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_q7.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_q7.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 1:fdd22bb7aa52 10 * Description: Correlation of Q7 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 Q7 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] *pScratch1 points to scratch buffer(of type q15_t) of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emilmont 1:fdd22bb7aa52 59 * @param[in] *pScratch2 points to scratch buffer (of type q15_t) of size min(srcALen, srcBLen).
emilmont 1:fdd22bb7aa52 60 * @return none.
emilmont 1:fdd22bb7aa52 61 *
emilmont 1:fdd22bb7aa52 62 *
emilmont 1:fdd22bb7aa52 63 * \par Restrictions
emilmont 1:fdd22bb7aa52 64 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 65 * In this case input, output, scratch1 and scratch2 buffers should be aligned by 32-bit
emilmont 1:fdd22bb7aa52 66 *
emilmont 1:fdd22bb7aa52 67 * @details
emilmont 1:fdd22bb7aa52 68 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 69 *
emilmont 1:fdd22bb7aa52 70 * \par
emilmont 1:fdd22bb7aa52 71 * The function is implemented using a 32-bit internal accumulator.
emilmont 1:fdd22bb7aa52 72 * Both the inputs are represented in 1.7 format and multiplications yield a 2.14 result.
emilmont 1:fdd22bb7aa52 73 * The 2.14 intermediate results are accumulated in a 32-bit accumulator in 18.14 format.
emilmont 1:fdd22bb7aa52 74 * This approach provides 17 guard bits and there is no risk of overflow as long as <code>max(srcALen, srcBLen)<131072</code>.
emilmont 1:fdd22bb7aa52 75 * The 18.14 result is then truncated to 18.7 format by discarding the low 7 bits and saturated to 1.7 format.
emilmont 1:fdd22bb7aa52 76 *
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_q7(
emilmont 1:fdd22bb7aa52 83 q7_t * pSrcA,
emilmont 1:fdd22bb7aa52 84 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 85 q7_t * pSrcB,
emilmont 1:fdd22bb7aa52 86 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 87 q7_t * pDst,
emilmont 1:fdd22bb7aa52 88 q15_t * pScratch1,
emilmont 1:fdd22bb7aa52 89 q15_t * pScratch2)
emilmont 1:fdd22bb7aa52 90 {
emilmont 1:fdd22bb7aa52 91 q7_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 92 q15_t *pScr1 = pScratch1; /* Temporary pointer for scratch */
emilmont 1:fdd22bb7aa52 93 q15_t *pScr2 = pScratch2; /* Temporary pointer for scratch */
emilmont 1:fdd22bb7aa52 94 q7_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 95 q7_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 96 q15_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 97 q31_t acc0, acc1, acc2, acc3; /* Accumulators */
emilmont 1:fdd22bb7aa52 98 uint32_t j, k = 0u, blkCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 99 int32_t inc = 1; /* output pointer increment */
emilmont 1:fdd22bb7aa52 100 uint32_t outBlockSize; /* loop counter */
emilmont 1:fdd22bb7aa52 101 q15_t x4; /* Temporary input variable */
emilmont 1:fdd22bb7aa52 102 uint32_t tapCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 103 q31_t x1, x2, x3, y1; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 104
emilmont 1:fdd22bb7aa52 105 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 106 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 107 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 108 /* But CORR(x, y) is reverse of CORR(y, x) */
emilmont 1:fdd22bb7aa52 109 /* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */
emilmont 1:fdd22bb7aa52 110 /* and the destination pointer modifier, inc is set to -1 */
emilmont 1:fdd22bb7aa52 111 /* If srcALen > srcBLen, zero pad has to be done to srcB to make the two inputs of same length */
emilmont 1:fdd22bb7aa52 112 /* But to improve the performance,
emilmont 1:fdd22bb7aa52 113 * we include zeroes in the output instead of zero padding either of the the inputs*/
emilmont 1:fdd22bb7aa52 114 /* If srcALen > srcBLen,
emilmont 1:fdd22bb7aa52 115 * (srcALen - srcBLen) zeroes has to included in the starting of the output buffer */
emilmont 1:fdd22bb7aa52 116 /* If srcALen < srcBLen,
emilmont 1:fdd22bb7aa52 117 * (srcALen - srcBLen) zeroes has to included in the ending of the output buffer */
emilmont 1:fdd22bb7aa52 118 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 119 {
emilmont 1:fdd22bb7aa52 120 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 121 pIn1 = (pSrcA);
emilmont 1:fdd22bb7aa52 122
emilmont 1:fdd22bb7aa52 123 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 124 pIn2 = (pSrcB);
emilmont 1:fdd22bb7aa52 125
emilmont 1:fdd22bb7aa52 126 /* Number of output samples is calculated */
emilmont 1:fdd22bb7aa52 127 outBlockSize = (2u * srcALen) - 1u;
emilmont 1:fdd22bb7aa52 128
emilmont 1:fdd22bb7aa52 129 /* When srcALen > srcBLen, zero padding is done to srcB
emilmont 1:fdd22bb7aa52 130 * to make their lengths equal.
emilmont 1:fdd22bb7aa52 131 * Instead, (outBlockSize - (srcALen + srcBLen - 1))
emilmont 1:fdd22bb7aa52 132 * number of output samples are made zero */
emilmont 1:fdd22bb7aa52 133 j = outBlockSize - (srcALen + (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 134
emilmont 1:fdd22bb7aa52 135 /* Updating the pointer position to non zero value */
emilmont 1:fdd22bb7aa52 136 pOut += j;
emilmont 1:fdd22bb7aa52 137
emilmont 1:fdd22bb7aa52 138 }
emilmont 1:fdd22bb7aa52 139 else
emilmont 1:fdd22bb7aa52 140 {
emilmont 1:fdd22bb7aa52 141 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 142 pIn1 = (pSrcB);
emilmont 1:fdd22bb7aa52 143
emilmont 1:fdd22bb7aa52 144 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 145 pIn2 = (pSrcA);
emilmont 1:fdd22bb7aa52 146
emilmont 1:fdd22bb7aa52 147 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 148 j = srcBLen;
emilmont 1:fdd22bb7aa52 149 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 150 srcALen = j;
emilmont 1:fdd22bb7aa52 151
emilmont 1:fdd22bb7aa52 152 /* CORR(x, y) = Reverse order(CORR(y, x)) */
emilmont 1:fdd22bb7aa52 153 /* Hence set the destination pointer to point to the last output sample */
emilmont 1:fdd22bb7aa52 154 pOut = pDst + ((srcALen + srcBLen) - 2u);
emilmont 1:fdd22bb7aa52 155
emilmont 1:fdd22bb7aa52 156 /* Destination address modifier is set to -1 */
emilmont 1:fdd22bb7aa52 157 inc = -1;
emilmont 1:fdd22bb7aa52 158
emilmont 1:fdd22bb7aa52 159 }
emilmont 1:fdd22bb7aa52 160
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* Copy (srcBLen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 163 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 166 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 167 while(k > 0u)
emilmont 1:fdd22bb7aa52 168 {
emilmont 1:fdd22bb7aa52 169 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 170 x4 = (q15_t) * pIn2++;
emilmont 1:fdd22bb7aa52 171 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 172 x4 = (q15_t) * pIn2++;
emilmont 1:fdd22bb7aa52 173 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 174 x4 = (q15_t) * pIn2++;
emilmont 1:fdd22bb7aa52 175 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 176 x4 = (q15_t) * pIn2++;
emilmont 1:fdd22bb7aa52 177 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 178
emilmont 1:fdd22bb7aa52 179 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 180 k--;
emilmont 1:fdd22bb7aa52 181 }
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 184 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 185 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 186
emilmont 1:fdd22bb7aa52 187 while(k > 0u)
emilmont 1:fdd22bb7aa52 188 {
emilmont 1:fdd22bb7aa52 189 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 190 x4 = (q15_t) * pIn2++;
emilmont 1:fdd22bb7aa52 191 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 194 k--;
emilmont 1:fdd22bb7aa52 195 }
emilmont 1:fdd22bb7aa52 196
emilmont 1:fdd22bb7aa52 197 /* Fill (srcBLen - 1u) zeros in scratch buffer */
emilmont 1:fdd22bb7aa52 198 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 199
emilmont 1:fdd22bb7aa52 200 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 201 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 204 k = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 205
emilmont 1:fdd22bb7aa52 206 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 207 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 208 while(k > 0u)
emilmont 1:fdd22bb7aa52 209 {
emilmont 1:fdd22bb7aa52 210 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 211 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 212 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 213 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 214 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 215 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 216 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 217 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 218 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 221 k--;
emilmont 1:fdd22bb7aa52 222 }
emilmont 1:fdd22bb7aa52 223
emilmont 1:fdd22bb7aa52 224 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 225 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 226 k = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 227
emilmont 1:fdd22bb7aa52 228 while(k > 0u)
emilmont 1:fdd22bb7aa52 229 {
emilmont 1:fdd22bb7aa52 230 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 231 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 232 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 233
emilmont 1:fdd22bb7aa52 234 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 235 k--;
emilmont 1:fdd22bb7aa52 236 }
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 239
emilmont 1:fdd22bb7aa52 240 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emilmont 1:fdd22bb7aa52 241 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 /* Update pointer */
emilmont 1:fdd22bb7aa52 244 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 245
emilmont 1:fdd22bb7aa52 246 #else
emilmont 1:fdd22bb7aa52 247
emilmont 1:fdd22bb7aa52 248 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 249 k = (srcBLen - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 252 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 253 while(k > 0u)
emilmont 1:fdd22bb7aa52 254 {
emilmont 1:fdd22bb7aa52 255 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 256 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 257 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 258 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 259 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 260
emilmont 1:fdd22bb7aa52 261 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 262 k--;
emilmont 1:fdd22bb7aa52 263 }
emilmont 1:fdd22bb7aa52 264
emilmont 1:fdd22bb7aa52 265 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 266 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 267 k = (srcBLen - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 268
emilmont 1:fdd22bb7aa52 269 while(k > 0u)
emilmont 1:fdd22bb7aa52 270 {
emilmont 1:fdd22bb7aa52 271 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 272 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 273
emilmont 1:fdd22bb7aa52 274 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 275 k--;
emilmont 1:fdd22bb7aa52 276 }
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 279
emilmont 1:fdd22bb7aa52 280 /* Temporary pointer for second sequence */
emilmont 1:fdd22bb7aa52 281 py = pScratch2;
emilmont 1:fdd22bb7aa52 282
emilmont 1:fdd22bb7aa52 283 /* Initialization of pScr2 pointer */
emilmont 1:fdd22bb7aa52 284 pScr2 = pScratch2;
emilmont 1:fdd22bb7aa52 285
emilmont 1:fdd22bb7aa52 286 /* Actual correlation process starts here */
emilmont 1:fdd22bb7aa52 287 blkCnt = (srcALen + srcBLen - 1u) >> 2;
emilmont 1:fdd22bb7aa52 288
emilmont 1:fdd22bb7aa52 289 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 290 {
emilmont 1:fdd22bb7aa52 291 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 292 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 293
emilmont 1:fdd22bb7aa52 294 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 295 acc0 = 0;
emilmont 1:fdd22bb7aa52 296 acc1 = 0;
emilmont 1:fdd22bb7aa52 297 acc2 = 0;
emilmont 1:fdd22bb7aa52 298 acc3 = 0;
emilmont 1:fdd22bb7aa52 299
emilmont 1:fdd22bb7aa52 300 /* Read two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 301 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 302
emilmont 1:fdd22bb7aa52 303 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 304 x2 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 305
emilmont 1:fdd22bb7aa52 306 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 307
emilmont 1:fdd22bb7aa52 308 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 309 {
emilmont 1:fdd22bb7aa52 310
emilmont 1:fdd22bb7aa52 311 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 312 y1 = _SIMD32_OFFSET(pScr2);
emilmont 1:fdd22bb7aa52 313
emilmont 1:fdd22bb7aa52 314 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 315 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 316 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 317
emilmont 1:fdd22bb7aa52 318 /* pack input data */
emilmont 1:fdd22bb7aa52 319 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 320 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 321 #else
emilmont 1:fdd22bb7aa52 322 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 323 #endif
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 326 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 327
emilmont 1:fdd22bb7aa52 328 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 329 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 330
emilmont 1:fdd22bb7aa52 331 /* pack input data */
emilmont 1:fdd22bb7aa52 332 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 333 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 334 #else
emilmont 1:fdd22bb7aa52 335 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 336 #endif
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 339
emilmont 1:fdd22bb7aa52 340 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 341 y1 = _SIMD32_OFFSET(pScr2 + 2u);
emilmont 1:fdd22bb7aa52 342
emilmont 1:fdd22bb7aa52 343 acc0 = __SMLAD(x2, y1, acc0);
emilmont 1:fdd22bb7aa52 344
emilmont 1:fdd22bb7aa52 345 acc2 = __SMLAD(x1, y1, acc2);
emilmont 1:fdd22bb7aa52 346
emilmont 1:fdd22bb7aa52 347 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 348
emilmont 1:fdd22bb7aa52 349 x2 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 350
emilmont 1:fdd22bb7aa52 351 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 352 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 353 #else
emilmont 1:fdd22bb7aa52 354 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 355 #endif
emilmont 1:fdd22bb7aa52 356
emilmont 1:fdd22bb7aa52 357 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 358
emilmont 1:fdd22bb7aa52 359 pScr2 += 4u;
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361
emilmont 1:fdd22bb7aa52 362 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 363 tapCnt--;
emilmont 1:fdd22bb7aa52 364 }
emilmont 1:fdd22bb7aa52 365
emilmont 1:fdd22bb7aa52 366
emilmont 1:fdd22bb7aa52 367
emilmont 1:fdd22bb7aa52 368 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 369 pScr1 -= 4u;
emilmont 1:fdd22bb7aa52 370
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 373 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 376 {
emilmont 1:fdd22bb7aa52 377
emilmont 1:fdd22bb7aa52 378 /* accumlate the results */
emilmont 1:fdd22bb7aa52 379 acc0 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 380 acc1 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 381 acc2 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 382 acc3 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 383
emilmont 1:fdd22bb7aa52 384 pScr1 -= 3u;
emilmont 1:fdd22bb7aa52 385
emilmont 1:fdd22bb7aa52 386 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 387 tapCnt--;
emilmont 1:fdd22bb7aa52 388 }
emilmont 1:fdd22bb7aa52 389
emilmont 1:fdd22bb7aa52 390 blkCnt--;
emilmont 1:fdd22bb7aa52 391
emilmont 1:fdd22bb7aa52 392 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 393 *pOut = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 394 pOut += inc;
emilmont 1:fdd22bb7aa52 395 *pOut = (q7_t) (__SSAT(acc1 >> 7u, 8));
emilmont 1:fdd22bb7aa52 396 pOut += inc;
emilmont 1:fdd22bb7aa52 397 *pOut = (q7_t) (__SSAT(acc2 >> 7u, 8));
emilmont 1:fdd22bb7aa52 398 pOut += inc;
emilmont 1:fdd22bb7aa52 399 *pOut = (q7_t) (__SSAT(acc3 >> 7u, 8));
emilmont 1:fdd22bb7aa52 400 pOut += inc;
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 403 pScr2 = py;
emilmont 1:fdd22bb7aa52 404
emilmont 1:fdd22bb7aa52 405 pScratch1 += 4u;
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407 }
emilmont 1:fdd22bb7aa52 408
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 blkCnt = (srcALen + srcBLen - 1u) & 0x3;
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412 /* Calculate correlation for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 413 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 414 {
emilmont 1:fdd22bb7aa52 415 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 416 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 417
emilmont 1:fdd22bb7aa52 418 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 419 acc0 = 0;
emilmont 1:fdd22bb7aa52 420
emilmont 1:fdd22bb7aa52 421 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 422
emilmont 1:fdd22bb7aa52 423 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 424 {
emilmont 1:fdd22bb7aa52 425 acc0 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 426 acc0 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 427
emilmont 1:fdd22bb7aa52 428 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 429 tapCnt--;
emilmont 1:fdd22bb7aa52 430 }
emilmont 1:fdd22bb7aa52 431
emilmont 1:fdd22bb7aa52 432 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 435 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 436 {
emilmont 1:fdd22bb7aa52 437
emilmont 1:fdd22bb7aa52 438 /* accumlate the results */
emilmont 1:fdd22bb7aa52 439 acc0 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 440
emilmont 1:fdd22bb7aa52 441 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 442 tapCnt--;
emilmont 1:fdd22bb7aa52 443 }
emilmont 1:fdd22bb7aa52 444
emilmont 1:fdd22bb7aa52 445 blkCnt--;
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 448 *pOut = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 449
emilmont 1:fdd22bb7aa52 450 pOut += inc;
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 453 pScr2 = py;
emilmont 1:fdd22bb7aa52 454
emilmont 1:fdd22bb7aa52 455 pScratch1 += 1u;
emilmont 1:fdd22bb7aa52 456
emilmont 1:fdd22bb7aa52 457 }
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 }
emilmont 1:fdd22bb7aa52 460
emilmont 1:fdd22bb7aa52 461 /**
emilmont 1:fdd22bb7aa52 462 * @} end of Corr group
emilmont 1:fdd22bb7aa52 463 */