tmk_core/tool/mbed/mbed-sdk/libraries/dsp/cmsis_dsp/FilteringFunctions/arm_conv_partial_fast_opt_q15.c

   1 /* ----------------------------------------------------------------------
   2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
   3 *
   4 * $Date:        17. January 2013
   5 * $Revision:    V1.4.1
   6 *
   7 * Project:          CMSIS DSP Library
   8 * Title:                arm_conv_partial_fast_opt_q15.c
   9 *
  10 * Description:  Fast Q15 Partial convolution.
  11 *
  12 * Target Processor: Cortex-M4/Cortex-M3
  13 *
  14 * Redistribution and use in source and binary forms, with or without
  15 * modification, are permitted provided that the following conditions
  16 * are met:
  17 *   - Redistributions of source code must retain the above copyright
  18 *     notice, this list of conditions and the following disclaimer.
  19 *   - Redistributions in binary form must reproduce the above copyright
  20 *     notice, this list of conditions and the following disclaimer in
  21 *     the documentation and/or other materials provided with the
  22 *     distribution.
  23 *   - Neither the name of ARM LIMITED nor the names of its contributors
  24 *     may be used to endorse or promote products derived from this
  25 *     software without specific prior written permission.
  26 *
  27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  38 * POSSIBILITY OF SUCH DAMAGE.
  39 * -------------------------------------------------------------------- */
  40
  41 #include "arm_math.h"
  42
  43 /**
  44  * @ingroup groupFilters
  45  */
  46
  47 /**
  48  * @addtogroup PartialConv
  49  * @{
  50  */
  51
  52 /**
  53  * @brief Partial convolution of Q15 sequences (fast version) for Cortex-M3 and Cortex-M4.
  54  * @param[in]       *pSrcA points to the first input sequence.
  55  * @param[in]       srcALen length of the first input sequence.
  56  * @param[in]       *pSrcB points to the second input sequence.
  57  * @param[in]       srcBLen length of the second input sequence.
  58  * @param[out]      *pDst points to the location where the output result is written.
  59  * @param[in]       firstIndex is the first output sample to start with.
  60  * @param[in]       numPoints is the number of output points to be computed.
  61  * @param[in]       *pScratch1 points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
  62  * @param[in]       *pScratch2 points to scratch buffer of size min(srcALen, srcBLen).
  63  * @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].
  64  *
  65  * See <code>arm_conv_partial_q15()</code> for a slower implementation of this function which uses a 64-bit accumulator to avoid wrap around distortion.
  66  *
  67  * \par Restrictions
  68  *  If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
  69  *      In this case input, output, scratch1 and scratch2 buffers should be aligned by 32-bit
  70  *
  71  */
  72
  73 #ifndef UNALIGNED_SUPPORT_DISABLE
  74
  75 arm_status arm_conv_partial_fast_opt_q15(
  76   q15_t * pSrcA,
  77   uint32_t srcALen,
  78   q15_t * pSrcB,
  79   uint32_t srcBLen,
  80   q15_t * pDst,
  81   uint32_t firstIndex,
  82   uint32_t numPoints,
  83   q15_t * pScratch1,
  84   q15_t * pScratch2)
  85 {
  86
  87   q15_t *pOut = pDst;                            /* output pointer */
  88   q15_t *pScr1 = pScratch1;                      /* Temporary pointer for scratch1 */
  89   q15_t *pScr2 = pScratch2;                      /* Temporary pointer for scratch1 */
  90   q31_t acc0, acc1, acc2, acc3;                  /* Accumulator */
  91   q31_t x1, x2, x3;                              /* Temporary variables to hold state and coefficient values */
  92   q31_t y1, y2;                                  /* State variables */
  93   q15_t *pIn1;                                   /* inputA pointer */
  94   q15_t *pIn2;                                   /* inputB pointer */
  95   q15_t *px;                                     /* Intermediate inputA pointer  */
  96   q15_t *py;                                     /* Intermediate inputB pointer  */
  97   uint32_t j, k, blkCnt;                         /* loop counter */
  98   arm_status status;
  99
 100   uint32_t tapCnt;                               /* loop count */
 101
 102   /* Check for range of output samples to be calculated */
 103   if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
 104   {
 105     /* Set status as ARM_MATH_ARGUMENT_ERROR */
 106     status = ARM_MATH_ARGUMENT_ERROR;
 107   }
 108   else
 109   {
 110
 111     /* The algorithm implementation is based on the lengths of the inputs. */
 112     /* srcB is always made to slide across srcA. */
 113     /* So srcBLen is always considered as shorter or equal to srcALen */
 114     if(srcALen >= srcBLen)
 115     {
 116       /* Initialization of inputA pointer */
 117       pIn1 = pSrcA;
 118
 119       /* Initialization of inputB pointer */
 120       pIn2 = pSrcB;
 121     }
 122     else
 123     {
 124       /* Initialization of inputA pointer */
 125       pIn1 = pSrcB;
 126
 127       /* Initialization of inputB pointer */
 128       pIn2 = pSrcA;
 129
 130       /* srcBLen is always considered as shorter or equal to srcALen */
 131       j = srcBLen;
 132       srcBLen = srcALen;
 133       srcALen = j;
 134     }
 135
 136     /* Temporary pointer for scratch2 */
 137     py = pScratch2;
 138
 139     /* pointer to take end of scratch2 buffer */
 140     pScr2 = pScratch2 + srcBLen - 1;
 141
 142     /* points to smaller length sequence */
 143     px = pIn2;
 144
 145     /* Apply loop unrolling and do 4 Copies simultaneously. */
 146     k = srcBLen >> 2u;
 147
 148     /* First part of the processing with loop unrolling copies 4 data points at a time.
 149      ** a second loop below copies for the remaining 1 to 3 samples. */
 150
 151     /* Copy smaller length input sequence in reverse order into second scratch buffer */
 152     while(k > 0u)
 153     {
 154       /* copy second buffer in reversal manner */
 155       *pScr2-- = *px++;
 156       *pScr2-- = *px++;
 157       *pScr2-- = *px++;
 158       *pScr2-- = *px++;
 159
 160       /* Decrement the loop counter */
 161       k--;
 162     }
 163
 164     /* If the count is not a multiple of 4, copy remaining samples here.
 165      ** No loop unrolling is used. */
 166     k = srcBLen % 0x4u;
 167
 168     while(k > 0u)
 169     {
 170       /* copy second buffer in reversal manner for remaining samples */
 171       *pScr2-- = *px++;
 172
 173       /* Decrement the loop counter */
 174       k--;
 175     }
 176
 177     /* Initialze temporary scratch pointer */
 178     pScr1 = pScratch1;
 179
 180     /* Assuming scratch1 buffer is aligned by 32-bit */
 181     /* Fill (srcBLen - 1u) zeros in scratch buffer */
 182     arm_fill_q15(0, pScr1, (srcBLen - 1u));
 183
 184     /* Update temporary scratch pointer */
 185     pScr1 += (srcBLen - 1u);
 186
 187     /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
 188
 189     /* Copy (srcALen) samples in scratch buffer */
 190     arm_copy_q15(pIn1, pScr1, srcALen);
 191
 192     /* Update pointers */
 193     pScr1 += srcALen;
 194
 195     /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
 196     arm_fill_q15(0, pScr1, (srcBLen - 1u));
 197
 198     /* Update pointer */
 199     pScr1 += (srcBLen - 1u);
 200
 201     /* Initialization of pIn2 pointer */
 202     pIn2 = py;
 203
 204     pScratch1 += firstIndex;
 205
 206     pOut = pDst + firstIndex;
 207
 208     /* First part of the processing with loop unrolling process 4 data points at a time.
 209      ** a second loop below process for the remaining 1 to 3 samples. */
 210
 211     /* Actual convolution process starts here */
 212     blkCnt = (numPoints) >> 2;
 213
 214     while(blkCnt > 0)
 215     {
 216       /* Initialze temporary scratch pointer as scratch1 */
 217       pScr1 = pScratch1;
 218
 219       /* Clear Accumlators */
 220       acc0 = 0;
 221       acc1 = 0;
 222       acc2 = 0;
 223       acc3 = 0;
 224
 225       /* Read two samples from scratch1 buffer */
 226       x1 = *__SIMD32(pScr1)++;
 227
 228       /* Read next two samples from scratch1 buffer */
 229       x2 = *__SIMD32(pScr1)++;
 230
 231       tapCnt = (srcBLen) >> 2u;
 232
 233       while(tapCnt > 0u)
 234       {
 235
 236         /* Read four samples from smaller buffer */
 237         y1 = _SIMD32_OFFSET(pIn2);
 238         y2 = _SIMD32_OFFSET(pIn2 + 2u);
 239
 240         /* multiply and accumlate */
 241         acc0 = __SMLAD(x1, y1, acc0);
 242         acc2 = __SMLAD(x2, y1, acc2);
 243
 244         /* pack input data */
 245 #ifndef ARM_MATH_BIG_ENDIAN
 246         x3 = __PKHBT(x2, x1, 0);
 247 #else
 248         x3 = __PKHBT(x1, x2, 0);
 249 #endif
 250
 251         /* multiply and accumlate */
 252         acc1 = __SMLADX(x3, y1, acc1);
 253
 254         /* Read next two samples from scratch1 buffer */
 255         x1 = _SIMD32_OFFSET(pScr1);
 256
 257         /* multiply and accumlate */
 258         acc0 = __SMLAD(x2, y2, acc0);
 259
 260         acc2 = __SMLAD(x1, y2, acc2);
 261
 262         /* pack input data */
 263 #ifndef ARM_MATH_BIG_ENDIAN
 264         x3 = __PKHBT(x1, x2, 0);
 265 #else
 266         x3 = __PKHBT(x2, x1, 0);
 267 #endif
 268
 269         acc3 = __SMLADX(x3, y1, acc3);
 270         acc1 = __SMLADX(x3, y2, acc1);
 271
 272         x2 = _SIMD32_OFFSET(pScr1 + 2u);
 273
 274 #ifndef ARM_MATH_BIG_ENDIAN
 275         x3 = __PKHBT(x2, x1, 0);
 276 #else
 277         x3 = __PKHBT(x1, x2, 0);
 278 #endif
 279
 280         acc3 = __SMLADX(x3, y2, acc3);
 281
 282         /* update scratch pointers */
 283         pIn2 += 4u;
 284         pScr1 += 4u;
 285
 286
 287         /* Decrement the loop counter */
 288         tapCnt--;
 289       }
 290
 291       /* Update scratch pointer for remaining samples of smaller length sequence */
 292       pScr1 -= 4u;
 293
 294       /* apply same above for remaining samples of smaller length sequence */
 295       tapCnt = (srcBLen) & 3u;
 296
 297       while(tapCnt > 0u)
 298       {
 299
 300         /* accumlate the results */
 301         acc0 += (*pScr1++ * *pIn2);
 302         acc1 += (*pScr1++ * *pIn2);
 303         acc2 += (*pScr1++ * *pIn2);
 304         acc3 += (*pScr1++ * *pIn2++);
 305
 306         pScr1 -= 3u;
 307
 308         /* Decrement the loop counter */
 309         tapCnt--;
 310       }
 311
 312       blkCnt--;
 313
 314
 315       /* Store the results in the accumulators in the destination buffer. */
 316
 317 #ifndef  ARM_MATH_BIG_ENDIAN
 318
 319       *__SIMD32(pOut)++ =
 320         __PKHBT(__SSAT((acc0 >> 15), 16), __SSAT((acc1 >> 15), 16), 16);
 321       *__SIMD32(pOut)++ =
 322         __PKHBT(__SSAT((acc2 >> 15), 16), __SSAT((acc3 >> 15), 16), 16);
 323
 324 #else
 325
 326       *__SIMD32(pOut)++ =
 327         __PKHBT(__SSAT((acc1 >> 15), 16), __SSAT((acc0 >> 15), 16), 16);
 328       *__SIMD32(pOut)++ =
 329         __PKHBT(__SSAT((acc3 >> 15), 16), __SSAT((acc2 >> 15), 16), 16);
 330
 331 #endif /*      #ifndef  ARM_MATH_BIG_ENDIAN    */
 332
 333       /* Initialization of inputB pointer */
 334       pIn2 = py;
 335
 336       pScratch1 += 4u;
 337
 338     }
 339
 340
 341     blkCnt = numPoints & 0x3;
 342
 343     /* Calculate convolution for remaining samples of Bigger length sequence */
 344     while(blkCnt > 0)
 345     {
 346       /* Initialze temporary scratch pointer as scratch1 */
 347       pScr1 = pScratch1;
 348
 349       /* Clear Accumlators */
 350       acc0 = 0;
 351
 352       tapCnt = (srcBLen) >> 1u;
 353
 354       while(tapCnt > 0u)
 355       {
 356
 357         /* Read next two samples from scratch1 buffer */
 358         x1 = *__SIMD32(pScr1)++;
 359
 360         /* Read two samples from smaller buffer */
 361         y1 = *__SIMD32(pIn2)++;
 362
 363         acc0 = __SMLAD(x1, y1, acc0);
 364
 365         /* Decrement the loop counter */
 366         tapCnt--;
 367       }
 368
 369       tapCnt = (srcBLen) & 1u;
 370
 371       /* apply same above for remaining samples of smaller length sequence */
 372       while(tapCnt > 0u)
 373       {
 374
 375         /* accumlate the results */
 376         acc0 += (*pScr1++ * *pIn2++);
 377
 378         /* Decrement the loop counter */
 379         tapCnt--;
 380       }
 381
 382       blkCnt--;
 383
 384       /* The result is in 2.30 format.  Convert to 1.15 with saturation.
 385        ** Then store the output in the destination buffer. */
 386       *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
 387
 388       /* Initialization of inputB pointer */
 389       pIn2 = py;
 390
 391       pScratch1 += 1u;
 392
 393     }
 394     /* set status as ARM_MATH_SUCCESS */
 395     status = ARM_MATH_SUCCESS;
 396   }
 397   /* Return to application */
 398   return (status);
 399 }
 400
 401 #else
 402
 403 arm_status arm_conv_partial_fast_opt_q15(
 404   q15_t * pSrcA,
 405   uint32_t srcALen,
 406   q15_t * pSrcB,
 407   uint32_t srcBLen,
 408   q15_t * pDst,
 409   uint32_t firstIndex,
 410   uint32_t numPoints,
 411   q15_t * pScratch1,
 412   q15_t * pScratch2)
 413 {
 414
 415   q15_t *pOut = pDst;                            /* output pointer */
 416   q15_t *pScr1 = pScratch1;                      /* Temporary pointer for scratch1 */
 417   q15_t *pScr2 = pScratch2;                      /* Temporary pointer for scratch1 */
 418   q31_t acc0, acc1, acc2, acc3;                  /* Accumulator */
 419   q15_t *pIn1;                                   /* inputA pointer */
 420   q15_t *pIn2;                                   /* inputB pointer */
 421   q15_t *px;                                     /* Intermediate inputA pointer  */
 422   q15_t *py;                                     /* Intermediate inputB pointer  */
 423   uint32_t j, k, blkCnt;                         /* loop counter */
 424   arm_status status;                             /* Status variable */
 425   uint32_t tapCnt;                               /* loop count */
 426   q15_t x10, x11, x20, x21;                      /* Temporary variables to hold srcA buffer */
 427   q15_t y10, y11;                                /* Temporary variables to hold srcB buffer */
 428
 429
 430   /* Check for range of output samples to be calculated */
 431   if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
 432   {
 433     /* Set status as ARM_MATH_ARGUMENT_ERROR */
 434     status = ARM_MATH_ARGUMENT_ERROR;
 435   }
 436   else
 437   {
 438
 439     /* The algorithm implementation is based on the lengths of the inputs. */
 440     /* srcB is always made to slide across srcA. */
 441     /* So srcBLen is always considered as shorter or equal to srcALen */
 442     if(srcALen >= srcBLen)
 443     {
 444       /* Initialization of inputA pointer */
 445       pIn1 = pSrcA;
 446
 447       /* Initialization of inputB pointer */
 448       pIn2 = pSrcB;
 449     }
 450     else
 451     {
 452       /* Initialization of inputA pointer */
 453       pIn1 = pSrcB;
 454
 455       /* Initialization of inputB pointer */
 456       pIn2 = pSrcA;
 457
 458       /* srcBLen is always considered as shorter or equal to srcALen */
 459       j = srcBLen;
 460       srcBLen = srcALen;
 461       srcALen = j;
 462     }
 463
 464     /* Temporary pointer for scratch2 */
 465     py = pScratch2;
 466
 467     /* pointer to take end of scratch2 buffer */
 468     pScr2 = pScratch2 + srcBLen - 1;
 469
 470     /* points to smaller length sequence */
 471     px = pIn2;
 472
 473     /* Apply loop unrolling and do 4 Copies simultaneously. */
 474     k = srcBLen >> 2u;
 475
 476     /* First part of the processing with loop unrolling copies 4 data points at a time.
 477      ** a second loop below copies for the remaining 1 to 3 samples. */
 478     while(k > 0u)
 479     {
 480       /* copy second buffer in reversal manner */
 481       *pScr2-- = *px++;
 482       *pScr2-- = *px++;
 483       *pScr2-- = *px++;
 484       *pScr2-- = *px++;
 485
 486       /* Decrement the loop counter */
 487       k--;
 488     }
 489
 490     /* If the count is not a multiple of 4, copy remaining samples here.
 491      ** No loop unrolling is used. */
 492     k = srcBLen % 0x4u;
 493
 494     while(k > 0u)
 495     {
 496       /* copy second buffer in reversal manner for remaining samples */
 497       *pScr2-- = *px++;
 498
 499       /* Decrement the loop counter */
 500       k--;
 501     }
 502
 503     /* Initialze temporary scratch pointer */
 504     pScr1 = pScratch1;
 505
 506     /* Fill (srcBLen - 1u) zeros in scratch buffer */
 507     arm_fill_q15(0, pScr1, (srcBLen - 1u));
 508
 509     /* Update temporary scratch pointer */
 510     pScr1 += (srcBLen - 1u);
 511
 512     /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
 513
 514
 515     /* Apply loop unrolling and do 4 Copies simultaneously. */
 516     k = srcALen >> 2u;
 517
 518     /* First part of the processing with loop unrolling copies 4 data points at a time.
 519      ** a second loop below copies for the remaining 1 to 3 samples. */
 520     while(k > 0u)
 521     {
 522       /* copy second buffer in reversal manner */
 523       *pScr1++ = *pIn1++;
 524       *pScr1++ = *pIn1++;
 525       *pScr1++ = *pIn1++;
 526       *pScr1++ = *pIn1++;
 527
 528       /* Decrement the loop counter */
 529       k--;
 530     }
 531
 532     /* If the count is not a multiple of 4, copy remaining samples here.
 533      ** No loop unrolling is used. */
 534     k = srcALen % 0x4u;
 535
 536     while(k > 0u)
 537     {
 538       /* copy second buffer in reversal manner for remaining samples */
 539       *pScr1++ = *pIn1++;
 540
 541       /* Decrement the loop counter */
 542       k--;
 543     }
 544
 545
 546     /* Apply loop unrolling and do 4 Copies simultaneously. */
 547     k = (srcBLen - 1u) >> 2u;
 548
 549     /* First part of the processing with loop unrolling copies 4 data points at a time.
 550      ** a second loop below copies for the remaining 1 to 3 samples. */
 551     while(k > 0u)
 552     {
 553       /* copy second buffer in reversal manner */
 554       *pScr1++ = 0;
 555       *pScr1++ = 0;
 556       *pScr1++ = 0;
 557       *pScr1++ = 0;
 558
 559       /* Decrement the loop counter */
 560       k--;
 561     }
 562
 563     /* If the count is not a multiple of 4, copy remaining samples here.
 564      ** No loop unrolling is used. */
 565     k = (srcBLen - 1u) % 0x4u;
 566
 567     while(k > 0u)
 568     {
 569       /* copy second buffer in reversal manner for remaining samples */
 570       *pScr1++ = 0;
 571
 572       /* Decrement the loop counter */
 573       k--;
 574     }
 575
 576
 577     /* Initialization of pIn2 pointer */
 578     pIn2 = py;
 579
 580     pScratch1 += firstIndex;
 581
 582     pOut = pDst + firstIndex;
 583
 584     /* Actual convolution process starts here */
 585     blkCnt = (numPoints) >> 2;
 586
 587     while(blkCnt > 0)
 588     {
 589       /* Initialze temporary scratch pointer as scratch1 */
 590       pScr1 = pScratch1;
 591
 592       /* Clear Accumlators */
 593       acc0 = 0;
 594       acc1 = 0;
 595       acc2 = 0;
 596       acc3 = 0;
 597
 598       /* Read two samples from scratch1 buffer */
 599       x10 = *pScr1++;
 600       x11 = *pScr1++;
 601
 602       /* Read next two samples from scratch1 buffer */
 603       x20 = *pScr1++;
 604       x21 = *pScr1++;
 605
 606       tapCnt = (srcBLen) >> 2u;
 607
 608       while(tapCnt > 0u)
 609       {
 610
 611         /* Read two samples from smaller buffer */
 612         y10 = *pIn2;
 613         y11 = *(pIn2 + 1u);
 614
 615         /* multiply and accumlate */
 616         acc0 += (q31_t) x10 *y10;
 617         acc0 += (q31_t) x11 *y11;
 618         acc2 += (q31_t) x20 *y10;
 619         acc2 += (q31_t) x21 *y11;
 620
 621         /* multiply and accumlate */
 622         acc1 += (q31_t) x11 *y10;
 623         acc1 += (q31_t) x20 *y11;
 624
 625         /* Read next two samples from scratch1 buffer */
 626         x10 = *pScr1;
 627         x11 = *(pScr1 + 1u);
 628
 629         /* multiply and accumlate */
 630         acc3 += (q31_t) x21 *y10;
 631         acc3 += (q31_t) x10 *y11;
 632
 633         /* Read next two samples from scratch2 buffer */
 634         y10 = *(pIn2 + 2u);
 635         y11 = *(pIn2 + 3u);
 636
 637         /* multiply and accumlate */
 638         acc0 += (q31_t) x20 *y10;
 639         acc0 += (q31_t) x21 *y11;
 640         acc2 += (q31_t) x10 *y10;
 641         acc2 += (q31_t) x11 *y11;
 642         acc1 += (q31_t) x21 *y10;
 643         acc1 += (q31_t) x10 *y11;
 644
 645         /* Read next two samples from scratch1 buffer */
 646         x20 = *(pScr1 + 2);
 647         x21 = *(pScr1 + 3);
 648
 649         /* multiply and accumlate */
 650         acc3 += (q31_t) x11 *y10;
 651         acc3 += (q31_t) x20 *y11;
 652
 653         /* update scratch pointers */
 654         pIn2 += 4u;
 655         pScr1 += 4u;
 656
 657         /* Decrement the loop counter */
 658         tapCnt--;
 659       }
 660
 661       /* Update scratch pointer for remaining samples of smaller length sequence */
 662       pScr1 -= 4u;
 663
 664       /* apply same above for remaining samples of smaller length sequence */
 665       tapCnt = (srcBLen) & 3u;
 666
 667       while(tapCnt > 0u)
 668       {
 669         /* accumlate the results */
 670         acc0 += (*pScr1++ * *pIn2);
 671         acc1 += (*pScr1++ * *pIn2);
 672         acc2 += (*pScr1++ * *pIn2);
 673         acc3 += (*pScr1++ * *pIn2++);
 674
 675         pScr1 -= 3u;
 676
 677         /* Decrement the loop counter */
 678         tapCnt--;
 679       }
 680
 681       blkCnt--;
 682
 683
 684       /* Store the results in the accumulators in the destination buffer. */
 685       *pOut++ = __SSAT((acc0 >> 15), 16);
 686       *pOut++ = __SSAT((acc1 >> 15), 16);
 687       *pOut++ = __SSAT((acc2 >> 15), 16);
 688       *pOut++ = __SSAT((acc3 >> 15), 16);
 689
 690       /* Initialization of inputB pointer */
 691       pIn2 = py;
 692
 693       pScratch1 += 4u;
 694
 695     }
 696
 697
 698     blkCnt = numPoints & 0x3;
 699
 700     /* Calculate convolution for remaining samples of Bigger length sequence */
 701     while(blkCnt > 0)
 702     {
 703       /* Initialze temporary scratch pointer as scratch1 */
 704       pScr1 = pScratch1;
 705
 706       /* Clear Accumlators */
 707       acc0 = 0;
 708
 709       tapCnt = (srcBLen) >> 1u;
 710
 711       while(tapCnt > 0u)
 712       {
 713
 714         /* Read next two samples from scratch1 buffer */
 715         x10 = *pScr1++;
 716         x11 = *pScr1++;
 717
 718         /* Read two samples from smaller buffer */
 719         y10 = *pIn2++;
 720         y11 = *pIn2++;
 721
 722         /* multiply and accumlate */
 723         acc0 += (q31_t) x10 *y10;
 724         acc0 += (q31_t) x11 *y11;
 725
 726         /* Decrement the loop counter */
 727         tapCnt--;
 728       }
 729
 730       tapCnt = (srcBLen) & 1u;
 731
 732       /* apply same above for remaining samples of smaller length sequence */
 733       while(tapCnt > 0u)
 734       {
 735
 736         /* accumlate the results */
 737         acc0 += (*pScr1++ * *pIn2++);
 738
 739         /* Decrement the loop counter */
 740         tapCnt--;
 741       }
 742
 743       blkCnt--;
 744
 745       /* Store the result in the accumulator in the destination buffer. */
 746       *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
 747
 748       /* Initialization of inputB pointer */
 749       pIn2 = py;
 750
 751       pScratch1 += 1u;
 752
 753     }
 754
 755     /* set status as ARM_MATH_SUCCESS */
 756     status = ARM_MATH_SUCCESS;
 757
 758   }
 759
 760   /* Return to application */
 761   return (status);
 762 }
 763
 764 #endif  /*      #ifndef UNALIGNED_SUPPORT_DISABLE       */
 765
 766 /**
 767  * @} end of PartialConv group
 768  */