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_correlate_opt_q15.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_correlate_opt_q15.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 1:fdd22bb7aa52 10 * Description: Correlation of Q15 sequences.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3
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 Corr
emilmont 1:fdd22bb7aa52 48 * @{
emilmont 1:fdd22bb7aa52 49 */
emilmont 1:fdd22bb7aa52 50
emilmont 1:fdd22bb7aa52 51 /**
emilmont 1:fdd22bb7aa52 52 * @brief Correlation of Q15 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. Length 2 * max(srcALen, srcBLen) - 1.
emilmont 1:fdd22bb7aa52 58 * @param[in] *pScratch points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emilmont 1:fdd22bb7aa52 59 * @return none.
emilmont 1:fdd22bb7aa52 60 *
emilmont 1:fdd22bb7aa52 61 * \par Restrictions
emilmont 1:fdd22bb7aa52 62 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 63 * In this case input, output, scratch buffers should be aligned by 32-bit
emilmont 1:fdd22bb7aa52 64 *
emilmont 1:fdd22bb7aa52 65 * @details
emilmont 1:fdd22bb7aa52 66 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 67 *
emilmont 1:fdd22bb7aa52 68 * \par
emilmont 1:fdd22bb7aa52 69 * The function is implemented using a 64-bit internal accumulator.
emilmont 1:fdd22bb7aa52 70 * Both inputs are in 1.15 format and multiplications yield a 2.30 result.
emilmont 1:fdd22bb7aa52 71 * The 2.30 intermediate results are accumulated in a 64-bit accumulator in 34.30 format.
emilmont 1:fdd22bb7aa52 72 * This approach provides 33 guard bits and there is no risk of overflow.
emilmont 1:fdd22bb7aa52 73 * The 34.30 result is then truncated to 34.15 format by discarding the low 15 bits and then saturated to 1.15 format.
emilmont 1:fdd22bb7aa52 74 *
emilmont 1:fdd22bb7aa52 75 * \par
emilmont 1:fdd22bb7aa52 76 * Refer to <code>arm_correlate_fast_q15()</code> for a faster but less precise version of this function for Cortex-M3 and Cortex-M4.
emilmont 1:fdd22bb7aa52 77 *
emilmont 1:fdd22bb7aa52 78 *
emilmont 1:fdd22bb7aa52 79 */
emilmont 1:fdd22bb7aa52 80
emilmont 1:fdd22bb7aa52 81
emilmont 1:fdd22bb7aa52 82 void arm_correlate_opt_q15(
emilmont 1:fdd22bb7aa52 83 q15_t * pSrcA,
emilmont 1:fdd22bb7aa52 84 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 85 q15_t * pSrcB,
emilmont 1:fdd22bb7aa52 86 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 87 q15_t * pDst,
emilmont 1:fdd22bb7aa52 88 q15_t * pScratch)
emilmont 1:fdd22bb7aa52 89 {
emilmont 1:fdd22bb7aa52 90 q15_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 91 q15_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 92 q63_t acc0, acc1, acc2, acc3; /* Accumulators */
emilmont 1:fdd22bb7aa52 93 q15_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 94 q31_t x1, x2, x3; /* temporary variables for holding input1 and input2 values */
emilmont 1:fdd22bb7aa52 95 uint32_t j, blkCnt, outBlockSize; /* loop counter */
emilmont 1:fdd22bb7aa52 96 int32_t inc = 1; /* output pointer increment */
emilmont 1:fdd22bb7aa52 97 uint32_t tapCnt;
emilmont 1:fdd22bb7aa52 98 q31_t y1, y2;
emilmont 1:fdd22bb7aa52 99 q15_t *pScr; /* Intermediate pointers */
emilmont 1:fdd22bb7aa52 100 q15_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 101 #ifdef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 102
emilmont 1:fdd22bb7aa52 103 q15_t a, b;
emilmont 1:fdd22bb7aa52 104
emilmont 1:fdd22bb7aa52 105 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 106
emilmont 1:fdd22bb7aa52 107 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 108 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 109 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 110 /* But CORR(x, y) is reverse of CORR(y, x) */
emilmont 1:fdd22bb7aa52 111 /* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */
emilmont 1:fdd22bb7aa52 112 /* and the destination pointer modifier, inc is set to -1 */
emilmont 1:fdd22bb7aa52 113 /* If srcALen > srcBLen, zero pad has to be done to srcB to make the two inputs of same length */
emilmont 1:fdd22bb7aa52 114 /* But to improve the performance,
emilmont 1:fdd22bb7aa52 115 * we include zeroes in the output instead of zero padding either of the the inputs*/
emilmont 1:fdd22bb7aa52 116 /* If srcALen > srcBLen,
emilmont 1:fdd22bb7aa52 117 * (srcALen - srcBLen) zeroes has to included in the starting of the output buffer */
emilmont 1:fdd22bb7aa52 118 /* If srcALen < srcBLen,
emilmont 1:fdd22bb7aa52 119 * (srcALen - srcBLen) zeroes has to included in the ending of the output buffer */
emilmont 1:fdd22bb7aa52 120 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 121 {
emilmont 1:fdd22bb7aa52 122 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 123 pIn1 = (pSrcA);
emilmont 1:fdd22bb7aa52 124
emilmont 1:fdd22bb7aa52 125 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 126 pIn2 = (pSrcB);
emilmont 1:fdd22bb7aa52 127
emilmont 1:fdd22bb7aa52 128 /* Number of output samples is calculated */
emilmont 1:fdd22bb7aa52 129 outBlockSize = (2u * srcALen) - 1u;
emilmont 1:fdd22bb7aa52 130
emilmont 1:fdd22bb7aa52 131 /* When srcALen > srcBLen, zero padding is done to srcB
emilmont 1:fdd22bb7aa52 132 * to make their lengths equal.
emilmont 1:fdd22bb7aa52 133 * Instead, (outBlockSize - (srcALen + srcBLen - 1))
emilmont 1:fdd22bb7aa52 134 * number of output samples are made zero */
emilmont 1:fdd22bb7aa52 135 j = outBlockSize - (srcALen + (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 136
emilmont 1:fdd22bb7aa52 137 /* Updating the pointer position to non zero value */
emilmont 1:fdd22bb7aa52 138 pOut += j;
emilmont 1:fdd22bb7aa52 139
emilmont 1:fdd22bb7aa52 140 }
emilmont 1:fdd22bb7aa52 141 else
emilmont 1:fdd22bb7aa52 142 {
emilmont 1:fdd22bb7aa52 143 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 144 pIn1 = (pSrcB);
emilmont 1:fdd22bb7aa52 145
emilmont 1:fdd22bb7aa52 146 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 147 pIn2 = (pSrcA);
emilmont 1:fdd22bb7aa52 148
emilmont 1:fdd22bb7aa52 149 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 150 j = srcBLen;
emilmont 1:fdd22bb7aa52 151 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 152 srcALen = j;
emilmont 1:fdd22bb7aa52 153
emilmont 1:fdd22bb7aa52 154 /* CORR(x, y) = Reverse order(CORR(y, x)) */
emilmont 1:fdd22bb7aa52 155 /* Hence set the destination pointer to point to the last output sample */
emilmont 1:fdd22bb7aa52 156 pOut = pDst + ((srcALen + srcBLen) - 2u);
emilmont 1:fdd22bb7aa52 157
emilmont 1:fdd22bb7aa52 158 /* Destination address modifier is set to -1 */
emilmont 1:fdd22bb7aa52 159 inc = -1;
emilmont 1:fdd22bb7aa52 160
emilmont 1:fdd22bb7aa52 161 }
emilmont 1:fdd22bb7aa52 162
emilmont 1:fdd22bb7aa52 163 pScr = pScratch;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* Fill (srcBLen - 1u) zeros in scratch buffer */
emilmont 1:fdd22bb7aa52 166 arm_fill_q15(0, pScr, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 167
emilmont 1:fdd22bb7aa52 168 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 169 pScr += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 174 arm_copy_q15(pIn1, pScr, srcALen);
emilmont 1:fdd22bb7aa52 175
emilmont 1:fdd22bb7aa52 176 /* Update pointers */
emilmont 1:fdd22bb7aa52 177 //pIn1 += srcALen;
emilmont 1:fdd22bb7aa52 178 pScr += srcALen;
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180 #else
emilmont 1:fdd22bb7aa52 181
emilmont 1:fdd22bb7aa52 182 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 183 j = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 184
emilmont 1:fdd22bb7aa52 185 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 186 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 187 while(j > 0u)
emilmont 1:fdd22bb7aa52 188 {
emilmont 1:fdd22bb7aa52 189 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 190 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 191 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 192 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 193 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 196 j--;
emilmont 1:fdd22bb7aa52 197 }
emilmont 1:fdd22bb7aa52 198
emilmont 1:fdd22bb7aa52 199 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 200 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 201 j = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 while(j > 0u)
emilmont 1:fdd22bb7aa52 204 {
emilmont 1:fdd22bb7aa52 205 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 206 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 207
emilmont 1:fdd22bb7aa52 208 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 209 j--;
emilmont 1:fdd22bb7aa52 210 }
emilmont 1:fdd22bb7aa52 211
emilmont 1:fdd22bb7aa52 212 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 213
emilmont 1:fdd22bb7aa52 214 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 215
emilmont 1:fdd22bb7aa52 216 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emilmont 1:fdd22bb7aa52 217 arm_fill_q15(0, pScr, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 218
emilmont 1:fdd22bb7aa52 219 /* Update pointer */
emilmont 1:fdd22bb7aa52 220 pScr += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 221
emilmont 1:fdd22bb7aa52 222 #else
emilmont 1:fdd22bb7aa52 223
emilmont 1:fdd22bb7aa52 224 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 225 j = (srcBLen - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 228 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 229 while(j > 0u)
emilmont 1:fdd22bb7aa52 230 {
emilmont 1:fdd22bb7aa52 231 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 232 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 233 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 234 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 235 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 236
emilmont 1:fdd22bb7aa52 237 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 238 j--;
emilmont 1:fdd22bb7aa52 239 }
emilmont 1:fdd22bb7aa52 240
emilmont 1:fdd22bb7aa52 241 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 242 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 243 j = (srcBLen - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 244
emilmont 1:fdd22bb7aa52 245 while(j > 0u)
emilmont 1:fdd22bb7aa52 246 {
emilmont 1:fdd22bb7aa52 247 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 248 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 249
emilmont 1:fdd22bb7aa52 250 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 251 j--;
emilmont 1:fdd22bb7aa52 252 }
emilmont 1:fdd22bb7aa52 253
emilmont 1:fdd22bb7aa52 254 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Temporary pointer for scratch2 */
emilmont 1:fdd22bb7aa52 257 py = pIn2;
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259
emilmont 1:fdd22bb7aa52 260 /* Actual correlation process starts here */
emilmont 1:fdd22bb7aa52 261 blkCnt = (srcALen + srcBLen - 1u) >> 2;
emilmont 1:fdd22bb7aa52 262
emilmont 1:fdd22bb7aa52 263 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 264 {
emilmont 1:fdd22bb7aa52 265 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 266 pScr = pScratch;
emilmont 1:fdd22bb7aa52 267
emilmont 1:fdd22bb7aa52 268 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 269 acc0 = 0;
emilmont 1:fdd22bb7aa52 270 acc1 = 0;
emilmont 1:fdd22bb7aa52 271 acc2 = 0;
emilmont 1:fdd22bb7aa52 272 acc3 = 0;
emilmont 1:fdd22bb7aa52 273
emilmont 1:fdd22bb7aa52 274 /* Read four samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 275 x1 = *__SIMD32(pScr)++;
emilmont 1:fdd22bb7aa52 276
emilmont 1:fdd22bb7aa52 277 /* Read next four samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 278 x2 = *__SIMD32(pScr)++;
emilmont 1:fdd22bb7aa52 279
emilmont 1:fdd22bb7aa52 280 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 281
emilmont 1:fdd22bb7aa52 282 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 283 {
emilmont 1:fdd22bb7aa52 284
emilmont 1:fdd22bb7aa52 285 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 286
emilmont 1:fdd22bb7aa52 287 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 288 y1 = _SIMD32_OFFSET(pIn2);
emilmont 1:fdd22bb7aa52 289 y2 = _SIMD32_OFFSET(pIn2 + 2u);
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 acc0 = __SMLALD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 292
emilmont 1:fdd22bb7aa52 293 acc2 = __SMLALD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 296 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 297 #else
emilmont 1:fdd22bb7aa52 298 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 299 #endif
emilmont 1:fdd22bb7aa52 300
emilmont 1:fdd22bb7aa52 301 acc1 = __SMLALDX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 302
emilmont 1:fdd22bb7aa52 303 x1 = _SIMD32_OFFSET(pScr);
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 acc0 = __SMLALD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 306
emilmont 1:fdd22bb7aa52 307 acc2 = __SMLALD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 308
emilmont 1:fdd22bb7aa52 309 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 310 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 311 #else
emilmont 1:fdd22bb7aa52 312 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 313 #endif
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 acc3 = __SMLALDX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 316
emilmont 1:fdd22bb7aa52 317 acc1 = __SMLALDX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 318
emilmont 1:fdd22bb7aa52 319 x2 = _SIMD32_OFFSET(pScr + 2u);
emilmont 1:fdd22bb7aa52 320
emilmont 1:fdd22bb7aa52 321 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 322 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 323 #else
emilmont 1:fdd22bb7aa52 324 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 325 #endif
emilmont 1:fdd22bb7aa52 326
emilmont 1:fdd22bb7aa52 327 acc3 = __SMLALDX(x3, y2, acc3);
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 #else
emilmont 1:fdd22bb7aa52 330
emilmont 1:fdd22bb7aa52 331 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 332 a = *pIn2;
emilmont 1:fdd22bb7aa52 333 b = *(pIn2 + 1);
emilmont 1:fdd22bb7aa52 334
emilmont 1:fdd22bb7aa52 335 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 336 y1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 337 #else
emilmont 1:fdd22bb7aa52 338 y1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 339 #endif
emilmont 1:fdd22bb7aa52 340
emilmont 1:fdd22bb7aa52 341 a = *(pIn2 + 2);
emilmont 1:fdd22bb7aa52 342 b = *(pIn2 + 3);
emilmont 1:fdd22bb7aa52 343 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 344 y2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 345 #else
emilmont 1:fdd22bb7aa52 346 y2 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 347 #endif
emilmont 1:fdd22bb7aa52 348
emilmont 1:fdd22bb7aa52 349 acc0 = __SMLALD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 350
emilmont 1:fdd22bb7aa52 351 acc2 = __SMLALD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 352
emilmont 1:fdd22bb7aa52 353 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 354 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 355 #else
emilmont 1:fdd22bb7aa52 356 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 357 #endif
emilmont 1:fdd22bb7aa52 358
emilmont 1:fdd22bb7aa52 359 acc1 = __SMLALDX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 a = *pScr;
emilmont 1:fdd22bb7aa52 362 b = *(pScr + 1);
emilmont 1:fdd22bb7aa52 363
emilmont 1:fdd22bb7aa52 364 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 365 x1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 366 #else
emilmont 1:fdd22bb7aa52 367 x1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 368 #endif
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 acc0 = __SMLALD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 acc2 = __SMLALD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 373
emilmont 1:fdd22bb7aa52 374 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 375 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 376 #else
emilmont 1:fdd22bb7aa52 377 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 378 #endif
emilmont 1:fdd22bb7aa52 379
emilmont 1:fdd22bb7aa52 380 acc3 = __SMLALDX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 381
emilmont 1:fdd22bb7aa52 382 acc1 = __SMLALDX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 383
emilmont 1:fdd22bb7aa52 384 a = *(pScr + 2);
emilmont 1:fdd22bb7aa52 385 b = *(pScr + 3);
emilmont 1:fdd22bb7aa52 386
emilmont 1:fdd22bb7aa52 387 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 388 x2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 389 #else
emilmont 1:fdd22bb7aa52 390 x2 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 391 #endif
emilmont 1:fdd22bb7aa52 392
emilmont 1:fdd22bb7aa52 393 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 394 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 395 #else
emilmont 1:fdd22bb7aa52 396 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 397 #endif
emilmont 1:fdd22bb7aa52 398
emilmont 1:fdd22bb7aa52 399 acc3 = __SMLALDX(x3, y2, acc3);
emilmont 1:fdd22bb7aa52 400
emilmont 1:fdd22bb7aa52 401 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 402
emilmont 1:fdd22bb7aa52 403 pIn2 += 4u;
emilmont 1:fdd22bb7aa52 404
emilmont 1:fdd22bb7aa52 405 pScr += 4u;
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407
emilmont 1:fdd22bb7aa52 408 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 409 tapCnt--;
emilmont 1:fdd22bb7aa52 410 }
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412
emilmont 1:fdd22bb7aa52 413
emilmont 1:fdd22bb7aa52 414 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 415 pScr -= 4u;
emilmont 1:fdd22bb7aa52 416
emilmont 1:fdd22bb7aa52 417
emilmont 1:fdd22bb7aa52 418 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 419 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 420
emilmont 1:fdd22bb7aa52 421 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 422 {
emilmont 1:fdd22bb7aa52 423
emilmont 1:fdd22bb7aa52 424 /* accumlate the results */
emilmont 1:fdd22bb7aa52 425 acc0 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 426 acc1 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 427 acc2 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 428 acc3 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 429
emilmont 1:fdd22bb7aa52 430 pScr -= 3u;
emilmont 1:fdd22bb7aa52 431
emilmont 1:fdd22bb7aa52 432 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 433 tapCnt--;
emilmont 1:fdd22bb7aa52 434 }
emilmont 1:fdd22bb7aa52 435
emilmont 1:fdd22bb7aa52 436 blkCnt--;
emilmont 1:fdd22bb7aa52 437
emilmont 1:fdd22bb7aa52 438
emilmont 1:fdd22bb7aa52 439 /* Store the results in the accumulators in the destination buffer. */
emilmont 1:fdd22bb7aa52 440 *pOut = (__SSAT(acc0 >> 15u, 16));
emilmont 1:fdd22bb7aa52 441 pOut += inc;
emilmont 1:fdd22bb7aa52 442 *pOut = (__SSAT(acc1 >> 15u, 16));
emilmont 1:fdd22bb7aa52 443 pOut += inc;
emilmont 1:fdd22bb7aa52 444 *pOut = (__SSAT(acc2 >> 15u, 16));
emilmont 1:fdd22bb7aa52 445 pOut += inc;
emilmont 1:fdd22bb7aa52 446 *pOut = (__SSAT(acc3 >> 15u, 16));
emilmont 1:fdd22bb7aa52 447 pOut += inc;
emilmont 1:fdd22bb7aa52 448
emilmont 1:fdd22bb7aa52 449 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 450 pIn2 = py;
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452 pScratch += 4u;
emilmont 1:fdd22bb7aa52 453
emilmont 1:fdd22bb7aa52 454 }
emilmont 1:fdd22bb7aa52 455
emilmont 1:fdd22bb7aa52 456
emilmont 1:fdd22bb7aa52 457 blkCnt = (srcALen + srcBLen - 1u) & 0x3;
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 /* Calculate correlation for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 460 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 461 {
emilmont 1:fdd22bb7aa52 462 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 463 pScr = pScratch;
emilmont 1:fdd22bb7aa52 464
emilmont 1:fdd22bb7aa52 465 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 466 acc0 = 0;
emilmont 1:fdd22bb7aa52 467
emilmont 1:fdd22bb7aa52 468 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 469
emilmont 1:fdd22bb7aa52 470 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 471 {
emilmont 1:fdd22bb7aa52 472
emilmont 1:fdd22bb7aa52 473 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 474 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 475
emilmont 1:fdd22bb7aa52 476 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 477 tapCnt--;
emilmont 1:fdd22bb7aa52 478 }
emilmont 1:fdd22bb7aa52 479
emilmont 1:fdd22bb7aa52 480 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 483 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 484 {
emilmont 1:fdd22bb7aa52 485
emilmont 1:fdd22bb7aa52 486 /* accumlate the results */
emilmont 1:fdd22bb7aa52 487 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 488
emilmont 1:fdd22bb7aa52 489 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 490 tapCnt--;
emilmont 1:fdd22bb7aa52 491 }
emilmont 1:fdd22bb7aa52 492
emilmont 1:fdd22bb7aa52 493 blkCnt--;
emilmont 1:fdd22bb7aa52 494
emilmont 1:fdd22bb7aa52 495 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 496 *pOut = (q15_t) (__SSAT((acc0 >> 15), 16));
emilmont 1:fdd22bb7aa52 497
emilmont 1:fdd22bb7aa52 498 pOut += inc;
emilmont 1:fdd22bb7aa52 499
emilmont 1:fdd22bb7aa52 500 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 501 pIn2 = py;
emilmont 1:fdd22bb7aa52 502
emilmont 1:fdd22bb7aa52 503 pScratch += 1u;
emilmont 1:fdd22bb7aa52 504
emilmont 1:fdd22bb7aa52 505 }
emilmont 1:fdd22bb7aa52 506
emilmont 1:fdd22bb7aa52 507
emilmont 1:fdd22bb7aa52 508 }
emilmont 1:fdd22bb7aa52 509
emilmont 1:fdd22bb7aa52 510 /**
emilmont 1:fdd22bb7aa52 511 * @} end of Corr group
emilmont 1:fdd22bb7aa52 512 */