00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2010 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. July 2011 00005 * $Revision: V1.0.10 00006 * 00007 * Project: CMSIS DSP Library 00008 * Title: arm_fir_decimate_fast_q31.c 00009 * 00010 * Description: Fast Q31 FIR Decimator. 00011 * 00012 * Target Processor: Cortex-M4/Cortex-M3 00013 * 00014 * Version 1.0.10 2011/7/15 00015 * Big Endian support added and Merged M0 and M3/M4 Source code. 00016 * 00017 * Version 1.0.3 2010/11/29 00018 * Re-organized the CMSIS folders and updated documentation. 00019 * 00020 * Version 1.0.2 2010/11/11 00021 * Documentation updated. 00022 * 00023 * Version 1.0.1 2010/10/05 00024 * Production release and review comments incorporated. 00025 * 00026 * Version 1.0.0 2010/09/20 00027 * Production release and review comments incorporated. 00028 * -------------------------------------------------------------------- */ 00029 00030 #include "arm_math.h" 00031 00065 void arm_fir_decimate_fast_q31( 00066 arm_fir_decimate_instance_q31 * S, 00067 q31_t * pSrc, 00068 q31_t * pDst, 00069 uint32_t blockSize) 00070 { 00071 q31_t *pState = S->pState; /* State pointer */ 00072 q31_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */ 00073 q31_t *pStateCurnt; /* Points to the current sample of the state */ 00074 q31_t x0, c0; /* Temporary variables to hold state and coefficient values */ 00075 q31_t *px; /* Temporary pointers for state buffer */ 00076 q31_t *pb; /* Temporary pointers for coefficient buffer */ 00077 q63_t sum0; /* Accumulator */ 00078 uint32_t numTaps = S->numTaps; /* Number of taps */ 00079 uint32_t i, tapCnt, blkCnt, outBlockSize = blockSize / S->M; /* Loop counters */ 00080 00081 00082 /* S->pState buffer contains previous frame (numTaps - 1) samples */ 00083 /* pStateCurnt points to the location where the new input data should be written */ 00084 pStateCurnt = S->pState + (numTaps - 1u); 00085 00086 /* Total number of output samples to be computed */ 00087 blkCnt = outBlockSize; 00088 00089 while(blkCnt > 0u) 00090 { 00091 /* Copy decimation factor number of new input samples into the state buffer */ 00092 i = S->M; 00093 00094 do 00095 { 00096 *pStateCurnt++ = *pSrc++; 00097 00098 } while(--i); 00099 00100 /* Set accumulator to zero */ 00101 sum0 = 0; 00102 00103 /* Initialize state pointer */ 00104 px = pState; 00105 00106 /* Initialize coeff pointer */ 00107 pb = pCoeffs; 00108 00109 /* Loop unrolling. Process 4 taps at a time. */ 00110 tapCnt = numTaps >> 2; 00111 00112 /* Loop over the number of taps. Unroll by a factor of 4. 00113 ** Repeat until we've computed numTaps-4 coefficients. */ 00114 while(tapCnt > 0u) 00115 { 00116 /* Read the b[numTaps-1] coefficient */ 00117 c0 = *(pb++); 00118 00119 /* Read x[n-numTaps-1] sample */ 00120 x0 = *(px++); 00121 00122 /* Perform the multiply-accumulate */ 00123 sum0 = (q31_t) ((((q63_t) x0 * c0) + (sum0 << 32)) >> 32); 00124 00125 /* Read the b[numTaps-2] coefficient */ 00126 c0 = *(pb++); 00127 00128 /* Read x[n-numTaps-2] sample */ 00129 x0 = *(px++); 00130 00131 /* Perform the multiply-accumulate */ 00132 sum0 = (q31_t) ((((q63_t) x0 * c0) + (sum0 << 32)) >> 32); 00133 00134 /* Read the b[numTaps-3] coefficient */ 00135 c0 = *(pb++); 00136 00137 /* Read x[n-numTaps-3] sample */ 00138 x0 = *(px++); 00139 00140 /* Perform the multiply-accumulate */ 00141 sum0 = (q31_t) ((((q63_t) x0 * c0) + (sum0 << 32)) >> 32); 00142 00143 /* Read the b[numTaps-4] coefficient */ 00144 c0 = *(pb++); 00145 00146 /* Read x[n-numTaps-4] sample */ 00147 x0 = *(px++); 00148 00149 /* Perform the multiply-accumulate */ 00150 sum0 = (q31_t) ((((q63_t) x0 * c0) + (sum0 << 32)) >> 32); 00151 00152 /* Decrement the loop counter */ 00153 tapCnt--; 00154 } 00155 00156 /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 00157 tapCnt = numTaps % 0x4u; 00158 00159 while(tapCnt > 0u) 00160 { 00161 /* Read coefficients */ 00162 c0 = *(pb++); 00163 00164 /* Fetch 1 state variable */ 00165 x0 = *(px++); 00166 00167 /* Perform the multiply-accumulate */ 00168 sum0 = (q31_t) ((((q63_t) x0 * c0) + (sum0 << 32)) >> 32); 00169 00170 /* Decrement the loop counter */ 00171 tapCnt--; 00172 } 00173 00174 /* Advance the state pointer by the decimation factor 00175 * to process the next group of decimation factor number samples */ 00176 pState = pState + S->M; 00177 00178 /* The result is in the accumulator, store in the destination buffer. */ 00179 *pDst++ = (q31_t) (sum0 << 1); 00180 00181 /* Decrement the loop counter */ 00182 blkCnt--; 00183 } 00184 00185 /* Processing is complete. 00186 ** Now copy the last numTaps - 1 samples to the satrt of the state buffer. 00187 ** This prepares the state buffer for the next function call. */ 00188 00189 /* Points to the start of the state buffer */ 00190 pStateCurnt = S->pState; 00191 00192 i = (numTaps - 1u) >> 2u; 00193 00194 /* copy data */ 00195 while(i > 0u) 00196 { 00197 *pStateCurnt++ = *pState++; 00198 *pStateCurnt++ = *pState++; 00199 *pStateCurnt++ = *pState++; 00200 *pStateCurnt++ = *pState++; 00201 00202 /* Decrement the loop counter */ 00203 i--; 00204 } 00205 00206 i = (numTaps - 1u) % 0x04u; 00207 00208 /* copy data */ 00209 while(i > 0u) 00210 { 00211 *pStateCurnt++ = *pState++; 00212 00213 /* Decrement the loop counter */ 00214 i--; 00215 } 00216 } 00217