arm_lms_q15.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_lms_q15.c  
00009 *  
00010 * Description:  Processing function for the Q15 LMS 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  * @ingroup groupFilters  
00033  */ 
00034  
00035 /**  
00036  * @addtogroup LMS  
00037  * @{  
00038  */ 
00039  
00040  /**  
00041  * @brief Processing function for Q15 LMS filter.  
00042  * @param[in] *S points to an instance of the Q15 LMS filter structure.  
00043  * @param[in] *pSrc points to the block of input data.  
00044  * @param[in] *pRef points to the block of reference data.  
00045  * @param[out] *pOut points to the block of output data.  
00046  * @param[out] *pErr points to the block of error data.  
00047  * @param[in] blockSize number of samples to process.  
00048  * @return none.  
00049  *  
00050  * \par Scaling and Overflow Behavior:  
00051  * The function is implemented using a 64-bit internal accumulator.  
00052  * Both coefficients and state variables are represented in 1.15 format and multiplications yield a 2.30 result.  
00053  * The 2.30 intermediate results are accumulated in a 64-bit accumulator in 34.30 format.  
00054  * There is no risk of internal overflow with this approach and the full precision of intermediate multiplications is preserved.  
00055  * After all additions have been performed, the accumulator is truncated to 34.15 format by discarding low 15 bits.  
00056  * Lastly, the accumulator is saturated to yield a result in 1.15 format.  
00057  * 
00058  * \par 
00059  *  In this filter, filter coefficients are updated for each sample and the updation of filter cofficients are saturted. 
00060  *  
00061  */ 
00062  
00063 void arm_lms_q15( 
00064   const arm_lms_instance_q15 * S, 
00065   q15_t * pSrc, 
00066   q15_t * pRef, 
00067   q15_t * pOut, 
00068   q15_t * pErr, 
00069   uint32_t blockSize) 
00070 { 
00071   q15_t *pState = S->pState;                     /* State pointer */ 
00072   uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */ 
00073   q15_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */ 
00074   q15_t *pStateCurnt;                            /* Points to the current sample of the state */ 
00075   q15_t mu = S->mu;                              /* Adaptive factor */ 
00076   q15_t *px;                                     /* Temporary pointer for state */ 
00077   q15_t *pb;                                     /* Temporary pointer for coefficient buffer */ 
00078   uint32_t tapCnt, blkCnt;                       /* Loop counters */ 
00079   q63_t acc;                                     /* Accumulator */ 
00080   q15_t e = 0;                                   /* error of data sample */ 
00081   q15_t alpha;                                   /* Intermediate constant for taps update */ 
00082   uint32_t shift = S->postShift + 1u;            /* Shift to be applied to the output */ 
00083   q31_t coef;                                    /* Teporary variable for coefficient */ 
00084  
00085   /* S->pState points to buffer which contains previous frame (numTaps - 1) samples */ 
00086   /* pStateCurnt points to the location where the new input data should be written */ 
00087   pStateCurnt = &(S->pState[(numTaps - 1u)]); 
00088  
00089   /* Initializing blkCnt with blockSize */ 
00090   blkCnt = blockSize; 
00091  
00092   while(blkCnt > 0u) 
00093   { 
00094     /* Copy the new input sample into the state buffer */ 
00095     *pStateCurnt++ = *pSrc++; 
00096  
00097     /* Initialize state pointer */ 
00098     px = pState; 
00099  
00100     /* Initialize coefficient pointer */ 
00101     pb = pCoeffs; 
00102  
00103     /* Set the accumulator to zero */ 
00104     acc = 0; 
00105  
00106     /* Loop unrolling.  Process 4 taps at a time. */ 
00107     tapCnt = numTaps >> 2u; 
00108  
00109     while(tapCnt > 0u) 
00110     { 
00111       /* acc +=  b[N] * x[n-N] + b[N-1] * x[n-N-1] */ 
00112       /* Perform the multiply-accumulate */ 
00113       acc = __SMLALD(*__SIMD32(px)++, (*__SIMD32(pb)++), acc); 
00114       acc = __SMLALD(*__SIMD32(px)++, (*__SIMD32(pb)++), acc); 
00115  
00116       /* Decrement the loop counter */ 
00117       tapCnt--; 
00118     } 
00119  
00120     /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 
00121     tapCnt = numTaps % 0x4u; 
00122  
00123     while(tapCnt > 0u) 
00124     { 
00125       /* Perform the multiply-accumulate */ 
00126       acc += (q63_t) (((q31_t) (*px++) * (*pb++))); 
00127  
00128       /* Decrement the loop counter */ 
00129       tapCnt--; 
00130     } 
00131  
00132     /* Converting the result to 1.15 format and saturate the output */ 
00133     acc = __SSAT((acc >> (16 - shift)), 16); 
00134  
00135     /* Store the result from accumulator into the destination buffer. */ 
00136     *pOut++ = (q15_t) acc; 
00137  
00138     /* Compute and store error */ 
00139     e = *pRef++ - (q15_t) acc; 
00140  
00141     *pErr++ = (q15_t) e; 
00142  
00143     /* Compute alpha i.e. intermediate constant for taps update */ 
00144     alpha = (q15_t) (((q31_t) e * (mu)) >> 15); 
00145  
00146     /* Initialize state pointer */ 
00147     /* Advance state pointer by 1 for the next sample */ 
00148     px = pState++; 
00149  
00150     /* Initialize coefficient pointer */ 
00151     pb = pCoeffs; 
00152  
00153     /* Loop unrolling.  Process 4 taps at a time. */ 
00154     tapCnt = numTaps >> 2u; 
00155  
00156     /* Update filter coefficients */ 
00157     while(tapCnt > 0u) 
00158     { 
00159       coef = (q31_t) *pb + (((q31_t) alpha * (*px++)) >> 15); 
00160       *pb++ = (q15_t) __SSAT((coef), 16); 
00161       coef = (q31_t) *pb + (((q31_t) alpha * (*px++)) >> 15); 
00162       *pb++ = (q15_t) __SSAT((coef), 16); 
00163       coef = (q31_t) *pb + (((q31_t) alpha * (*px++)) >> 15); 
00164       *pb++ = (q15_t) __SSAT((coef), 16); 
00165       coef = (q31_t) *pb + (((q31_t) alpha * (*px++)) >> 15); 
00166       *pb++ = (q15_t) __SSAT((coef), 16); 
00167  
00168       /* Decrement the loop counter */ 
00169       tapCnt--; 
00170     } 
00171  
00172     /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 
00173     tapCnt = numTaps % 0x4u; 
00174  
00175     while(tapCnt > 0u) 
00176     { 
00177       /* Perform the multiply-accumulate */ 
00178       coef = (q31_t) *pb + (((q31_t) alpha * (*px++)) >> 15); 
00179       *pb++ = (q15_t) __SSAT((coef), 16); 
00180  
00181       /* Decrement the loop counter */ 
00182       tapCnt--; 
00183     } 
00184  
00185     /* Decrement the loop counter */ 
00186     blkCnt--; 
00187  
00188   } 
00189  
00190   /* Processing is complete. Now copy the last numTaps - 1 samples to the  
00191      satrt of the state buffer. This prepares the state buffer for the  
00192      next function call. */ 
00193  
00194   /* Points to the start of the pState buffer */ 
00195   pStateCurnt = S->pState; 
00196  
00197   /* Calculation of count for copying integer writes */ 
00198   tapCnt = (numTaps - 1u) >> 2; 
00199  
00200   while(tapCnt > 0u) 
00201   { 
00202  
00203     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++; 
00204     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++; 
00205  
00206     tapCnt--; 
00207  
00208   } 
00209  
00210   /* Calculation of count for remaining q15_t data */ 
00211   tapCnt = (numTaps - 1u) % 0x4u; 
00212  
00213   /* copy data */ 
00214   while(tapCnt > 0u) 
00215   { 
00216     *pStateCurnt++ = *pState++; 
00217  
00218     /* Decrement the loop counter */ 
00219     tapCnt--; 
00220   } 
00221  
00222 } 
00223  
00224 /**  
00225    * @} end of LMS group  
00226    */