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_conv_f32.c@2:da51fb522205, 2013-05-30 (annotated)

Committer:
emilmont
Date:
Thu May 30 17:10:11 2013 +0100
Revision:
2:da51fb522205
Parent:
1:fdd22bb7aa52
Child:
3:7a284390b0ce
Keep "cmsis-dsp" module in synch with its source

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------------
emilmont 1:fdd22bb7aa52 2 * Copyright (C) 2010 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
emilmont 1:fdd22bb7aa52 4 * $Date: 15. February 2012
emilmont 2:da51fb522205 5 * $Revision: V1.1.0
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_conv_f32.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Convolution of floating-point sequences.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 13 *
emilmont 1:fdd22bb7aa52 14 * Version 1.1.0 2012/02/15
emilmont 1:fdd22bb7aa52 15 * Updated with more optimizations, bug fixes and minor API changes.
emilmont 1:fdd22bb7aa52 16 *
emilmont 1:fdd22bb7aa52 17 * Version 1.0.11 2011/10/18
emilmont 1:fdd22bb7aa52 18 * Bug Fix in conv, correlation, partial convolution.
emilmont 1:fdd22bb7aa52 19 *
emilmont 1:fdd22bb7aa52 20 * Version 1.0.10 2011/7/15
emilmont 1:fdd22bb7aa52 21 * Big Endian support added and Merged M0 and M3/M4 Source code.
emilmont 1:fdd22bb7aa52 22 *
emilmont 1:fdd22bb7aa52 23 * Version 1.0.3 2010/11/29
emilmont 1:fdd22bb7aa52 24 * Re-organized the CMSIS folders and updated documentation.
emilmont 1:fdd22bb7aa52 25 *
emilmont 1:fdd22bb7aa52 26 * Version 1.0.2 2010/11/11
emilmont 1:fdd22bb7aa52 27 * Documentation updated.
emilmont 1:fdd22bb7aa52 28 *
emilmont 1:fdd22bb7aa52 29 * Version 1.0.1 2010/10/05
emilmont 1:fdd22bb7aa52 30 * Production release and review comments incorporated.
emilmont 1:fdd22bb7aa52 31 *
emilmont 1:fdd22bb7aa52 32 * Version 1.0.0 2010/09/20
emilmont 1:fdd22bb7aa52 33 * Production release and review comments incorporated
emilmont 1:fdd22bb7aa52 34 *
emilmont 1:fdd22bb7aa52 35 * Version 0.0.7 2010/06/10
emilmont 1:fdd22bb7aa52 36 * Misra-C changes done
emilmont 1:fdd22bb7aa52 37 *
emilmont 1:fdd22bb7aa52 38 * -------------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 39
emilmont 1:fdd22bb7aa52 40 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 41
emilmont 1:fdd22bb7aa52 42 /**
emilmont 1:fdd22bb7aa52 43 * @ingroup groupFilters
emilmont 1:fdd22bb7aa52 44 */
emilmont 1:fdd22bb7aa52 45
emilmont 1:fdd22bb7aa52 46 /**
emilmont 1:fdd22bb7aa52 47 * @defgroup Conv Convolution
emilmont 1:fdd22bb7aa52 48 *
emilmont 1:fdd22bb7aa52 49 * Convolution is a mathematical operation that operates on two finite length vectors to generate a finite length output vector.
emilmont 1:fdd22bb7aa52 50 * Convolution is similar to correlation and is frequently used in filtering and data analysis.
emilmont 1:fdd22bb7aa52 51 * The CMSIS DSP library contains functions for convolving Q7, Q15, Q31, and floating-point data types.
emilmont 1:fdd22bb7aa52 52 * The library also provides fast versions of the Q15 and Q31 functions on Cortex-M4 and Cortex-M3.
emilmont 1:fdd22bb7aa52 53 *
emilmont 1:fdd22bb7aa52 54 * \par Algorithm
emilmont 1:fdd22bb7aa52 55 * Let <code>a[n]</code> and <code>b[n]</code> be sequences of length <code>srcALen</code> and <code>srcBLen</code> samples respectively.
emilmont 1:fdd22bb7aa52 56 * Then the convolution
emilmont 1:fdd22bb7aa52 57 *
emilmont 1:fdd22bb7aa52 58 * <pre>
emilmont 1:fdd22bb7aa52 59 * c[n] = a[n] * b[n]
emilmont 1:fdd22bb7aa52 60 * </pre>
emilmont 1:fdd22bb7aa52 61 *
emilmont 1:fdd22bb7aa52 62 * \par
emilmont 1:fdd22bb7aa52 63 * is defined as
emilmont 1:fdd22bb7aa52 64 * \image html ConvolutionEquation.gif
emilmont 1:fdd22bb7aa52 65 * \par
emilmont 1:fdd22bb7aa52 66 * Note that <code>c[n]</code> is of length <code>srcALen + srcBLen - 1</code> and is defined over the interval <code>n=0, 1, 2, ..., srcALen + srcBLen - 2</code>.
emilmont 1:fdd22bb7aa52 67 * <code>pSrcA</code> points to the first input vector of length <code>srcALen</code> and
emilmont 1:fdd22bb7aa52 68 * <code>pSrcB</code> points to the second input vector of length <code>srcBLen</code>.
emilmont 1:fdd22bb7aa52 69 * The output result is written to <code>pDst</code> and the calling function must allocate <code>srcALen+srcBLen-1</code> words for the result.
emilmont 1:fdd22bb7aa52 70 *
emilmont 1:fdd22bb7aa52 71 * \par
emilmont 1:fdd22bb7aa52 72 * Conceptually, when two signals <code>a[n]</code> and <code>b[n]</code> are convolved,
emilmont 1:fdd22bb7aa52 73 * the signal <code>b[n]</code> slides over <code>a[n]</code>.
emilmont 1:fdd22bb7aa52 74 * For each offset \c n, the overlapping portions of a[n] and b[n] are multiplied and summed together.
emilmont 1:fdd22bb7aa52 75 *
emilmont 1:fdd22bb7aa52 76 * \par
emilmont 1:fdd22bb7aa52 77 * Note that convolution is a commutative operation:
emilmont 1:fdd22bb7aa52 78 *
emilmont 1:fdd22bb7aa52 79 * <pre>
emilmont 1:fdd22bb7aa52 80 * a[n] * b[n] = b[n] * a[n].
emilmont 1:fdd22bb7aa52 81 * </pre>
emilmont 1:fdd22bb7aa52 82 *
emilmont 1:fdd22bb7aa52 83 * \par
emilmont 1:fdd22bb7aa52 84 * This means that switching the A and B arguments to the convolution functions has no effect.
emilmont 1:fdd22bb7aa52 85 *
emilmont 1:fdd22bb7aa52 86 * <b>Fixed-Point Behavior</b>
emilmont 1:fdd22bb7aa52 87 *
emilmont 1:fdd22bb7aa52 88 * \par
emilmont 1:fdd22bb7aa52 89 * Convolution requires summing up a large number of intermediate products.
emilmont 1:fdd22bb7aa52 90 * As such, the Q7, Q15, and Q31 functions run a risk of overflow and saturation.
emilmont 1:fdd22bb7aa52 91 * Refer to the function specific documentation below for further details of the particular algorithm used.
emilmont 1:fdd22bb7aa52 92 *
emilmont 1:fdd22bb7aa52 93 *
emilmont 1:fdd22bb7aa52 94 * <b>Fast Versions</b>
emilmont 1:fdd22bb7aa52 95 *
emilmont 1:fdd22bb7aa52 96 * \par
emilmont 1:fdd22bb7aa52 97 * Fast versions are supported for Q31 and Q15. Cycles for Fast versions are less compared to Q31 and Q15 of conv and the design requires
emilmont 1:fdd22bb7aa52 98 * the input signals should be scaled down to avoid intermediate overflows.
emilmont 1:fdd22bb7aa52 99 *
emilmont 1:fdd22bb7aa52 100 *
emilmont 1:fdd22bb7aa52 101 * <b>Opt Versions</b>
emilmont 1:fdd22bb7aa52 102 *
emilmont 1:fdd22bb7aa52 103 * \par
emilmont 1:fdd22bb7aa52 104 * Opt versions are supported for Q15 and Q7. Design uses internal scratch buffer for getting good optimisation.
emilmont 1:fdd22bb7aa52 105 * These versions are optimised in cycles and consumes more memory(Scratch memory) compared to Q15 and Q7 versions
emilmont 1:fdd22bb7aa52 106 */
emilmont 1:fdd22bb7aa52 107
emilmont 1:fdd22bb7aa52 108 /**
emilmont 1:fdd22bb7aa52 109 * @addtogroup Conv
emilmont 1:fdd22bb7aa52 110 * @{
emilmont 1:fdd22bb7aa52 111 */
emilmont 1:fdd22bb7aa52 112
emilmont 1:fdd22bb7aa52 113 /**
emilmont 1:fdd22bb7aa52 114 * @brief Convolution of floating-point sequences.
emilmont 1:fdd22bb7aa52 115 * @param[in] *pSrcA points to the first input sequence.
emilmont 1:fdd22bb7aa52 116 * @param[in] srcALen length of the first input sequence.
emilmont 1:fdd22bb7aa52 117 * @param[in] *pSrcB points to the second input sequence.
emilmont 1:fdd22bb7aa52 118 * @param[in] srcBLen length of the second input sequence.
emilmont 1:fdd22bb7aa52 119 * @param[out] *pDst points to the location where the output result is written. Length srcALen+srcBLen-1.
emilmont 1:fdd22bb7aa52 120 * @return none.
emilmont 1:fdd22bb7aa52 121 */
emilmont 1:fdd22bb7aa52 122
emilmont 1:fdd22bb7aa52 123 void arm_conv_f32(
emilmont 1:fdd22bb7aa52 124 float32_t * pSrcA,
emilmont 1:fdd22bb7aa52 125 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 126 float32_t * pSrcB,
emilmont 1:fdd22bb7aa52 127 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 128 float32_t * pDst)
emilmont 1:fdd22bb7aa52 129 {
emilmont 1:fdd22bb7aa52 130
emilmont 1:fdd22bb7aa52 131
emilmont 1:fdd22bb7aa52 132 #ifndef ARM_MATH_CM0
emilmont 1:fdd22bb7aa52 133
emilmont 1:fdd22bb7aa52 134 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 135
emilmont 1:fdd22bb7aa52 136 float32_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 137 float32_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 138 float32_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 139 float32_t *px; /* Intermediate inputA pointer */
emilmont 1:fdd22bb7aa52 140 float32_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 141 float32_t *pSrc1, *pSrc2; /* Intermediate pointers */
emilmont 1:fdd22bb7aa52 142 float32_t sum, acc0, acc1, acc2, acc3; /* Accumulator */
emilmont 1:fdd22bb7aa52 143 float32_t x0, x1, x2, x3, c0; /* Temporary variables to hold state and coefficient values */
emilmont 1:fdd22bb7aa52 144 uint32_t j, k, count, blkCnt, blockSize1, blockSize2, blockSize3; /* loop counters */
emilmont 1:fdd22bb7aa52 145
emilmont 1:fdd22bb7aa52 146 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 147 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 148 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 149 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 150 {
emilmont 1:fdd22bb7aa52 151 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 152 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 153
emilmont 1:fdd22bb7aa52 154 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 155 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 156 }
emilmont 1:fdd22bb7aa52 157 else
emilmont 1:fdd22bb7aa52 158 {
emilmont 1:fdd22bb7aa52 159 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 160 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 163 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 166 j = srcBLen;
emilmont 1:fdd22bb7aa52 167 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 168 srcALen = j;
emilmont 1:fdd22bb7aa52 169 }
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* conv(x,y) at n = x[n] * y[0] + x[n-1] * y[1] + x[n-2] * y[2] + ...+ x[n-N+1] * y[N -1] */
emilmont 1:fdd22bb7aa52 172 /* The function is internally
emilmont 1:fdd22bb7aa52 173 * divided into three stages according to the number of multiplications that has to be
emilmont 1:fdd22bb7aa52 174 * taken place between inputA samples and inputB samples. In the first stage of the
emilmont 1:fdd22bb7aa52 175 * algorithm, the multiplications increase by one for every iteration.
emilmont 1:fdd22bb7aa52 176 * In the second stage of the algorithm, srcBLen number of multiplications are done.
emilmont 1:fdd22bb7aa52 177 * In the third stage of the algorithm, the multiplications decrease by one
emilmont 1:fdd22bb7aa52 178 * for every iteration. */
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180 /* The algorithm is implemented in three stages.
emilmont 1:fdd22bb7aa52 181 The loop counters of each stage is initiated here. */
emilmont 1:fdd22bb7aa52 182 blockSize1 = srcBLen - 1u;
emilmont 1:fdd22bb7aa52 183 blockSize2 = srcALen - (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 184 blockSize3 = blockSize1;
emilmont 1:fdd22bb7aa52 185
emilmont 1:fdd22bb7aa52 186 /* --------------------------
emilmont 1:fdd22bb7aa52 187 * initializations of stage1
emilmont 1:fdd22bb7aa52 188 * -------------------------*/
emilmont 1:fdd22bb7aa52 189
emilmont 1:fdd22bb7aa52 190 /* sum = x[0] * y[0]
emilmont 1:fdd22bb7aa52 191 * sum = x[0] * y[1] + x[1] * y[0]
emilmont 1:fdd22bb7aa52 192 * ....
emilmont 1:fdd22bb7aa52 193 * sum = x[0] * y[srcBlen - 1] + x[1] * y[srcBlen - 2] +...+ x[srcBLen - 1] * y[0]
emilmont 1:fdd22bb7aa52 194 */
emilmont 1:fdd22bb7aa52 195
emilmont 1:fdd22bb7aa52 196 /* In this stage the MAC operations are increased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 197 The count variable holds the number of MAC operations performed */
emilmont 1:fdd22bb7aa52 198 count = 1u;
emilmont 1:fdd22bb7aa52 199
emilmont 1:fdd22bb7aa52 200 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 201 px = pIn1;
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 204 py = pIn2;
emilmont 1:fdd22bb7aa52 205
emilmont 1:fdd22bb7aa52 206
emilmont 1:fdd22bb7aa52 207 /* ------------------------
emilmont 1:fdd22bb7aa52 208 * Stage1 process
emilmont 1:fdd22bb7aa52 209 * ----------------------*/
emilmont 1:fdd22bb7aa52 210
emilmont 1:fdd22bb7aa52 211 /* The first stage starts here */
emilmont 1:fdd22bb7aa52 212 while(blockSize1 > 0u)
emilmont 1:fdd22bb7aa52 213 {
emilmont 1:fdd22bb7aa52 214 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 215 sum = 0.0f;
emilmont 1:fdd22bb7aa52 216
emilmont 1:fdd22bb7aa52 217 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 218 k = count >> 2u;
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 221 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 222 while(k > 0u)
emilmont 1:fdd22bb7aa52 223 {
emilmont 1:fdd22bb7aa52 224 /* x[0] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 225 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 228 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 229
emilmont 1:fdd22bb7aa52 230 /* x[2] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 231 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 /* x[3] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 234 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 235
emilmont 1:fdd22bb7aa52 236 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 237 k--;
emilmont 1:fdd22bb7aa52 238 }
emilmont 1:fdd22bb7aa52 239
emilmont 1:fdd22bb7aa52 240 /* If the count is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 241 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 242 k = count % 0x4u;
emilmont 1:fdd22bb7aa52 243
emilmont 1:fdd22bb7aa52 244 while(k > 0u)
emilmont 1:fdd22bb7aa52 245 {
emilmont 1:fdd22bb7aa52 246 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 247 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 248
emilmont 1:fdd22bb7aa52 249 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 250 k--;
emilmont 1:fdd22bb7aa52 251 }
emilmont 1:fdd22bb7aa52 252
emilmont 1:fdd22bb7aa52 253 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 254 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 257 py = pIn2 + count;
emilmont 1:fdd22bb7aa52 258 px = pIn1;
emilmont 1:fdd22bb7aa52 259
emilmont 1:fdd22bb7aa52 260 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 261 count++;
emilmont 1:fdd22bb7aa52 262
emilmont 1:fdd22bb7aa52 263 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 264 blockSize1--;
emilmont 1:fdd22bb7aa52 265 }
emilmont 1:fdd22bb7aa52 266
emilmont 1:fdd22bb7aa52 267 /* --------------------------
emilmont 1:fdd22bb7aa52 268 * Initializations of stage2
emilmont 1:fdd22bb7aa52 269 * ------------------------*/
emilmont 1:fdd22bb7aa52 270
emilmont 1:fdd22bb7aa52 271 /* sum = x[0] * y[srcBLen-1] + x[1] * y[srcBLen-2] +...+ x[srcBLen-1] * y[0]
emilmont 1:fdd22bb7aa52 272 * sum = x[1] * y[srcBLen-1] + x[2] * y[srcBLen-2] +...+ x[srcBLen] * y[0]
emilmont 1:fdd22bb7aa52 273 * ....
emilmont 1:fdd22bb7aa52 274 * sum = x[srcALen-srcBLen-2] * y[srcBLen-1] + x[srcALen] * y[srcBLen-2] +...+ x[srcALen-1] * y[0]
emilmont 1:fdd22bb7aa52 275 */
emilmont 1:fdd22bb7aa52 276
emilmont 1:fdd22bb7aa52 277 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 278 px = pIn1;
emilmont 1:fdd22bb7aa52 279
emilmont 1:fdd22bb7aa52 280 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 281 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 282 py = pSrc2;
emilmont 1:fdd22bb7aa52 283
emilmont 1:fdd22bb7aa52 284 /* count is index by which the pointer pIn1 to be incremented */
emilmont 1:fdd22bb7aa52 285 count = 0u;
emilmont 1:fdd22bb7aa52 286
emilmont 1:fdd22bb7aa52 287 /* -------------------
emilmont 1:fdd22bb7aa52 288 * Stage2 process
emilmont 1:fdd22bb7aa52 289 * ------------------*/
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed.
emilmont 1:fdd22bb7aa52 292 * So, to loop unroll over blockSize2,
emilmont 1:fdd22bb7aa52 293 * srcBLen should be greater than or equal to 4 */
emilmont 1:fdd22bb7aa52 294 if(srcBLen >= 4u)
emilmont 1:fdd22bb7aa52 295 {
emilmont 1:fdd22bb7aa52 296 /* Loop unroll over blockSize2, by 4 */
emilmont 1:fdd22bb7aa52 297 blkCnt = blockSize2 >> 2u;
emilmont 1:fdd22bb7aa52 298
emilmont 1:fdd22bb7aa52 299 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 300 {
emilmont 1:fdd22bb7aa52 301 /* Set all accumulators to zero */
emilmont 1:fdd22bb7aa52 302 acc0 = 0.0f;
emilmont 1:fdd22bb7aa52 303 acc1 = 0.0f;
emilmont 1:fdd22bb7aa52 304 acc2 = 0.0f;
emilmont 1:fdd22bb7aa52 305 acc3 = 0.0f;
emilmont 1:fdd22bb7aa52 306
emilmont 1:fdd22bb7aa52 307 /* read x[0], x[1], x[2] samples */
emilmont 1:fdd22bb7aa52 308 x0 = *(px++);
emilmont 1:fdd22bb7aa52 309 x1 = *(px++);
emilmont 1:fdd22bb7aa52 310 x2 = *(px++);
emilmont 1:fdd22bb7aa52 311
emilmont 1:fdd22bb7aa52 312 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 313 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 316 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 317 do
emilmont 1:fdd22bb7aa52 318 {
emilmont 1:fdd22bb7aa52 319 /* Read y[srcBLen - 1] sample */
emilmont 1:fdd22bb7aa52 320 c0 = *(py--);
emilmont 1:fdd22bb7aa52 321
emilmont 1:fdd22bb7aa52 322 /* Read x[3] sample */
emilmont 1:fdd22bb7aa52 323 x3 = *(px);
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 326 /* acc0 += x[0] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 327 acc0 += x0 * c0;
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 /* acc1 += x[1] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 330 acc1 += x1 * c0;
emilmont 1:fdd22bb7aa52 331
emilmont 1:fdd22bb7aa52 332 /* acc2 += x[2] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 333 acc2 += x2 * c0;
emilmont 1:fdd22bb7aa52 334
emilmont 1:fdd22bb7aa52 335 /* acc3 += x[3] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 336 acc3 += x3 * c0;
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 /* Read y[srcBLen - 2] sample */
emilmont 1:fdd22bb7aa52 339 c0 = *(py--);
emilmont 1:fdd22bb7aa52 340
emilmont 1:fdd22bb7aa52 341 /* Read x[4] sample */
emilmont 1:fdd22bb7aa52 342 x0 = *(px + 1u);
emilmont 1:fdd22bb7aa52 343
emilmont 1:fdd22bb7aa52 344 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 345 /* acc0 += x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 346 acc0 += x1 * c0;
emilmont 1:fdd22bb7aa52 347 /* acc1 += x[2] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 348 acc1 += x2 * c0;
emilmont 1:fdd22bb7aa52 349 /* acc2 += x[3] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 350 acc2 += x3 * c0;
emilmont 1:fdd22bb7aa52 351 /* acc3 += x[4] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 352 acc3 += x0 * c0;
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 /* Read y[srcBLen - 3] sample */
emilmont 1:fdd22bb7aa52 355 c0 = *(py--);
emilmont 1:fdd22bb7aa52 356
emilmont 1:fdd22bb7aa52 357 /* Read x[5] sample */
emilmont 1:fdd22bb7aa52 358 x1 = *(px + 2u);
emilmont 1:fdd22bb7aa52 359
emilmont 1:fdd22bb7aa52 360 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 361 /* acc0 += x[2] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 362 acc0 += x2 * c0;
emilmont 1:fdd22bb7aa52 363 /* acc1 += x[3] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 364 acc1 += x3 * c0;
emilmont 1:fdd22bb7aa52 365 /* acc2 += x[4] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 366 acc2 += x0 * c0;
emilmont 1:fdd22bb7aa52 367 /* acc3 += x[5] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 368 acc3 += x1 * c0;
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 /* Read y[srcBLen - 4] sample */
emilmont 1:fdd22bb7aa52 371 c0 = *(py--);
emilmont 1:fdd22bb7aa52 372
emilmont 1:fdd22bb7aa52 373 /* Read x[6] sample */
emilmont 1:fdd22bb7aa52 374 x2 = *(px + 3u);
emilmont 1:fdd22bb7aa52 375 px += 4u;
emilmont 1:fdd22bb7aa52 376
emilmont 1:fdd22bb7aa52 377 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 378 /* acc0 += x[3] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 379 acc0 += x3 * c0;
emilmont 1:fdd22bb7aa52 380 /* acc1 += x[4] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 381 acc1 += x0 * c0;
emilmont 1:fdd22bb7aa52 382 /* acc2 += x[5] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 383 acc2 += x1 * c0;
emilmont 1:fdd22bb7aa52 384 /* acc3 += x[6] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 385 acc3 += x2 * c0;
emilmont 1:fdd22bb7aa52 386
emilmont 1:fdd22bb7aa52 387
emilmont 1:fdd22bb7aa52 388 } while(--k);
emilmont 1:fdd22bb7aa52 389
emilmont 1:fdd22bb7aa52 390 /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 391 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 392 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 393
emilmont 1:fdd22bb7aa52 394 while(k > 0u)
emilmont 1:fdd22bb7aa52 395 {
emilmont 1:fdd22bb7aa52 396 /* Read y[srcBLen - 5] sample */
emilmont 1:fdd22bb7aa52 397 c0 = *(py--);
emilmont 1:fdd22bb7aa52 398
emilmont 1:fdd22bb7aa52 399 /* Read x[7] sample */
emilmont 1:fdd22bb7aa52 400 x3 = *(px++);
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 403 /* acc0 += x[4] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 404 acc0 += x0 * c0;
emilmont 1:fdd22bb7aa52 405 /* acc1 += x[5] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 406 acc1 += x1 * c0;
emilmont 1:fdd22bb7aa52 407 /* acc2 += x[6] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 408 acc2 += x2 * c0;
emilmont 1:fdd22bb7aa52 409 /* acc3 += x[7] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 410 acc3 += x3 * c0;
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412 /* Reuse the present samples for the next MAC */
emilmont 1:fdd22bb7aa52 413 x0 = x1;
emilmont 1:fdd22bb7aa52 414 x1 = x2;
emilmont 1:fdd22bb7aa52 415 x2 = x3;
emilmont 1:fdd22bb7aa52 416
emilmont 1:fdd22bb7aa52 417 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 418 k--;
emilmont 1:fdd22bb7aa52 419 }
emilmont 1:fdd22bb7aa52 420
emilmont 1:fdd22bb7aa52 421 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 422 *pOut++ = acc0;
emilmont 1:fdd22bb7aa52 423 *pOut++ = acc1;
emilmont 1:fdd22bb7aa52 424 *pOut++ = acc2;
emilmont 1:fdd22bb7aa52 425 *pOut++ = acc3;
emilmont 1:fdd22bb7aa52 426
emilmont 1:fdd22bb7aa52 427 /* Increment the pointer pIn1 index, count by 4 */
emilmont 1:fdd22bb7aa52 428 count += 4u;
emilmont 1:fdd22bb7aa52 429
emilmont 1:fdd22bb7aa52 430 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 431 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 432 py = pSrc2;
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434
emilmont 1:fdd22bb7aa52 435 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 436 blkCnt--;
emilmont 1:fdd22bb7aa52 437 }
emilmont 1:fdd22bb7aa52 438
emilmont 1:fdd22bb7aa52 439
emilmont 1:fdd22bb7aa52 440 /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here.
emilmont 1:fdd22bb7aa52 441 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 442 blkCnt = blockSize2 % 0x4u;
emilmont 1:fdd22bb7aa52 443
emilmont 1:fdd22bb7aa52 444 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 445 {
emilmont 1:fdd22bb7aa52 446 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 447 sum = 0.0f;
emilmont 1:fdd22bb7aa52 448
emilmont 1:fdd22bb7aa52 449 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 450 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 453 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 454 while(k > 0u)
emilmont 1:fdd22bb7aa52 455 {
emilmont 1:fdd22bb7aa52 456 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 457 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 458 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 459 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 460 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 461
emilmont 1:fdd22bb7aa52 462 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 463 k--;
emilmont 1:fdd22bb7aa52 464 }
emilmont 1:fdd22bb7aa52 465
emilmont 1:fdd22bb7aa52 466 /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 467 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 468 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 469
emilmont 1:fdd22bb7aa52 470 while(k > 0u)
emilmont 1:fdd22bb7aa52 471 {
emilmont 1:fdd22bb7aa52 472 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 473 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 474
emilmont 1:fdd22bb7aa52 475 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 476 k--;
emilmont 1:fdd22bb7aa52 477 }
emilmont 1:fdd22bb7aa52 478
emilmont 1:fdd22bb7aa52 479 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 480 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 483 count++;
emilmont 1:fdd22bb7aa52 484
emilmont 1:fdd22bb7aa52 485 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 486 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 487 py = pSrc2;
emilmont 1:fdd22bb7aa52 488
emilmont 1:fdd22bb7aa52 489 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 490 blkCnt--;
emilmont 1:fdd22bb7aa52 491 }
emilmont 1:fdd22bb7aa52 492 }
emilmont 1:fdd22bb7aa52 493 else
emilmont 1:fdd22bb7aa52 494 {
emilmont 1:fdd22bb7aa52 495 /* If the srcBLen is not a multiple of 4,
emilmont 1:fdd22bb7aa52 496 * the blockSize2 loop cannot be unrolled by 4 */
emilmont 1:fdd22bb7aa52 497 blkCnt = blockSize2;
emilmont 1:fdd22bb7aa52 498
emilmont 1:fdd22bb7aa52 499 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 500 {
emilmont 1:fdd22bb7aa52 501 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 502 sum = 0.0f;
emilmont 1:fdd22bb7aa52 503
emilmont 1:fdd22bb7aa52 504 /* srcBLen number of MACS should be performed */
emilmont 1:fdd22bb7aa52 505 k = srcBLen;
emilmont 1:fdd22bb7aa52 506
emilmont 1:fdd22bb7aa52 507 while(k > 0u)
emilmont 1:fdd22bb7aa52 508 {
emilmont 1:fdd22bb7aa52 509 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 510 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 511
emilmont 1:fdd22bb7aa52 512 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 513 k--;
emilmont 1:fdd22bb7aa52 514 }
emilmont 1:fdd22bb7aa52 515
emilmont 1:fdd22bb7aa52 516 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 517 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 518
emilmont 1:fdd22bb7aa52 519 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 520 count++;
emilmont 1:fdd22bb7aa52 521
emilmont 1:fdd22bb7aa52 522 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 523 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 524 py = pSrc2;
emilmont 1:fdd22bb7aa52 525
emilmont 1:fdd22bb7aa52 526 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 527 blkCnt--;
emilmont 1:fdd22bb7aa52 528 }
emilmont 1:fdd22bb7aa52 529 }
emilmont 1:fdd22bb7aa52 530
emilmont 1:fdd22bb7aa52 531
emilmont 1:fdd22bb7aa52 532 /* --------------------------
emilmont 1:fdd22bb7aa52 533 * Initializations of stage3
emilmont 1:fdd22bb7aa52 534 * -------------------------*/
emilmont 1:fdd22bb7aa52 535
emilmont 1:fdd22bb7aa52 536 /* sum += x[srcALen-srcBLen+1] * y[srcBLen-1] + x[srcALen-srcBLen+2] * y[srcBLen-2] +...+ x[srcALen-1] * y[1]
emilmont 1:fdd22bb7aa52 537 * sum += x[srcALen-srcBLen+2] * y[srcBLen-1] + x[srcALen-srcBLen+3] * y[srcBLen-2] +...+ x[srcALen-1] * y[2]
emilmont 1:fdd22bb7aa52 538 * ....
emilmont 1:fdd22bb7aa52 539 * sum += x[srcALen-2] * y[srcBLen-1] + x[srcALen-1] * y[srcBLen-2]
emilmont 1:fdd22bb7aa52 540 * sum += x[srcALen-1] * y[srcBLen-1]
emilmont 1:fdd22bb7aa52 541 */
emilmont 1:fdd22bb7aa52 542
emilmont 1:fdd22bb7aa52 543 /* In this stage the MAC operations are decreased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 544 The blockSize3 variable holds the number of MAC operations performed */
emilmont 1:fdd22bb7aa52 545
emilmont 1:fdd22bb7aa52 546 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 547 pSrc1 = (pIn1 + srcALen) - (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 548 px = pSrc1;
emilmont 1:fdd22bb7aa52 549
emilmont 1:fdd22bb7aa52 550 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 551 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 552 py = pSrc2;
emilmont 1:fdd22bb7aa52 553
emilmont 1:fdd22bb7aa52 554 /* -------------------
emilmont 1:fdd22bb7aa52 555 * Stage3 process
emilmont 1:fdd22bb7aa52 556 * ------------------*/
emilmont 1:fdd22bb7aa52 557
emilmont 1:fdd22bb7aa52 558 while(blockSize3 > 0u)
emilmont 1:fdd22bb7aa52 559 {
emilmont 1:fdd22bb7aa52 560 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 561 sum = 0.0f;
emilmont 1:fdd22bb7aa52 562
emilmont 1:fdd22bb7aa52 563 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 564 k = blockSize3 >> 2u;
emilmont 1:fdd22bb7aa52 565
emilmont 1:fdd22bb7aa52 566 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 567 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 568 while(k > 0u)
emilmont 1:fdd22bb7aa52 569 {
emilmont 1:fdd22bb7aa52 570 /* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 571 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 572
emilmont 1:fdd22bb7aa52 573 /* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 574 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 575
emilmont 1:fdd22bb7aa52 576 /* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 577 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 578
emilmont 1:fdd22bb7aa52 579 /* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 580 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 581
emilmont 1:fdd22bb7aa52 582 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 583 k--;
emilmont 1:fdd22bb7aa52 584 }
emilmont 1:fdd22bb7aa52 585
emilmont 1:fdd22bb7aa52 586 /* If the blockSize3 is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 587 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 588 k = blockSize3 % 0x4u;
emilmont 1:fdd22bb7aa52 589
emilmont 1:fdd22bb7aa52 590 while(k > 0u)
emilmont 1:fdd22bb7aa52 591 {
emilmont 1:fdd22bb7aa52 592 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 593 /* sum += x[srcALen-1] * y[srcBLen-1] */
emilmont 1:fdd22bb7aa52 594 sum += *px++ * *py--;
emilmont 1:fdd22bb7aa52 595
emilmont 1:fdd22bb7aa52 596 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 597 k--;
emilmont 1:fdd22bb7aa52 598 }
emilmont 1:fdd22bb7aa52 599
emilmont 1:fdd22bb7aa52 600 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 601 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 602
emilmont 1:fdd22bb7aa52 603 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 604 px = ++pSrc1;
emilmont 1:fdd22bb7aa52 605 py = pSrc2;
emilmont 1:fdd22bb7aa52 606
emilmont 1:fdd22bb7aa52 607 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 608 blockSize3--;
emilmont 1:fdd22bb7aa52 609 }
emilmont 1:fdd22bb7aa52 610
emilmont 1:fdd22bb7aa52 611 #else
emilmont 1:fdd22bb7aa52 612
emilmont 1:fdd22bb7aa52 613 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 614
emilmont 1:fdd22bb7aa52 615 float32_t *pIn1 = pSrcA; /* inputA pointer */
emilmont 1:fdd22bb7aa52 616 float32_t *pIn2 = pSrcB; /* inputB pointer */
emilmont 1:fdd22bb7aa52 617 float32_t sum; /* Accumulator */
emilmont 1:fdd22bb7aa52 618 uint32_t i, j; /* loop counters */
emilmont 1:fdd22bb7aa52 619
emilmont 1:fdd22bb7aa52 620 /* Loop to calculate convolution for output length number of times */
emilmont 1:fdd22bb7aa52 621 for (i = 0u; i < ((srcALen + srcBLen) - 1u); i++)
emilmont 1:fdd22bb7aa52 622 {
emilmont 1:fdd22bb7aa52 623 /* Initialize sum with zero to carry out MAC operations */
emilmont 1:fdd22bb7aa52 624 sum = 0.0f;
emilmont 1:fdd22bb7aa52 625
emilmont 1:fdd22bb7aa52 626 /* Loop to perform MAC operations according to convolution equation */
emilmont 1:fdd22bb7aa52 627 for (j = 0u; j <= i; j++)
emilmont 1:fdd22bb7aa52 628 {
emilmont 1:fdd22bb7aa52 629 /* Check the array limitations */
emilmont 1:fdd22bb7aa52 630 if((((i - j) < srcBLen) && (j < srcALen)))
emilmont 1:fdd22bb7aa52 631 {
emilmont 1:fdd22bb7aa52 632 /* z[i] += x[i-j] * y[j] */
emilmont 1:fdd22bb7aa52 633 sum += pIn1[j] * pIn2[i - j];
emilmont 1:fdd22bb7aa52 634 }
emilmont 1:fdd22bb7aa52 635 }
emilmont 1:fdd22bb7aa52 636 /* Store the output in the destination buffer */
emilmont 1:fdd22bb7aa52 637 pDst[i] = sum;
emilmont 1:fdd22bb7aa52 638 }
emilmont 1:fdd22bb7aa52 639
emilmont 1:fdd22bb7aa52 640 #endif /* #ifndef ARM_MATH_CM0 */
emilmont 1:fdd22bb7aa52 641
emilmont 1:fdd22bb7aa52 642 }
emilmont 1:fdd22bb7aa52 643
emilmont 1:fdd22bb7aa52 644 /**
emilmont 1:fdd22bb7aa52 645 * @} end of Conv group
emilmont 1:fdd22bb7aa52 646 */