arm_lms_norm_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_lms_norm_q31.c    
00009 *    
00010 * Description:  Processing function for the Q31 NLMS filter.    
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 LMS_NORM    
00049  * @{    
00050  */
00051 
00052 /**    
00053 * @brief Processing function for Q31 normalized LMS filter.    
00054 * @param[in] *S points to an instance of the Q31 normalized LMS filter structure.    
00055 * @param[in] *pSrc points to the block of input data.    
00056 * @param[in] *pRef points to the block of reference data.    
00057 * @param[out] *pOut points to the block of output data.    
00058 * @param[out] *pErr points to the block of error data.    
00059 * @param[in] blockSize number of samples to process.    
00060 * @return none.    
00061 *    
00062 * <b>Scaling and Overflow Behavior:</b>     
00063 * \par     
00064 * The function is implemented using an internal 64-bit accumulator.     
00065 * The accumulator has a 2.62 format and maintains full precision of the intermediate   
00066 * multiplication results but provides only a single guard bit.     
00067 * Thus, if the accumulator result overflows it wraps around rather than clip.     
00068 * In order to avoid overflows completely the input signal must be scaled down by    
00069 * log2(numTaps) bits. The reference signal should not be scaled down.     
00070 * After all multiply-accumulates are performed, the 2.62 accumulator is shifted    
00071 * and saturated to 1.31 format to yield the final result.     
00072 * The output signal and error signal are in 1.31 format.     
00073 *    
00074 * \par    
00075 *   In this filter, filter coefficients are updated for each sample and the    
00076 * updation of filter cofficients are saturted.    
00077 *     
00078 */
00079 
00080 void arm_lms_norm_q31(
00081   arm_lms_norm_instance_q31 * S,
00082   q31_t * pSrc,
00083   q31_t * pRef,
00084   q31_t * pOut,
00085   q31_t * pErr,
00086   uint32_t blockSize)
00087 {
00088   q31_t *pState = S->pState;                     /* State pointer */
00089   q31_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */
00090   q31_t *pStateCurnt;                            /* Points to the current sample of the state */
00091   q31_t *px, *pb;                                /* Temporary pointers for state and coefficient buffers */
00092   q31_t mu = S->mu;                              /* Adaptive factor */
00093   uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */
00094   uint32_t tapCnt, blkCnt;                       /* Loop counters */
00095   q63_t energy;                                  /* Energy of the input */
00096   q63_t acc;                                     /* Accumulator */
00097   q31_t e = 0, d = 0;                            /* error, reference data sample */
00098   q31_t w = 0, in;                               /* weight factor and state */
00099   q31_t x0;                                      /* temporary variable to hold input sample */
00100 //  uint32_t shift = 32u - ((uint32_t) S->postShift + 1u);        /* Shift to be applied to the output */      
00101   q31_t errorXmu, oneByEnergy;                   /* Temporary variables to store error and mu product and reciprocal of energy */
00102   q31_t postShift;                               /* Post shift to be applied to weight after reciprocal calculation */
00103   q31_t coef;                                    /* Temporary variable for coef */
00104   q31_t acc_l, acc_h;                            /*  temporary input */
00105   uint32_t uShift = ((uint32_t) S->postShift + 1u);
00106   uint32_t lShift = 32u - uShift;                /*  Shift to be applied to the output */
00107 
00108   energy = S->energy;
00109   x0 = S->x0;
00110 
00111   /* S->pState points to buffer which contains previous frame (numTaps - 1) samples */
00112   /* pStateCurnt points to the location where the new input data should be written */
00113   pStateCurnt = &(S->pState[(numTaps - 1u)]);
00114 
00115   /* Loop over blockSize number of values */
00116   blkCnt = blockSize;
00117 
00118 
00119 #ifndef ARM_MATH_CM0_FAMILY
00120 
00121   /* Run the below code for Cortex-M4 and Cortex-M3 */
00122 
00123   while(blkCnt > 0u)
00124   {
00125 
00126     /* Copy the new input sample into the state buffer */
00127     *pStateCurnt++ = *pSrc;
00128 
00129     /* Initialize pState pointer */
00130     px = pState;
00131 
00132     /* Initialize coeff pointer */
00133     pb = (pCoeffs);
00134 
00135     /* Read the sample from input buffer */
00136     in = *pSrc++;
00137 
00138     /* Update the energy calculation */
00139     energy = (q31_t) ((((q63_t) energy << 32) -
00140                        (((q63_t) x0 * x0) << 1)) >> 32);
00141     energy = (q31_t) (((((q63_t) in * in) << 1) + (energy << 32)) >> 32);
00142 
00143     /* Set the accumulator to zero */
00144     acc = 0;
00145 
00146     /* Loop unrolling.  Process 4 taps at a time. */
00147     tapCnt = numTaps >> 2;
00148 
00149     while(tapCnt > 0u)
00150     {
00151       /* Perform the multiply-accumulate */
00152       acc += ((q63_t) (*px++)) * (*pb++);
00153       acc += ((q63_t) (*px++)) * (*pb++);
00154       acc += ((q63_t) (*px++)) * (*pb++);
00155       acc += ((q63_t) (*px++)) * (*pb++);
00156 
00157       /* Decrement the loop counter */
00158       tapCnt--;
00159     }
00160 
00161     /* If the filter length is not a multiple of 4, compute the remaining filter taps */
00162     tapCnt = numTaps % 0x4u;
00163 
00164     while(tapCnt > 0u)
00165     {
00166       /* Perform the multiply-accumulate */
00167       acc += ((q63_t) (*px++)) * (*pb++);
00168 
00169       /* Decrement the loop counter */
00170       tapCnt--;
00171     }
00172 
00173     /* Converting the result to 1.31 format */
00174     /* Calc lower part of acc */
00175     acc_l = acc & 0xffffffff;
00176 
00177     /* Calc upper part of acc */
00178     acc_h = (acc >> 32) & 0xffffffff;
00179 
00180     acc = (uint32_t) acc_l >> lShift | acc_h << uShift;
00181 
00182     /* Store the result from accumulator into the destination buffer. */
00183     *pOut++ = (q31_t) acc;
00184 
00185     /* Compute and store error */
00186     d = *pRef++;
00187     e = d - (q31_t) acc;
00188     *pErr++ = e;
00189 
00190     /* Calculates the reciprocal of energy */
00191     postShift = arm_recip_q31(energy + DELTA_Q31,
00192                               &oneByEnergy, &S->recipTable[0]);
00193 
00194     /* Calculation of product of (e * mu) */
00195     errorXmu = (q31_t) (((q63_t) e * mu) >> 31);
00196 
00197     /* Weighting factor for the normalized version */
00198     w = clip_q63_to_q31(((q63_t) errorXmu * oneByEnergy) >> (31 - postShift));
00199 
00200     /* Initialize pState pointer */
00201     px = pState;
00202 
00203     /* Initialize coeff pointer */
00204     pb = (pCoeffs);
00205 
00206     /* Loop unrolling.  Process 4 taps at a time. */
00207     tapCnt = numTaps >> 2;
00208 
00209     /* Update filter coefficients */
00210     while(tapCnt > 0u)
00211     {
00212       /* Perform the multiply-accumulate */
00213 
00214       /* coef is in 2.30 format */
00215       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00216       /* get coef in 1.31 format by left shifting */
00217       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00218       /* update coefficient buffer to next coefficient */
00219       pb++;
00220 
00221       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00222       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00223       pb++;
00224 
00225       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00226       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00227       pb++;
00228 
00229       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00230       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00231       pb++;
00232 
00233       /* Decrement the loop counter */
00234       tapCnt--;
00235     }
00236 
00237     /* If the filter length is not a multiple of 4, compute the remaining filter taps */
00238     tapCnt = numTaps % 0x4u;
00239 
00240     while(tapCnt > 0u)
00241     {
00242       /* Perform the multiply-accumulate */
00243       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00244       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00245       pb++;
00246 
00247       /* Decrement the loop counter */
00248       tapCnt--;
00249     }
00250 
00251     /* Read the sample from state buffer */
00252     x0 = *pState;
00253 
00254     /* Advance state pointer by 1 for the next sample */
00255     pState = pState + 1;
00256 
00257     /* Decrement the loop counter */
00258     blkCnt--;
00259   }
00260 
00261   /* Save energy and x0 values for the next frame */
00262   S->energy = (q31_t) energy;
00263   S->x0 = x0;
00264 
00265   /* Processing is complete. Now copy the last numTaps - 1 samples to the    
00266      satrt of the state buffer. This prepares the state buffer for the    
00267      next function call. */
00268 
00269   /* Points to the start of the pState buffer */
00270   pStateCurnt = S->pState;
00271 
00272   /* Loop unrolling for (numTaps - 1u) samples copy */
00273   tapCnt = (numTaps - 1u) >> 2u;
00274 
00275   /* copy data */
00276   while(tapCnt > 0u)
00277   {
00278     *pStateCurnt++ = *pState++;
00279     *pStateCurnt++ = *pState++;
00280     *pStateCurnt++ = *pState++;
00281     *pStateCurnt++ = *pState++;
00282 
00283     /* Decrement the loop counter */
00284     tapCnt--;
00285   }
00286 
00287   /* Calculate remaining number of copies */
00288   tapCnt = (numTaps - 1u) % 0x4u;
00289 
00290   /* Copy the remaining q31_t data */
00291   while(tapCnt > 0u)
00292   {
00293     *pStateCurnt++ = *pState++;
00294 
00295     /* Decrement the loop counter */
00296     tapCnt--;
00297   }
00298 
00299 #else
00300 
00301   /* Run the below code for Cortex-M0 */
00302 
00303   while(blkCnt > 0u)
00304   {
00305 
00306     /* Copy the new input sample into the state buffer */
00307     *pStateCurnt++ = *pSrc;
00308 
00309     /* Initialize pState pointer */
00310     px = pState;
00311 
00312     /* Initialize pCoeffs pointer */
00313     pb = pCoeffs;
00314 
00315     /* Read the sample from input buffer */
00316     in = *pSrc++;
00317 
00318     /* Update the energy calculation */
00319     energy =
00320       (q31_t) ((((q63_t) energy << 32) - (((q63_t) x0 * x0) << 1)) >> 32);
00321     energy = (q31_t) (((((q63_t) in * in) << 1) + (energy << 32)) >> 32);
00322 
00323     /* Set the accumulator to zero */
00324     acc = 0;
00325 
00326     /* Loop over numTaps number of values */
00327     tapCnt = numTaps;
00328 
00329     while(tapCnt > 0u)
00330     {
00331       /* Perform the multiply-accumulate */
00332       acc += ((q63_t) (*px++)) * (*pb++);
00333 
00334       /* Decrement the loop counter */
00335       tapCnt--;
00336     }
00337 
00338     /* Converting the result to 1.31 format */
00339     /* Converting the result to 1.31 format */
00340     /* Calc lower part of acc */
00341     acc_l = acc & 0xffffffff;
00342 
00343     /* Calc upper part of acc */
00344     acc_h = (acc >> 32) & 0xffffffff;
00345 
00346     acc = (uint32_t) acc_l >> lShift | acc_h << uShift;
00347 
00348 
00349     //acc = (q31_t) (acc >> shift); 
00350 
00351     /* Store the result from accumulator into the destination buffer. */
00352     *pOut++ = (q31_t) acc;
00353 
00354     /* Compute and store error */
00355     d = *pRef++;
00356     e = d - (q31_t) acc;
00357     *pErr++ = e;
00358 
00359     /* Calculates the reciprocal of energy */
00360     postShift =
00361       arm_recip_q31(energy + DELTA_Q31, &oneByEnergy, &S->recipTable[0]);
00362 
00363     /* Calculation of product of (e * mu) */
00364     errorXmu = (q31_t) (((q63_t) e * mu) >> 31);
00365 
00366     /* Weighting factor for the normalized version */
00367     w = clip_q63_to_q31(((q63_t) errorXmu * oneByEnergy) >> (31 - postShift));
00368 
00369     /* Initialize pState pointer */
00370     px = pState;
00371 
00372     /* Initialize coeff pointer */
00373     pb = (pCoeffs);
00374 
00375     /* Loop over numTaps number of values */
00376     tapCnt = numTaps;
00377 
00378     while(tapCnt > 0u)
00379     {
00380       /* Perform the multiply-accumulate */
00381       /* coef is in 2.30 format */
00382       coef = (q31_t) (((q63_t) w * (*px++)) >> (32));
00383       /* get coef in 1.31 format by left shifting */
00384       *pb = clip_q63_to_q31((q63_t) * pb + (coef << 1u));
00385       /* update coefficient buffer to next coefficient */
00386       pb++;
00387 
00388       /* Decrement the loop counter */
00389       tapCnt--;
00390     }
00391 
00392     /* Read the sample from state buffer */
00393     x0 = *pState;
00394 
00395     /* Advance state pointer by 1 for the next sample */
00396     pState = pState + 1;
00397 
00398     /* Decrement the loop counter */
00399     blkCnt--;
00400   }
00401 
00402   /* Save energy and x0 values for the next frame */
00403   S->energy = (q31_t) energy;
00404   S->x0 = x0;
00405 
00406   /* Processing is complete. Now copy the last numTaps - 1 samples to the     
00407      start of the state buffer. This prepares the state buffer for the        
00408      next function call. */
00409 
00410   /* Points to the start of the pState buffer */
00411   pStateCurnt = S->pState;
00412 
00413   /* Loop for (numTaps - 1u) samples copy */
00414   tapCnt = (numTaps - 1u);
00415 
00416   /* Copy the remaining q31_t data */
00417   while(tapCnt > 0u)
00418   {
00419     *pStateCurnt++ = *pState++;
00420 
00421     /* Decrement the loop counter */
00422     tapCnt--;
00423   }
00424 
00425 #endif /*   #ifndef ARM_MATH_CM0_FAMILY */
00426 
00427 }
00428 
00429 /**    
00430  * @} end of LMS_NORM group    
00431  */