arm_fir_decimate_fast_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_fir_decimate_fast_q15.c  
00009 *  
00010 * Description:  Fast Q15 FIR Decimator.  
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  
00027 #include "arm_math.h" 
00028  
00029 /**  
00030  * @ingroup groupFilters  
00031  */ 
00032  
00033 /**  
00034  * @addtogroup FIR_decimate  
00035  * @{  
00036  */ 
00037  
00038 /**  
00039  * @brief Processing function for the Q15 FIR decimator (fast variant).  
00040  * @param[in] *S points to an instance of the Q15 FIR decimator structure.  
00041  * @param[in] *pSrc points to the block of input data.  
00042  * @param[out] *pDst points to the block of output data  
00043  * @param[in] blockSize number of input samples to process per call.  
00044  * @return none  
00045  *  
00046  * <b>Scaling and Overflow Behavior:</b>  
00047  * \par  
00048  * This fast version uses a 32-bit accumulator with 2.30 format.  
00049  * The accumulator maintains full precision of the intermediate multiplication results but provides only a single guard bit.  
00050  * Thus, if the accumulator result overflows it wraps around and distorts the result.  
00051  * In order to avoid overflows completely the input signal must be scaled down by log2(numTaps) bits (log2 is read as log to the base 2).  
00052  * The 2.30 accumulator is then truncated to 2.15 format and saturated to yield the 1.15 result.  
00053  *  
00054  * \par  
00055  * Refer to the function <code>arm_fir_decimate_q15()</code> for a slower implementation of this function which uses 64-bit accumulation to avoid wrap around distortion.  
00056  * Both the slow and the fast versions use the same instance structure.  
00057  * Use the function <code>arm_fir_decimate_init_q15()</code> to initialize the filter structure.  
00058  */ 
00059  
00060 void arm_fir_decimate_fast_q15( 
00061   const arm_fir_decimate_instance_q15 * S, 
00062   q15_t * pSrc, 
00063   q15_t * pDst, 
00064   uint32_t blockSize) 
00065 { 
00066   q15_t *pState = S->pState;                     /* State pointer */ 
00067   q15_t *pCoeffs = S->pCoeffs;                   /* Coefficient pointer */ 
00068   q15_t *pStateCurnt;                            /* Points to the current sample of the state */ 
00069   q15_t *px;                                     /* Temporary pointer for state buffer */ 
00070   q15_t *pb;                                     /* Temporary pointer coefficient buffer */ 
00071   q31_t x0, c0;                                  /* Temporary variables to hold state and coefficient values */ 
00072   q31_t sum0;                                    /* Accumulators */ 
00073   uint32_t numTaps = S->numTaps;                 /* Number of taps */ 
00074   uint32_t i, blkCnt, tapCnt, outBlockSize = blockSize / S->M;  /* Loop counters */ 
00075  
00076  
00077   /* S->pState buffer contains previous frame (numTaps - 1) samples */ 
00078   /* pStateCurnt points to the location where the new input data should be written */ 
00079   pStateCurnt = S->pState + (numTaps - 1u); 
00080  
00081   /* Total number of output samples to be computed */ 
00082   blkCnt = outBlockSize; 
00083  
00084   while(blkCnt > 0u) 
00085   { 
00086     /* Copy decimation factor number of new input samples into the state buffer */ 
00087     i = S->M; 
00088  
00089     do 
00090     { 
00091       *pStateCurnt++ = *pSrc++; 
00092  
00093     } while(--i); 
00094  
00095     /*Set sum to zero */ 
00096     sum0 = 0; 
00097  
00098     /* Initialize state pointer */ 
00099     px = pState; 
00100  
00101     /* Initialize coeff pointer */ 
00102     pb = pCoeffs; 
00103  
00104     /* Loop unrolling.  Process 4 taps at a time. */ 
00105     tapCnt = numTaps >> 2; 
00106  
00107     /* Loop over the number of taps.  Unroll by a factor of 4.  
00108      ** Repeat until we've computed numTaps-4 coefficients. */ 
00109     while(tapCnt > 0u) 
00110     { 
00111       /* Read the Read b[numTaps-1] and b[numTaps-2]  coefficients */ 
00112       c0 = *__SIMD32(pb)++; 
00113  
00114       /* Read x[n-numTaps-1] and x[n-numTaps-2]sample */ 
00115       x0 = *__SIMD32(px)++; 
00116  
00117       /* Perform the multiply-accumulate */ 
00118       sum0 = __SMLAD(x0, c0, sum0); 
00119  
00120       /* Read the b[numTaps-3] and b[numTaps-4] coefficient */ 
00121       c0 = *__SIMD32(pb)++; 
00122  
00123       /* Read x[n-numTaps-2] and x[n-numTaps-3] sample */ 
00124       x0 = *__SIMD32(px)++; 
00125  
00126       /* Perform the multiply-accumulate */ 
00127       sum0 = __SMLAD(x0, c0, sum0); 
00128  
00129       /* Decrement the loop counter */ 
00130       tapCnt--; 
00131     } 
00132  
00133     /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 
00134     tapCnt = numTaps % 0x4u; 
00135  
00136     while(tapCnt > 0u) 
00137     { 
00138       /* Read coefficients */ 
00139       c0 = *pb++; 
00140  
00141       /* Fetch 1 state variable */ 
00142       x0 = *px++; 
00143  
00144       /* Perform the multiply-accumulate */ 
00145       sum0 = __SMLAD(x0, c0, sum0); 
00146  
00147       /* Decrement the loop counter */ 
00148       tapCnt--; 
00149     } 
00150  
00151     /* Advance the state pointer by the decimation factor  
00152      * to process the next group of decimation factor number samples */ 
00153     pState = pState + S->M; 
00154  
00155     /* Store filter output , smlad returns the values in 2.14 format */ 
00156     /* so downsacle by 15 to get output in 1.15 */ 
00157     *pDst++ = (q15_t) ((sum0 >> 15)); 
00158  
00159     /* Decrement the loop counter */ 
00160     blkCnt--; 
00161   } 
00162  
00163   /* Processing is complete.  
00164    ** Now copy the last numTaps - 1 samples to the satrt of the state buffer.  
00165    ** This prepares the state buffer for the next function call. */ 
00166  
00167   /* Points to the start of the state buffer */ 
00168   pStateCurnt = S->pState; 
00169  
00170   i = (numTaps - 1u) >> 2u; 
00171  
00172   /* copy data */ 
00173   while(i > 0u) 
00174   { 
00175     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++; 
00176     *__SIMD32(pStateCurnt)++ = *__SIMD32(pState)++; 
00177  
00178     /* Decrement the loop counter */ 
00179     i--; 
00180   } 
00181  
00182   i = (numTaps - 1u) % 0x04u; 
00183  
00184   /* copy data */ 
00185   while(i > 0u) 
00186   { 
00187     *pStateCurnt++ = *pState++; 
00188  
00189     /* Decrement the loop counter */ 
00190     i--; 
00191   } 
00192 } 
00193  
00194 /**  
00195  * @} end of FIR_decimate group  
00196  */