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_fast_opt_q15.c@3:7a284390b0ce, 2013-11-08 (annotated)

Committer:
mbed_official
Date:
Fri Nov 08 13:45:10 2013 +0000
Revision:
3:7a284390b0ce
Parent:
2:da51fb522205
Child:
5:3762170b6d4d
Synchronized with git revision e69956aba2f68a2a26ac26b051f8d349deaa1ce8

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 3:7a284390b0ce 2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 3:7a284390b0ce 4 * $Date: 17. January 2013
mbed_official 3:7a284390b0ce 5 * $Revision: V1.4.1
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_conv_fast_opt_q15.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Fast Q15 Convolution.
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 Conv
emilmont 1:fdd22bb7aa52 49 * @{
emilmont 1:fdd22bb7aa52 50 */
emilmont 1:fdd22bb7aa52 51
emilmont 1:fdd22bb7aa52 52 /**
emilmont 1:fdd22bb7aa52 53 * @brief Convolution of Q15 sequences (fast version) for Cortex-M3 and Cortex-M4.
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. Length srcALen+srcBLen-1.
emilmont 1:fdd22bb7aa52 59 * @param[in] *pScratch1 points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emilmont 1:fdd22bb7aa52 60 * @param[in] *pScratch2 points to scratch buffer of size min(srcALen, srcBLen).
emilmont 1:fdd22bb7aa52 61 * @return none.
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 2:da51fb522205 65 * In this case input, output, scratch1 and scratch2 buffers should be aligned by 32-bit
emilmont 1:fdd22bb7aa52 66 *
emilmont 1:fdd22bb7aa52 67 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 68 *
emilmont 1:fdd22bb7aa52 69 * \par
emilmont 1:fdd22bb7aa52 70 * This fast version uses a 32-bit accumulator with 2.30 format.
emilmont 1:fdd22bb7aa52 71 * The accumulator maintains full precision of the intermediate multiplication results
emilmont 1:fdd22bb7aa52 72 * but provides only a single guard bit. There is no saturation on intermediate additions.
emilmont 1:fdd22bb7aa52 73 * Thus, if the accumulator overflows it wraps around and distorts the result.
emilmont 1:fdd22bb7aa52 74 * The input signals should be scaled down to avoid intermediate overflows.
emilmont 1:fdd22bb7aa52 75 * Scale down the inputs by log2(min(srcALen, srcBLen)) (log2 is read as log to the base 2) times to avoid overflows,
emilmont 1:fdd22bb7aa52 76 * as maximum of min(srcALen, srcBLen) number of additions are carried internally.
emilmont 1:fdd22bb7aa52 77 * The 2.30 accumulator is right shifted by 15 bits and then saturated to 1.15 format to yield the final result.
emilmont 1:fdd22bb7aa52 78 *
emilmont 1:fdd22bb7aa52 79 * \par
emilmont 1:fdd22bb7aa52 80 * See <code>arm_conv_q15()</code> for a slower implementation of this function which uses 64-bit accumulation to avoid wrap around distortion.
emilmont 1:fdd22bb7aa52 81 */
emilmont 1:fdd22bb7aa52 82
emilmont 1:fdd22bb7aa52 83 void arm_conv_fast_opt_q15(
emilmont 1:fdd22bb7aa52 84 q15_t * pSrcA,
emilmont 1:fdd22bb7aa52 85 uint32_t srcALen,
emilmont 1:fdd22bb7aa52 86 q15_t * pSrcB,
emilmont 1:fdd22bb7aa52 87 uint32_t srcBLen,
emilmont 1:fdd22bb7aa52 88 q15_t * pDst,
emilmont 1:fdd22bb7aa52 89 q15_t * pScratch1,
emilmont 1:fdd22bb7aa52 90 q15_t * pScratch2)
emilmont 1:fdd22bb7aa52 91 {
emilmont 1:fdd22bb7aa52 92 q31_t acc0, acc1, acc2, acc3; /* Accumulators */
emilmont 1:fdd22bb7aa52 93 q31_t x1, x2, x3; /* Temporary variables to hold state and coefficient values */
emilmont 1:fdd22bb7aa52 94 q31_t y1, y2; /* State variables */
emilmont 1:fdd22bb7aa52 95 q15_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 96 q15_t *pScr1 = pScratch1; /* Temporary pointer for scratch1 */
emilmont 1:fdd22bb7aa52 97 q15_t *pScr2 = pScratch2; /* Temporary pointer for scratch1 */
emilmont 1:fdd22bb7aa52 98 q15_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 99 q15_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 100 q15_t *px; /* Intermediate inputA pointer */
emilmont 1:fdd22bb7aa52 101 q15_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 102 uint32_t j, k, blkCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 103 uint32_t tapCnt; /* loop count */
emilmont 1:fdd22bb7aa52 104 #ifdef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 105
emilmont 1:fdd22bb7aa52 106 q15_t a, b;
emilmont 1:fdd22bb7aa52 107
emilmont 2:da51fb522205 108 #endif /* #ifdef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 109
emilmont 1:fdd22bb7aa52 110 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 111 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 112 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 113 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 114 {
emilmont 1:fdd22bb7aa52 115 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 116 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 117
emilmont 1:fdd22bb7aa52 118 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 119 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 120 }
emilmont 1:fdd22bb7aa52 121 else
emilmont 1:fdd22bb7aa52 122 {
emilmont 1:fdd22bb7aa52 123 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 124 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 125
emilmont 1:fdd22bb7aa52 126 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 127 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 128
emilmont 1:fdd22bb7aa52 129 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 130 j = srcBLen;
emilmont 1:fdd22bb7aa52 131 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 132 srcALen = j;
emilmont 1:fdd22bb7aa52 133 }
emilmont 1:fdd22bb7aa52 134
emilmont 1:fdd22bb7aa52 135 /* Pointer to take end of scratch2 buffer */
emilmont 1:fdd22bb7aa52 136 pScr2 = pScratch2 + srcBLen - 1;
emilmont 1:fdd22bb7aa52 137
emilmont 1:fdd22bb7aa52 138 /* points to smaller length sequence */
emilmont 1:fdd22bb7aa52 139 px = pIn2;
emilmont 1:fdd22bb7aa52 140
emilmont 1:fdd22bb7aa52 141 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 142 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 143
emilmont 1:fdd22bb7aa52 144 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 145 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 146
emilmont 1:fdd22bb7aa52 147 /* Copy smaller length input sequence in reverse order into second scratch buffer */
emilmont 1:fdd22bb7aa52 148 while(k > 0u)
emilmont 1:fdd22bb7aa52 149 {
emilmont 1:fdd22bb7aa52 150 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 151 *pScr2-- = *px++;
emilmont 1:fdd22bb7aa52 152 *pScr2-- = *px++;
emilmont 1:fdd22bb7aa52 153 *pScr2-- = *px++;
emilmont 1:fdd22bb7aa52 154 *pScr2-- = *px++;
emilmont 1:fdd22bb7aa52 155
emilmont 1:fdd22bb7aa52 156 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 157 k--;
emilmont 1:fdd22bb7aa52 158 }
emilmont 1:fdd22bb7aa52 159
emilmont 1:fdd22bb7aa52 160 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 161 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 162 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 163
emilmont 1:fdd22bb7aa52 164 while(k > 0u)
emilmont 1:fdd22bb7aa52 165 {
emilmont 1:fdd22bb7aa52 166 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 167 *pScr2-- = *px++;
emilmont 1:fdd22bb7aa52 168
emilmont 1:fdd22bb7aa52 169 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 170 k--;
emilmont 1:fdd22bb7aa52 171 }
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 /* Initialze temporary scratch pointer */
emilmont 1:fdd22bb7aa52 174 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 175
emilmont 1:fdd22bb7aa52 176 /* Assuming scratch1 buffer is aligned by 32-bit */
emilmont 1:fdd22bb7aa52 177 /* Fill (srcBLen - 1u) zeros in scratch1 buffer */
emilmont 1:fdd22bb7aa52 178 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180 /* Update temporary scratch pointer */
emilmont 1:fdd22bb7aa52 181 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
emilmont 1:fdd22bb7aa52 184
emilmont 1:fdd22bb7aa52 185 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 186
emilmont 1:fdd22bb7aa52 187 /* Copy (srcALen) samples in scratch buffer */
emilmont 1:fdd22bb7aa52 188 arm_copy_q15(pIn1, pScr1, srcALen);
emilmont 1:fdd22bb7aa52 189
emilmont 1:fdd22bb7aa52 190 /* Update pointers */
emilmont 1:fdd22bb7aa52 191 pScr1 += srcALen;
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 #else
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 196 k = srcALen >> 2u;
emilmont 1:fdd22bb7aa52 197
emilmont 1:fdd22bb7aa52 198 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 199 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 200 while(k > 0u)
emilmont 1:fdd22bb7aa52 201 {
emilmont 1:fdd22bb7aa52 202 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 203 *pScr1++ = *pIn1++;
emilmont 1:fdd22bb7aa52 204 *pScr1++ = *pIn1++;
emilmont 1:fdd22bb7aa52 205 *pScr1++ = *pIn1++;
emilmont 1:fdd22bb7aa52 206 *pScr1++ = *pIn1++;
emilmont 1:fdd22bb7aa52 207
emilmont 1:fdd22bb7aa52 208 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 209 k--;
emilmont 1:fdd22bb7aa52 210 }
emilmont 1:fdd22bb7aa52 211
emilmont 1:fdd22bb7aa52 212 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 213 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 214 k = srcALen % 0x4u;
emilmont 1:fdd22bb7aa52 215
emilmont 1:fdd22bb7aa52 216 while(k > 0u)
emilmont 1:fdd22bb7aa52 217 {
emilmont 1:fdd22bb7aa52 218 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 219 *pScr1++ = *pIn1++;
emilmont 1:fdd22bb7aa52 220
emilmont 1:fdd22bb7aa52 221 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 222 k--;
emilmont 1:fdd22bb7aa52 223 }
emilmont 1:fdd22bb7aa52 224
emilmont 2:da51fb522205 225 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227
emilmont 1:fdd22bb7aa52 228 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 229
emilmont 1:fdd22bb7aa52 230 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emilmont 1:fdd22bb7aa52 231 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 /* Update pointer */
emilmont 1:fdd22bb7aa52 234 pScr1 += (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 235
emilmont 1:fdd22bb7aa52 236 #else
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 /* Apply loop unrolling and do 4 Copies simultaneously. */
emilmont 1:fdd22bb7aa52 239 k = (srcBLen - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 240
emilmont 1:fdd22bb7aa52 241 /* First part of the processing with loop unrolling copies 4 data points at a time.
emilmont 1:fdd22bb7aa52 242 ** a second loop below copies for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 243 while(k > 0u)
emilmont 1:fdd22bb7aa52 244 {
emilmont 1:fdd22bb7aa52 245 /* copy second buffer in reversal manner */
emilmont 1:fdd22bb7aa52 246 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 247 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 248 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 249 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 252 k--;
emilmont 1:fdd22bb7aa52 253 }
emilmont 1:fdd22bb7aa52 254
emilmont 1:fdd22bb7aa52 255 /* If the count is not a multiple of 4, copy remaining samples here.
emilmont 1:fdd22bb7aa52 256 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 257 k = (srcBLen - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259 while(k > 0u)
emilmont 1:fdd22bb7aa52 260 {
emilmont 1:fdd22bb7aa52 261 /* copy second buffer in reversal manner for remaining samples */
emilmont 1:fdd22bb7aa52 262 *pScr1++ = 0;
emilmont 1:fdd22bb7aa52 263
emilmont 1:fdd22bb7aa52 264 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 265 k--;
emilmont 1:fdd22bb7aa52 266 }
emilmont 1:fdd22bb7aa52 267
emilmont 2:da51fb522205 268 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 269
emilmont 1:fdd22bb7aa52 270 /* Temporary pointer for scratch2 */
emilmont 1:fdd22bb7aa52 271 py = pScratch2;
emilmont 1:fdd22bb7aa52 272
emilmont 1:fdd22bb7aa52 273
emilmont 1:fdd22bb7aa52 274 /* Initialization of pIn2 pointer */
emilmont 1:fdd22bb7aa52 275 pIn2 = py;
emilmont 1:fdd22bb7aa52 276
emilmont 1:fdd22bb7aa52 277 /* First part of the processing with loop unrolling process 4 data points at a time.
emilmont 1:fdd22bb7aa52 278 ** a second loop below process for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 279
emilmont 1:fdd22bb7aa52 280 /* Actual convolution process starts here */
emilmont 1:fdd22bb7aa52 281 blkCnt = (srcALen + srcBLen - 1u) >> 2;
emilmont 1:fdd22bb7aa52 282
emilmont 1:fdd22bb7aa52 283 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 284 {
emilmont 1:fdd22bb7aa52 285 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 286 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 287
emilmont 1:fdd22bb7aa52 288 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 289 acc0 = 0;
emilmont 1:fdd22bb7aa52 290 acc1 = 0;
emilmont 1:fdd22bb7aa52 291 acc2 = 0;
emilmont 1:fdd22bb7aa52 292 acc3 = 0;
emilmont 1:fdd22bb7aa52 293
emilmont 1:fdd22bb7aa52 294 /* Read two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 295 x1 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 296
emilmont 1:fdd22bb7aa52 297 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 298 x2 = *__SIMD32(pScr1)++;
emilmont 1:fdd22bb7aa52 299
emilmont 1:fdd22bb7aa52 300 tapCnt = (srcBLen) >> 2u;
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 303 {
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 306
emilmont 1:fdd22bb7aa52 307 /* Read four samples from smaller buffer */
emilmont 1:fdd22bb7aa52 308 y1 = _SIMD32_OFFSET(pIn2);
emilmont 1:fdd22bb7aa52 309 y2 = _SIMD32_OFFSET(pIn2 + 2u);
emilmont 1:fdd22bb7aa52 310
emilmont 1:fdd22bb7aa52 311 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 312 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 313 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 /* pack input data */
emilmont 1:fdd22bb7aa52 316 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 317 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 318 #else
emilmont 1:fdd22bb7aa52 319 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 320 #endif
emilmont 1:fdd22bb7aa52 321
emilmont 1:fdd22bb7aa52 322 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 323 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 /* Read next two samples from scratch1 buffer */
emilmont 1:fdd22bb7aa52 326 x1 = _SIMD32_OFFSET(pScr1);
emilmont 1:fdd22bb7aa52 327
emilmont 1:fdd22bb7aa52 328 /* multiply and accumlate */
emilmont 1:fdd22bb7aa52 329 acc0 = __SMLAD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 330 acc2 = __SMLAD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 331
emilmont 1:fdd22bb7aa52 332 /* pack input data */
emilmont 1:fdd22bb7aa52 333 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 334 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 335 #else
emilmont 1:fdd22bb7aa52 336 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 337 #endif
emilmont 1:fdd22bb7aa52 338
emilmont 1:fdd22bb7aa52 339 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 340 acc1 = __SMLADX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 341
emilmont 1:fdd22bb7aa52 342 x2 = _SIMD32_OFFSET(pScr1 + 2u);
emilmont 1:fdd22bb7aa52 343
emilmont 1:fdd22bb7aa52 344 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 345 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 346 #else
emilmont 1:fdd22bb7aa52 347 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 348 #endif
emilmont 1:fdd22bb7aa52 349
emilmont 1:fdd22bb7aa52 350 acc3 = __SMLADX(x3, y2, acc3);
emilmont 1:fdd22bb7aa52 351
emilmont 2:da51fb522205 352 #else
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 /* Read four samples from smaller buffer */
emilmont 2:da51fb522205 355 a = *pIn2;
emilmont 2:da51fb522205 356 b = *(pIn2 + 1);
emilmont 1:fdd22bb7aa52 357
emilmont 1:fdd22bb7aa52 358 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 359 y1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 360 #else
emilmont 1:fdd22bb7aa52 361 y1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 362 #endif
emilmont 2:da51fb522205 363
emilmont 2:da51fb522205 364 a = *(pIn2 + 2);
emilmont 2:da51fb522205 365 b = *(pIn2 + 3);
emilmont 1:fdd22bb7aa52 366 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 367 y2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 368 #else
emilmont 1:fdd22bb7aa52 369 y2 = __PKHBT(b, a, 16);
emilmont 2:da51fb522205 370 #endif
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 acc0 = __SMLAD(x1, y1, acc0);
emilmont 1:fdd22bb7aa52 373
emilmont 1:fdd22bb7aa52 374 acc2 = __SMLAD(x2, y1, acc2);
emilmont 1:fdd22bb7aa52 375
emilmont 1:fdd22bb7aa52 376 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 377 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 378 #else
emilmont 1:fdd22bb7aa52 379 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 380 #endif
emilmont 1:fdd22bb7aa52 381
emilmont 1:fdd22bb7aa52 382 acc1 = __SMLADX(x3, y1, acc1);
emilmont 1:fdd22bb7aa52 383
emilmont 2:da51fb522205 384 a = *pScr1;
emilmont 2:da51fb522205 385 b = *(pScr1 + 1);
emilmont 1:fdd22bb7aa52 386
emilmont 1:fdd22bb7aa52 387 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 388 x1 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 389 #else
emilmont 1:fdd22bb7aa52 390 x1 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 391 #endif
emilmont 1:fdd22bb7aa52 392
emilmont 1:fdd22bb7aa52 393 acc0 = __SMLAD(x2, y2, acc0);
emilmont 1:fdd22bb7aa52 394
emilmont 1:fdd22bb7aa52 395 acc2 = __SMLAD(x1, y2, acc2);
emilmont 1:fdd22bb7aa52 396
emilmont 1:fdd22bb7aa52 397 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 398 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 399 #else
emilmont 1:fdd22bb7aa52 400 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 401 #endif
emilmont 1:fdd22bb7aa52 402
emilmont 1:fdd22bb7aa52 403 acc3 = __SMLADX(x3, y1, acc3);
emilmont 1:fdd22bb7aa52 404
emilmont 1:fdd22bb7aa52 405 acc1 = __SMLADX(x3, y2, acc1);
emilmont 1:fdd22bb7aa52 406
emilmont 2:da51fb522205 407 a = *(pScr1 + 2);
emilmont 2:da51fb522205 408 b = *(pScr1 + 3);
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 411 x2 = __PKHBT(a, b, 16);
emilmont 1:fdd22bb7aa52 412 #else
emilmont 1:fdd22bb7aa52 413 x2 = __PKHBT(b, a, 16);
emilmont 1:fdd22bb7aa52 414 #endif
emilmont 1:fdd22bb7aa52 415
emilmont 1:fdd22bb7aa52 416 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 417 x3 = __PKHBT(x2, x1, 0);
emilmont 1:fdd22bb7aa52 418 #else
emilmont 1:fdd22bb7aa52 419 x3 = __PKHBT(x1, x2, 0);
emilmont 1:fdd22bb7aa52 420 #endif
emilmont 1:fdd22bb7aa52 421
emilmont 1:fdd22bb7aa52 422 acc3 = __SMLADX(x3, y2, acc3);
emilmont 1:fdd22bb7aa52 423
emilmont 2:da51fb522205 424 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 425
emilmont 1:fdd22bb7aa52 426 /* update scratch pointers */
emilmont 1:fdd22bb7aa52 427 pIn2 += 4u;
emilmont 1:fdd22bb7aa52 428 pScr1 += 4u;
emilmont 1:fdd22bb7aa52 429
emilmont 1:fdd22bb7aa52 430
emilmont 1:fdd22bb7aa52 431 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 432 tapCnt--;
emilmont 1:fdd22bb7aa52 433 }
emilmont 1:fdd22bb7aa52 434
emilmont 1:fdd22bb7aa52 435 /* Update scratch pointer for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 436 pScr1 -= 4u;
emilmont 1:fdd22bb7aa52 437
emilmont 1:fdd22bb7aa52 438 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 439 tapCnt = (srcBLen) & 3u;
emilmont 1:fdd22bb7aa52 440
emilmont 1:fdd22bb7aa52 441 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 442 {
emilmont 1:fdd22bb7aa52 443
emilmont 1:fdd22bb7aa52 444 /* accumlate the results */
emilmont 1:fdd22bb7aa52 445 acc0 += (*pScr1++ * *pIn2);
emilmont 1:fdd22bb7aa52 446 acc1 += (*pScr1++ * *pIn2);
emilmont 1:fdd22bb7aa52 447 acc2 += (*pScr1++ * *pIn2);
emilmont 1:fdd22bb7aa52 448 acc3 += (*pScr1++ * *pIn2++);
emilmont 1:fdd22bb7aa52 449
emilmont 1:fdd22bb7aa52 450 pScr1 -= 3u;
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 453 tapCnt--;
emilmont 1:fdd22bb7aa52 454 }
emilmont 1:fdd22bb7aa52 455
emilmont 1:fdd22bb7aa52 456 blkCnt--;
emilmont 1:fdd22bb7aa52 457
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 /* Store the results in the accumulators in the destination buffer. */
emilmont 1:fdd22bb7aa52 460
emilmont 1:fdd22bb7aa52 461 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 462
emilmont 1:fdd22bb7aa52 463 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 464 __PKHBT(__SSAT((acc0 >> 15), 16), __SSAT((acc1 >> 15), 16), 16);
emilmont 1:fdd22bb7aa52 465
emilmont 1:fdd22bb7aa52 466 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 467 __PKHBT(__SSAT((acc2 >> 15), 16), __SSAT((acc3 >> 15), 16), 16);
emilmont 1:fdd22bb7aa52 468
emilmont 1:fdd22bb7aa52 469
emilmont 1:fdd22bb7aa52 470 #else
emilmont 1:fdd22bb7aa52 471
emilmont 1:fdd22bb7aa52 472 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 473 __PKHBT(__SSAT((acc1 >> 15), 16), __SSAT((acc0 >> 15), 16), 16);
emilmont 1:fdd22bb7aa52 474
emilmont 1:fdd22bb7aa52 475 *__SIMD32(pOut)++ =
emilmont 1:fdd22bb7aa52 476 __PKHBT(__SSAT((acc3 >> 15), 16), __SSAT((acc2 >> 15), 16), 16);
emilmont 1:fdd22bb7aa52 477
emilmont 1:fdd22bb7aa52 478
emilmont 1:fdd22bb7aa52 479
emilmont 1:fdd22bb7aa52 480 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 483 pIn2 = py;
emilmont 1:fdd22bb7aa52 484
emilmont 1:fdd22bb7aa52 485 pScratch1 += 4u;
emilmont 1:fdd22bb7aa52 486
emilmont 1:fdd22bb7aa52 487 }
emilmont 1:fdd22bb7aa52 488
emilmont 1:fdd22bb7aa52 489
emilmont 1:fdd22bb7aa52 490 blkCnt = (srcALen + srcBLen - 1u) & 0x3;
emilmont 1:fdd22bb7aa52 491
emilmont 1:fdd22bb7aa52 492 /* Calculate convolution for remaining samples of Bigger length sequence */
emilmont 1:fdd22bb7aa52 493 while(blkCnt > 0)
emilmont 1:fdd22bb7aa52 494 {
emilmont 1:fdd22bb7aa52 495 /* Initialze temporary scratch pointer as scratch1 */
emilmont 1:fdd22bb7aa52 496 pScr1 = pScratch1;
emilmont 1:fdd22bb7aa52 497
emilmont 1:fdd22bb7aa52 498 /* Clear Accumlators */
emilmont 1:fdd22bb7aa52 499 acc0 = 0;
emilmont 1:fdd22bb7aa52 500
emilmont 1:fdd22bb7aa52 501 tapCnt = (srcBLen) >> 1u;
emilmont 1:fdd22bb7aa52 502
emilmont 1:fdd22bb7aa52 503 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 504 {
emilmont 1:fdd22bb7aa52 505
emilmont 1:fdd22bb7aa52 506 acc0 += (*pScr1++ * *pIn2++);
emilmont 1:fdd22bb7aa52 507 acc0 += (*pScr1++ * *pIn2++);
emilmont 1:fdd22bb7aa52 508
emilmont 1:fdd22bb7aa52 509 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 510 tapCnt--;
emilmont 1:fdd22bb7aa52 511 }
emilmont 1:fdd22bb7aa52 512
emilmont 1:fdd22bb7aa52 513 tapCnt = (srcBLen) & 1u;
emilmont 1:fdd22bb7aa52 514
emilmont 1:fdd22bb7aa52 515 /* apply same above for remaining samples of smaller length sequence */
emilmont 1:fdd22bb7aa52 516 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 517 {
emilmont 1:fdd22bb7aa52 518
emilmont 1:fdd22bb7aa52 519 /* accumlate the results */
emilmont 1:fdd22bb7aa52 520 acc0 += (*pScr1++ * *pIn2++);
emilmont 1:fdd22bb7aa52 521
emilmont 1:fdd22bb7aa52 522 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 523 tapCnt--;
emilmont 1:fdd22bb7aa52 524 }
emilmont 1:fdd22bb7aa52 525
emilmont 1:fdd22bb7aa52 526 blkCnt--;
emilmont 1:fdd22bb7aa52 527
emilmont 1:fdd22bb7aa52 528 /* The result is in 2.30 format. Convert to 1.15 with saturation.
emilmont 1:fdd22bb7aa52 529 ** Then store the output in the destination buffer. */
emilmont 1:fdd22bb7aa52 530 *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
emilmont 1:fdd22bb7aa52 531
emilmont 1:fdd22bb7aa52 532 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 533 pIn2 = py;
emilmont 1:fdd22bb7aa52 534
emilmont 1:fdd22bb7aa52 535 pScratch1 += 1u;
emilmont 1:fdd22bb7aa52 536
emilmont 1:fdd22bb7aa52 537 }
emilmont 1:fdd22bb7aa52 538
emilmont 1:fdd22bb7aa52 539 }
emilmont 1:fdd22bb7aa52 540
emilmont 1:fdd22bb7aa52 541 /**
emilmont 1:fdd22bb7aa52 542 * @} end of Conv group
emilmont 1:fdd22bb7aa52 543 */