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_partial_q31.c@1:fdd22bb7aa52, 2012-11-28 (annotated)

Committer:
emilmont
Date:
Wed Nov 28 12:30:09 2012 +0000
Revision:
1:fdd22bb7aa52
Child:
2:da51fb522205
DSP library code

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 1:fdd22bb7aa52 5 * $Revision: V1.1.0
emilmont 1:fdd22bb7aa52 6 *
emilmont 1:fdd22bb7aa52 7 * Project: CMSIS DSP Library
emilmont 1:fdd22bb7aa52 8 * Title: arm_conv_partial_q31.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 1:fdd22bb7aa52 10 * Description: Partial convolution of Q31 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 * @addtogroup PartialConv
emilmont 1:fdd22bb7aa52 48 * @{
emilmont 1:fdd22bb7aa52 49 */
emilmont 1:fdd22bb7aa52 50
emilmont 1:fdd22bb7aa52 51 /**
emilmont 1:fdd22bb7aa52 52 * @brief Partial convolution of Q31 sequences.
emilmont 1:fdd22bb7aa52 53 * @param[in] *pSrcA points to the first input sequence.
emilmont 1:fdd22bb7aa52 54 * @param[in] srcALen length of the first input sequence.
emilmont 1:fdd22bb7aa52 55 * @param[in] *pSrcB points to the second input sequence.
emilmont 1:fdd22bb7aa52 56 * @param[in] srcBLen length of the second input sequence.
emilmont 1:fdd22bb7aa52 57 * @param[out] *pDst points to the location where the output result is written.
emilmont 1:fdd22bb7aa52 58 * @param[in] firstIndex is the first output sample to start with.
emilmont 1:fdd22bb7aa52 59 * @param[in] numPoints is the number of output points to be computed.
emilmont 1:fdd22bb7aa52 60 * @return Returns either ARM_MATH_SUCCESS if the function completed correctly or ARM_MATH_ARGUMENT_ERROR if the requested subset is not in the range [0 srcALen+srcBLen-2].
emilmont 1:fdd22bb7aa52 61 *
emilmont 1:fdd22bb7aa52 62 * See <code>arm_conv_partial_fast_q31()</code> for a faster but less precise implementation of this function for Cortex-M3 and Cortex-M4.
emilmont 1:fdd22bb7aa52 63 */
emilmont 1:fdd22bb7aa52 64
emilmont 1:fdd22bb7aa52 65 arm_status arm_conv_partial_q31(
emilmont 1:fdd22bb7aa52 66 q31_t * pSrcA,
emilmont 1:fdd22bb7aa52 67 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 68 q31_t * pSrcB,
emilmont 1:fdd22bb7aa52 69 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 70 q31_t * pDst,
emilmont 1:fdd22bb7aa52 71 uint32_t firstIndex,
emilmont 1:fdd22bb7aa52 72 uint32_t numPoints)
emilmont 1:fdd22bb7aa52 73 {
emilmont 1:fdd22bb7aa52 74
emilmont 1:fdd22bb7aa52 75
emilmont 1:fdd22bb7aa52 76 #ifndef ARM_MATH_CM0
emilmont 1:fdd22bb7aa52 77
emilmont 1:fdd22bb7aa52 78 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 79
emilmont 1:fdd22bb7aa52 80 q31_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 81 q31_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 82 q31_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 83 q31_t *px; /* Intermediate inputA pointer */
emilmont 1:fdd22bb7aa52 84 q31_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 85 q31_t *pSrc1, *pSrc2; /* Intermediate pointers */
emilmont 1:fdd22bb7aa52 86 q63_t sum, acc0, acc1, acc2; /* Accumulator */
emilmont 1:fdd22bb7aa52 87 q31_t x0, x1, x2, c0;
emilmont 1:fdd22bb7aa52 88 uint32_t j, k, count, check, blkCnt;
emilmont 1:fdd22bb7aa52 89 int32_t blockSize1, blockSize2, blockSize3; /* loop counter */
emilmont 1:fdd22bb7aa52 90 arm_status status; /* status of Partial convolution */
emilmont 1:fdd22bb7aa52 91
emilmont 1:fdd22bb7aa52 92
emilmont 1:fdd22bb7aa52 93 /* Check for range of output samples to be calculated */
emilmont 1:fdd22bb7aa52 94 if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
emilmont 1:fdd22bb7aa52 95 {
emilmont 1:fdd22bb7aa52 96 /* Set status as ARM_MATH_ARGUMENT_ERROR */
emilmont 1:fdd22bb7aa52 97 status = ARM_MATH_ARGUMENT_ERROR;
emilmont 1:fdd22bb7aa52 98 }
emilmont 1:fdd22bb7aa52 99 else
emilmont 1:fdd22bb7aa52 100 {
emilmont 1:fdd22bb7aa52 101
emilmont 1:fdd22bb7aa52 102 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 103 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 104 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 105 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 106 {
emilmont 1:fdd22bb7aa52 107 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 108 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 109
emilmont 1:fdd22bb7aa52 110 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 111 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 112 }
emilmont 1:fdd22bb7aa52 113 else
emilmont 1:fdd22bb7aa52 114 {
emilmont 1:fdd22bb7aa52 115 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 116 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 117
emilmont 1:fdd22bb7aa52 118 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 119 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 120
emilmont 1:fdd22bb7aa52 121 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 122 j = srcBLen;
emilmont 1:fdd22bb7aa52 123 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 124 srcALen = j;
emilmont 1:fdd22bb7aa52 125 }
emilmont 1:fdd22bb7aa52 126
emilmont 1:fdd22bb7aa52 127 /* Conditions to check which loopCounter holds
emilmont 1:fdd22bb7aa52 128 * the first and last indices of the output samples to be calculated. */
emilmont 1:fdd22bb7aa52 129 check = firstIndex + numPoints;
emilmont 1:fdd22bb7aa52 130 blockSize3 = ((int32_t) check - (int32_t) srcALen);
emilmont 1:fdd22bb7aa52 131 blockSize3 = (blockSize3 > 0) ? blockSize3 : 0;
emilmont 1:fdd22bb7aa52 132 blockSize1 = (((int32_t) srcBLen - 1) - (int32_t) firstIndex);
emilmont 1:fdd22bb7aa52 133 blockSize1 = (blockSize1 > 0) ? ((check > (srcBLen - 1u)) ? blockSize1 :
emilmont 1:fdd22bb7aa52 134 (int32_t) numPoints) : 0;
emilmont 1:fdd22bb7aa52 135 blockSize2 = (int32_t) check - ((blockSize3 + blockSize1) +
emilmont 1:fdd22bb7aa52 136 (int32_t) firstIndex);
emilmont 1:fdd22bb7aa52 137 blockSize2 = (blockSize2 > 0) ? blockSize2 : 0;
emilmont 1:fdd22bb7aa52 138
emilmont 1:fdd22bb7aa52 139 /* 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 140 /* The function is internally
emilmont 1:fdd22bb7aa52 141 * divided into three stages according to the number of multiplications that has to be
emilmont 1:fdd22bb7aa52 142 * taken place between inputA samples and inputB samples. In the first stage of the
emilmont 1:fdd22bb7aa52 143 * algorithm, the multiplications increase by one for every iteration.
emilmont 1:fdd22bb7aa52 144 * In the second stage of the algorithm, srcBLen number of multiplications are done.
emilmont 1:fdd22bb7aa52 145 * In the third stage of the algorithm, the multiplications decrease by one
emilmont 1:fdd22bb7aa52 146 * for every iteration. */
emilmont 1:fdd22bb7aa52 147
emilmont 1:fdd22bb7aa52 148 /* Set the output pointer to point to the firstIndex
emilmont 1:fdd22bb7aa52 149 * of the output sample to be calculated. */
emilmont 1:fdd22bb7aa52 150 pOut = pDst + firstIndex;
emilmont 1:fdd22bb7aa52 151
emilmont 1:fdd22bb7aa52 152 /* --------------------------
emilmont 1:fdd22bb7aa52 153 * Initializations of stage1
emilmont 1:fdd22bb7aa52 154 * -------------------------*/
emilmont 1:fdd22bb7aa52 155
emilmont 1:fdd22bb7aa52 156 /* sum = x[0] * y[0]
emilmont 1:fdd22bb7aa52 157 * sum = x[0] * y[1] + x[1] * y[0]
emilmont 1:fdd22bb7aa52 158 * ....
emilmont 1:fdd22bb7aa52 159 * sum = x[0] * y[srcBlen - 1] + x[1] * y[srcBlen - 2] +...+ x[srcBLen - 1] * y[0]
emilmont 1:fdd22bb7aa52 160 */
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* In this stage the MAC operations are increased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 163 The count variable holds the number of MAC operations performed.
emilmont 1:fdd22bb7aa52 164 Since the partial convolution starts from firstIndex
emilmont 1:fdd22bb7aa52 165 Number of Macs to be performed is firstIndex + 1 */
emilmont 1:fdd22bb7aa52 166 count = 1u + firstIndex;
emilmont 1:fdd22bb7aa52 167
emilmont 1:fdd22bb7aa52 168 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 169 px = pIn1;
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 172 pSrc2 = pIn2 + firstIndex;
emilmont 1:fdd22bb7aa52 173 py = pSrc2;
emilmont 1:fdd22bb7aa52 174
emilmont 1:fdd22bb7aa52 175 /* ------------------------
emilmont 1:fdd22bb7aa52 176 * Stage1 process
emilmont 1:fdd22bb7aa52 177 * ----------------------*/
emilmont 1:fdd22bb7aa52 178
emilmont 1:fdd22bb7aa52 179 /* The first loop starts here */
emilmont 1:fdd22bb7aa52 180 while(blockSize1 > 0)
emilmont 1:fdd22bb7aa52 181 {
emilmont 1:fdd22bb7aa52 182 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 183 sum = 0;
emilmont 1:fdd22bb7aa52 184
emilmont 1:fdd22bb7aa52 185 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 186 k = count >> 2u;
emilmont 1:fdd22bb7aa52 187
emilmont 1:fdd22bb7aa52 188 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 189 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 190 while(k > 0u)
emilmont 1:fdd22bb7aa52 191 {
emilmont 1:fdd22bb7aa52 192 /* x[0] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 193 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 194 /* x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 195 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 196 /* x[2] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 197 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 198 /* x[3] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 199 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 200
emilmont 1:fdd22bb7aa52 201 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 202 k--;
emilmont 1:fdd22bb7aa52 203 }
emilmont 1:fdd22bb7aa52 204
emilmont 1:fdd22bb7aa52 205 /* If the count is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 206 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 207 k = count % 0x4u;
emilmont 1:fdd22bb7aa52 208
emilmont 1:fdd22bb7aa52 209 while(k > 0u)
emilmont 1:fdd22bb7aa52 210 {
emilmont 1:fdd22bb7aa52 211 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 212 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 213
emilmont 1:fdd22bb7aa52 214 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 215 k--;
emilmont 1:fdd22bb7aa52 216 }
emilmont 1:fdd22bb7aa52 217
emilmont 1:fdd22bb7aa52 218 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 219 *pOut++ = (q31_t) (sum >> 31);
emilmont 1:fdd22bb7aa52 220
emilmont 1:fdd22bb7aa52 221 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 222 py = ++pSrc2;
emilmont 1:fdd22bb7aa52 223 px = pIn1;
emilmont 1:fdd22bb7aa52 224
emilmont 1:fdd22bb7aa52 225 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 226 count++;
emilmont 1:fdd22bb7aa52 227
emilmont 1:fdd22bb7aa52 228 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 229 blockSize1--;
emilmont 1:fdd22bb7aa52 230 }
emilmont 1:fdd22bb7aa52 231
emilmont 1:fdd22bb7aa52 232 /* --------------------------
emilmont 1:fdd22bb7aa52 233 * Initializations of stage2
emilmont 1:fdd22bb7aa52 234 * ------------------------*/
emilmont 1:fdd22bb7aa52 235
emilmont 1:fdd22bb7aa52 236 /* sum = x[0] * y[srcBLen-1] + x[1] * y[srcBLen-2] +...+ x[srcBLen-1] * y[0]
emilmont 1:fdd22bb7aa52 237 * sum = x[1] * y[srcBLen-1] + x[2] * y[srcBLen-2] +...+ x[srcBLen] * y[0]
emilmont 1:fdd22bb7aa52 238 * ....
emilmont 1:fdd22bb7aa52 239 * sum = x[srcALen-srcBLen-2] * y[srcBLen-1] + x[srcALen] * y[srcBLen-2] +...+ x[srcALen-1] * y[0]
emilmont 1:fdd22bb7aa52 240 */
emilmont 1:fdd22bb7aa52 241
emilmont 1:fdd22bb7aa52 242 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 243 px = pIn1;
emilmont 1:fdd22bb7aa52 244
emilmont 1:fdd22bb7aa52 245 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 246 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 247 py = pSrc2;
emilmont 1:fdd22bb7aa52 248
emilmont 1:fdd22bb7aa52 249 /* count is index by which the pointer pIn1 to be incremented */
emilmont 1:fdd22bb7aa52 250 count = 0u;
emilmont 1:fdd22bb7aa52 251
emilmont 1:fdd22bb7aa52 252 /* -------------------
emilmont 1:fdd22bb7aa52 253 * Stage2 process
emilmont 1:fdd22bb7aa52 254 * ------------------*/
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed.
emilmont 1:fdd22bb7aa52 257 * So, to loop unroll over blockSize2,
emilmont 1:fdd22bb7aa52 258 * srcBLen should be greater than or equal to 4 */
emilmont 1:fdd22bb7aa52 259 if(srcBLen >= 4u)
emilmont 1:fdd22bb7aa52 260 {
emilmont 1:fdd22bb7aa52 261 /* Loop unroll over blkCnt */
emilmont 1:fdd22bb7aa52 262
emilmont 1:fdd22bb7aa52 263 blkCnt = blockSize2 / 3;
emilmont 1:fdd22bb7aa52 264 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 265 {
emilmont 1:fdd22bb7aa52 266 /* Set all accumulators to zero */
emilmont 1:fdd22bb7aa52 267 acc0 = 0;
emilmont 1:fdd22bb7aa52 268 acc1 = 0;
emilmont 1:fdd22bb7aa52 269 acc2 = 0;
emilmont 1:fdd22bb7aa52 270
emilmont 1:fdd22bb7aa52 271 /* read x[0], x[1] samples */
emilmont 1:fdd22bb7aa52 272 x0 = *(px++);
emilmont 1:fdd22bb7aa52 273 x1 = *(px++);
emilmont 1:fdd22bb7aa52 274
emilmont 1:fdd22bb7aa52 275 /* Apply loop unrolling and compute 3 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 276 k = srcBLen / 3;
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 /* First part of the processing with loop unrolling. Compute 3 MACs at a time.
emilmont 1:fdd22bb7aa52 279 ** a second loop below computes MACs for the remaining 1 to 2 samples. */
emilmont 1:fdd22bb7aa52 280 do
emilmont 1:fdd22bb7aa52 281 {
emilmont 1:fdd22bb7aa52 282 /* Read y[srcBLen - 1] sample */
emilmont 1:fdd22bb7aa52 283 c0 = *(py);
emilmont 1:fdd22bb7aa52 284
emilmont 1:fdd22bb7aa52 285 /* Read x[2] sample */
emilmont 1:fdd22bb7aa52 286 x2 = *(px);
emilmont 1:fdd22bb7aa52 287
emilmont 1:fdd22bb7aa52 288 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 289 /* acc0 += x[0] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 290 acc0 += (q63_t) x0 *c0;
emilmont 1:fdd22bb7aa52 291 /* acc1 += x[1] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 292 acc1 += (q63_t) x1 *c0;
emilmont 1:fdd22bb7aa52 293 /* acc2 += x[2] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 294 acc2 += (q63_t) x2 *c0;
emilmont 1:fdd22bb7aa52 295
emilmont 1:fdd22bb7aa52 296 /* Read y[srcBLen - 2] sample */
emilmont 1:fdd22bb7aa52 297 c0 = *(py - 1u);
emilmont 1:fdd22bb7aa52 298
emilmont 1:fdd22bb7aa52 299 /* Read x[3] sample */
emilmont 1:fdd22bb7aa52 300 x0 = *(px + 1u);
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 303 /* acc0 += x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 304 acc0 += (q63_t) x1 *c0;
emilmont 1:fdd22bb7aa52 305 /* acc1 += x[2] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 306 acc1 += (q63_t) x2 *c0;
emilmont 1:fdd22bb7aa52 307 /* acc2 += x[3] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 308 acc2 += (q63_t) x0 *c0;
emilmont 1:fdd22bb7aa52 309
emilmont 1:fdd22bb7aa52 310 /* Read y[srcBLen - 3] sample */
emilmont 1:fdd22bb7aa52 311 c0 = *(py - 2u);
emilmont 1:fdd22bb7aa52 312
emilmont 1:fdd22bb7aa52 313 /* Read x[4] sample */
emilmont 1:fdd22bb7aa52 314 x1 = *(px + 2u);
emilmont 1:fdd22bb7aa52 315
emilmont 1:fdd22bb7aa52 316 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 317 /* acc0 += x[2] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 318 acc0 += (q63_t) x2 *c0;
emilmont 1:fdd22bb7aa52 319 /* acc1 += x[3] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 320 acc1 += (q63_t) x0 *c0;
emilmont 1:fdd22bb7aa52 321 /* acc2 += x[4] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 322 acc2 += (q63_t) x1 *c0;
emilmont 1:fdd22bb7aa52 323
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 px += 3u;
emilmont 1:fdd22bb7aa52 326
emilmont 1:fdd22bb7aa52 327 py -= 3u;
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 } while(--k);
emilmont 1:fdd22bb7aa52 330
emilmont 1:fdd22bb7aa52 331 /* If the srcBLen is not a multiple of 3, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 332 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 333 k = srcBLen - (3 * (srcBLen / 3));
emilmont 1:fdd22bb7aa52 334
emilmont 1:fdd22bb7aa52 335 while(k > 0u)
emilmont 1:fdd22bb7aa52 336 {
emilmont 1:fdd22bb7aa52 337 /* Read y[srcBLen - 5] sample */
emilmont 1:fdd22bb7aa52 338 c0 = *(py--);
emilmont 1:fdd22bb7aa52 339
emilmont 1:fdd22bb7aa52 340 /* Read x[7] sample */
emilmont 1:fdd22bb7aa52 341 x2 = *(px++);
emilmont 1:fdd22bb7aa52 342
emilmont 1:fdd22bb7aa52 343 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 344 /* acc0 += x[4] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 345 acc0 += (q63_t) x0 *c0;
emilmont 1:fdd22bb7aa52 346 /* acc1 += x[5] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 347 acc1 += (q63_t) x1 *c0;
emilmont 1:fdd22bb7aa52 348 /* acc2 += x[6] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 349 acc2 += (q63_t) x2 *c0;
emilmont 1:fdd22bb7aa52 350
emilmont 1:fdd22bb7aa52 351 /* Reuse the present samples for the next MAC */
emilmont 1:fdd22bb7aa52 352 x0 = x1;
emilmont 1:fdd22bb7aa52 353 x1 = x2;
emilmont 1:fdd22bb7aa52 354
emilmont 1:fdd22bb7aa52 355 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 356 k--;
emilmont 1:fdd22bb7aa52 357 }
emilmont 1:fdd22bb7aa52 358
emilmont 1:fdd22bb7aa52 359 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 360 *pOut++ = (q31_t) (acc0 >> 31);
emilmont 1:fdd22bb7aa52 361 *pOut++ = (q31_t) (acc1 >> 31);
emilmont 1:fdd22bb7aa52 362 *pOut++ = (q31_t) (acc2 >> 31);
emilmont 1:fdd22bb7aa52 363
emilmont 1:fdd22bb7aa52 364 /* Increment the pointer pIn1 index, count by 3 */
emilmont 1:fdd22bb7aa52 365 count += 3u;
emilmont 1:fdd22bb7aa52 366
emilmont 1:fdd22bb7aa52 367 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 368 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 369 py = pSrc2;
emilmont 1:fdd22bb7aa52 370
emilmont 1:fdd22bb7aa52 371 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 372 blkCnt--;
emilmont 1:fdd22bb7aa52 373 }
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 /* If the blockSize2 is not a multiple of 3, compute any remaining output samples here.
emilmont 1:fdd22bb7aa52 376 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 377 blkCnt = blockSize2 - 3 * (blockSize2 / 3);
emilmont 1:fdd22bb7aa52 378
emilmont 1:fdd22bb7aa52 379 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 380 {
emilmont 1:fdd22bb7aa52 381 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 382 sum = 0;
emilmont 1:fdd22bb7aa52 383
emilmont 1:fdd22bb7aa52 384 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 385 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 386
emilmont 1:fdd22bb7aa52 387 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 388 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 389 while(k > 0u)
emilmont 1:fdd22bb7aa52 390 {
emilmont 1:fdd22bb7aa52 391 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 392 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 393 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 394 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 395 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 396
emilmont 1:fdd22bb7aa52 397 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 398 k--;
emilmont 1:fdd22bb7aa52 399 }
emilmont 1:fdd22bb7aa52 400
emilmont 1:fdd22bb7aa52 401 /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 402 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 403 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 404
emilmont 1:fdd22bb7aa52 405 while(k > 0u)
emilmont 1:fdd22bb7aa52 406 {
emilmont 1:fdd22bb7aa52 407 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 408 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 411 k--;
emilmont 1:fdd22bb7aa52 412 }
emilmont 1:fdd22bb7aa52 413
emilmont 1:fdd22bb7aa52 414 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 415 *pOut++ = (q31_t) (sum >> 31);
emilmont 1:fdd22bb7aa52 416
emilmont 1:fdd22bb7aa52 417 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 418 count++;
emilmont 1:fdd22bb7aa52 419
emilmont 1:fdd22bb7aa52 420 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 421 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 422 py = pSrc2;
emilmont 1:fdd22bb7aa52 423
emilmont 1:fdd22bb7aa52 424 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 425 blkCnt--;
emilmont 1:fdd22bb7aa52 426 }
emilmont 1:fdd22bb7aa52 427 }
emilmont 1:fdd22bb7aa52 428 else
emilmont 1:fdd22bb7aa52 429 {
emilmont 1:fdd22bb7aa52 430 /* If the srcBLen is not a multiple of 4,
emilmont 1:fdd22bb7aa52 431 * the blockSize2 loop cannot be unrolled by 4 */
emilmont 1:fdd22bb7aa52 432 blkCnt = (uint32_t) blockSize2;
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 435 {
emilmont 1:fdd22bb7aa52 436 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 437 sum = 0;
emilmont 1:fdd22bb7aa52 438
emilmont 1:fdd22bb7aa52 439 /* srcBLen number of MACS should be performed */
emilmont 1:fdd22bb7aa52 440 k = srcBLen;
emilmont 1:fdd22bb7aa52 441
emilmont 1:fdd22bb7aa52 442 while(k > 0u)
emilmont 1:fdd22bb7aa52 443 {
emilmont 1:fdd22bb7aa52 444 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 445 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 448 k--;
emilmont 1:fdd22bb7aa52 449 }
emilmont 1:fdd22bb7aa52 450
emilmont 1:fdd22bb7aa52 451 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 452 *pOut++ = (q31_t) (sum >> 31);
emilmont 1:fdd22bb7aa52 453
emilmont 1:fdd22bb7aa52 454 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 455 count++;
emilmont 1:fdd22bb7aa52 456
emilmont 1:fdd22bb7aa52 457 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 458 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 459 py = pSrc2;
emilmont 1:fdd22bb7aa52 460
emilmont 1:fdd22bb7aa52 461 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 462 blkCnt--;
emilmont 1:fdd22bb7aa52 463 }
emilmont 1:fdd22bb7aa52 464 }
emilmont 1:fdd22bb7aa52 465
emilmont 1:fdd22bb7aa52 466
emilmont 1:fdd22bb7aa52 467 /* --------------------------
emilmont 1:fdd22bb7aa52 468 * Initializations of stage3
emilmont 1:fdd22bb7aa52 469 * -------------------------*/
emilmont 1:fdd22bb7aa52 470
emilmont 1:fdd22bb7aa52 471 /* sum += x[srcALen-srcBLen+1] * y[srcBLen-1] + x[srcALen-srcBLen+2] * y[srcBLen-2] +...+ x[srcALen-1] * y[1]
emilmont 1:fdd22bb7aa52 472 * sum += x[srcALen-srcBLen+2] * y[srcBLen-1] + x[srcALen-srcBLen+3] * y[srcBLen-2] +...+ x[srcALen-1] * y[2]
emilmont 1:fdd22bb7aa52 473 * ....
emilmont 1:fdd22bb7aa52 474 * sum += x[srcALen-2] * y[srcBLen-1] + x[srcALen-1] * y[srcBLen-2]
emilmont 1:fdd22bb7aa52 475 * sum += x[srcALen-1] * y[srcBLen-1]
emilmont 1:fdd22bb7aa52 476 */
emilmont 1:fdd22bb7aa52 477
emilmont 1:fdd22bb7aa52 478 /* In this stage the MAC operations are decreased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 479 The blockSize3 variable holds the number of MAC operations performed */
emilmont 1:fdd22bb7aa52 480 count = srcBLen - 1u;
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 483 pSrc1 = (pIn1 + srcALen) - (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 484 px = pSrc1;
emilmont 1:fdd22bb7aa52 485
emilmont 1:fdd22bb7aa52 486 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 487 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 488 py = pSrc2;
emilmont 1:fdd22bb7aa52 489
emilmont 1:fdd22bb7aa52 490 /* -------------------
emilmont 1:fdd22bb7aa52 491 * Stage3 process
emilmont 1:fdd22bb7aa52 492 * ------------------*/
emilmont 1:fdd22bb7aa52 493
emilmont 1:fdd22bb7aa52 494 while(blockSize3 > 0)
emilmont 1:fdd22bb7aa52 495 {
emilmont 1:fdd22bb7aa52 496 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 497 sum = 0;
emilmont 1:fdd22bb7aa52 498
emilmont 1:fdd22bb7aa52 499 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 500 k = count >> 2u;
emilmont 1:fdd22bb7aa52 501
emilmont 1:fdd22bb7aa52 502 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 503 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 504 while(k > 0u)
emilmont 1:fdd22bb7aa52 505 {
emilmont 1:fdd22bb7aa52 506 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 507 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 508 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 509 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 510
emilmont 1:fdd22bb7aa52 511 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 512 k--;
emilmont 1:fdd22bb7aa52 513 }
emilmont 1:fdd22bb7aa52 514
emilmont 1:fdd22bb7aa52 515 /* If the blockSize3 is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 516 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 517 k = count % 0x4u;
emilmont 1:fdd22bb7aa52 518
emilmont 1:fdd22bb7aa52 519 while(k > 0u)
emilmont 1:fdd22bb7aa52 520 {
emilmont 1:fdd22bb7aa52 521 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 522 sum += (q63_t) * px++ * (*py--);
emilmont 1:fdd22bb7aa52 523
emilmont 1:fdd22bb7aa52 524 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 525 k--;
emilmont 1:fdd22bb7aa52 526 }
emilmont 1:fdd22bb7aa52 527
emilmont 1:fdd22bb7aa52 528 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 529 *pOut++ = (q31_t) (sum >> 31);
emilmont 1:fdd22bb7aa52 530
emilmont 1:fdd22bb7aa52 531 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 532 px = ++pSrc1;
emilmont 1:fdd22bb7aa52 533 py = pSrc2;
emilmont 1:fdd22bb7aa52 534
emilmont 1:fdd22bb7aa52 535 /* Decrement the MAC count */
emilmont 1:fdd22bb7aa52 536 count--;
emilmont 1:fdd22bb7aa52 537
emilmont 1:fdd22bb7aa52 538 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 539 blockSize3--;
emilmont 1:fdd22bb7aa52 540
emilmont 1:fdd22bb7aa52 541 }
emilmont 1:fdd22bb7aa52 542
emilmont 1:fdd22bb7aa52 543 /* set status as ARM_MATH_SUCCESS */
emilmont 1:fdd22bb7aa52 544 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 545 }
emilmont 1:fdd22bb7aa52 546
emilmont 1:fdd22bb7aa52 547 /* Return to application */
emilmont 1:fdd22bb7aa52 548 return (status);
emilmont 1:fdd22bb7aa52 549
emilmont 1:fdd22bb7aa52 550 #else
emilmont 1:fdd22bb7aa52 551
emilmont 1:fdd22bb7aa52 552 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 553
emilmont 1:fdd22bb7aa52 554 q31_t *pIn1 = pSrcA; /* inputA pointer */
emilmont 1:fdd22bb7aa52 555 q31_t *pIn2 = pSrcB; /* inputB pointer */
emilmont 1:fdd22bb7aa52 556 q63_t sum; /* Accumulator */
emilmont 1:fdd22bb7aa52 557 uint32_t i, j; /* loop counters */
emilmont 1:fdd22bb7aa52 558 arm_status status; /* status of Partial convolution */
emilmont 1:fdd22bb7aa52 559
emilmont 1:fdd22bb7aa52 560 /* Check for range of output samples to be calculated */
emilmont 1:fdd22bb7aa52 561 if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
emilmont 1:fdd22bb7aa52 562 {
emilmont 1:fdd22bb7aa52 563 /* Set status as ARM_ARGUMENT_ERROR */
emilmont 1:fdd22bb7aa52 564 status = ARM_MATH_ARGUMENT_ERROR;
emilmont 1:fdd22bb7aa52 565 }
emilmont 1:fdd22bb7aa52 566 else
emilmont 1:fdd22bb7aa52 567 {
emilmont 1:fdd22bb7aa52 568 /* Loop to calculate convolution for output length number of values */
emilmont 1:fdd22bb7aa52 569 for (i = firstIndex; i <= (firstIndex + numPoints - 1); i++)
emilmont 1:fdd22bb7aa52 570 {
emilmont 1:fdd22bb7aa52 571 /* Initialize sum with zero to carry on MAC operations */
emilmont 1:fdd22bb7aa52 572 sum = 0;
emilmont 1:fdd22bb7aa52 573
emilmont 1:fdd22bb7aa52 574 /* Loop to perform MAC operations according to convolution equation */
emilmont 1:fdd22bb7aa52 575 for (j = 0; j <= i; j++)
emilmont 1:fdd22bb7aa52 576 {
emilmont 1:fdd22bb7aa52 577 /* Check the array limitations */
emilmont 1:fdd22bb7aa52 578 if(((i - j) < srcBLen) && (j < srcALen))
emilmont 1:fdd22bb7aa52 579 {
emilmont 1:fdd22bb7aa52 580 /* z[i] += x[i-j] * y[j] */
emilmont 1:fdd22bb7aa52 581 sum += ((q63_t) pIn1[j] * (pIn2[i - j]));
emilmont 1:fdd22bb7aa52 582 }
emilmont 1:fdd22bb7aa52 583 }
emilmont 1:fdd22bb7aa52 584
emilmont 1:fdd22bb7aa52 585 /* Store the output in the destination buffer */
emilmont 1:fdd22bb7aa52 586 pDst[i] = (q31_t) (sum >> 31u);
emilmont 1:fdd22bb7aa52 587 }
emilmont 1:fdd22bb7aa52 588 /* set status as ARM_SUCCESS as there are no argument errors */
emilmont 1:fdd22bb7aa52 589 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 590 }
emilmont 1:fdd22bb7aa52 591 return (status);
emilmont 1:fdd22bb7aa52 592
emilmont 1:fdd22bb7aa52 593 #endif /* #ifndef ARM_MATH_CM0 */
emilmont 1:fdd22bb7aa52 594
emilmont 1:fdd22bb7aa52 595 }
emilmont 1:fdd22bb7aa52 596
emilmont 1:fdd22bb7aa52 597 /**
emilmont 1:fdd22bb7aa52 598 * @} end of PartialConv group
emilmont 1:fdd22bb7aa52 599 */