arm_conv_partial_opt_q15.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        17. January 2013
00005 * $Revision:    V1.4.1
00006 *    
00007 * Project:      CMSIS DSP Library    
00008 * Title:        arm_conv_partial_opt_q15.c    
00009 *    
00010 * Description:  Partial convolution of Q15 sequences.   
00011 *    
00012 * Target Processor: Cortex-M4/Cortex-M3
00013 *  
00014 * Redistribution and use in source and binary forms, with or without 
00015 * modification, are permitted provided that the following conditions
00016 * are met:
00017 *   - Redistributions of source code must retain the above copyright
00018 *     notice, this list of conditions and the following disclaimer.
00019 *   - Redistributions in binary form must reproduce the above copyright
00020 *     notice, this list of conditions and the following disclaimer in
00021 *     the documentation and/or other materials provided with the 
00022 *     distribution.
00023 *   - Neither the name of ARM LIMITED nor the names of its contributors
00024 *     may be used to endorse or promote products derived from this
00025 *     software without specific prior written permission.
00026 *
00027 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
00028 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
00029 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
00030 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 
00031 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
00032 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
00033 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
00034 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
00035 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
00036 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
00037 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
00038 * POSSIBILITY OF SUCH DAMAGE.  
00039 * -------------------------------------------------------------------- */
00040 
00041 #include "arm_math.h"
00042 
00043 /**    
00044  * @ingroup groupFilters    
00045  */
00046 
00047 /**    
00048  * @addtogroup PartialConv    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief Partial convolution of Q15 sequences.    
00054  * @param[in]       *pSrcA points to the first input sequence.    
00055  * @param[in]       srcALen length of the first input sequence.    
00056  * @param[in]       *pSrcB points to the second input sequence.    
00057  * @param[in]       srcBLen length of the second input sequence.    
00058  * @param[out]      *pDst points to the location where the output result is written.    
00059  * @param[in]       firstIndex is the first output sample to start with.    
00060  * @param[in]       numPoints is the number of output points to be computed.    
00061  * @param[in]       *pScratch1 points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.   
00062  * @param[in]       *pScratch2 points to scratch buffer of size min(srcALen, srcBLen).   
00063  * @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].    
00064  *    
00065  * \par Restrictions    
00066  *  If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE    
00067  *  In this case input, output, state buffers should be aligned by 32-bit    
00068  *    
00069  * Refer to <code>arm_conv_partial_fast_q15()</code> for a faster but less precise version of this function for Cortex-M3 and Cortex-M4.   
00070  *  
00071  * 
00072  */
00073 
00074 #ifndef UNALIGNED_SUPPORT_DISABLE
00075 
00076 arm_status arm_conv_partial_opt_q15(
00077   q15_t * pSrcA,
00078   uint32_t srcALen,
00079   q15_t * pSrcB,
00080   uint32_t srcBLen,
00081   q15_t * pDst,
00082   uint32_t firstIndex,
00083   uint32_t numPoints,
00084   q15_t * pScratch1,
00085   q15_t * pScratch2)
00086 {
00087 
00088   q15_t *pOut = pDst;                            /* output pointer */
00089   q15_t *pScr1 = pScratch1;                      /* Temporary pointer for scratch1 */
00090   q15_t *pScr2 = pScratch2;                      /* Temporary pointer for scratch1 */
00091   q63_t acc0, acc1, acc2, acc3;                  /* Accumulator */
00092   q31_t x1, x2, x3;                              /* Temporary variables to hold state and coefficient values */
00093   q31_t y1, y2;                                  /* State variables */
00094   q15_t *pIn1;                                   /* inputA pointer */
00095   q15_t *pIn2;                                   /* inputB pointer */
00096   q15_t *px;                                     /* Intermediate inputA pointer  */
00097   q15_t *py;                                     /* Intermediate inputB pointer  */
00098   uint32_t j, k, blkCnt;                         /* loop counter */
00099   arm_status status;                             /* Status variable */
00100   uint32_t tapCnt;                               /* loop count */
00101 
00102   /* Check for range of output samples to be calculated */
00103   if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
00104   {
00105     /* Set status as ARM_MATH_ARGUMENT_ERROR */
00106     status = ARM_MATH_ARGUMENT_ERROR;
00107   }
00108   else
00109   {
00110 
00111     /* The algorithm implementation is based on the lengths of the inputs. */
00112     /* srcB is always made to slide across srcA. */
00113     /* So srcBLen is always considered as shorter or equal to srcALen */
00114     if(srcALen >= srcBLen)
00115     {
00116       /* Initialization of inputA pointer */
00117       pIn1 = pSrcA;
00118 
00119       /* Initialization of inputB pointer */
00120       pIn2 = pSrcB;
00121     }
00122     else
00123     {
00124       /* Initialization of inputA pointer */
00125       pIn1 = pSrcB;
00126 
00127       /* Initialization of inputB pointer */
00128       pIn2 = pSrcA;
00129 
00130       /* srcBLen is always considered as shorter or equal to srcALen */
00131       j = srcBLen;
00132       srcBLen = srcALen;
00133       srcALen = j;
00134     }
00135 
00136     /* Temporary pointer for scratch2 */
00137     py = pScratch2;
00138 
00139     /* pointer to take end of scratch2 buffer */
00140     pScr2 = pScratch2 + srcBLen - 1;
00141 
00142     /* points to smaller length sequence */
00143     px = pIn2;
00144 
00145     /* Apply loop unrolling and do 4 Copies simultaneously. */
00146     k = srcBLen >> 2u;
00147 
00148     /* First part of the processing with loop unrolling copies 4 data points at a time.       
00149      ** a second loop below copies for the remaining 1 to 3 samples. */
00150     while(k > 0u)
00151     {
00152       /* copy second buffer in reversal manner */
00153       *pScr2-- = *px++;
00154       *pScr2-- = *px++;
00155       *pScr2-- = *px++;
00156       *pScr2-- = *px++;
00157 
00158       /* Decrement the loop counter */
00159       k--;
00160     }
00161 
00162     /* If the count is not a multiple of 4, copy remaining samples here.       
00163      ** No loop unrolling is used. */
00164     k = srcBLen % 0x4u;
00165 
00166     while(k > 0u)
00167     {
00168       /* copy second buffer in reversal manner for remaining samples */
00169       *pScr2-- = *px++;
00170 
00171       /* Decrement the loop counter */
00172       k--;
00173     }
00174 
00175     /* Initialze temporary scratch pointer */
00176     pScr1 = pScratch1;
00177 
00178     /* Fill (srcBLen - 1u) zeros in scratch buffer */
00179     arm_fill_q15(0, pScr1, (srcBLen - 1u));
00180 
00181     /* Update temporary scratch pointer */
00182     pScr1 += (srcBLen - 1u);
00183 
00184     /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
00185 
00186     /* Copy (srcALen) samples in scratch buffer */
00187     arm_copy_q15(pIn1, pScr1, srcALen);
00188 
00189     /* Update pointers */
00190     pScr1 += srcALen;
00191 
00192     /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
00193     arm_fill_q15(0, pScr1, (srcBLen - 1u));
00194 
00195     /* Update pointer */
00196     pScr1 += (srcBLen - 1u);
00197 
00198     /* Initialization of pIn2 pointer */
00199     pIn2 = py;
00200 
00201     pScratch1 += firstIndex;
00202 
00203     pOut = pDst + firstIndex;
00204 
00205     /* Actual convolution process starts here */
00206     blkCnt = (numPoints) >> 2;
00207 
00208     while(blkCnt > 0)
00209     {
00210       /* Initialze temporary scratch pointer as scratch1 */
00211       pScr1 = pScratch1;
00212 
00213       /* Clear Accumlators */
00214       acc0 = 0;
00215       acc1 = 0;
00216       acc2 = 0;
00217       acc3 = 0;
00218 
00219       /* Read two samples from scratch1 buffer */
00220       x1 = *__SIMD32(pScr1)++;
00221 
00222       /* Read next two samples from scratch1 buffer */
00223       x2 = *__SIMD32(pScr1)++;
00224 
00225       tapCnt = (srcBLen) >> 2u;
00226 
00227       while(tapCnt > 0u)
00228       {
00229 
00230         /* Read four samples from smaller buffer */
00231         y1 = _SIMD32_OFFSET(pIn2);
00232         y2 = _SIMD32_OFFSET(pIn2 + 2u);
00233 
00234         /* multiply and accumlate */
00235         acc0 = __SMLALD(x1, y1, acc0);
00236         acc2 = __SMLALD(x2, y1, acc2);
00237 
00238         /* pack input data */
00239 #ifndef ARM_MATH_BIG_ENDIAN
00240         x3 = __PKHBT(x2, x1, 0);
00241 #else
00242         x3 = __PKHBT(x1, x2, 0);
00243 #endif
00244 
00245         /* multiply and accumlate */
00246         acc1 = __SMLALDX(x3, y1, acc1);
00247 
00248         /* Read next two samples from scratch1 buffer */
00249         x1 = _SIMD32_OFFSET(pScr1);
00250 
00251         /* multiply and accumlate */
00252         acc0 = __SMLALD(x2, y2, acc0);
00253         acc2 = __SMLALD(x1, y2, acc2);
00254 
00255         /* pack input data */
00256 #ifndef ARM_MATH_BIG_ENDIAN
00257         x3 = __PKHBT(x1, x2, 0);
00258 #else
00259         x3 = __PKHBT(x2, x1, 0);
00260 #endif
00261 
00262         acc3 = __SMLALDX(x3, y1, acc3);
00263         acc1 = __SMLALDX(x3, y2, acc1);
00264 
00265         x2 = _SIMD32_OFFSET(pScr1 + 2u);
00266 
00267 #ifndef ARM_MATH_BIG_ENDIAN
00268         x3 = __PKHBT(x2, x1, 0);
00269 #else
00270         x3 = __PKHBT(x1, x2, 0);
00271 #endif
00272 
00273         acc3 = __SMLALDX(x3, y2, acc3);
00274 
00275         /* update scratch pointers */
00276         pIn2 += 4u;
00277         pScr1 += 4u;
00278 
00279 
00280         /* Decrement the loop counter */
00281         tapCnt--;
00282       }
00283 
00284       /* Update scratch pointer for remaining samples of smaller length sequence */
00285       pScr1 -= 4u;
00286 
00287       /* apply same above for remaining samples of smaller length sequence */
00288       tapCnt = (srcBLen) & 3u;
00289 
00290       while(tapCnt > 0u)
00291       {
00292         /* accumlate the results */
00293         acc0 += (*pScr1++ * *pIn2);
00294         acc1 += (*pScr1++ * *pIn2);
00295         acc2 += (*pScr1++ * *pIn2);
00296         acc3 += (*pScr1++ * *pIn2++);
00297 
00298         pScr1 -= 3u;
00299 
00300         /* Decrement the loop counter */
00301         tapCnt--;
00302       }
00303 
00304       blkCnt--;
00305 
00306 
00307       /* Store the results in the accumulators in the destination buffer. */
00308 
00309 #ifndef  ARM_MATH_BIG_ENDIAN
00310 
00311       *__SIMD32(pOut)++ =
00312         __PKHBT(__SSAT((acc0 >> 15), 16), __SSAT((acc1 >> 15), 16), 16);
00313       *__SIMD32(pOut)++ =
00314         __PKHBT(__SSAT((acc2 >> 15), 16), __SSAT((acc3 >> 15), 16), 16);
00315 
00316 #else
00317 
00318       *__SIMD32(pOut)++ =
00319         __PKHBT(__SSAT((acc1 >> 15), 16), __SSAT((acc0 >> 15), 16), 16);
00320       *__SIMD32(pOut)++ =
00321         __PKHBT(__SSAT((acc3 >> 15), 16), __SSAT((acc2 >> 15), 16), 16);
00322 
00323 #endif /*      #ifndef  ARM_MATH_BIG_ENDIAN    */
00324 
00325       /* Initialization of inputB pointer */
00326       pIn2 = py;
00327 
00328       pScratch1 += 4u;
00329 
00330     }
00331 
00332 
00333     blkCnt = numPoints & 0x3;
00334 
00335     /* Calculate convolution for remaining samples of Bigger length sequence */
00336     while(blkCnt > 0)
00337     {
00338       /* Initialze temporary scratch pointer as scratch1 */
00339       pScr1 = pScratch1;
00340 
00341       /* Clear Accumlators */
00342       acc0 = 0;
00343 
00344       tapCnt = (srcBLen) >> 1u;
00345 
00346       while(tapCnt > 0u)
00347       {
00348 
00349         /* Read next two samples from scratch1 buffer */
00350         x1 = *__SIMD32(pScr1)++;
00351 
00352         /* Read two samples from smaller buffer */
00353         y1 = *__SIMD32(pIn2)++;
00354 
00355         acc0 = __SMLALD(x1, y1, acc0);
00356 
00357         /* Decrement the loop counter */
00358         tapCnt--;
00359       }
00360 
00361       tapCnt = (srcBLen) & 1u;
00362 
00363       /* apply same above for remaining samples of smaller length sequence */
00364       while(tapCnt > 0u)
00365       {
00366 
00367         /* accumlate the results */
00368         acc0 += (*pScr1++ * *pIn2++);
00369 
00370         /* Decrement the loop counter */
00371         tapCnt--;
00372       }
00373 
00374       blkCnt--;
00375 
00376       /* Store the result in the accumulator in the destination buffer. */
00377       *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
00378 
00379       /* Initialization of inputB pointer */
00380       pIn2 = py;
00381 
00382       pScratch1 += 1u;
00383 
00384     }
00385 
00386     /* set status as ARM_MATH_SUCCESS */
00387     status = ARM_MATH_SUCCESS;
00388 
00389   }
00390 
00391   /* Return to application */
00392   return (status);
00393 }
00394 
00395 #else
00396 
00397 arm_status arm_conv_partial_opt_q15(
00398   q15_t * pSrcA,
00399   uint32_t srcALen,
00400   q15_t * pSrcB,
00401   uint32_t srcBLen,
00402   q15_t * pDst,
00403   uint32_t firstIndex,
00404   uint32_t numPoints,
00405   q15_t * pScratch1,
00406   q15_t * pScratch2)
00407 {
00408 
00409   q15_t *pOut = pDst;                            /* output pointer */
00410   q15_t *pScr1 = pScratch1;                      /* Temporary pointer for scratch1 */
00411   q15_t *pScr2 = pScratch2;                      /* Temporary pointer for scratch1 */
00412   q63_t acc0, acc1, acc2, acc3;                  /* Accumulator */
00413   q15_t *pIn1;                                   /* inputA pointer */
00414   q15_t *pIn2;                                   /* inputB pointer */
00415   q15_t *px;                                     /* Intermediate inputA pointer  */
00416   q15_t *py;                                     /* Intermediate inputB pointer  */
00417   uint32_t j, k, blkCnt;                         /* loop counter */
00418   arm_status status;                             /* Status variable */
00419   uint32_t tapCnt;                               /* loop count */
00420   q15_t x10, x11, x20, x21;                      /* Temporary variables to hold srcA buffer */
00421   q15_t y10, y11;                                /* Temporary variables to hold srcB buffer */
00422 
00423 
00424   /* Check for range of output samples to be calculated */
00425   if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
00426   {
00427     /* Set status as ARM_MATH_ARGUMENT_ERROR */
00428     status = ARM_MATH_ARGUMENT_ERROR;
00429   }
00430   else
00431   {
00432 
00433     /* The algorithm implementation is based on the lengths of the inputs. */
00434     /* srcB is always made to slide across srcA. */
00435     /* So srcBLen is always considered as shorter or equal to srcALen */
00436     if(srcALen >= srcBLen)
00437     {
00438       /* Initialization of inputA pointer */
00439       pIn1 = pSrcA;
00440 
00441       /* Initialization of inputB pointer */
00442       pIn2 = pSrcB;
00443     }
00444     else
00445     {
00446       /* Initialization of inputA pointer */
00447       pIn1 = pSrcB;
00448 
00449       /* Initialization of inputB pointer */
00450       pIn2 = pSrcA;
00451 
00452       /* srcBLen is always considered as shorter or equal to srcALen */
00453       j = srcBLen;
00454       srcBLen = srcALen;
00455       srcALen = j;
00456     }
00457 
00458     /* Temporary pointer for scratch2 */
00459     py = pScratch2;
00460 
00461     /* pointer to take end of scratch2 buffer */
00462     pScr2 = pScratch2 + srcBLen - 1;
00463 
00464     /* points to smaller length sequence */
00465     px = pIn2;
00466 
00467     /* Apply loop unrolling and do 4 Copies simultaneously. */
00468     k = srcBLen >> 2u;
00469 
00470     /* First part of the processing with loop unrolling copies 4 data points at a time.       
00471      ** a second loop below copies for the remaining 1 to 3 samples. */
00472     while(k > 0u)
00473     {
00474       /* copy second buffer in reversal manner */
00475       *pScr2-- = *px++;
00476       *pScr2-- = *px++;
00477       *pScr2-- = *px++;
00478       *pScr2-- = *px++;
00479 
00480       /* Decrement the loop counter */
00481       k--;
00482     }
00483 
00484     /* If the count is not a multiple of 4, copy remaining samples here.       
00485      ** No loop unrolling is used. */
00486     k = srcBLen % 0x4u;
00487 
00488     while(k > 0u)
00489     {
00490       /* copy second buffer in reversal manner for remaining samples */
00491       *pScr2-- = *px++;
00492 
00493       /* Decrement the loop counter */
00494       k--;
00495     }
00496 
00497     /* Initialze temporary scratch pointer */
00498     pScr1 = pScratch1;
00499 
00500     /* Fill (srcBLen - 1u) zeros in scratch buffer */
00501     arm_fill_q15(0, pScr1, (srcBLen - 1u));
00502 
00503     /* Update temporary scratch pointer */
00504     pScr1 += (srcBLen - 1u);
00505 
00506     /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
00507 
00508 
00509     /* Apply loop unrolling and do 4 Copies simultaneously. */
00510     k = srcALen >> 2u;
00511 
00512     /* First part of the processing with loop unrolling copies 4 data points at a time.       
00513      ** a second loop below copies for the remaining 1 to 3 samples. */
00514     while(k > 0u)
00515     {
00516       /* copy second buffer in reversal manner */
00517       *pScr1++ = *pIn1++;
00518       *pScr1++ = *pIn1++;
00519       *pScr1++ = *pIn1++;
00520       *pScr1++ = *pIn1++;
00521 
00522       /* Decrement the loop counter */
00523       k--;
00524     }
00525 
00526     /* If the count is not a multiple of 4, copy remaining samples here.       
00527      ** No loop unrolling is used. */
00528     k = srcALen % 0x4u;
00529 
00530     while(k > 0u)
00531     {
00532       /* copy second buffer in reversal manner for remaining samples */
00533       *pScr1++ = *pIn1++;
00534 
00535       /* Decrement the loop counter */
00536       k--;
00537     }
00538 
00539 
00540     /* Apply loop unrolling and do 4 Copies simultaneously. */
00541     k = (srcBLen - 1u) >> 2u;
00542 
00543     /* First part of the processing with loop unrolling copies 4 data points at a time.       
00544      ** a second loop below copies for the remaining 1 to 3 samples. */
00545     while(k > 0u)
00546     {
00547       /* copy second buffer in reversal manner */
00548       *pScr1++ = 0;
00549       *pScr1++ = 0;
00550       *pScr1++ = 0;
00551       *pScr1++ = 0;
00552 
00553       /* Decrement the loop counter */
00554       k--;
00555     }
00556 
00557     /* If the count is not a multiple of 4, copy remaining samples here.       
00558      ** No loop unrolling is used. */
00559     k = (srcBLen - 1u) % 0x4u;
00560 
00561     while(k > 0u)
00562     {
00563       /* copy second buffer in reversal manner for remaining samples */
00564       *pScr1++ = 0;
00565 
00566       /* Decrement the loop counter */
00567       k--;
00568     }
00569 
00570 
00571     /* Initialization of pIn2 pointer */
00572     pIn2 = py;
00573 
00574     pScratch1 += firstIndex;
00575 
00576     pOut = pDst + firstIndex;
00577 
00578     /* Actual convolution process starts here */
00579     blkCnt = (numPoints) >> 2;
00580 
00581     while(blkCnt > 0)
00582     {
00583       /* Initialze temporary scratch pointer as scratch1 */
00584       pScr1 = pScratch1;
00585 
00586       /* Clear Accumlators */
00587       acc0 = 0;
00588       acc1 = 0;
00589       acc2 = 0;
00590       acc3 = 0;
00591 
00592       /* Read two samples from scratch1 buffer */
00593       x10 = *pScr1++;
00594       x11 = *pScr1++;
00595 
00596       /* Read next two samples from scratch1 buffer */
00597       x20 = *pScr1++;
00598       x21 = *pScr1++;
00599 
00600       tapCnt = (srcBLen) >> 2u;
00601 
00602       while(tapCnt > 0u)
00603       {
00604 
00605         /* Read two samples from smaller buffer */
00606         y10 = *pIn2;
00607         y11 = *(pIn2 + 1u);
00608 
00609         /* multiply and accumlate */
00610         acc0 += (q63_t) x10 *y10;
00611         acc0 += (q63_t) x11 *y11;
00612         acc2 += (q63_t) x20 *y10;
00613         acc2 += (q63_t) x21 *y11;
00614 
00615         /* multiply and accumlate */
00616         acc1 += (q63_t) x11 *y10;
00617         acc1 += (q63_t) x20 *y11;
00618 
00619         /* Read next two samples from scratch1 buffer */
00620         x10 = *pScr1;
00621         x11 = *(pScr1 + 1u);
00622 
00623         /* multiply and accumlate */
00624         acc3 += (q63_t) x21 *y10;
00625         acc3 += (q63_t) x10 *y11;
00626 
00627         /* Read next two samples from scratch2 buffer */
00628         y10 = *(pIn2 + 2u);
00629         y11 = *(pIn2 + 3u);
00630 
00631         /* multiply and accumlate */
00632         acc0 += (q63_t) x20 *y10;
00633         acc0 += (q63_t) x21 *y11;
00634         acc2 += (q63_t) x10 *y10;
00635         acc2 += (q63_t) x11 *y11;
00636         acc1 += (q63_t) x21 *y10;
00637         acc1 += (q63_t) x10 *y11;
00638 
00639         /* Read next two samples from scratch1 buffer */
00640         x20 = *(pScr1 + 2);
00641         x21 = *(pScr1 + 3);
00642 
00643         /* multiply and accumlate */
00644         acc3 += (q63_t) x11 *y10;
00645         acc3 += (q63_t) x20 *y11;
00646 
00647         /* update scratch pointers */
00648         pIn2 += 4u;
00649         pScr1 += 4u;
00650 
00651         /* Decrement the loop counter */
00652         tapCnt--;
00653       }
00654 
00655       /* Update scratch pointer for remaining samples of smaller length sequence */
00656       pScr1 -= 4u;
00657 
00658       /* apply same above for remaining samples of smaller length sequence */
00659       tapCnt = (srcBLen) & 3u;
00660 
00661       while(tapCnt > 0u)
00662       {
00663         /* accumlate the results */
00664         acc0 += (*pScr1++ * *pIn2);
00665         acc1 += (*pScr1++ * *pIn2);
00666         acc2 += (*pScr1++ * *pIn2);
00667         acc3 += (*pScr1++ * *pIn2++);
00668 
00669         pScr1 -= 3u;
00670 
00671         /* Decrement the loop counter */
00672         tapCnt--;
00673       }
00674 
00675       blkCnt--;
00676 
00677 
00678       /* Store the results in the accumulators in the destination buffer. */
00679       *pOut++ = __SSAT((acc0 >> 15), 16);
00680       *pOut++ = __SSAT((acc1 >> 15), 16);
00681       *pOut++ = __SSAT((acc2 >> 15), 16);
00682       *pOut++ = __SSAT((acc3 >> 15), 16);
00683 
00684 
00685       /* Initialization of inputB pointer */
00686       pIn2 = py;
00687 
00688       pScratch1 += 4u;
00689 
00690     }
00691 
00692 
00693     blkCnt = numPoints & 0x3;
00694 
00695     /* Calculate convolution for remaining samples of Bigger length sequence */
00696     while(blkCnt > 0)
00697     {
00698       /* Initialze temporary scratch pointer as scratch1 */
00699       pScr1 = pScratch1;
00700 
00701       /* Clear Accumlators */
00702       acc0 = 0;
00703 
00704       tapCnt = (srcBLen) >> 1u;
00705 
00706       while(tapCnt > 0u)
00707       {
00708 
00709         /* Read next two samples from scratch1 buffer */
00710         x10 = *pScr1++;
00711         x11 = *pScr1++;
00712 
00713         /* Read two samples from smaller buffer */
00714         y10 = *pIn2++;
00715         y11 = *pIn2++;
00716 
00717         /* multiply and accumlate */
00718         acc0 += (q63_t) x10 *y10;
00719         acc0 += (q63_t) x11 *y11;
00720 
00721         /* Decrement the loop counter */
00722         tapCnt--;
00723       }
00724 
00725       tapCnt = (srcBLen) & 1u;
00726 
00727       /* apply same above for remaining samples of smaller length sequence */
00728       while(tapCnt > 0u)
00729       {
00730 
00731         /* accumlate the results */
00732         acc0 += (*pScr1++ * *pIn2++);
00733 
00734         /* Decrement the loop counter */
00735         tapCnt--;
00736       }
00737 
00738       blkCnt--;
00739 
00740       /* Store the result in the accumulator in the destination buffer. */
00741       *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
00742 
00743 
00744       /* Initialization of inputB pointer */
00745       pIn2 = py;
00746 
00747       pScratch1 += 1u;
00748 
00749     }
00750 
00751     /* set status as ARM_MATH_SUCCESS */
00752     status = ARM_MATH_SUCCESS;
00753 
00754   }
00755 
00756   /* Return to application */
00757   return (status);
00758 }
00759 
00760 #endif  /*  #ifndef UNALIGNED_SUPPORT_DISABLE   */
00761 
00762 
00763 /**    
00764  * @} end of PartialConv group    
00765  */