arm_lms_norm_q15.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_q15.c    
00009 *    
00010 * Description:  Q15 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 Q15 normalized LMS filter.    
00054 * @param[in] *S points to an instance of the Q15 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 a 64-bit internal accumulator.     
00065 * Both coefficients and state variables are represented in 1.15 format and    
00066 * multiplications yield a 2.30 result. The 2.30 intermediate results are    
00067 * accumulated in a 64-bit accumulator in 34.30 format.     
00068 * There is no risk of internal overflow with this approach and the full    
00069 * precision of intermediate multiplications is preserved. After all additions    
00070 * have been performed, the accumulator is truncated to 34.15 format by    
00071 * discarding low 15 bits. Lastly, the accumulator is saturated to yield a    
00072 * result in 1.15 format.    
00073 *    
00074 * \par   
00075 *   In this filter, filter coefficients are updated for each sample and the updation of filter cofficients are saturted.    
00076 *    
00077  */
00078 
00079 void arm_lms_norm_q15(
00080   arm_lms_norm_instance_q15 * S,
00081   q15_t * pSrc,
00082   q15_t * pRef,
00083   q15_t * pOut,
00084   q15_t * pErr,
00085   uint32_t blockSize)
00086 {
00087   q15_t *pState = S->pState;                     /* State pointer */
00088   q15_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */
00089   q15_t *pStateCurnt;                            /* Points to the current sample of the state */
00090   q15_t *px, *pb;                                /* Temporary pointers for state and coefficient buffers */
00091   q15_t mu = S->mu;                              /* Adaptive factor */
00092   uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */
00093   uint32_t tapCnt, blkCnt;                       /* Loop counters */
00094   q31_t energy;                                  /* Energy of the input */
00095   q63_t acc;                                     /* Accumulator */
00096   q15_t e = 0, d = 0;                            /* error, reference data sample */
00097   q15_t w = 0, in;                               /* weight factor and state */
00098   q15_t x0;                                      /* temporary variable to hold input sample */
00099   //uint32_t shift = (uint32_t) S->postShift + 1u; /* Shift to be applied to the output */ 
00100   q15_t errorXmu, oneByEnergy;                   /* Temporary variables to store error and mu product and reciprocal of energy */
00101   q15_t postShift;                               /* Post shift to be applied to weight after reciprocal calculation */
00102   q31_t coef;                                    /* Teporary variable for coefficient */
00103   q31_t acc_l, acc_h;
00104   int32_t lShift = (15 - (int32_t) S->postShift);       /*  Post shift  */
00105   int32_t uShift = (32 - lShift);
00106 
00107   energy = S->energy;
00108   x0 = S->x0;
00109 
00110   /* S->pState points to buffer which contains previous frame (numTaps - 1) samples */
00111   /* pStateCurnt points to the location where the new input data should be written */
00112   pStateCurnt = &(S->pState[(numTaps - 1u)]);
00113 
00114   /* Loop over blockSize number of values */
00115   blkCnt = blockSize;
00116 
00117 
00118 #ifndef ARM_MATH_CM0_FAMILY
00119 
00120   /* Run the below code for Cortex-M4 and Cortex-M3 */
00121 
00122   while(blkCnt > 0u)
00123   {
00124     /* Copy the new input sample into the state buffer */
00125     *pStateCurnt++ = *pSrc;
00126 
00127     /* Initialize pState pointer */
00128     px = pState;
00129 
00130     /* Initialize coeff pointer */
00131     pb = (pCoeffs);
00132 
00133     /* Read the sample from input buffer */
00134     in = *pSrc++;
00135 
00136     /* Update the energy calculation */
00137     energy -= (((q31_t) x0 * (x0)) >> 15);
00138     energy += (((q31_t) in * (in)) >> 15);
00139 
00140     /* Set the accumulator to zero */
00141     acc = 0;
00142 
00143     /* Loop unrolling.  Process 4 taps at a time. */
00144     tapCnt = numTaps >> 2;
00145 
00146     while(tapCnt > 0u)
00147     {
00148 
00149       /* Perform the multiply-accumulate */
00150 #ifndef UNALIGNED_SUPPORT_DISABLE
00151 
00152       acc = __SMLALD(*__SIMD32(px)++, (*__SIMD32(pb)++), acc);
00153       acc = __SMLALD(*__SIMD32(px)++, (*__SIMD32(pb)++), acc);
00154 
00155 #else
00156 
00157       acc += (((q31_t) * px++ * (*pb++)));
00158       acc += (((q31_t) * px++ * (*pb++)));
00159       acc += (((q31_t) * px++ * (*pb++)));
00160       acc += (((q31_t) * px++ * (*pb++)));
00161 
00162 #endif  /*  #ifndef UNALIGNED_SUPPORT_DISABLE   */
00163 
00164       /* Decrement the loop counter */
00165       tapCnt--;
00166     }
00167 
00168     /* If the filter length is not a multiple of 4, compute the remaining filter taps */
00169     tapCnt = numTaps % 0x4u;
00170 
00171     while(tapCnt > 0u)
00172     {
00173       /* Perform the multiply-accumulate */
00174       acc += (((q31_t) * px++ * (*pb++)));
00175 
00176       /* Decrement the loop counter */
00177       tapCnt--;
00178     }
00179 
00180     /* Calc lower part of acc */
00181     acc_l = acc & 0xffffffff;
00182 
00183     /* Calc upper part of acc */
00184     acc_h = (acc >> 32) & 0xffffffff;
00185 
00186     /* Apply shift for lower part of acc and upper part of acc */
00187     acc = (uint32_t) acc_l >> lShift | acc_h << uShift;
00188 
00189     /* Converting the result to 1.15 format and saturate the output */
00190     acc = __SSAT(acc, 16u);
00191 
00192     /* Store the result from accumulator into the destination buffer. */
00193     *pOut++ = (q15_t) acc;
00194 
00195     /* Compute and store error */
00196     d = *pRef++;
00197     e = d - (q15_t) acc;
00198     *pErr++ = e;
00199 
00200     /* Calculation of 1/energy */
00201     postShift = arm_recip_q15((q15_t) energy + DELTA_Q15,
00202                               &oneByEnergy, S->recipTable);
00203 
00204     /* Calculation of e * mu value */
00205     errorXmu = (q15_t) (((q31_t) e * mu) >> 15);
00206 
00207     /* Calculation of (e * mu) * (1/energy) value */
00208     acc = (((q31_t) errorXmu * oneByEnergy) >> (15 - postShift));
00209 
00210     /* Weighting factor for the normalized version */
00211     w = (q15_t) __SSAT((q31_t) acc, 16);
00212 
00213     /* Initialize pState pointer */
00214     px = pState;
00215 
00216     /* Initialize coeff pointer */
00217     pb = (pCoeffs);
00218 
00219     /* Loop unrolling.  Process 4 taps at a time. */
00220     tapCnt = numTaps >> 2;
00221 
00222     /* Update filter coefficients */
00223     while(tapCnt > 0u)
00224     {
00225       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00226       *pb++ = (q15_t) __SSAT((coef), 16);
00227       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00228       *pb++ = (q15_t) __SSAT((coef), 16);
00229       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00230       *pb++ = (q15_t) __SSAT((coef), 16);
00231       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00232       *pb++ = (q15_t) __SSAT((coef), 16);
00233 
00234       /* Decrement the loop counter */
00235       tapCnt--;
00236     }
00237 
00238     /* If the filter length is not a multiple of 4, compute the remaining filter taps */
00239     tapCnt = numTaps % 0x4u;
00240 
00241     while(tapCnt > 0u)
00242     {
00243       /* Perform the multiply-accumulate */
00244       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00245       *pb++ = (q15_t) __SSAT((coef), 16);
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 + 1u;
00256 
00257     /* Decrement the loop counter */
00258     blkCnt--;
00259   }
00260 
00261   /* Save energy and x0 values for the next frame */
00262   S->energy = (q15_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   /* Calculation of count for copying integer writes */
00273   tapCnt = (numTaps - 1u) >> 2;
00274 
00275   while(tapCnt > 0u)
00276   {
00277 
00278 #ifndef UNALIGNED_SUPPORT_DISABLE
00279 
00280     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++;
00281     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++;
00282 
00283 #else
00284 
00285     *pStateCurnt++ = *pState++;
00286     *pStateCurnt++ = *pState++;
00287     *pStateCurnt++ = *pState++;
00288     *pStateCurnt++ = *pState++;
00289 
00290 #endif
00291 
00292     tapCnt--;
00293 
00294   }
00295 
00296   /* Calculation of count for remaining q15_t data */
00297   tapCnt = (numTaps - 1u) % 0x4u;
00298 
00299   /* copy data */
00300   while(tapCnt > 0u)
00301   {
00302     *pStateCurnt++ = *pState++;
00303 
00304     /* Decrement the loop counter */
00305     tapCnt--;
00306   }
00307 
00308 #else
00309 
00310   /* Run the below code for Cortex-M0 */
00311 
00312   while(blkCnt > 0u)
00313   {
00314     /* Copy the new input sample into the state buffer */
00315     *pStateCurnt++ = *pSrc;
00316 
00317     /* Initialize pState pointer */
00318     px = pState;
00319 
00320     /* Initialize pCoeffs pointer */
00321     pb = pCoeffs;
00322 
00323     /* Read the sample from input buffer */
00324     in = *pSrc++;
00325 
00326     /* Update the energy calculation */
00327     energy -= (((q31_t) x0 * (x0)) >> 15);
00328     energy += (((q31_t) in * (in)) >> 15);
00329 
00330     /* Set the accumulator to zero */
00331     acc = 0;
00332 
00333     /* Loop over numTaps number of values */
00334     tapCnt = numTaps;
00335 
00336     while(tapCnt > 0u)
00337     {
00338       /* Perform the multiply-accumulate */
00339       acc += (((q31_t) * px++ * (*pb++)));
00340 
00341       /* Decrement the loop counter */
00342       tapCnt--;
00343     }
00344 
00345     /* Calc lower part of acc */
00346     acc_l = acc & 0xffffffff;
00347 
00348     /* Calc upper part of acc */
00349     acc_h = (acc >> 32) & 0xffffffff;
00350 
00351     /* Apply shift for lower part of acc and upper part of acc */
00352     acc = (uint32_t) acc_l >> lShift | acc_h << uShift;
00353 
00354     /* Converting the result to 1.15 format and saturate the output */
00355     acc = __SSAT(acc, 16u);
00356 
00357     /* Converting the result to 1.15 format */
00358     //acc = __SSAT((acc >> (16u - shift)), 16u); 
00359 
00360     /* Store the result from accumulator into the destination buffer. */
00361     *pOut++ = (q15_t) acc;
00362 
00363     /* Compute and store error */
00364     d = *pRef++;
00365     e = d - (q15_t) acc;
00366     *pErr++ = e;
00367 
00368     /* Calculation of 1/energy */
00369     postShift = arm_recip_q15((q15_t) energy + DELTA_Q15,
00370                               &oneByEnergy, S->recipTable);
00371 
00372     /* Calculation of e * mu value */
00373     errorXmu = (q15_t) (((q31_t) e * mu) >> 15);
00374 
00375     /* Calculation of (e * mu) * (1/energy) value */
00376     acc = (((q31_t) errorXmu * oneByEnergy) >> (15 - postShift));
00377 
00378     /* Weighting factor for the normalized version */
00379     w = (q15_t) __SSAT((q31_t) acc, 16);
00380 
00381     /* Initialize pState pointer */
00382     px = pState;
00383 
00384     /* Initialize coeff pointer */
00385     pb = (pCoeffs);
00386 
00387     /* Loop over numTaps number of values */
00388     tapCnt = numTaps;
00389 
00390     while(tapCnt > 0u)
00391     {
00392       /* Perform the multiply-accumulate */
00393       coef = *pb + (((q31_t) w * (*px++)) >> 15);
00394       *pb++ = (q15_t) __SSAT((coef), 16);
00395 
00396       /* Decrement the loop counter */
00397       tapCnt--;
00398     }
00399 
00400     /* Read the sample from state buffer */
00401     x0 = *pState;
00402 
00403     /* Advance state pointer by 1 for the next sample */
00404     pState = pState + 1u;
00405 
00406     /* Decrement the loop counter */
00407     blkCnt--;
00408   }
00409 
00410   /* Save energy and x0 values for the next frame */
00411   S->energy = (q15_t) energy;
00412   S->x0 = x0;
00413 
00414   /* Processing is complete. Now copy the last numTaps - 1 samples to the        
00415      satrt of the state buffer. This prepares the state buffer for the        
00416      next function call. */
00417 
00418   /* Points to the start of the pState buffer */
00419   pStateCurnt = S->pState;
00420 
00421   /* copy (numTaps - 1u) data */
00422   tapCnt = (numTaps - 1u);
00423 
00424   /* copy data */
00425   while(tapCnt > 0u)
00426   {
00427     *pStateCurnt++ = *pState++;
00428 
00429     /* Decrement the loop counter */
00430     tapCnt--;
00431   }
00432 
00433 #endif /*   #ifndef ARM_MATH_CM0_FAMILY */
00434 
00435 }
00436 
00437 
00438 /**    
00439    * @} end of LMS_NORM group    
00440    */