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_fast_opt_q15.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_correlate_fast_opt_q15.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Fast Q15 Correlation.
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
emilmont 1:fdd22bb7aa52 36 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 37
emilmont 1:fdd22bb7aa52 38 /**
emilmont 1:fdd22bb7aa52 39 * @ingroup groupFilters
emilmont 1:fdd22bb7aa52 40 */
emilmont 1:fdd22bb7aa52 41
emilmont 1:fdd22bb7aa52 42 /**
emilmont 1:fdd22bb7aa52 43 * @addtogroup Corr
emilmont 1:fdd22bb7aa52 44 * @{
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
emilmont 1:fdd22bb7aa52 47 /**
emilmont 1:fdd22bb7aa52 48 * @brief Correlation of Q15 sequences (fast version) for Cortex-M3 and Cortex-M4.
emilmont 1:fdd22bb7aa52 49 * @param[in] *pSrcA points to the first input sequence.
emilmont 1:fdd22bb7aa52 50 * @param[in] srcALen length of the first input sequence.
emilmont 1:fdd22bb7aa52 51 * @param[in] *pSrcB points to the second input sequence.
emilmont 1:fdd22bb7aa52 52 * @param[in] srcBLen length of the second input sequence.
emilmont 1:fdd22bb7aa52 53 * @param[out] *pDst points to the location where the output result is written. Length 2 * max(srcALen, srcBLen) - 1.
emilmont 1:fdd22bb7aa52 54 * @param[in] *pScratch points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emilmont 1:fdd22bb7aa52 55 * @return none.
emilmont 1:fdd22bb7aa52 56 *
emilmont 1:fdd22bb7aa52 57 *
emilmont 1:fdd22bb7aa52 58 * \par Restrictions
emilmont 1:fdd22bb7aa52 59 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
emilmont 2:da51fb522205 60 * In this case input, output, scratch buffers should be aligned by 32-bit
emilmont 1:fdd22bb7aa52 61 *
emilmont 1:fdd22bb7aa52 62 *
emilmont 1:fdd22bb7aa52 63 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 64 *
emilmont 1:fdd22bb7aa52 65 * \par
emilmont 1:fdd22bb7aa52 66 * This fast version uses a 32-bit accumulator with 2.30 format.
emilmont 1:fdd22bb7aa52 67 * The accumulator maintains full precision of the intermediate multiplication results but provides only a single guard bit.
emilmont 1:fdd22bb7aa52 68 * There is no saturation on intermediate additions.
emilmont 1:fdd22bb7aa52 69 * Thus, if the accumulator overflows it wraps around and distorts the result.
emilmont 1:fdd22bb7aa52 70 * The input signals should be scaled down to avoid intermediate overflows.
emilmont 1:fdd22bb7aa52 71 * Scale down one of the inputs by 1/min(srcALen, srcBLen) to avoid overflow since a
emilmont 1:fdd22bb7aa52 72 * maximum of min(srcALen, srcBLen) number of additions is carried internally.
emilmont 1:fdd22bb7aa52 73 * The 2.30 accumulator is right shifted by 15 bits and then saturated to 1.15 format to yield the final result.
emilmont 1:fdd22bb7aa52 74 *
emilmont 1:fdd22bb7aa52 75 * \par
emilmont 1:fdd22bb7aa52 76 * See <code>arm_correlate_q15()</code> for a slower implementation of this function which uses a 64-bit accumulator to avoid wrap around distortion.
emilmont 1:fdd22bb7aa52 77 */
emilmont 1:fdd22bb7aa52 78
emilmont 1:fdd22bb7aa52 79 void arm_correlate_fast_opt_q15(
emilmont 1:fdd22bb7aa52 80 q15_t * pSrcA,
emilmont 1:fdd22bb7aa52 81 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 82 q15_t * pSrcB,
emilmont 1:fdd22bb7aa52 83 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 84 q15_t * pDst,
emilmont 1:fdd22bb7aa52 85 q15_t * pScratch)
emilmont 1:fdd22bb7aa52 86 {
emilmont 1:fdd22bb7aa52 87 q15_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 88 q15_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 89 q31_t acc0, acc1, acc2, acc3; /* Accumulators */
emilmont 1:fdd22bb7aa52 90 q15_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 91 q31_t x1, x2, x3; /* temporary variables for holding input and coefficient values */
emilmont 1:fdd22bb7aa52 92 uint32_t j, blkCnt, outBlockSize; /* loop counter */
emilmont 1:fdd22bb7aa52 93 int32_t inc = 1; /* Destination address modifier */
emilmont 1:fdd22bb7aa52 94 uint32_t tapCnt;
emilmont 1:fdd22bb7aa52 95 q31_t y1, y2;
emilmont 1:fdd22bb7aa52 96 q15_t *pScr; /* Intermediate pointers */
emilmont 1:fdd22bb7aa52 97 q15_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 98 #ifdef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 99
emilmont 1:fdd22bb7aa52 100 q15_t a, b;
emilmont 1:fdd22bb7aa52 101
emilmont 2:da51fb522205 102 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 103
emilmont 1:fdd22bb7aa52 104 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 105 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 106 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 107 /* But CORR(x, y) is reverse of CORR(y, x) */
emilmont 1:fdd22bb7aa52 108 /* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */
emilmont 1:fdd22bb7aa52 109 /* and the destination pointer modifier, inc is set to -1 */
emilmont 1:fdd22bb7aa52 110 /* If srcALen > srcBLen, zero pad has to be done to srcB to make the two inputs of same length */
emilmont 1:fdd22bb7aa52 111 /* But to improve the performance,
emilmont 1:fdd22bb7aa52 112 * we include zeroes in the output instead of zero padding either of the the inputs*/
emilmont 1:fdd22bb7aa52 113 /* If srcALen > srcBLen,
emilmont 1:fdd22bb7aa52 114 * (srcALen - srcBLen) zeroes has to included in the starting of the output buffer */
emilmont 1:fdd22bb7aa52 115 /* If srcALen < srcBLen,
emilmont 1:fdd22bb7aa52 116 * (srcALen - srcBLen) zeroes has to included in the ending of the output buffer */
emilmont 1:fdd22bb7aa52 117 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 118 {
emilmont 1:fdd22bb7aa52 119 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 120 pIn1 = (pSrcA);
emilmont 1:fdd22bb7aa52 121
emilmont 1:fdd22bb7aa52 122 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 123 pIn2 = (pSrcB);
emilmont 1:fdd22bb7aa52 124
emilmont 1:fdd22bb7aa52 125 /* Number of output samples is calculated */
emilmont 1:fdd22bb7aa52 126 outBlockSize = (2u * srcALen) - 1u;
emilmont 1:fdd22bb7aa52 127
emilmont 1:fdd22bb7aa52 128 /* When srcALen > srcBLen, zero padding is done to srcB
emilmont 1:fdd22bb7aa52 129 * to make their lengths equal.
emilmont 1:fdd22bb7aa52 130 * Instead, (outBlockSize - (srcALen + srcBLen - 1))
emilmont 1:fdd22bb7aa52 131 * number of output samples are made zero */
emilmont 1:fdd22bb7aa52 132 j = outBlockSize - (srcALen + (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 133
emilmont 1:fdd22bb7aa52 134 /* Updating the pointer position to non zero value */
emilmont 1:fdd22bb7aa52 135 pOut += j;
emilmont 1:fdd22bb7aa52 136
emilmont 1:fdd22bb7aa52 137 }
emilmont 1:fdd22bb7aa52 138 else
emilmont 1:fdd22bb7aa52 139 {
emilmont 1:fdd22bb7aa52 140 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 141 pIn1 = (pSrcB);
emilmont 1:fdd22bb7aa52 142
emilmont 1:fdd22bb7aa52 143 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 144 pIn2 = (pSrcA);
emilmont 1:fdd22bb7aa52 145
emilmont 1:fdd22bb7aa52 146 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 147 j = srcBLen;
emilmont 1:fdd22bb7aa52 148 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 149 srcALen = j;
emilmont 1:fdd22bb7aa52 150
emilmont 1:fdd22bb7aa52 151 /* CORR(x, y) = Reverse order(CORR(y, x)) */
emilmont 1:fdd22bb7aa52 152 /* Hence set the destination pointer to point to the last output sample */
emilmont 1:fdd22bb7aa52 153 pOut = pDst + ((srcALen + srcBLen) - 2u);
emilmont 1:fdd22bb7aa52 154
emilmont 1:fdd22bb7aa52 155 /* Destination address modifier is set to -1 */
emilmont 1:fdd22bb7aa52 156 inc = -1;
emilmont 1:fdd22bb7aa52 157
emilmont 1:fdd22bb7aa52 158 }
emilmont 1:fdd22bb7aa52 159
emilmont 1:fdd22bb7aa52 160 pScr = pScratch;
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* Fill (srcBLen - 1u) zeros in scratch buffer */
emilmont 1:fdd22bb7aa52 163 arm_fill_q15(0, pScr, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 166 pScr += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 167
emilmont 1:fdd22bb7aa52 168 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 169
emilmont 1:fdd22bb7aa52 170 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 171 arm_copy_q15(pIn1, pScr, srcALen);
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 /* Update pointers */
emilmont 1:fdd22bb7aa52 174 pScr += srcALen;
emilmont 1:fdd22bb7aa52 175
emilmont 1:fdd22bb7aa52 176 #else
emilmont 1:fdd22bb7aa52 177
emilmont 1:fdd22bb7aa52 178 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 179 j = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 180
emilmont 1:fdd22bb7aa52 181 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 182 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 183 while(j > 0u)
emilmont 1:fdd22bb7aa52 184 {
emilmont 1:fdd22bb7aa52 185 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 186 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 187 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 188 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 189 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 190
emilmont 1:fdd22bb7aa52 191 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 192 j--;
emilmont 1:fdd22bb7aa52 193 }
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 196 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 197 j = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 198
emilmont 1:fdd22bb7aa52 199 while(j > 0u)
emilmont 1:fdd22bb7aa52 200 {
emilmont 1:fdd22bb7aa52 201 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 202 *pScr++ = *pIn1++;
emilmont 1:fdd22bb7aa52 203
emilmont 1:fdd22bb7aa52 204 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 205 j--;
emilmont 1:fdd22bb7aa52 206 }
emilmont 1:fdd22bb7aa52 207
emilmont 2:da51fb522205 208 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 209
emilmont 1:fdd22bb7aa52 210 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 211
emilmont 1:fdd22bb7aa52 212 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emilmont 1:fdd22bb7aa52 213 arm_fill_q15(0, pScr, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 214
emilmont 1:fdd22bb7aa52 215 /* Update pointer */
emilmont 1:fdd22bb7aa52 216 pScr += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 217
emilmont 1:fdd22bb7aa52 218 #else
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 221 j = (srcBLen - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 222
emilmont 1:fdd22bb7aa52 223 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 224 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 225 while(j > 0u)
emilmont 1:fdd22bb7aa52 226 {
emilmont 1:fdd22bb7aa52 227 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 228 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 229 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 230 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 231 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 234 j--;
emilmont 1:fdd22bb7aa52 235 }
emilmont 1:fdd22bb7aa52 236
emilmont 1:fdd22bb7aa52 237 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 238 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 239 j = (srcBLen - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 240
emilmont 1:fdd22bb7aa52 241 while(j > 0u)
emilmont 1:fdd22bb7aa52 242 {
emilmont 1:fdd22bb7aa52 243 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 244 *pScr++ = 0;
emilmont 1:fdd22bb7aa52 245
emilmont 1:fdd22bb7aa52 246 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 247 j--;
emilmont 1:fdd22bb7aa52 248 }
emilmont 1:fdd22bb7aa52 249
emilmont 2:da51fb522205 250 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 251
emilmont 1:fdd22bb7aa52 252 /* Temporary pointer for scratch2 */
emilmont 1:fdd22bb7aa52 253 py = pIn2;
emilmont 1:fdd22bb7aa52 254
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Actual correlation process starts here */
emilmont 1:fdd22bb7aa52 257 blkCnt = (srcALen + srcBLen - 1u) >> 2;
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 260 {
emilmont 1:fdd22bb7aa52 261 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 262 pScr = pScratch;
emilmont 1:fdd22bb7aa52 263
emilmont 1:fdd22bb7aa52 264 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 265 acc0 = 0;
emilmont 1:fdd22bb7aa52 266 acc1 = 0;
emilmont 1:fdd22bb7aa52 267 acc2 = 0;
emilmont 1:fdd22bb7aa52 268 acc3 = 0;
emilmont 1:fdd22bb7aa52 269
emilmont 1:fdd22bb7aa52 270 /* Read four samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 271 x1 = *__SIMD32(pScr)++;
emilmont 1:fdd22bb7aa52 272
emilmont 1:fdd22bb7aa52 273 /* Read next four samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 274 x2 = *__SIMD32(pScr)++;
emilmont 1:fdd22bb7aa52 275
emilmont 1:fdd22bb7aa52 276 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 279 {
emilmont 1:fdd22bb7aa52 280
emilmont 1:fdd22bb7aa52 281 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 282
emilmont 1:fdd22bb7aa52 283 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 284 y1 = _SIMD32_OFFSET(pIn2);
emilmont 1:fdd22bb7aa52 285 y2 = _SIMD32_OFFSET(pIn2 + 2u);
emilmont 1:fdd22bb7aa52 286
emilmont 1:fdd22bb7aa52 287 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 288
emilmont 1:fdd22bb7aa52 289 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 292 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 293 #else
emilmont 1:fdd22bb7aa52 294 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 295 #endif
emilmont 1:fdd22bb7aa52 296
emilmont 1:fdd22bb7aa52 297 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 298
emilmont 1:fdd22bb7aa52 299 x1 = _SIMD32_OFFSET(pScr);
emilmont 1:fdd22bb7aa52 300
emilmont 1:fdd22bb7aa52 301 acc0 = __SMLAD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 302
emilmont 1:fdd22bb7aa52 303 acc2 = __SMLAD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 306 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 307 #else
emilmont 1:fdd22bb7aa52 308 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 309 #endif
emilmont 1:fdd22bb7aa52 310
emilmont 1:fdd22bb7aa52 311 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 312
emilmont 1:fdd22bb7aa52 313 acc1 = __SMLADX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 x2 = _SIMD32_OFFSET(pScr + 2u);
emilmont 1:fdd22bb7aa52 316
emilmont 1:fdd22bb7aa52 317 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 318 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 319 #else
emilmont 1:fdd22bb7aa52 320 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 321 #endif
emilmont 1:fdd22bb7aa52 322
emilmont 1:fdd22bb7aa52 323 acc3 = __SMLADX(x3, y2, acc3);
emilmont 2:da51fb522205 324 #else
emilmont 1:fdd22bb7aa52 325
emilmont 1:fdd22bb7aa52 326 /* Read four samples from smaller buffer */
emilmont 2:da51fb522205 327 a = *pIn2;
emilmont 2:da51fb522205 328 b = *(pIn2 + 1);
emilmont 1:fdd22bb7aa52 329
emilmont 1:fdd22bb7aa52 330 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 331 y1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 332 #else
emilmont 1:fdd22bb7aa52 333 y1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 334 #endif
emilmont 2:da51fb522205 335
emilmont 2:da51fb522205 336 a = *(pIn2 + 2);
emilmont 2:da51fb522205 337 b = *(pIn2 + 3);
emilmont 1:fdd22bb7aa52 338 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 339 y2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 340 #else
emilmont 1:fdd22bb7aa52 341 y2 = __PKHBT(b, a, 16);
emilmont 2:da51fb522205 342 #endif
emilmont 1:fdd22bb7aa52 343
emilmont 1:fdd22bb7aa52 344 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 345
emilmont 1:fdd22bb7aa52 346 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 347
emilmont 1:fdd22bb7aa52 348 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 349 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 350 #else
emilmont 1:fdd22bb7aa52 351 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 352 #endif
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 355
emilmont 2:da51fb522205 356 a = *pScr;
emilmont 2:da51fb522205 357 b = *(pScr + 1);
emilmont 1:fdd22bb7aa52 358
emilmont 1:fdd22bb7aa52 359 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 360 x1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 361 #else
emilmont 1:fdd22bb7aa52 362 x1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 363 #endif
emilmont 1:fdd22bb7aa52 364
emilmont 1:fdd22bb7aa52 365 acc0 = __SMLAD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 366
emilmont 1:fdd22bb7aa52 367 acc2 = __SMLAD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 368
emilmont 1:fdd22bb7aa52 369 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 370 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 371 #else
emilmont 1:fdd22bb7aa52 372 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 373 #endif
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 376
emilmont 1:fdd22bb7aa52 377 acc1 = __SMLADX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 378
emilmont 2:da51fb522205 379 a = *(pScr + 2);
emilmont 2:da51fb522205 380 b = *(pScr + 3);
emilmont 1:fdd22bb7aa52 381
emilmont 1:fdd22bb7aa52 382 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 383 x2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 384 #else
emilmont 1:fdd22bb7aa52 385 x2 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 386 #endif
emilmont 1:fdd22bb7aa52 387
emilmont 1:fdd22bb7aa52 388 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 389 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 390 #else
emilmont 1:fdd22bb7aa52 391 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 392 #endif
emilmont 1:fdd22bb7aa52 393
emilmont 1:fdd22bb7aa52 394 acc3 = __SMLADX(x3, y2, acc3);
emilmont 1:fdd22bb7aa52 395
emilmont 2:da51fb522205 396 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 pIn2 += 4u;
emilmont 1:fdd22bb7aa52 399
emilmont 1:fdd22bb7aa52 400 pScr += 4u;
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402
emilmont 1:fdd22bb7aa52 403 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 404 tapCnt--;
emilmont 1:fdd22bb7aa52 405 }
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407
emilmont 1:fdd22bb7aa52 408
emilmont 1:fdd22bb7aa52 409 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 410 pScr -= 4u;
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412
emilmont 1:fdd22bb7aa52 413 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 414 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 415
emilmont 1:fdd22bb7aa52 416 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 417 {
emilmont 1:fdd22bb7aa52 418
emilmont 1:fdd22bb7aa52 419 /* accumlate the results */
emilmont 1:fdd22bb7aa52 420 acc0 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 421 acc1 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 422 acc2 += (*pScr++ * *pIn2);
emilmont 1:fdd22bb7aa52 423 acc3 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 pScr -= 3u;
emilmont 1:fdd22bb7aa52 426
emilmont 1:fdd22bb7aa52 427 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 428 tapCnt--;
emilmont 1:fdd22bb7aa52 429 }
emilmont 1:fdd22bb7aa52 430
emilmont 1:fdd22bb7aa52 431 blkCnt--;
emilmont 1:fdd22bb7aa52 432
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 /* Store the results in the accumulators in the destination buffer. */
emilmont 1:fdd22bb7aa52 435 *pOut = (__SSAT(acc0 >> 15u, 16));
emilmont 1:fdd22bb7aa52 436 pOut += inc;
emilmont 1:fdd22bb7aa52 437 *pOut = (__SSAT(acc1 >> 15u, 16));
emilmont 1:fdd22bb7aa52 438 pOut += inc;
emilmont 1:fdd22bb7aa52 439 *pOut = (__SSAT(acc2 >> 15u, 16));
emilmont 1:fdd22bb7aa52 440 pOut += inc;
emilmont 1:fdd22bb7aa52 441 *pOut = (__SSAT(acc3 >> 15u, 16));
emilmont 1:fdd22bb7aa52 442 pOut += inc;
emilmont 1:fdd22bb7aa52 443
emilmont 1:fdd22bb7aa52 444
emilmont 1:fdd22bb7aa52 445 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 446 pIn2 = py;
emilmont 1:fdd22bb7aa52 447
emilmont 1:fdd22bb7aa52 448 pScratch += 4u;
emilmont 1:fdd22bb7aa52 449
emilmont 1:fdd22bb7aa52 450 }
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452
emilmont 1:fdd22bb7aa52 453 blkCnt = (srcALen + srcBLen - 1u) & 0x3;
emilmont 1:fdd22bb7aa52 454
emilmont 1:fdd22bb7aa52 455 /* Calculate correlation for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 456 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 457 {
emilmont 1:fdd22bb7aa52 458 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 459 pScr = pScratch;
emilmont 1:fdd22bb7aa52 460
emilmont 1:fdd22bb7aa52 461 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 462 acc0 = 0;
emilmont 1:fdd22bb7aa52 463
emilmont 1:fdd22bb7aa52 464 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 465
emilmont 1:fdd22bb7aa52 466 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 467 {
emilmont 1:fdd22bb7aa52 468
emilmont 1:fdd22bb7aa52 469 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 470 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 471
emilmont 1:fdd22bb7aa52 472 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 473 tapCnt--;
emilmont 1:fdd22bb7aa52 474 }
emilmont 1:fdd22bb7aa52 475
emilmont 1:fdd22bb7aa52 476 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 477
emilmont 1:fdd22bb7aa52 478 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 479 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 480 {
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* accumlate the results */
emilmont 1:fdd22bb7aa52 483 acc0 += (*pScr++ * *pIn2++);
emilmont 1:fdd22bb7aa52 484
emilmont 1:fdd22bb7aa52 485 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 486 tapCnt--;
emilmont 1:fdd22bb7aa52 487 }
emilmont 1:fdd22bb7aa52 488
emilmont 1:fdd22bb7aa52 489 blkCnt--;
emilmont 1:fdd22bb7aa52 490
emilmont 1:fdd22bb7aa52 491 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 492
emilmont 1:fdd22bb7aa52 493 *pOut = (q15_t) (__SSAT((acc0 >> 15), 16));
emilmont 1:fdd22bb7aa52 494
emilmont 1:fdd22bb7aa52 495 pOut += inc;
emilmont 1:fdd22bb7aa52 496
emilmont 1:fdd22bb7aa52 497 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 498 pIn2 = py;
emilmont 1:fdd22bb7aa52 499
emilmont 1:fdd22bb7aa52 500 pScratch += 1u;
emilmont 1:fdd22bb7aa52 501
emilmont 1:fdd22bb7aa52 502 }
emilmont 1:fdd22bb7aa52 503 }
emilmont 1:fdd22bb7aa52 504
emilmont 1:fdd22bb7aa52 505 /**
emilmont 1:fdd22bb7aa52 506 * @} end of Corr group
emilmont 1:fdd22bb7aa52 507 */