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_conv_partial_opt_q7.c@5:3762170b6d4d, 2015-11-20 (annotated)

Committer:
mbed_official
Date:
Fri Nov 20 08:45:18 2015 +0000
Revision:
5:3762170b6d4d
Parent:
3:7a284390b0ce 
Synchronized with git revision 2eb940b9a73af188d3004a2575fdfbb05febe62b



Full URL: https://github.com/mbedmicro/mbed/commit/2eb940b9a73af188d3004a2575fdfbb05febe62b/



Added option to build rpc library. closes #1426

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 5:3762170b6d4d 2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 5:3762170b6d4d 4 * $Date: 19. March 2015
mbed_official 5:3762170b6d4d 5 * $Revision: V.1.4.5
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_conv_partial_opt_q7.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Partial convolution of Q7 sequences.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3
emilmont 1:fdd22bb7aa52 13 *
mbed_official 3:7a284390b0ce 14 * Redistribution and use in source and binary forms, with or without
mbed_official 3:7a284390b0ce 15 * modification, are permitted provided that the following conditions
mbed_official 3:7a284390b0ce 16 * are met:
mbed_official 3:7a284390b0ce 17 * - Redistributions of source code must retain the above copyright
mbed_official 3:7a284390b0ce 18 * notice, this list of conditions and the following disclaimer.
mbed_official 3:7a284390b0ce 19 * - Redistributions in binary form must reproduce the above copyright
mbed_official 3:7a284390b0ce 20 * notice, this list of conditions and the following disclaimer in
mbed_official 3:7a284390b0ce 21 * the documentation and/or other materials provided with the
mbed_official 3:7a284390b0ce 22 * distribution.
mbed_official 3:7a284390b0ce 23 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 3:7a284390b0ce 24 * may be used to endorse or promote products derived from this
mbed_official 3:7a284390b0ce 25 * software without specific prior written permission.
mbed_official 3:7a284390b0ce 26 *
mbed_official 3:7a284390b0ce 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 3:7a284390b0ce 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 3:7a284390b0ce 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 3:7a284390b0ce 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 3:7a284390b0ce 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 3:7a284390b0ce 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 3:7a284390b0ce 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 3:7a284390b0ce 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 3:7a284390b0ce 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 3:7a284390b0ce 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 3:7a284390b0ce 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 3:7a284390b0ce 38 * POSSIBILITY OF SUCH DAMAGE.
emilmont 1:fdd22bb7aa52 39 * -------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 40
emilmont 1:fdd22bb7aa52 41 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 42
emilmont 1:fdd22bb7aa52 43 /**
emilmont 1:fdd22bb7aa52 44 * @ingroup groupFilters
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
emilmont 1:fdd22bb7aa52 47 /**
emilmont 1:fdd22bb7aa52 48 * @addtogroup PartialConv
emilmont 1:fdd22bb7aa52 49 * @{
emilmont 1:fdd22bb7aa52 50 */
emilmont 1:fdd22bb7aa52 51
emilmont 1:fdd22bb7aa52 52 /**
emilmont 1:fdd22bb7aa52 53 * @brief Partial convolution of Q7 sequences.
emilmont 1:fdd22bb7aa52 54 * @param[in] *pSrcA points to the first input sequence.
emilmont 1:fdd22bb7aa52 55 * @param[in] srcALen length of the first input sequence.
emilmont 1:fdd22bb7aa52 56 * @param[in] *pSrcB points to the second input sequence.
emilmont 1:fdd22bb7aa52 57 * @param[in] srcBLen length of the second input sequence.
emilmont 1:fdd22bb7aa52 58 * @param[out] *pDst points to the location where the output result is written.
emilmont 1:fdd22bb7aa52 59 * @param[in] firstIndex is the first output sample to start with.
emilmont 1:fdd22bb7aa52 60 * @param[in] numPoints is the number of output points to be computed.
emilmont 1:fdd22bb7aa52 61 * @param[in] *pScratch1 points to scratch buffer(of type q15_t) of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emilmont 1:fdd22bb7aa52 62 * @param[in] *pScratch2 points to scratch buffer (of type q15_t) of size min(srcALen, srcBLen).
emilmont 1:fdd22bb7aa52 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].
emilmont 1:fdd22bb7aa52 64 *
emilmont 1:fdd22bb7aa52 65 * \par Restrictions
emilmont 1:fdd22bb7aa52 66 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
emilmont 2:da51fb522205 67 * In this case input, output, scratch1 and scratch2 buffers should be aligned by 32-bit
emilmont 1:fdd22bb7aa52 68 *
emilmont 1:fdd22bb7aa52 69 *
emilmont 1:fdd22bb7aa52 70 *
emilmont 1:fdd22bb7aa52 71 */
emilmont 1:fdd22bb7aa52 72
emilmont 1:fdd22bb7aa52 73
emilmont 1:fdd22bb7aa52 74 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 75
emilmont 1:fdd22bb7aa52 76 arm_status arm_conv_partial_opt_q7(
emilmont 1:fdd22bb7aa52 77 q7_t * pSrcA,
emilmont 1:fdd22bb7aa52 78 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 79 q7_t * pSrcB,
emilmont 1:fdd22bb7aa52 80 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 81 q7_t * pDst,
emilmont 1:fdd22bb7aa52 82 uint32_t firstIndex,
emilmont 1:fdd22bb7aa52 83 uint32_t numPoints,
emilmont 1:fdd22bb7aa52 84 q15_t * pScratch1,
emilmont 1:fdd22bb7aa52 85 q15_t * pScratch2)
emilmont 1:fdd22bb7aa52 86 {
emilmont 1:fdd22bb7aa52 87
emilmont 1:fdd22bb7aa52 88 q15_t *pScr2, *pScr1; /* Intermediate pointers for scratch pointers */
emilmont 1:fdd22bb7aa52 89 q15_t x4; /* Temporary input variable */
emilmont 1:fdd22bb7aa52 90 q7_t *pIn1, *pIn2; /* inputA and inputB pointer */
emilmont 1:fdd22bb7aa52 91 uint32_t j, k, blkCnt, tapCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 92 q7_t *px; /* Temporary input1 pointer */
emilmont 1:fdd22bb7aa52 93 q15_t *py; /* Temporary input2 pointer */
emilmont 1:fdd22bb7aa52 94 q31_t acc0, acc1, acc2, acc3; /* Accumulator */
emilmont 1:fdd22bb7aa52 95 q31_t x1, x2, x3, y1; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 96 arm_status status;
emilmont 1:fdd22bb7aa52 97 q7_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 98 q7_t out0, out1, out2, out3; /* temporary variables */
emilmont 1:fdd22bb7aa52 99
emilmont 1:fdd22bb7aa52 100 /* Check for range of output samples to be calculated */
emilmont 1:fdd22bb7aa52 101 if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
emilmont 1:fdd22bb7aa52 102 {
emilmont 1:fdd22bb7aa52 103 /* Set status as ARM_MATH_ARGUMENT_ERROR */
emilmont 1:fdd22bb7aa52 104 status = ARM_MATH_ARGUMENT_ERROR;
emilmont 1:fdd22bb7aa52 105 }
emilmont 1:fdd22bb7aa52 106 else
emilmont 1:fdd22bb7aa52 107 {
emilmont 1:fdd22bb7aa52 108
emilmont 1:fdd22bb7aa52 109 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 110 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 111 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 112 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 113 {
emilmont 1:fdd22bb7aa52 114 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 115 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 116
emilmont 1:fdd22bb7aa52 117 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 118 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 119 }
emilmont 1:fdd22bb7aa52 120 else
emilmont 1:fdd22bb7aa52 121 {
emilmont 1:fdd22bb7aa52 122 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 123 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 124
emilmont 1:fdd22bb7aa52 125 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 126 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 127
emilmont 1:fdd22bb7aa52 128 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 129 j = srcBLen;
emilmont 1:fdd22bb7aa52 130 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 131 srcALen = j;
emilmont 1:fdd22bb7aa52 132 }
emilmont 1:fdd22bb7aa52 133
emilmont 1:fdd22bb7aa52 134 /* pointer to take end of scratch2 buffer */
emilmont 1:fdd22bb7aa52 135 pScr2 = pScratch2;
emilmont 1:fdd22bb7aa52 136
emilmont 1:fdd22bb7aa52 137 /* points to smaller length sequence */
emilmont 1:fdd22bb7aa52 138 px = pIn2 + srcBLen - 1;
emilmont 1:fdd22bb7aa52 139
emilmont 1:fdd22bb7aa52 140 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 141 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 142
emilmont 1:fdd22bb7aa52 143 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 144 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 145 while(k > 0u)
emilmont 1:fdd22bb7aa52 146 {
emilmont 1:fdd22bb7aa52 147 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 148 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 149 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 150 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 151 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 152 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 153 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 154 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 155 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 156
emilmont 1:fdd22bb7aa52 157 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 158 k--;
emilmont 1:fdd22bb7aa52 159 }
emilmont 1:fdd22bb7aa52 160
emilmont 1:fdd22bb7aa52 161 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 162 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 163 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 while(k > 0u)
emilmont 1:fdd22bb7aa52 166 {
emilmont 1:fdd22bb7aa52 167 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 168 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 169 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 172 k--;
emilmont 1:fdd22bb7aa52 173 }
emilmont 1:fdd22bb7aa52 174
emilmont 1:fdd22bb7aa52 175 /* Initialze temporary scratch pointer */
emilmont 1:fdd22bb7aa52 176 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 177
emilmont 1:fdd22bb7aa52 178 /* Fill (srcBLen - 1u) zeros in scratch buffer */
emilmont 1:fdd22bb7aa52 179 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 180
emilmont 1:fdd22bb7aa52 181 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 182 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 183
emilmont 1:fdd22bb7aa52 184 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 185 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 186 k = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 187
emilmont 1:fdd22bb7aa52 188 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 189 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 190 while(k > 0u)
emilmont 1:fdd22bb7aa52 191 {
emilmont 1:fdd22bb7aa52 192 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 193 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 194 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 195 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 196 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 197 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 198 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 199 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 200 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 201
emilmont 1:fdd22bb7aa52 202 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 203 k--;
emilmont 1:fdd22bb7aa52 204 }
emilmont 1:fdd22bb7aa52 205
emilmont 1:fdd22bb7aa52 206 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 207 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 208 k = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 209
emilmont 1:fdd22bb7aa52 210 while(k > 0u)
emilmont 1:fdd22bb7aa52 211 {
emilmont 1:fdd22bb7aa52 212 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 213 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 214 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 215
emilmont 1:fdd22bb7aa52 216 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 217 k--;
emilmont 1:fdd22bb7aa52 218 }
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emilmont 1:fdd22bb7aa52 221 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 222
emilmont 1:fdd22bb7aa52 223 /* Update pointer */
emilmont 1:fdd22bb7aa52 224 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 225
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* Temporary pointer for scratch2 */
emilmont 1:fdd22bb7aa52 228 py = pScratch2;
emilmont 1:fdd22bb7aa52 229
emilmont 1:fdd22bb7aa52 230 /* Initialization of pIn2 pointer */
emilmont 1:fdd22bb7aa52 231 pIn2 = (q7_t *) py;
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 pScr2 = py;
emilmont 1:fdd22bb7aa52 234
emilmont 1:fdd22bb7aa52 235 pOut = pDst + firstIndex;
emilmont 1:fdd22bb7aa52 236
emilmont 1:fdd22bb7aa52 237 pScratch1 += firstIndex;
emilmont 1:fdd22bb7aa52 238
emilmont 1:fdd22bb7aa52 239 /* Actual convolution process starts here */
emilmont 1:fdd22bb7aa52 240 blkCnt = (numPoints) >> 2;
emilmont 1:fdd22bb7aa52 241
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 244 {
emilmont 1:fdd22bb7aa52 245 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 246 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 247
emilmont 1:fdd22bb7aa52 248 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 249 acc0 = 0;
emilmont 1:fdd22bb7aa52 250 acc1 = 0;
emilmont 1:fdd22bb7aa52 251 acc2 = 0;
emilmont 1:fdd22bb7aa52 252 acc3 = 0;
emilmont 1:fdd22bb7aa52 253
emilmont 1:fdd22bb7aa52 254 /* Read two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 255 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 256
emilmont 1:fdd22bb7aa52 257 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 258 x2 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 259
emilmont 1:fdd22bb7aa52 260 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 261
emilmont 1:fdd22bb7aa52 262 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 263 {
emilmont 1:fdd22bb7aa52 264
emilmont 1:fdd22bb7aa52 265 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 266 y1 = _SIMD32_OFFSET(pScr2);
emilmont 1:fdd22bb7aa52 267
emilmont 1:fdd22bb7aa52 268 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 269 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 270 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 271
emilmont 1:fdd22bb7aa52 272 /* pack input data */
emilmont 1:fdd22bb7aa52 273 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 274 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 275 #else
emilmont 1:fdd22bb7aa52 276 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 277 #endif
emilmont 1:fdd22bb7aa52 278
emilmont 1:fdd22bb7aa52 279 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 280 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 281
emilmont 1:fdd22bb7aa52 282 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 283 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 284
emilmont 1:fdd22bb7aa52 285 /* pack input data */
emilmont 1:fdd22bb7aa52 286 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 287 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 288 #else
emilmont 1:fdd22bb7aa52 289 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 290 #endif
emilmont 1:fdd22bb7aa52 291
emilmont 1:fdd22bb7aa52 292 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 293
emilmont 1:fdd22bb7aa52 294 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 295 y1 = _SIMD32_OFFSET(pScr2 + 2u);
emilmont 1:fdd22bb7aa52 296
emilmont 1:fdd22bb7aa52 297 acc0 = __SMLAD(x2, y1, acc0);
emilmont 1:fdd22bb7aa52 298
emilmont 1:fdd22bb7aa52 299 acc2 = __SMLAD(x1, y1, acc2);
emilmont 1:fdd22bb7aa52 300
emilmont 1:fdd22bb7aa52 301 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 302
emilmont 1:fdd22bb7aa52 303 x2 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 306 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 307 #else
emilmont 1:fdd22bb7aa52 308 x3 = __PKHBT(x1, x2, 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 pScr2 += 4u;
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315
emilmont 1:fdd22bb7aa52 316 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 317 tapCnt--;
emilmont 1:fdd22bb7aa52 318 }
emilmont 1:fdd22bb7aa52 319
emilmont 1:fdd22bb7aa52 320
emilmont 1:fdd22bb7aa52 321
emilmont 1:fdd22bb7aa52 322 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 323 pScr1 -= 4u;
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325
emilmont 1:fdd22bb7aa52 326 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 327 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 330 {
emilmont 1:fdd22bb7aa52 331
emilmont 1:fdd22bb7aa52 332 /* accumlate the results */
emilmont 1:fdd22bb7aa52 333 acc0 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 334 acc1 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 335 acc2 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 336 acc3 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 pScr1 -= 3u;
emilmont 1:fdd22bb7aa52 339
emilmont 1:fdd22bb7aa52 340 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 341 tapCnt--;
emilmont 1:fdd22bb7aa52 342 }
emilmont 1:fdd22bb7aa52 343
emilmont 1:fdd22bb7aa52 344 blkCnt--;
emilmont 1:fdd22bb7aa52 345
emilmont 1:fdd22bb7aa52 346 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 347 out0 = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 348 out1 = (q7_t) (__SSAT(acc1 >> 7u, 8));
emilmont 1:fdd22bb7aa52 349 out2 = (q7_t) (__SSAT(acc2 >> 7u, 8));
emilmont 1:fdd22bb7aa52 350 out3 = (q7_t) (__SSAT(acc3 >> 7u, 8));
emilmont 1:fdd22bb7aa52 351
emilmont 1:fdd22bb7aa52 352 *__SIMD32(pOut)++ = __PACKq7(out0, out1, out2, out3);
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 355 pScr2 = py;
emilmont 1:fdd22bb7aa52 356
emilmont 1:fdd22bb7aa52 357 pScratch1 += 4u;
emilmont 1:fdd22bb7aa52 358
emilmont 1:fdd22bb7aa52 359 }
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 blkCnt = (numPoints) & 0x3;
emilmont 1:fdd22bb7aa52 362
emilmont 1:fdd22bb7aa52 363 /* Calculate convolution for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 364 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 365 {
emilmont 1:fdd22bb7aa52 366 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 367 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 368
emilmont 1:fdd22bb7aa52 369 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 370 acc0 = 0;
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 373
emilmont 1:fdd22bb7aa52 374 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 375 {
emilmont 1:fdd22bb7aa52 376
emilmont 1:fdd22bb7aa52 377 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 378 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 379
emilmont 1:fdd22bb7aa52 380 /* Read two samples from smaller buffer */
emilmont 1:fdd22bb7aa52 381 y1 = *__SIMD32(pScr2)++;
emilmont 1:fdd22bb7aa52 382
emilmont 1:fdd22bb7aa52 383 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 384
emilmont 1:fdd22bb7aa52 385 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 386 tapCnt--;
emilmont 1:fdd22bb7aa52 387 }
emilmont 1:fdd22bb7aa52 388
emilmont 1:fdd22bb7aa52 389 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 390
emilmont 1:fdd22bb7aa52 391 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 392 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 393 {
emilmont 1:fdd22bb7aa52 394
emilmont 1:fdd22bb7aa52 395 /* accumlate the results */
emilmont 1:fdd22bb7aa52 396 acc0 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 399 tapCnt--;
emilmont 1:fdd22bb7aa52 400 }
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402 blkCnt--;
emilmont 1:fdd22bb7aa52 403
emilmont 1:fdd22bb7aa52 404 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 405 *pOut++ = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 408 pScr2 = py;
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 pScratch1 += 1u;
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412 }
emilmont 1:fdd22bb7aa52 413
emilmont 1:fdd22bb7aa52 414 /* set status as ARM_MATH_SUCCESS */
emilmont 1:fdd22bb7aa52 415 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 416
emilmont 1:fdd22bb7aa52 417
emilmont 1:fdd22bb7aa52 418 }
emilmont 1:fdd22bb7aa52 419
emilmont 1:fdd22bb7aa52 420 return (status);
emilmont 1:fdd22bb7aa52 421
emilmont 1:fdd22bb7aa52 422 }
emilmont 1:fdd22bb7aa52 423
emilmont 1:fdd22bb7aa52 424 #else
emilmont 1:fdd22bb7aa52 425
emilmont 1:fdd22bb7aa52 426 arm_status arm_conv_partial_opt_q7(
emilmont 1:fdd22bb7aa52 427 q7_t * pSrcA,
emilmont 1:fdd22bb7aa52 428 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 429 q7_t * pSrcB,
emilmont 1:fdd22bb7aa52 430 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 431 q7_t * pDst,
emilmont 1:fdd22bb7aa52 432 uint32_t firstIndex,
emilmont 1:fdd22bb7aa52 433 uint32_t numPoints,
emilmont 1:fdd22bb7aa52 434 q15_t * pScratch1,
emilmont 1:fdd22bb7aa52 435 q15_t * pScratch2)
emilmont 1:fdd22bb7aa52 436 {
emilmont 1:fdd22bb7aa52 437
emilmont 1:fdd22bb7aa52 438 q15_t *pScr2, *pScr1; /* Intermediate pointers for scratch pointers */
emilmont 1:fdd22bb7aa52 439 q15_t x4; /* Temporary input variable */
emilmont 1:fdd22bb7aa52 440 q7_t *pIn1, *pIn2; /* inputA and inputB pointer */
emilmont 1:fdd22bb7aa52 441 uint32_t j, k, blkCnt, tapCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 442 q7_t *px; /* Temporary input1 pointer */
emilmont 1:fdd22bb7aa52 443 q15_t *py; /* Temporary input2 pointer */
emilmont 1:fdd22bb7aa52 444 q31_t acc0, acc1, acc2, acc3; /* Accumulator */
emilmont 1:fdd22bb7aa52 445 arm_status status;
emilmont 1:fdd22bb7aa52 446 q7_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 447 q15_t x10, x11, x20, x21; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 448 q15_t y10, y11; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 449
emilmont 1:fdd22bb7aa52 450 /* Check for range of output samples to be calculated */
emilmont 1:fdd22bb7aa52 451 if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
emilmont 1:fdd22bb7aa52 452 {
emilmont 1:fdd22bb7aa52 453 /* Set status as ARM_MATH_ARGUMENT_ERROR */
emilmont 1:fdd22bb7aa52 454 status = ARM_MATH_ARGUMENT_ERROR;
emilmont 1:fdd22bb7aa52 455 }
emilmont 1:fdd22bb7aa52 456 else
emilmont 1:fdd22bb7aa52 457 {
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 460 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 461 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 462 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 463 {
emilmont 1:fdd22bb7aa52 464 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 465 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 466
emilmont 1:fdd22bb7aa52 467 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 468 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 469 }
emilmont 1:fdd22bb7aa52 470 else
emilmont 1:fdd22bb7aa52 471 {
emilmont 1:fdd22bb7aa52 472 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 473 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 474
emilmont 1:fdd22bb7aa52 475 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 476 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 477
emilmont 1:fdd22bb7aa52 478 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 479 j = srcBLen;
emilmont 1:fdd22bb7aa52 480 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 481 srcALen = j;
emilmont 1:fdd22bb7aa52 482 }
emilmont 1:fdd22bb7aa52 483
emilmont 1:fdd22bb7aa52 484 /* pointer to take end of scratch2 buffer */
emilmont 1:fdd22bb7aa52 485 pScr2 = pScratch2;
emilmont 1:fdd22bb7aa52 486
emilmont 1:fdd22bb7aa52 487 /* points to smaller length sequence */
emilmont 1:fdd22bb7aa52 488 px = pIn2 + srcBLen - 1;
emilmont 1:fdd22bb7aa52 489
emilmont 1:fdd22bb7aa52 490 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 491 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 492
emilmont 1:fdd22bb7aa52 493 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 494 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 495 while(k > 0u)
emilmont 1:fdd22bb7aa52 496 {
emilmont 1:fdd22bb7aa52 497 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 498 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 499 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 500 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 501 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 502 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 503 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 504 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 505 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 506
emilmont 1:fdd22bb7aa52 507 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 508 k--;
emilmont 1:fdd22bb7aa52 509 }
emilmont 1:fdd22bb7aa52 510
emilmont 1:fdd22bb7aa52 511 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 512 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 513 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 514
emilmont 1:fdd22bb7aa52 515 while(k > 0u)
emilmont 1:fdd22bb7aa52 516 {
emilmont 1:fdd22bb7aa52 517 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 518 x4 = (q15_t) * px--;
emilmont 1:fdd22bb7aa52 519 *pScr2++ = x4;
emilmont 1:fdd22bb7aa52 520
emilmont 1:fdd22bb7aa52 521 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 522 k--;
emilmont 1:fdd22bb7aa52 523 }
emilmont 1:fdd22bb7aa52 524
emilmont 1:fdd22bb7aa52 525 /* Initialze temporary scratch pointer */
emilmont 1:fdd22bb7aa52 526 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 527
emilmont 1:fdd22bb7aa52 528 /* Fill (srcBLen - 1u) zeros in scratch buffer */
emilmont 1:fdd22bb7aa52 529 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 530
emilmont 1:fdd22bb7aa52 531 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 532 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 533
emilmont 1:fdd22bb7aa52 534 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 535 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 536 k = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 537
emilmont 1:fdd22bb7aa52 538 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 539 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 540 while(k > 0u)
emilmont 1:fdd22bb7aa52 541 {
emilmont 1:fdd22bb7aa52 542 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 543 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 544 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 545 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 546 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 547 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 548 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 549 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 550 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 551
emilmont 1:fdd22bb7aa52 552 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 553 k--;
emilmont 1:fdd22bb7aa52 554 }
emilmont 1:fdd22bb7aa52 555
emilmont 1:fdd22bb7aa52 556 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 557 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 558 k = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 559
emilmont 1:fdd22bb7aa52 560 while(k > 0u)
emilmont 1:fdd22bb7aa52 561 {
emilmont 1:fdd22bb7aa52 562 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 563 x4 = (q15_t) * pIn1++;
emilmont 1:fdd22bb7aa52 564 *pScr1++ = x4;
emilmont 1:fdd22bb7aa52 565
emilmont 1:fdd22bb7aa52 566 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 567 k--;
emilmont 1:fdd22bb7aa52 568 }
emilmont 1:fdd22bb7aa52 569
emilmont 1:fdd22bb7aa52 570 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 571 k = (srcBLen - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 572
emilmont 1:fdd22bb7aa52 573 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 574 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 575 while(k > 0u)
emilmont 1:fdd22bb7aa52 576 {
emilmont 1:fdd22bb7aa52 577 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 578 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 579 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 580 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 581 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 582
emilmont 1:fdd22bb7aa52 583 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 584 k--;
emilmont 1:fdd22bb7aa52 585 }
emilmont 1:fdd22bb7aa52 586
emilmont 1:fdd22bb7aa52 587 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 588 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 589 k = (srcBLen - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 590
emilmont 1:fdd22bb7aa52 591 while(k > 0u)
emilmont 1:fdd22bb7aa52 592 {
emilmont 1:fdd22bb7aa52 593 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 594 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 595
emilmont 1:fdd22bb7aa52 596 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 597 k--;
emilmont 1:fdd22bb7aa52 598 }
emilmont 1:fdd22bb7aa52 599
emilmont 1:fdd22bb7aa52 600
emilmont 1:fdd22bb7aa52 601 /* Temporary pointer for scratch2 */
emilmont 1:fdd22bb7aa52 602 py = pScratch2;
emilmont 1:fdd22bb7aa52 603
emilmont 1:fdd22bb7aa52 604 /* Initialization of pIn2 pointer */
emilmont 1:fdd22bb7aa52 605 pIn2 = (q7_t *) py;
emilmont 1:fdd22bb7aa52 606
emilmont 1:fdd22bb7aa52 607 pScr2 = py;
emilmont 1:fdd22bb7aa52 608
emilmont 1:fdd22bb7aa52 609 pOut = pDst + firstIndex;
emilmont 1:fdd22bb7aa52 610
emilmont 1:fdd22bb7aa52 611 pScratch1 += firstIndex;
emilmont 1:fdd22bb7aa52 612
emilmont 1:fdd22bb7aa52 613 /* Actual convolution process starts here */
emilmont 1:fdd22bb7aa52 614 blkCnt = (numPoints) >> 2;
emilmont 1:fdd22bb7aa52 615
emilmont 1:fdd22bb7aa52 616
emilmont 1:fdd22bb7aa52 617 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 618 {
emilmont 1:fdd22bb7aa52 619 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 620 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 621
emilmont 1:fdd22bb7aa52 622 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 623 acc0 = 0;
emilmont 1:fdd22bb7aa52 624 acc1 = 0;
emilmont 1:fdd22bb7aa52 625 acc2 = 0;
emilmont 1:fdd22bb7aa52 626 acc3 = 0;
emilmont 1:fdd22bb7aa52 627
emilmont 1:fdd22bb7aa52 628 /* Read two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 629 x10 = *pScr1++;
emilmont 1:fdd22bb7aa52 630 x11 = *pScr1++;
emilmont 1:fdd22bb7aa52 631
emilmont 1:fdd22bb7aa52 632 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 633 x20 = *pScr1++;
emilmont 1:fdd22bb7aa52 634 x21 = *pScr1++;
emilmont 1:fdd22bb7aa52 635
emilmont 1:fdd22bb7aa52 636 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 637
emilmont 1:fdd22bb7aa52 638 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 639 {
emilmont 1:fdd22bb7aa52 640
emilmont 1:fdd22bb7aa52 641 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 642 y10 = *pScr2;
emilmont 1:fdd22bb7aa52 643 y11 = *(pScr2 + 1u);
emilmont 1:fdd22bb7aa52 644
emilmont 1:fdd22bb7aa52 645 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 646 acc0 += (q31_t) x10 *y10;
emilmont 1:fdd22bb7aa52 647 acc0 += (q31_t) x11 *y11;
emilmont 1:fdd22bb7aa52 648 acc2 += (q31_t) x20 *y10;
emilmont 1:fdd22bb7aa52 649 acc2 += (q31_t) x21 *y11;
emilmont 1:fdd22bb7aa52 650
emilmont 1:fdd22bb7aa52 651
emilmont 1:fdd22bb7aa52 652 acc1 += (q31_t) x11 *y10;
emilmont 1:fdd22bb7aa52 653 acc1 += (q31_t) x20 *y11;
emilmont 1:fdd22bb7aa52 654
emilmont 1:fdd22bb7aa52 655 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 656 x10 = *pScr1;
emilmont 1:fdd22bb7aa52 657 x11 = *(pScr1 + 1u);
emilmont 1:fdd22bb7aa52 658
emilmont 1:fdd22bb7aa52 659 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 660 acc3 += (q31_t) x21 *y10;
emilmont 1:fdd22bb7aa52 661 acc3 += (q31_t) x10 *y11;
emilmont 1:fdd22bb7aa52 662
emilmont 1:fdd22bb7aa52 663 /* Read next two samples from scratch2 buffer */
emilmont 1:fdd22bb7aa52 664 y10 = *(pScr2 + 2u);
emilmont 1:fdd22bb7aa52 665 y11 = *(pScr2 + 3u);
emilmont 1:fdd22bb7aa52 666
emilmont 1:fdd22bb7aa52 667 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 668 acc0 += (q31_t) x20 *y10;
emilmont 1:fdd22bb7aa52 669 acc0 += (q31_t) x21 *y11;
emilmont 1:fdd22bb7aa52 670 acc2 += (q31_t) x10 *y10;
emilmont 1:fdd22bb7aa52 671 acc2 += (q31_t) x11 *y11;
emilmont 1:fdd22bb7aa52 672 acc1 += (q31_t) x21 *y10;
emilmont 1:fdd22bb7aa52 673 acc1 += (q31_t) x10 *y11;
emilmont 1:fdd22bb7aa52 674
emilmont 1:fdd22bb7aa52 675 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 676 x20 = *(pScr1 + 2);
emilmont 1:fdd22bb7aa52 677 x21 = *(pScr1 + 3);
emilmont 1:fdd22bb7aa52 678
emilmont 1:fdd22bb7aa52 679 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 680 acc3 += (q31_t) x11 *y10;
emilmont 1:fdd22bb7aa52 681 acc3 += (q31_t) x20 *y11;
emilmont 1:fdd22bb7aa52 682
emilmont 1:fdd22bb7aa52 683 /* update scratch pointers */
emilmont 1:fdd22bb7aa52 684
emilmont 1:fdd22bb7aa52 685 pScr1 += 4u;
emilmont 1:fdd22bb7aa52 686 pScr2 += 4u;
emilmont 1:fdd22bb7aa52 687
emilmont 1:fdd22bb7aa52 688 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 689 tapCnt--;
emilmont 1:fdd22bb7aa52 690 }
emilmont 1:fdd22bb7aa52 691
emilmont 1:fdd22bb7aa52 692
emilmont 1:fdd22bb7aa52 693
emilmont 1:fdd22bb7aa52 694 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 695 pScr1 -= 4u;
emilmont 1:fdd22bb7aa52 696
emilmont 1:fdd22bb7aa52 697
emilmont 1:fdd22bb7aa52 698 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 699 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 700
emilmont 1:fdd22bb7aa52 701 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 702 {
emilmont 1:fdd22bb7aa52 703
emilmont 1:fdd22bb7aa52 704 /* accumlate the results */
emilmont 1:fdd22bb7aa52 705 acc0 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 706 acc1 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 707 acc2 += (*pScr1++ * *pScr2);
emilmont 1:fdd22bb7aa52 708 acc3 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 709
emilmont 1:fdd22bb7aa52 710 pScr1 -= 3u;
emilmont 1:fdd22bb7aa52 711
emilmont 1:fdd22bb7aa52 712 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 713 tapCnt--;
emilmont 1:fdd22bb7aa52 714 }
emilmont 1:fdd22bb7aa52 715
emilmont 1:fdd22bb7aa52 716 blkCnt--;
emilmont 1:fdd22bb7aa52 717
emilmont 1:fdd22bb7aa52 718 /* Store the result in the accumulator in the destination buffer. */
mbed_official 5:3762170b6d4d 719 *pOut++ = (q7_t) (__SSAT(acc0 >> 7u, 8));
mbed_official 5:3762170b6d4d 720 *pOut++ = (q7_t) (__SSAT(acc1 >> 7u, 8));
mbed_official 5:3762170b6d4d 721 *pOut++ = (q7_t) (__SSAT(acc2 >> 7u, 8));
mbed_official 5:3762170b6d4d 722 *pOut++ = (q7_t) (__SSAT(acc3 >> 7u, 8));
emilmont 1:fdd22bb7aa52 723
emilmont 1:fdd22bb7aa52 724 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 725 pScr2 = py;
emilmont 1:fdd22bb7aa52 726
emilmont 1:fdd22bb7aa52 727 pScratch1 += 4u;
emilmont 1:fdd22bb7aa52 728
emilmont 1:fdd22bb7aa52 729 }
emilmont 1:fdd22bb7aa52 730
emilmont 1:fdd22bb7aa52 731 blkCnt = (numPoints) & 0x3;
emilmont 1:fdd22bb7aa52 732
emilmont 1:fdd22bb7aa52 733 /* Calculate convolution for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 734 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 735 {
emilmont 1:fdd22bb7aa52 736 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 737 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 738
emilmont 1:fdd22bb7aa52 739 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 740 acc0 = 0;
emilmont 1:fdd22bb7aa52 741
emilmont 1:fdd22bb7aa52 742 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 743
emilmont 1:fdd22bb7aa52 744 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 745 {
emilmont 1:fdd22bb7aa52 746
emilmont 1:fdd22bb7aa52 747 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 748 x10 = *pScr1++;
emilmont 1:fdd22bb7aa52 749 x11 = *pScr1++;
emilmont 1:fdd22bb7aa52 750
emilmont 1:fdd22bb7aa52 751 /* Read two samples from smaller buffer */
emilmont 1:fdd22bb7aa52 752 y10 = *pScr2++;
emilmont 1:fdd22bb7aa52 753 y11 = *pScr2++;
emilmont 1:fdd22bb7aa52 754
emilmont 1:fdd22bb7aa52 755 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 756 acc0 += (q31_t) x10 *y10;
emilmont 1:fdd22bb7aa52 757 acc0 += (q31_t) x11 *y11;
emilmont 1:fdd22bb7aa52 758
emilmont 1:fdd22bb7aa52 759 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 760 tapCnt--;
emilmont 1:fdd22bb7aa52 761 }
emilmont 1:fdd22bb7aa52 762
emilmont 1:fdd22bb7aa52 763 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 764
emilmont 1:fdd22bb7aa52 765 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 766 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 767 {
emilmont 1:fdd22bb7aa52 768
emilmont 1:fdd22bb7aa52 769 /* accumlate the results */
emilmont 1:fdd22bb7aa52 770 acc0 += (*pScr1++ * *pScr2++);
emilmont 1:fdd22bb7aa52 771
emilmont 1:fdd22bb7aa52 772 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 773 tapCnt--;
emilmont 1:fdd22bb7aa52 774 }
emilmont 1:fdd22bb7aa52 775
emilmont 1:fdd22bb7aa52 776 blkCnt--;
emilmont 1:fdd22bb7aa52 777
emilmont 1:fdd22bb7aa52 778 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 779 *pOut++ = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 780
emilmont 1:fdd22bb7aa52 781 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 782 pScr2 = py;
emilmont 1:fdd22bb7aa52 783
emilmont 1:fdd22bb7aa52 784 pScratch1 += 1u;
emilmont 1:fdd22bb7aa52 785
emilmont 1:fdd22bb7aa52 786 }
emilmont 1:fdd22bb7aa52 787
emilmont 1:fdd22bb7aa52 788 /* set status as ARM_MATH_SUCCESS */
emilmont 1:fdd22bb7aa52 789 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 790
emilmont 1:fdd22bb7aa52 791 }
emilmont 1:fdd22bb7aa52 792
emilmont 1:fdd22bb7aa52 793 return (status);
emilmont 1:fdd22bb7aa52 794
emilmont 1:fdd22bb7aa52 795 }
emilmont 1:fdd22bb7aa52 796
emilmont 2:da51fb522205 797 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 798
emilmont 1:fdd22bb7aa52 799
emilmont 1:fdd22bb7aa52 800
emilmont 1:fdd22bb7aa52 801 /**
emilmont 1:fdd22bb7aa52 802 * @} end of PartialConv group
emilmont 1:fdd22bb7aa52 803 */