mbed official / mbed-dsp

CMSIS DSP library

Dependents:   performance_timer Surfboard_ gps2rtty Capstone ... more

Legacy Warning

This is an mbed 2 library. To learn more about mbed OS 5, visit the docs.

cmsis_dsp/FilteringFunctions/arm_fir_sparse_q15.c@5:3762170b6d4d, 2015-11-20 (annotated)

Committer:
mbed_official
Date:
Fri Nov 20 08:45:18 2015 +0000
Revision:
5:3762170b6d4d
Parent:
3:7a284390b0ce 
Synchronized with git revision 2eb940b9a73af188d3004a2575fdfbb05febe62b



Full URL: https://github.com/mbedmicro/mbed/commit/2eb940b9a73af188d3004a2575fdfbb05febe62b/



Added option to build rpc library. closes #1426

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 5:3762170b6d4d 2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 5:3762170b6d4d 4 * $Date: 19. March 2015
mbed_official 5:3762170b6d4d 5 * $Revision: V.1.4.5
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_fir_sparse_q15.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Q15 sparse FIR filter processing function.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 13 *
mbed_official 3:7a284390b0ce 14 * Redistribution and use in source and binary forms, with or without
mbed_official 3:7a284390b0ce 15 * modification, are permitted provided that the following conditions
mbed_official 3:7a284390b0ce 16 * are met:
mbed_official 3:7a284390b0ce 17 * - Redistributions of source code must retain the above copyright
mbed_official 3:7a284390b0ce 18 * notice, this list of conditions and the following disclaimer.
mbed_official 3:7a284390b0ce 19 * - Redistributions in binary form must reproduce the above copyright
mbed_official 3:7a284390b0ce 20 * notice, this list of conditions and the following disclaimer in
mbed_official 3:7a284390b0ce 21 * the documentation and/or other materials provided with the
mbed_official 3:7a284390b0ce 22 * distribution.
mbed_official 3:7a284390b0ce 23 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 3:7a284390b0ce 24 * may be used to endorse or promote products derived from this
mbed_official 3:7a284390b0ce 25 * software without specific prior written permission.
mbed_official 3:7a284390b0ce 26 *
mbed_official 3:7a284390b0ce 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 3:7a284390b0ce 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 3:7a284390b0ce 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 3:7a284390b0ce 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 3:7a284390b0ce 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 3:7a284390b0ce 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 3:7a284390b0ce 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 3:7a284390b0ce 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 3:7a284390b0ce 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 3:7a284390b0ce 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 3:7a284390b0ce 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 3:7a284390b0ce 38 * POSSIBILITY OF SUCH DAMAGE.
emilmont 1:fdd22bb7aa52 39 * ------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 40 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 41
emilmont 1:fdd22bb7aa52 42 /**
emilmont 1:fdd22bb7aa52 43 * @addtogroup FIR_Sparse
emilmont 1:fdd22bb7aa52 44 * @{
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
emilmont 1:fdd22bb7aa52 47 /**
emilmont 1:fdd22bb7aa52 48 * @brief Processing function for the Q15 sparse FIR filter.
emilmont 1:fdd22bb7aa52 49 * @param[in] *S points to an instance of the Q15 sparse FIR structure.
emilmont 1:fdd22bb7aa52 50 * @param[in] *pSrc points to the block of input data.
emilmont 1:fdd22bb7aa52 51 * @param[out] *pDst points to the block of output data
emilmont 1:fdd22bb7aa52 52 * @param[in] *pScratchIn points to a temporary buffer of size blockSize.
emilmont 1:fdd22bb7aa52 53 * @param[in] *pScratchOut points to a temporary buffer of size blockSize.
emilmont 1:fdd22bb7aa52 54 * @param[in] blockSize number of input samples to process per call.
emilmont 1:fdd22bb7aa52 55 * @return none.
emilmont 1:fdd22bb7aa52 56 *
emilmont 1:fdd22bb7aa52 57 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 58 * \par
emilmont 1:fdd22bb7aa52 59 * The function is implemented using an internal 32-bit accumulator.
emilmont 1:fdd22bb7aa52 60 * The 1.15 x 1.15 multiplications yield a 2.30 result and these are added to a 2.30 accumulator.
emilmont 1:fdd22bb7aa52 61 * Thus the full precision of the multiplications is maintained but there is only a single guard bit in the accumulator.
emilmont 1:fdd22bb7aa52 62 * If the accumulator result overflows it will wrap around rather than saturate.
emilmont 1:fdd22bb7aa52 63 * After all multiply-accumulates are performed, the 2.30 accumulator is truncated to 2.15 format and then saturated to 1.15 format.
emilmont 1:fdd22bb7aa52 64 * In order to avoid overflows the input signal or coefficients must be scaled down by log2(numTaps) bits.
emilmont 1:fdd22bb7aa52 65 */
emilmont 1:fdd22bb7aa52 66
emilmont 1:fdd22bb7aa52 67
emilmont 1:fdd22bb7aa52 68 void arm_fir_sparse_q15(
emilmont 1:fdd22bb7aa52 69 arm_fir_sparse_instance_q15 * S,
emilmont 1:fdd22bb7aa52 70 q15_t * pSrc,
emilmont 1:fdd22bb7aa52 71 q15_t * pDst,
emilmont 1:fdd22bb7aa52 72 q15_t * pScratchIn,
emilmont 1:fdd22bb7aa52 73 q31_t * pScratchOut,
emilmont 1:fdd22bb7aa52 74 uint32_t blockSize)
emilmont 1:fdd22bb7aa52 75 {
emilmont 1:fdd22bb7aa52 76
emilmont 1:fdd22bb7aa52 77 q15_t *pState = S->pState; /* State pointer */
emilmont 1:fdd22bb7aa52 78 q15_t *pIn = pSrc; /* Working pointer for input */
emilmont 1:fdd22bb7aa52 79 q15_t *pOut = pDst; /* Working pointer for output */
emilmont 1:fdd22bb7aa52 80 q15_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */
emilmont 1:fdd22bb7aa52 81 q15_t *px; /* Temporary pointers for scratch buffer */
emilmont 1:fdd22bb7aa52 82 q15_t *pb = pScratchIn; /* Temporary pointers for scratch buffer */
emilmont 1:fdd22bb7aa52 83 q15_t *py = pState; /* Temporary pointers for state buffer */
emilmont 1:fdd22bb7aa52 84 int32_t *pTapDelay = S->pTapDelay; /* Pointer to the array containing offset of the non-zero tap values. */
emilmont 1:fdd22bb7aa52 85 uint32_t delaySize = S->maxDelay + blockSize; /* state length */
emilmont 1:fdd22bb7aa52 86 uint16_t numTaps = S->numTaps; /* Filter order */
emilmont 1:fdd22bb7aa52 87 int32_t readIndex; /* Read index of the state buffer */
emilmont 1:fdd22bb7aa52 88 uint32_t tapCnt, blkCnt; /* loop counters */
emilmont 1:fdd22bb7aa52 89 q15_t coeff = *pCoeffs++; /* Read the first coefficient value */
emilmont 1:fdd22bb7aa52 90 q31_t *pScr2 = pScratchOut; /* Working pointer for pScratchOut */
emilmont 1:fdd22bb7aa52 91
emilmont 1:fdd22bb7aa52 92
mbed_official 3:7a284390b0ce 93 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 94
emilmont 1:fdd22bb7aa52 95 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 96
emilmont 1:fdd22bb7aa52 97 q31_t in1, in2; /* Temporary variables */
emilmont 1:fdd22bb7aa52 98
emilmont 1:fdd22bb7aa52 99
emilmont 1:fdd22bb7aa52 100 /* BlockSize of Input samples are copied into the state buffer */
emilmont 1:fdd22bb7aa52 101 /* StateIndex points to the starting position to write in the state buffer */
emilmont 1:fdd22bb7aa52 102 arm_circularWrite_q15(py, delaySize, &S->stateIndex, 1, pIn, 1, blockSize);
emilmont 1:fdd22bb7aa52 103
emilmont 1:fdd22bb7aa52 104 /* Loop over the number of taps. */
emilmont 1:fdd22bb7aa52 105 tapCnt = numTaps;
emilmont 1:fdd22bb7aa52 106
emilmont 1:fdd22bb7aa52 107 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 108 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 109
emilmont 1:fdd22bb7aa52 110 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 111 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 112 {
emilmont 1:fdd22bb7aa52 113 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 114 }
emilmont 1:fdd22bb7aa52 115
emilmont 1:fdd22bb7aa52 116 /* Working pointer for state buffer is updated */
emilmont 1:fdd22bb7aa52 117 py = pState;
emilmont 1:fdd22bb7aa52 118
emilmont 1:fdd22bb7aa52 119 /* blockSize samples are read from the state buffer */
emilmont 1:fdd22bb7aa52 120 arm_circularRead_q15(py, delaySize, &readIndex, 1,
emilmont 1:fdd22bb7aa52 121 pb, pb, blockSize, 1, blockSize);
emilmont 1:fdd22bb7aa52 122
emilmont 1:fdd22bb7aa52 123 /* Working pointer for the scratch buffer of state values */
emilmont 1:fdd22bb7aa52 124 px = pb;
emilmont 1:fdd22bb7aa52 125
emilmont 1:fdd22bb7aa52 126 /* Working pointer for scratch buffer of output values */
emilmont 1:fdd22bb7aa52 127 pScratchOut = pScr2;
emilmont 1:fdd22bb7aa52 128
emilmont 1:fdd22bb7aa52 129 /* Loop over the blockSize. Unroll by a factor of 4.
emilmont 1:fdd22bb7aa52 130 * Compute 4 multiplications at a time. */
emilmont 1:fdd22bb7aa52 131 blkCnt = blockSize >> 2;
emilmont 1:fdd22bb7aa52 132
emilmont 1:fdd22bb7aa52 133 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 134 {
emilmont 1:fdd22bb7aa52 135 /* Perform multiplication and store in the scratch buffer */
emilmont 1:fdd22bb7aa52 136 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 137 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 138 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 139 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 140
emilmont 1:fdd22bb7aa52 141 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 142 blkCnt--;
emilmont 1:fdd22bb7aa52 143 }
emilmont 1:fdd22bb7aa52 144
emilmont 1:fdd22bb7aa52 145 /* If the blockSize is not a multiple of 4,
emilmont 1:fdd22bb7aa52 146 * compute the remaining samples */
emilmont 1:fdd22bb7aa52 147 blkCnt = blockSize % 0x4u;
emilmont 1:fdd22bb7aa52 148
emilmont 1:fdd22bb7aa52 149 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 150 {
emilmont 1:fdd22bb7aa52 151 /* Perform multiplication and store in the scratch buffer */
emilmont 1:fdd22bb7aa52 152 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 153
emilmont 1:fdd22bb7aa52 154 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 155 blkCnt--;
emilmont 1:fdd22bb7aa52 156 }
emilmont 1:fdd22bb7aa52 157
emilmont 1:fdd22bb7aa52 158 /* Load the coefficient value and
emilmont 1:fdd22bb7aa52 159 * increment the coefficient buffer for the next set of state values */
emilmont 1:fdd22bb7aa52 160 coeff = *pCoeffs++;
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 163 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 166 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 167 {
emilmont 1:fdd22bb7aa52 168 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 169 }
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* Loop over the number of taps. */
mbed_official 5:3762170b6d4d 172 tapCnt = (uint32_t) numTaps - 2u;
emilmont 1:fdd22bb7aa52 173
emilmont 1:fdd22bb7aa52 174 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 175 {
emilmont 1:fdd22bb7aa52 176 /* Working pointer for state buffer is updated */
emilmont 1:fdd22bb7aa52 177 py = pState;
emilmont 1:fdd22bb7aa52 178
emilmont 1:fdd22bb7aa52 179 /* blockSize samples are read from the state buffer */
emilmont 1:fdd22bb7aa52 180 arm_circularRead_q15(py, delaySize, &readIndex, 1,
emilmont 1:fdd22bb7aa52 181 pb, pb, blockSize, 1, blockSize);
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 /* Working pointer for the scratch buffer of state values */
emilmont 1:fdd22bb7aa52 184 px = pb;
emilmont 1:fdd22bb7aa52 185
emilmont 1:fdd22bb7aa52 186 /* Working pointer for scratch buffer of output values */
emilmont 1:fdd22bb7aa52 187 pScratchOut = pScr2;
emilmont 1:fdd22bb7aa52 188
emilmont 1:fdd22bb7aa52 189 /* Loop over the blockSize. Unroll by a factor of 4.
emilmont 1:fdd22bb7aa52 190 * Compute 4 MACS at a time. */
emilmont 1:fdd22bb7aa52 191 blkCnt = blockSize >> 2;
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 194 {
emilmont 1:fdd22bb7aa52 195 /* Perform Multiply-Accumulate */
emilmont 1:fdd22bb7aa52 196 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 197 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 198 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 199 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 200
emilmont 1:fdd22bb7aa52 201 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 202 blkCnt--;
emilmont 1:fdd22bb7aa52 203 }
emilmont 1:fdd22bb7aa52 204
emilmont 1:fdd22bb7aa52 205 /* If the blockSize is not a multiple of 4,
emilmont 1:fdd22bb7aa52 206 * compute the remaining samples */
emilmont 1:fdd22bb7aa52 207 blkCnt = blockSize % 0x4u;
emilmont 1:fdd22bb7aa52 208
emilmont 1:fdd22bb7aa52 209 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 210 {
emilmont 1:fdd22bb7aa52 211 /* Perform Multiply-Accumulate */
emilmont 1:fdd22bb7aa52 212 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 213
emilmont 1:fdd22bb7aa52 214 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 215 blkCnt--;
emilmont 1:fdd22bb7aa52 216 }
emilmont 1:fdd22bb7aa52 217
emilmont 1:fdd22bb7aa52 218 /* Load the coefficient value and
emilmont 1:fdd22bb7aa52 219 * increment the coefficient buffer for the next set of state values */
emilmont 1:fdd22bb7aa52 220 coeff = *pCoeffs++;
emilmont 1:fdd22bb7aa52 221
emilmont 1:fdd22bb7aa52 222 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 223 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 224
emilmont 1:fdd22bb7aa52 225 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 226 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 227 {
emilmont 1:fdd22bb7aa52 228 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 229 }
emilmont 1:fdd22bb7aa52 230
emilmont 1:fdd22bb7aa52 231 /* Decrement the tap loop counter */
emilmont 1:fdd22bb7aa52 232 tapCnt--;
emilmont 1:fdd22bb7aa52 233 }
mbed_official 5:3762170b6d4d 234
mbed_official 5:3762170b6d4d 235 /* Compute last tap without the final read of pTapDelay */
mbed_official 5:3762170b6d4d 236
mbed_official 5:3762170b6d4d 237 /* Working pointer for state buffer is updated */
mbed_official 5:3762170b6d4d 238 py = pState;
mbed_official 5:3762170b6d4d 239
mbed_official 5:3762170b6d4d 240 /* blockSize samples are read from the state buffer */
mbed_official 5:3762170b6d4d 241 arm_circularRead_q15(py, delaySize, &readIndex, 1,
mbed_official 5:3762170b6d4d 242 pb, pb, blockSize, 1, blockSize);
mbed_official 5:3762170b6d4d 243
mbed_official 5:3762170b6d4d 244 /* Working pointer for the scratch buffer of state values */
mbed_official 5:3762170b6d4d 245 px = pb;
mbed_official 5:3762170b6d4d 246
mbed_official 5:3762170b6d4d 247 /* Working pointer for scratch buffer of output values */
mbed_official 5:3762170b6d4d 248 pScratchOut = pScr2;
mbed_official 5:3762170b6d4d 249
mbed_official 5:3762170b6d4d 250 /* Loop over the blockSize. Unroll by a factor of 4.
mbed_official 5:3762170b6d4d 251 * Compute 4 MACS at a time. */
mbed_official 5:3762170b6d4d 252 blkCnt = blockSize >> 2;
mbed_official 5:3762170b6d4d 253
mbed_official 5:3762170b6d4d 254 while(blkCnt > 0u)
mbed_official 5:3762170b6d4d 255 {
mbed_official 5:3762170b6d4d 256 /* Perform Multiply-Accumulate */
mbed_official 5:3762170b6d4d 257 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 258 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 259 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 260 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 261
mbed_official 5:3762170b6d4d 262 /* Decrement the loop counter */
mbed_official 5:3762170b6d4d 263 blkCnt--;
mbed_official 5:3762170b6d4d 264 }
mbed_official 5:3762170b6d4d 265
mbed_official 5:3762170b6d4d 266 /* If the blockSize is not a multiple of 4,
mbed_official 5:3762170b6d4d 267 * compute the remaining samples */
mbed_official 5:3762170b6d4d 268 blkCnt = blockSize % 0x4u;
mbed_official 5:3762170b6d4d 269
mbed_official 5:3762170b6d4d 270 while(blkCnt > 0u)
mbed_official 5:3762170b6d4d 271 {
mbed_official 5:3762170b6d4d 272 /* Perform Multiply-Accumulate */
mbed_official 5:3762170b6d4d 273 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 274
mbed_official 5:3762170b6d4d 275 /* Decrement the loop counter */
mbed_official 5:3762170b6d4d 276 blkCnt--;
mbed_official 5:3762170b6d4d 277 }
emilmont 1:fdd22bb7aa52 278
emilmont 1:fdd22bb7aa52 279 /* All the output values are in pScratchOut buffer.
emilmont 1:fdd22bb7aa52 280 Convert them into 1.15 format, saturate and store in the destination buffer. */
emilmont 1:fdd22bb7aa52 281 /* Loop over the blockSize. */
emilmont 1:fdd22bb7aa52 282 blkCnt = blockSize >> 2;
emilmont 1:fdd22bb7aa52 283
emilmont 1:fdd22bb7aa52 284 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 285 {
emilmont 1:fdd22bb7aa52 286 in1 = *pScr2++;
emilmont 1:fdd22bb7aa52 287 in2 = *pScr2++;
emilmont 1:fdd22bb7aa52 288
emilmont 1:fdd22bb7aa52 289 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 292 __PKHBT((q15_t) __SSAT(in1 >> 15, 16), (q15_t) __SSAT(in2 >> 15, 16),
emilmont 1:fdd22bb7aa52 293 16);
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 #else
emilmont 1:fdd22bb7aa52 296 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 297 __PKHBT((q15_t) __SSAT(in2 >> 15, 16), (q15_t) __SSAT(in1 >> 15, 16),
emilmont 1:fdd22bb7aa52 298 16);
emilmont 1:fdd22bb7aa52 299
emilmont 1:fdd22bb7aa52 300 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 in1 = *pScr2++;
emilmont 1:fdd22bb7aa52 303
emilmont 1:fdd22bb7aa52 304 in2 = *pScr2++;
emilmont 1:fdd22bb7aa52 305
emilmont 1:fdd22bb7aa52 306 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 307
emilmont 1:fdd22bb7aa52 308 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 309 __PKHBT((q15_t) __SSAT(in1 >> 15, 16), (q15_t) __SSAT(in2 >> 15, 16),
emilmont 1:fdd22bb7aa52 310 16);
emilmont 1:fdd22bb7aa52 311
emilmont 1:fdd22bb7aa52 312 #else
emilmont 1:fdd22bb7aa52 313
emilmont 1:fdd22bb7aa52 314 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 315 __PKHBT((q15_t) __SSAT(in2 >> 15, 16), (q15_t) __SSAT(in1 >> 15, 16),
emilmont 1:fdd22bb7aa52 316 16);
emilmont 1:fdd22bb7aa52 317
emilmont 1:fdd22bb7aa52 318 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 319
emilmont 1:fdd22bb7aa52 320
emilmont 1:fdd22bb7aa52 321 blkCnt--;
emilmont 1:fdd22bb7aa52 322
emilmont 1:fdd22bb7aa52 323 }
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 /* If the blockSize is not a multiple of 4,
emilmont 1:fdd22bb7aa52 326 remaining samples are processed in the below loop */
emilmont 1:fdd22bb7aa52 327 blkCnt = blockSize % 0x4u;
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 330 {
emilmont 1:fdd22bb7aa52 331 *pOut++ = (q15_t) __SSAT(*pScr2++ >> 15, 16);
emilmont 1:fdd22bb7aa52 332 blkCnt--;
emilmont 1:fdd22bb7aa52 333 }
emilmont 1:fdd22bb7aa52 334
emilmont 1:fdd22bb7aa52 335 #else
emilmont 1:fdd22bb7aa52 336
emilmont 1:fdd22bb7aa52 337 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 338
emilmont 1:fdd22bb7aa52 339 /* BlockSize of Input samples are copied into the state buffer */
emilmont 1:fdd22bb7aa52 340 /* StateIndex points to the starting position to write in the state buffer */
emilmont 1:fdd22bb7aa52 341 arm_circularWrite_q15(py, delaySize, &S->stateIndex, 1, pIn, 1, blockSize);
emilmont 1:fdd22bb7aa52 342
emilmont 1:fdd22bb7aa52 343 /* Loop over the number of taps. */
emilmont 1:fdd22bb7aa52 344 tapCnt = numTaps;
emilmont 1:fdd22bb7aa52 345
emilmont 1:fdd22bb7aa52 346 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 347 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 348
emilmont 1:fdd22bb7aa52 349 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 350 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 351 {
emilmont 1:fdd22bb7aa52 352 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 353 }
emilmont 1:fdd22bb7aa52 354
emilmont 1:fdd22bb7aa52 355 /* Working pointer for state buffer is updated */
emilmont 1:fdd22bb7aa52 356 py = pState;
emilmont 1:fdd22bb7aa52 357
emilmont 1:fdd22bb7aa52 358 /* blockSize samples are read from the state buffer */
emilmont 1:fdd22bb7aa52 359 arm_circularRead_q15(py, delaySize, &readIndex, 1,
emilmont 1:fdd22bb7aa52 360 pb, pb, blockSize, 1, blockSize);
emilmont 1:fdd22bb7aa52 361
emilmont 1:fdd22bb7aa52 362 /* Working pointer for the scratch buffer of state values */
emilmont 1:fdd22bb7aa52 363 px = pb;
emilmont 1:fdd22bb7aa52 364
emilmont 1:fdd22bb7aa52 365 /* Working pointer for scratch buffer of output values */
emilmont 1:fdd22bb7aa52 366 pScratchOut = pScr2;
emilmont 1:fdd22bb7aa52 367
emilmont 1:fdd22bb7aa52 368 blkCnt = blockSize;
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 371 {
emilmont 1:fdd22bb7aa52 372 /* Perform multiplication and store in the scratch buffer */
emilmont 1:fdd22bb7aa52 373 *pScratchOut++ = ((q31_t) * px++ * coeff);
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 376 blkCnt--;
emilmont 1:fdd22bb7aa52 377 }
emilmont 1:fdd22bb7aa52 378
emilmont 1:fdd22bb7aa52 379 /* Load the coefficient value and
emilmont 1:fdd22bb7aa52 380 * increment the coefficient buffer for the next set of state values */
emilmont 1:fdd22bb7aa52 381 coeff = *pCoeffs++;
emilmont 1:fdd22bb7aa52 382
emilmont 1:fdd22bb7aa52 383 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 384 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 385
emilmont 1:fdd22bb7aa52 386 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 387 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 388 {
emilmont 1:fdd22bb7aa52 389 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 390 }
emilmont 1:fdd22bb7aa52 391
emilmont 1:fdd22bb7aa52 392 /* Loop over the number of taps. */
mbed_official 5:3762170b6d4d 393 tapCnt = (uint32_t) numTaps - 2u;
emilmont 1:fdd22bb7aa52 394
emilmont 1:fdd22bb7aa52 395 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 396 {
emilmont 1:fdd22bb7aa52 397 /* Working pointer for state buffer is updated */
emilmont 1:fdd22bb7aa52 398 py = pState;
emilmont 1:fdd22bb7aa52 399
emilmont 1:fdd22bb7aa52 400 /* blockSize samples are read from the state buffer */
emilmont 1:fdd22bb7aa52 401 arm_circularRead_q15(py, delaySize, &readIndex, 1,
emilmont 1:fdd22bb7aa52 402 pb, pb, blockSize, 1, blockSize);
emilmont 1:fdd22bb7aa52 403
emilmont 1:fdd22bb7aa52 404 /* Working pointer for the scratch buffer of state values */
emilmont 1:fdd22bb7aa52 405 px = pb;
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407 /* Working pointer for scratch buffer of output values */
emilmont 1:fdd22bb7aa52 408 pScratchOut = pScr2;
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 blkCnt = blockSize;
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 413 {
emilmont 1:fdd22bb7aa52 414 /* Perform Multiply-Accumulate */
emilmont 1:fdd22bb7aa52 415 *pScratchOut++ += (q31_t) * px++ * coeff;
emilmont 1:fdd22bb7aa52 416
emilmont 1:fdd22bb7aa52 417 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 418 blkCnt--;
emilmont 1:fdd22bb7aa52 419 }
emilmont 1:fdd22bb7aa52 420
emilmont 1:fdd22bb7aa52 421 /* Load the coefficient value and
emilmont 1:fdd22bb7aa52 422 * increment the coefficient buffer for the next set of state values */
emilmont 1:fdd22bb7aa52 423 coeff = *pCoeffs++;
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* Read Index, from where the state buffer should be read, is calculated. */
emilmont 1:fdd22bb7aa52 426 readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
emilmont 1:fdd22bb7aa52 427
emilmont 1:fdd22bb7aa52 428 /* Wraparound of readIndex */
emilmont 1:fdd22bb7aa52 429 if(readIndex < 0)
emilmont 1:fdd22bb7aa52 430 {
emilmont 1:fdd22bb7aa52 431 readIndex += (int32_t) delaySize;
emilmont 1:fdd22bb7aa52 432 }
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 /* Decrement the tap loop counter */
emilmont 1:fdd22bb7aa52 435 tapCnt--;
emilmont 1:fdd22bb7aa52 436 }
mbed_official 5:3762170b6d4d 437
mbed_official 5:3762170b6d4d 438 /* Compute last tap without the final read of pTapDelay */
mbed_official 5:3762170b6d4d 439
mbed_official 5:3762170b6d4d 440 /* Working pointer for state buffer is updated */
mbed_official 5:3762170b6d4d 441 py = pState;
mbed_official 5:3762170b6d4d 442
mbed_official 5:3762170b6d4d 443 /* blockSize samples are read from the state buffer */
mbed_official 5:3762170b6d4d 444 arm_circularRead_q15(py, delaySize, &readIndex, 1,
mbed_official 5:3762170b6d4d 445 pb, pb, blockSize, 1, blockSize);
mbed_official 5:3762170b6d4d 446
mbed_official 5:3762170b6d4d 447 /* Working pointer for the scratch buffer of state values */
mbed_official 5:3762170b6d4d 448 px = pb;
mbed_official 5:3762170b6d4d 449
mbed_official 5:3762170b6d4d 450 /* Working pointer for scratch buffer of output values */
mbed_official 5:3762170b6d4d 451 pScratchOut = pScr2;
mbed_official 5:3762170b6d4d 452
mbed_official 5:3762170b6d4d 453 blkCnt = blockSize;
mbed_official 5:3762170b6d4d 454
mbed_official 5:3762170b6d4d 455 while(blkCnt > 0u)
mbed_official 5:3762170b6d4d 456 {
mbed_official 5:3762170b6d4d 457 /* Perform Multiply-Accumulate */
mbed_official 5:3762170b6d4d 458 *pScratchOut++ += (q31_t) * px++ * coeff;
mbed_official 5:3762170b6d4d 459
mbed_official 5:3762170b6d4d 460 /* Decrement the loop counter */
mbed_official 5:3762170b6d4d 461 blkCnt--;
mbed_official 5:3762170b6d4d 462 }
emilmont 1:fdd22bb7aa52 463
emilmont 1:fdd22bb7aa52 464 /* All the output values are in pScratchOut buffer.
emilmont 1:fdd22bb7aa52 465 Convert them into 1.15 format, saturate and store in the destination buffer. */
emilmont 1:fdd22bb7aa52 466 /* Loop over the blockSize. */
emilmont 1:fdd22bb7aa52 467 blkCnt = blockSize;
emilmont 1:fdd22bb7aa52 468
emilmont 1:fdd22bb7aa52 469 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 470 {
emilmont 1:fdd22bb7aa52 471 *pOut++ = (q15_t) __SSAT(*pScr2++ >> 15, 16);
emilmont 1:fdd22bb7aa52 472 blkCnt--;
emilmont 1:fdd22bb7aa52 473 }
emilmont 1:fdd22bb7aa52 474
mbed_official 3:7a284390b0ce 475 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 476
emilmont 1:fdd22bb7aa52 477 }
emilmont 1:fdd22bb7aa52 478
emilmont 1:fdd22bb7aa52 479 /**
emilmont 1:fdd22bb7aa52 480 * @} end of FIR_Sparse group
emilmont 1:fdd22bb7aa52 481 */