSize;
00085 
00086   pState = &S->pState[0];
00087 
00088 
00089 #ifndef ARM_MATH_CM0_FAMILY
00090 
00091   /* Run the below code for Cortex-M4 and Cortex-M3 */
00092 
00093   /* Sample processing */
00094   while(blkCnt > 0u)
00095   {
00096     /* Read Sample from input buffer */
00097     /* fN(n) = x(n) */
00098     fcurr = *pSrc++;
00099 
00100     /* Initialize state read pointer */
00101     px1 = pState;
00102     /* Initialize state write pointer */
00103     px2 = pState;
00104     /* Set accumulator to zero */
00105     acc = 0;
00106     /* Initialize Ladder coeff pointer */
00107     pv = &S->pvCoeffs[0];
00108     /* Initialize Reflection coeff pointer */
00109     pk = &S->pkCoeffs[0];
00110 
00111 
00112     /* Process sample for first tap */
00113     gcurr = *px1++;
00114     /* fN-1(n) = fN(n) - kN * gN-1(n-1) */
00115     fnext = __QSUB(fcurr, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00116     /* gN(n) = kN * fN-1(n) + gN-1(n-1) */
00117     gnext = __QADD(gcurr, (q31_t) (((q63_t) fnext * (*pk++)) >> 31));
00118     /* write gN-1(n-1) into state for next sample processing */
00119     *px2++ = gnext;
00120     /* y(n) += gN(n) * vN  */
00121     acc += ((q63_t) gnext * *pv++);
00122 
00123     /* Update f values for next coefficient processing */
00124     fcurr = fnext;
00125 
00126     /* Loop unrolling.  Process 4 taps at a time. */
00127     tapCnt = (numStages - 1u) >> 2;
00128 
00129     while(tapCnt > 0u)
00130     {
00131 
00132       /* Process sample for 2nd, 6th .. taps */
00133       /* Read gN-2(n-1) from state buffer */
00134       gcurr = *px1++;
00135       /* fN-2(n) = fN-1(n) - kN-1 * gN-2(n-1) */
00136       fnext = __QSUB(fcurr, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00137       /* gN-1(n) = kN-1 * fN-2(n) + gN-2(n-1) */
00138       gnext = __QADD(gcurr, (q31_t) (((q63_t) fnext * (*pk++)) >> 31));
00139       /* y(n) += gN-1(n) * vN-1  */
00140       /* process for gN-5(n) * vN-5, gN-9(n) * vN-9 ... */
00141       acc += ((q63_t) gnext * *pv++);
00142       /* write gN-1(n) into state for next sample processing */
00143       *px2++ = gnext;
00144 
00145       /* Process sample for 3nd, 7th ...taps */
00146       /* Read gN-3(n-1) from state buffer */
00147       gcurr = *px1++;
00148       /* Process sample for 3rd, 7th .. taps */
00149       /* fN-3(n) = fN-2(n) - kN-2 * gN-3(n-1) */
00150       fcurr = __QSUB(fnext, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00151       /* gN-2(n) = kN-2 * fN-3(n) + gN-3(n-1) */
00152       gnext = __QADD(gcurr, (q31_t) (((q63_t) fcurr * (*pk++)) >> 31));
00153       /* y(n) += gN-2(n) * vN-2  */
00154       /* process for gN-6(n) * vN-6, gN-10(n) * vN-10 ... */
00155       acc += ((q63_t) gnext * *pv++);
00156       /* write gN-2(n) into state for next sample processing */
00157       *px2++ = gnext;
00158 
00159 
00160       /* Process sample for 4th, 8th ...taps */
00161       /* Read gN-4(n-1) from state buffer */
00162       gcurr = *px1++;
00163       /* Process sample for 4th, 8th .. taps */
00164       /* fN-4(n) = fN-3(n) - kN-3 * gN-4(n-1) */
00165       fnext = __QSUB(fcurr, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00166       /* gN-3(n) = kN-3 * fN-4(n) + gN-4(n-1) */
00167       gnext = __QADD(gcurr, (q31_t) (((q63_t) fnext * (*pk++)) >> 31));
00168       /* y(n) += gN-3(n) * vN-3  */
00169       /* process for gN-7(n) * vN-7, gN-11(n) * vN-11 ... */
00170       acc += ((q63_t) gnext * *pv++);
00171       /* write gN-3(n) into state for next sample processing */
00172       *px2++ = gnext;
00173 
00174 
00175       /* Process sample for 5th, 9th ...taps */
00176       /* Read gN-5(n-1) from state buffer */
00177       gcurr = *px1++;
00178       /* Process sample for 5th, 9th .. taps */
00179       /* fN-5(n) = fN-4(n) - kN-4 * gN-1(n-1) */
00180       fcurr = __QSUB(fnext, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00181       /* gN-4(n) = kN-4 * fN-5(n) + gN-5(n-1) */
00182       gnext = __QADD(gcurr, (q31_t) (((q63_t) fcurr * (*pk++)) >> 31));
00183       /* y(n) += gN-4(n) * vN-4  */
00184       /* process for gN-8(n) * vN-8, gN-12(n) * vN-12 ... */
00185       acc += ((q63_t) gnext * *pv++);
00186       /* write gN-4(n) into state for next sample processing */
00187       *px2++ = gnext;
00188 
00189       tapCnt--;
00190 
00191     }
00192 
00193     fnext = fcurr;
00194 
00195     /* If the filter length is not a multiple of 4, compute the remaining filter taps */
00196     tapCnt = (numStages - 1u) % 0x4u;
00197 
00198     while(tapCnt > 0u)
00199     {
00200       gcurr = *px1++;
00201       /* Process sample for last taps */
00202       fnext = __QSUB(fcurr, (q31_t) (((q63_t) gcurr * (*pk)) >> 31));
00203       gnext = __QADD(gcurr, (q31_t) (((q63_t) fnext * (*pk++)) >> 31));
00204       /* Output samples for last taps */
00205       acc += ((q63_t) gnext * *pv++);
00206       *px2++ = gnext;
00207       fcurr = fnext;
00208 
00209       tapCnt--;
00210 
00211     }
00212 
00213     /* y(n) += g0(n) * v0 */
00214     acc += (q63_t) fnext *(
00215   *pv++);
00216 
00217     *px2++ = fnext;
00218 
00219     /* write out into pDst */
00220     *pDst++ = (q31_t) (acc >> 31u);
00221 
00222     /* Advance the state pointer by 4 to process the next group of 4 samples */
00223     pState = pState + 1u;
00224     blkCnt--;
00225 
00226   }
00227 
00228   /* Processing is complete. Now copy last S->numStages samples to start of the buffer    
00229      for the preperation of next frame process */
00230 
00231   /* Points to the start of the state buffer */
00232   pStateCurnt = &S->pState[0];
00233   pState = &S->pState[blockSize];
00234 
00235   tapCnt = numStages >> 2u;
00236 
00237   /* copy data */
00238   while(tapCnt > 0u)
00239   {
00240     *pStateCurnt++ = *pState++;
00241     *pStateCurnt++ = *pState++;
00242     *pStateCurnt++ = *pState++;
00243     *pStateCurnt++ = *pState++;
00244 
00245     /* Decrement the loop counter */
00246     tapCnt--;
00247 
00248   }
00249 
00250   /* Calculate remaining number of copies */
00251   tapCnt = (numStages) % 0x4u;
00252 
00253   /* Copy the remaining q31_t data */
00254   while(tapCnt > 0u)
00255   {
00256     *pStateCurnt++ = *pState++;
00257 
00258     /* Decrement the loop counter */
00259     tapCnt--;
00260   };
00261 
00262 #else
00263 
00264   /* Run the below code for Cortex-M0 */
00265   /* Sample processing */
00266   while(blkCnt > 0u)
00267   {
00268     /* Read Sample from input buffer */
00269     /* fN(n) = x(n) */
00270     fcurr = *pSrc++;
00271 
00272     /* Initialize state read pointer */
00273     px1 = pState;
00274     /* Initialize state write pointer */
00275     px2 = pState;
00276     /* Set accumulator to zero */
00277     acc = 0;
00278     /* Initialize Ladder coeff pointer */
00279     pv = &S->pvCoeffs[0];
00280     /* Initialize Reflection coeff pointer */
00281     pk = &S->pkCoeffs[0];
00282 
00283     tapCnt = numStages;
00284 
00285     while(tapCnt > 0u)
00286     {
00287       gcurr = *px1++;
00288       /* Process sample */
00289       /* fN-1(n) = fN(n) - kN * gN-1(n-1) */
00290       fnext =
00291         clip_q63_to_q31(((q63_t) fcurr -
00292                          ((q31_t) (((q63_t) gcurr * (*pk)) >> 31))));
00293       /* gN(n) = kN * fN-1(n) + gN-1(n-1) */
00294       gnext =
00295         clip_q63_to_q31(((q63_t) gcurr +
00296                          ((q31_t) (((q63_t) fnext * (*pk++)) >> 31))));
00297       /* Output samples */
00298       /* y(n) += gN(n) * vN  */
00299       acc += ((q63_t) gnext * *pv++);
00300       /* write gN-1(n-1) into state for next sample processing */
00301       *px2++ = gnext;
00302       /* Update f values for next coefficient processing */
00303       fcurr = fnext;
00304 
00305       tapCnt--;
00306     }
00307 
00308     /* y(n) += g0(n) * v0 */
00309     acc += (q63_t) fnext *(
00310   *pv++);
00311 
00312     *px2++ = fnext;
00313 
00314     /* write out into pDst */
00315     *pDst++ = (q31_t) (acc >> 31u);
00316 
00317     /* Advance the state pointer by 1 to process the next group of samples */
00318     pState = pState + 1u;
00319     blkCnt--;
00320 
00321   }
00322 
00323   /* Processing is complete. Now copy last S->numStages samples to start of the buffer           
00324      for the preperation of next frame process */
00325 
00326   /* Points to the start of the state buffer */
00327   pStateCurnt = &S->pState[0];
00328   pState = &S->pState[blockSize];
00329 
00330   tapCnt = numStages;
00331 
00332   /* Copy the remaining q31_t data */
00333   while(tapCnt > 0u)
00334   {
00335     *pStateCurnt++ = *pState++;
00336 
00337     /* Decrement the loop counter */
00338     tapCnt--;
00339   }
00340 
00341 #endif /*   #ifndef ARM_MATH_CM0_FAMILY */
00342 
00343 }
00344 
00345 
00346 
00347 
00348 /**    
00349  * @} end of IIR_Lattice group    
00350  */