V4.0.1 of the ARM CMSIS DSP libraries. Note that arm_bitreversal2.s, arm_cfft_f32.c and arm_rfft_fast_f32.c had to be removed. arm_bitreversal2.s will not assemble with the online tools. So, the fast f32 FFT functions are not yet available. All the other FFT functions are available.

Dependents:   MPU9150_Example fir_f32 fir_f32 MPU9150_nucleo_noni2cdev ... more

Committer:
emh203
Date:
Mon Jul 28 15:03:15 2014 +0000
Revision:
0:3d9c67d97d6f
1st working commit.   Had to remove arm_bitreversal2.s     arm_cfft_f32.c and arm_rfft_fast_f32.c.    The .s will not assemble.      For now I removed these functions so we could at least have a library for the other functions.

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emh203 0:3d9c67d97d6f 1 /* ----------------------------------------------------------------------
emh203 0:3d9c67d97d6f 2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
emh203 0:3d9c67d97d6f 3 *
emh203 0:3d9c67d97d6f 4 * $Date: 12. March 2014
emh203 0:3d9c67d97d6f 5 * $Revision: V1.4.3
emh203 0:3d9c67d97d6f 6 *
emh203 0:3d9c67d97d6f 7 * Project: CMSIS DSP Library
emh203 0:3d9c67d97d6f 8 * Title: arm_conv_fast_opt_q15.c
emh203 0:3d9c67d97d6f 9 *
emh203 0:3d9c67d97d6f 10 * Description: Fast Q15 Convolution.
emh203 0:3d9c67d97d6f 11 *
emh203 0:3d9c67d97d6f 12 * Target Processor: Cortex-M4/Cortex-M3
emh203 0:3d9c67d97d6f 13 *
emh203 0:3d9c67d97d6f 14 * Redistribution and use in source and binary forms, with or without
emh203 0:3d9c67d97d6f 15 * modification, are permitted provided that the following conditions
emh203 0:3d9c67d97d6f 16 * are met:
emh203 0:3d9c67d97d6f 17 * - Redistributions of source code must retain the above copyright
emh203 0:3d9c67d97d6f 18 * notice, this list of conditions and the following disclaimer.
emh203 0:3d9c67d97d6f 19 * - Redistributions in binary form must reproduce the above copyright
emh203 0:3d9c67d97d6f 20 * notice, this list of conditions and the following disclaimer in
emh203 0:3d9c67d97d6f 21 * the documentation and/or other materials provided with the
emh203 0:3d9c67d97d6f 22 * distribution.
emh203 0:3d9c67d97d6f 23 * - Neither the name of ARM LIMITED nor the names of its contributors
emh203 0:3d9c67d97d6f 24 * may be used to endorse or promote products derived from this
emh203 0:3d9c67d97d6f 25 * software without specific prior written permission.
emh203 0:3d9c67d97d6f 26 *
emh203 0:3d9c67d97d6f 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
emh203 0:3d9c67d97d6f 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
emh203 0:3d9c67d97d6f 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
emh203 0:3d9c67d97d6f 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
emh203 0:3d9c67d97d6f 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
emh203 0:3d9c67d97d6f 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
emh203 0:3d9c67d97d6f 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
emh203 0:3d9c67d97d6f 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
emh203 0:3d9c67d97d6f 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
emh203 0:3d9c67d97d6f 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
emh203 0:3d9c67d97d6f 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
emh203 0:3d9c67d97d6f 38 * POSSIBILITY OF SUCH DAMAGE.
emh203 0:3d9c67d97d6f 39 * -------------------------------------------------------------------- */
emh203 0:3d9c67d97d6f 40
emh203 0:3d9c67d97d6f 41 #include "arm_math.h"
emh203 0:3d9c67d97d6f 42
emh203 0:3d9c67d97d6f 43 /**
emh203 0:3d9c67d97d6f 44 * @ingroup groupFilters
emh203 0:3d9c67d97d6f 45 */
emh203 0:3d9c67d97d6f 46
emh203 0:3d9c67d97d6f 47 /**
emh203 0:3d9c67d97d6f 48 * @addtogroup Conv
emh203 0:3d9c67d97d6f 49 * @{
emh203 0:3d9c67d97d6f 50 */
emh203 0:3d9c67d97d6f 51
emh203 0:3d9c67d97d6f 52 /**
emh203 0:3d9c67d97d6f 53 * @brief Convolution of Q15 sequences (fast version) for Cortex-M3 and Cortex-M4.
emh203 0:3d9c67d97d6f 54 * @param[in] *pSrcA points to the first input sequence.
emh203 0:3d9c67d97d6f 55 * @param[in] srcALen length of the first input sequence.
emh203 0:3d9c67d97d6f 56 * @param[in] *pSrcB points to the second input sequence.
emh203 0:3d9c67d97d6f 57 * @param[in] srcBLen length of the second input sequence.
emh203 0:3d9c67d97d6f 58 * @param[out] *pDst points to the location where the output result is written. Length srcALen+srcBLen-1.
emh203 0:3d9c67d97d6f 59 * @param[in] *pScratch1 points to scratch buffer of size max(srcALen, srcBLen) + 2*min(srcALen, srcBLen) - 2.
emh203 0:3d9c67d97d6f 60 * @param[in] *pScratch2 points to scratch buffer of size min(srcALen, srcBLen).
emh203 0:3d9c67d97d6f 61 * @return none.
emh203 0:3d9c67d97d6f 62 *
emh203 0:3d9c67d97d6f 63 * \par Restrictions
emh203 0:3d9c67d97d6f 64 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
emh203 0:3d9c67d97d6f 65 * In this case input, output, scratch1 and scratch2 buffers should be aligned by 32-bit
emh203 0:3d9c67d97d6f 66 *
emh203 0:3d9c67d97d6f 67 * <b>Scaling and Overflow Behavior:</b>
emh203 0:3d9c67d97d6f 68 *
emh203 0:3d9c67d97d6f 69 * \par
emh203 0:3d9c67d97d6f 70 * This fast version uses a 32-bit accumulator with 2.30 format.
emh203 0:3d9c67d97d6f 71 * The accumulator maintains full precision of the intermediate multiplication results
emh203 0:3d9c67d97d6f 72 * but provides only a single guard bit. There is no saturation on intermediate additions.
emh203 0:3d9c67d97d6f 73 * Thus, if the accumulator overflows it wraps around and distorts the result.
emh203 0:3d9c67d97d6f 74 * The input signals should be scaled down to avoid intermediate overflows.
emh203 0:3d9c67d97d6f 75 * Scale down the inputs by log2(min(srcALen, srcBLen)) (log2 is read as log to the base 2) times to avoid overflows,
emh203 0:3d9c67d97d6f 76 * as maximum of min(srcALen, srcBLen) number of additions are carried internally.
emh203 0:3d9c67d97d6f 77 * The 2.30 accumulator is right shifted by 15 bits and then saturated to 1.15 format to yield the final result.
emh203 0:3d9c67d97d6f 78 *
emh203 0:3d9c67d97d6f 79 * \par
emh203 0:3d9c67d97d6f 80 * See <code>arm_conv_q15()</code> for a slower implementation of this function which uses 64-bit accumulation to avoid wrap around distortion.
emh203 0:3d9c67d97d6f 81 */
emh203 0:3d9c67d97d6f 82
emh203 0:3d9c67d97d6f 83 void arm_conv_fast_opt_q15(
emh203 0:3d9c67d97d6f 84 q15_t * pSrcA,
emh203 0:3d9c67d97d6f 85 uint32_t srcALen,
emh203 0:3d9c67d97d6f 86 q15_t * pSrcB,
emh203 0:3d9c67d97d6f 87 uint32_t srcBLen,
emh203 0:3d9c67d97d6f 88 q15_t * pDst,
emh203 0:3d9c67d97d6f 89 q15_t * pScratch1,
emh203 0:3d9c67d97d6f 90 q15_t * pScratch2)
emh203 0:3d9c67d97d6f 91 {
emh203 0:3d9c67d97d6f 92 q31_t acc0, acc1, acc2, acc3; /* Accumulators */
emh203 0:3d9c67d97d6f 93 q31_t x1, x2, x3; /* Temporary variables to hold state and coefficient values */
emh203 0:3d9c67d97d6f 94 q31_t y1, y2; /* State variables */
emh203 0:3d9c67d97d6f 95 q15_t *pOut = pDst; /* output pointer */
emh203 0:3d9c67d97d6f 96 q15_t *pScr1 = pScratch1; /* Temporary pointer for scratch1 */
emh203 0:3d9c67d97d6f 97 q15_t *pScr2 = pScratch2; /* Temporary pointer for scratch1 */
emh203 0:3d9c67d97d6f 98 q15_t *pIn1; /* inputA pointer */
emh203 0:3d9c67d97d6f 99 q15_t *pIn2; /* inputB pointer */
emh203 0:3d9c67d97d6f 100 q15_t *px; /* Intermediate inputA pointer */
emh203 0:3d9c67d97d6f 101 q15_t *py; /* Intermediate inputB pointer */
emh203 0:3d9c67d97d6f 102 uint32_t j, k, blkCnt; /* loop counter */
emh203 0:3d9c67d97d6f 103 uint32_t tapCnt; /* loop count */
emh203 0:3d9c67d97d6f 104 #ifdef UNALIGNED_SUPPORT_DISABLE
emh203 0:3d9c67d97d6f 105
emh203 0:3d9c67d97d6f 106 q15_t a, b;
emh203 0:3d9c67d97d6f 107
emh203 0:3d9c67d97d6f 108 #endif /* #ifdef UNALIGNED_SUPPORT_DISABLE */
emh203 0:3d9c67d97d6f 109
emh203 0:3d9c67d97d6f 110 /* The algorithm implementation is based on the lengths of the inputs. */
emh203 0:3d9c67d97d6f 111 /* srcB is always made to slide across srcA. */
emh203 0:3d9c67d97d6f 112 /* So srcBLen is always considered as shorter or equal to srcALen */
emh203 0:3d9c67d97d6f 113 if(srcALen >= srcBLen)
emh203 0:3d9c67d97d6f 114 {
emh203 0:3d9c67d97d6f 115 /* Initialization of inputA pointer */
emh203 0:3d9c67d97d6f 116 pIn1 = pSrcA;
emh203 0:3d9c67d97d6f 117
emh203 0:3d9c67d97d6f 118 /* Initialization of inputB pointer */
emh203 0:3d9c67d97d6f 119 pIn2 = pSrcB;
emh203 0:3d9c67d97d6f 120 }
emh203 0:3d9c67d97d6f 121 else
emh203 0:3d9c67d97d6f 122 {
emh203 0:3d9c67d97d6f 123 /* Initialization of inputA pointer */
emh203 0:3d9c67d97d6f 124 pIn1 = pSrcB;
emh203 0:3d9c67d97d6f 125
emh203 0:3d9c67d97d6f 126 /* Initialization of inputB pointer */
emh203 0:3d9c67d97d6f 127 pIn2 = pSrcA;
emh203 0:3d9c67d97d6f 128
emh203 0:3d9c67d97d6f 129 /* srcBLen is always considered as shorter or equal to srcALen */
emh203 0:3d9c67d97d6f 130 j = srcBLen;
emh203 0:3d9c67d97d6f 131 srcBLen = srcALen;
emh203 0:3d9c67d97d6f 132 srcALen = j;
emh203 0:3d9c67d97d6f 133 }
emh203 0:3d9c67d97d6f 134
emh203 0:3d9c67d97d6f 135 /* Pointer to take end of scratch2 buffer */
emh203 0:3d9c67d97d6f 136 pScr2 = pScratch2 + srcBLen - 1;
emh203 0:3d9c67d97d6f 137
emh203 0:3d9c67d97d6f 138 /* points to smaller length sequence */
emh203 0:3d9c67d97d6f 139 px = pIn2;
emh203 0:3d9c67d97d6f 140
emh203 0:3d9c67d97d6f 141 /* Apply loop unrolling and do 4 Copies simultaneously. */
emh203 0:3d9c67d97d6f 142 k = srcBLen >> 2u;
emh203 0:3d9c67d97d6f 143
emh203 0:3d9c67d97d6f 144 /* First part of the processing with loop unrolling copies 4 data points at a time.
emh203 0:3d9c67d97d6f 145 ** a second loop below copies for the remaining 1 to 3 samples. */
emh203 0:3d9c67d97d6f 146
emh203 0:3d9c67d97d6f 147 /* Copy smaller length input sequence in reverse order into second scratch buffer */
emh203 0:3d9c67d97d6f 148 while(k > 0u)
emh203 0:3d9c67d97d6f 149 {
emh203 0:3d9c67d97d6f 150 /* copy second buffer in reversal manner */
emh203 0:3d9c67d97d6f 151 *pScr2-- = *px++;
emh203 0:3d9c67d97d6f 152 *pScr2-- = *px++;
emh203 0:3d9c67d97d6f 153 *pScr2-- = *px++;
emh203 0:3d9c67d97d6f 154 *pScr2-- = *px++;
emh203 0:3d9c67d97d6f 155
emh203 0:3d9c67d97d6f 156 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 157 k--;
emh203 0:3d9c67d97d6f 158 }
emh203 0:3d9c67d97d6f 159
emh203 0:3d9c67d97d6f 160 /* If the count is not a multiple of 4, copy remaining samples here.
emh203 0:3d9c67d97d6f 161 ** No loop unrolling is used. */
emh203 0:3d9c67d97d6f 162 k = srcBLen % 0x4u;
emh203 0:3d9c67d97d6f 163
emh203 0:3d9c67d97d6f 164 while(k > 0u)
emh203 0:3d9c67d97d6f 165 {
emh203 0:3d9c67d97d6f 166 /* copy second buffer in reversal manner for remaining samples */
emh203 0:3d9c67d97d6f 167 *pScr2-- = *px++;
emh203 0:3d9c67d97d6f 168
emh203 0:3d9c67d97d6f 169 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 170 k--;
emh203 0:3d9c67d97d6f 171 }
emh203 0:3d9c67d97d6f 172
emh203 0:3d9c67d97d6f 173 /* Initialze temporary scratch pointer */
emh203 0:3d9c67d97d6f 174 pScr1 = pScratch1;
emh203 0:3d9c67d97d6f 175
emh203 0:3d9c67d97d6f 176 /* Assuming scratch1 buffer is aligned by 32-bit */
emh203 0:3d9c67d97d6f 177 /* Fill (srcBLen - 1u) zeros in scratch1 buffer */
emh203 0:3d9c67d97d6f 178 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emh203 0:3d9c67d97d6f 179
emh203 0:3d9c67d97d6f 180 /* Update temporary scratch pointer */
emh203 0:3d9c67d97d6f 181 pScr1 += (srcBLen - 1u);
emh203 0:3d9c67d97d6f 182
emh203 0:3d9c67d97d6f 183 /* Copy bigger length sequence(srcALen) samples in scratch1 buffer */
emh203 0:3d9c67d97d6f 184
emh203 0:3d9c67d97d6f 185 #ifndef UNALIGNED_SUPPORT_DISABLE
emh203 0:3d9c67d97d6f 186
emh203 0:3d9c67d97d6f 187 /* Copy (srcALen) samples in scratch buffer */
emh203 0:3d9c67d97d6f 188 arm_copy_q15(pIn1, pScr1, srcALen);
emh203 0:3d9c67d97d6f 189
emh203 0:3d9c67d97d6f 190 /* Update pointers */
emh203 0:3d9c67d97d6f 191 pScr1 += srcALen;
emh203 0:3d9c67d97d6f 192
emh203 0:3d9c67d97d6f 193 #else
emh203 0:3d9c67d97d6f 194
emh203 0:3d9c67d97d6f 195 /* Apply loop unrolling and do 4 Copies simultaneously. */
emh203 0:3d9c67d97d6f 196 k = srcALen >> 2u;
emh203 0:3d9c67d97d6f 197
emh203 0:3d9c67d97d6f 198 /* First part of the processing with loop unrolling copies 4 data points at a time.
emh203 0:3d9c67d97d6f 199 ** a second loop below copies for the remaining 1 to 3 samples. */
emh203 0:3d9c67d97d6f 200 while(k > 0u)
emh203 0:3d9c67d97d6f 201 {
emh203 0:3d9c67d97d6f 202 /* copy second buffer in reversal manner */
emh203 0:3d9c67d97d6f 203 *pScr1++ = *pIn1++;
emh203 0:3d9c67d97d6f 204 *pScr1++ = *pIn1++;
emh203 0:3d9c67d97d6f 205 *pScr1++ = *pIn1++;
emh203 0:3d9c67d97d6f 206 *pScr1++ = *pIn1++;
emh203 0:3d9c67d97d6f 207
emh203 0:3d9c67d97d6f 208 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 209 k--;
emh203 0:3d9c67d97d6f 210 }
emh203 0:3d9c67d97d6f 211
emh203 0:3d9c67d97d6f 212 /* If the count is not a multiple of 4, copy remaining samples here.
emh203 0:3d9c67d97d6f 213 ** No loop unrolling is used. */
emh203 0:3d9c67d97d6f 214 k = srcALen % 0x4u;
emh203 0:3d9c67d97d6f 215
emh203 0:3d9c67d97d6f 216 while(k > 0u)
emh203 0:3d9c67d97d6f 217 {
emh203 0:3d9c67d97d6f 218 /* copy second buffer in reversal manner for remaining samples */
emh203 0:3d9c67d97d6f 219 *pScr1++ = *pIn1++;
emh203 0:3d9c67d97d6f 220
emh203 0:3d9c67d97d6f 221 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 222 k--;
emh203 0:3d9c67d97d6f 223 }
emh203 0:3d9c67d97d6f 224
emh203 0:3d9c67d97d6f 225 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emh203 0:3d9c67d97d6f 226
emh203 0:3d9c67d97d6f 227
emh203 0:3d9c67d97d6f 228 #ifndef UNALIGNED_SUPPORT_DISABLE
emh203 0:3d9c67d97d6f 229
emh203 0:3d9c67d97d6f 230 /* Fill (srcBLen - 1u) zeros at end of scratch buffer */
emh203 0:3d9c67d97d6f 231 arm_fill_q15(0, pScr1, (srcBLen - 1u));
emh203 0:3d9c67d97d6f 232
emh203 0:3d9c67d97d6f 233 /* Update pointer */
emh203 0:3d9c67d97d6f 234 pScr1 += (srcBLen - 1u);
emh203 0:3d9c67d97d6f 235
emh203 0:3d9c67d97d6f 236 #else
emh203 0:3d9c67d97d6f 237
emh203 0:3d9c67d97d6f 238 /* Apply loop unrolling and do 4 Copies simultaneously. */
emh203 0:3d9c67d97d6f 239 k = (srcBLen - 1u) >> 2u;
emh203 0:3d9c67d97d6f 240
emh203 0:3d9c67d97d6f 241 /* First part of the processing with loop unrolling copies 4 data points at a time.
emh203 0:3d9c67d97d6f 242 ** a second loop below copies for the remaining 1 to 3 samples. */
emh203 0:3d9c67d97d6f 243 while(k > 0u)
emh203 0:3d9c67d97d6f 244 {
emh203 0:3d9c67d97d6f 245 /* copy second buffer in reversal manner */
emh203 0:3d9c67d97d6f 246 *pScr1++ = 0;
emh203 0:3d9c67d97d6f 247 *pScr1++ = 0;
emh203 0:3d9c67d97d6f 248 *pScr1++ = 0;
emh203 0:3d9c67d97d6f 249 *pScr1++ = 0;
emh203 0:3d9c67d97d6f 250
emh203 0:3d9c67d97d6f 251 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 252 k--;
emh203 0:3d9c67d97d6f 253 }
emh203 0:3d9c67d97d6f 254
emh203 0:3d9c67d97d6f 255 /* If the count is not a multiple of 4, copy remaining samples here.
emh203 0:3d9c67d97d6f 256 ** No loop unrolling is used. */
emh203 0:3d9c67d97d6f 257 k = (srcBLen - 1u) % 0x4u;
emh203 0:3d9c67d97d6f 258
emh203 0:3d9c67d97d6f 259 while(k > 0u)
emh203 0:3d9c67d97d6f 260 {
emh203 0:3d9c67d97d6f 261 /* copy second buffer in reversal manner for remaining samples */
emh203 0:3d9c67d97d6f 262 *pScr1++ = 0;
emh203 0:3d9c67d97d6f 263
emh203 0:3d9c67d97d6f 264 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 265 k--;
emh203 0:3d9c67d97d6f 266 }
emh203 0:3d9c67d97d6f 267
emh203 0:3d9c67d97d6f 268 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emh203 0:3d9c67d97d6f 269
emh203 0:3d9c67d97d6f 270 /* Temporary pointer for scratch2 */
emh203 0:3d9c67d97d6f 271 py = pScratch2;
emh203 0:3d9c67d97d6f 272
emh203 0:3d9c67d97d6f 273
emh203 0:3d9c67d97d6f 274 /* Initialization of pIn2 pointer */
emh203 0:3d9c67d97d6f 275 pIn2 = py;
emh203 0:3d9c67d97d6f 276
emh203 0:3d9c67d97d6f 277 /* First part of the processing with loop unrolling process 4 data points at a time.
emh203 0:3d9c67d97d6f 278 ** a second loop below process for the remaining 1 to 3 samples. */
emh203 0:3d9c67d97d6f 279
emh203 0:3d9c67d97d6f 280 /* Actual convolution process starts here */
emh203 0:3d9c67d97d6f 281 blkCnt = (srcALen + srcBLen - 1u) >> 2;
emh203 0:3d9c67d97d6f 282
emh203 0:3d9c67d97d6f 283 while(blkCnt > 0)
emh203 0:3d9c67d97d6f 284 {
emh203 0:3d9c67d97d6f 285 /* Initialze temporary scratch pointer as scratch1 */
emh203 0:3d9c67d97d6f 286 pScr1 = pScratch1;
emh203 0:3d9c67d97d6f 287
emh203 0:3d9c67d97d6f 288 /* Clear Accumlators */
emh203 0:3d9c67d97d6f 289 acc0 = 0;
emh203 0:3d9c67d97d6f 290 acc1 = 0;
emh203 0:3d9c67d97d6f 291 acc2 = 0;
emh203 0:3d9c67d97d6f 292 acc3 = 0;
emh203 0:3d9c67d97d6f 293
emh203 0:3d9c67d97d6f 294 /* Read two samples from scratch1 buffer */
emh203 0:3d9c67d97d6f 295 x1 = *__SIMD32(pScr1)++;
emh203 0:3d9c67d97d6f 296
emh203 0:3d9c67d97d6f 297 /* Read next two samples from scratch1 buffer */
emh203 0:3d9c67d97d6f 298 x2 = *__SIMD32(pScr1)++;
emh203 0:3d9c67d97d6f 299
emh203 0:3d9c67d97d6f 300 tapCnt = (srcBLen) >> 2u;
emh203 0:3d9c67d97d6f 301
emh203 0:3d9c67d97d6f 302 while(tapCnt > 0u)
emh203 0:3d9c67d97d6f 303 {
emh203 0:3d9c67d97d6f 304
emh203 0:3d9c67d97d6f 305 #ifndef UNALIGNED_SUPPORT_DISABLE
emh203 0:3d9c67d97d6f 306
emh203 0:3d9c67d97d6f 307 /* Read four samples from smaller buffer */
emh203 0:3d9c67d97d6f 308 y1 = _SIMD32_OFFSET(pIn2);
emh203 0:3d9c67d97d6f 309 y2 = _SIMD32_OFFSET(pIn2 + 2u);
emh203 0:3d9c67d97d6f 310
emh203 0:3d9c67d97d6f 311 /* multiply and accumlate */
emh203 0:3d9c67d97d6f 312 acc0 = __SMLAD(x1, y1, acc0);
emh203 0:3d9c67d97d6f 313 acc2 = __SMLAD(x2, y1, acc2);
emh203 0:3d9c67d97d6f 314
emh203 0:3d9c67d97d6f 315 /* pack input data */
emh203 0:3d9c67d97d6f 316 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 317 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 318 #else
emh203 0:3d9c67d97d6f 319 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 320 #endif
emh203 0:3d9c67d97d6f 321
emh203 0:3d9c67d97d6f 322 /* multiply and accumlate */
emh203 0:3d9c67d97d6f 323 acc1 = __SMLADX(x3, y1, acc1);
emh203 0:3d9c67d97d6f 324
emh203 0:3d9c67d97d6f 325 /* Read next two samples from scratch1 buffer */
emh203 0:3d9c67d97d6f 326 x1 = _SIMD32_OFFSET(pScr1);
emh203 0:3d9c67d97d6f 327
emh203 0:3d9c67d97d6f 328 /* multiply and accumlate */
emh203 0:3d9c67d97d6f 329 acc0 = __SMLAD(x2, y2, acc0);
emh203 0:3d9c67d97d6f 330 acc2 = __SMLAD(x1, y2, acc2);
emh203 0:3d9c67d97d6f 331
emh203 0:3d9c67d97d6f 332 /* pack input data */
emh203 0:3d9c67d97d6f 333 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 334 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 335 #else
emh203 0:3d9c67d97d6f 336 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 337 #endif
emh203 0:3d9c67d97d6f 338
emh203 0:3d9c67d97d6f 339 acc3 = __SMLADX(x3, y1, acc3);
emh203 0:3d9c67d97d6f 340 acc1 = __SMLADX(x3, y2, acc1);
emh203 0:3d9c67d97d6f 341
emh203 0:3d9c67d97d6f 342 x2 = _SIMD32_OFFSET(pScr1 + 2u);
emh203 0:3d9c67d97d6f 343
emh203 0:3d9c67d97d6f 344 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 345 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 346 #else
emh203 0:3d9c67d97d6f 347 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 348 #endif
emh203 0:3d9c67d97d6f 349
emh203 0:3d9c67d97d6f 350 acc3 = __SMLADX(x3, y2, acc3);
emh203 0:3d9c67d97d6f 351
emh203 0:3d9c67d97d6f 352 #else
emh203 0:3d9c67d97d6f 353
emh203 0:3d9c67d97d6f 354 /* Read four samples from smaller buffer */
emh203 0:3d9c67d97d6f 355 a = *pIn2;
emh203 0:3d9c67d97d6f 356 b = *(pIn2 + 1);
emh203 0:3d9c67d97d6f 357
emh203 0:3d9c67d97d6f 358 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 359 y1 = __PKHBT(a, b, 16);
emh203 0:3d9c67d97d6f 360 #else
emh203 0:3d9c67d97d6f 361 y1 = __PKHBT(b, a, 16);
emh203 0:3d9c67d97d6f 362 #endif
emh203 0:3d9c67d97d6f 363
emh203 0:3d9c67d97d6f 364 a = *(pIn2 + 2);
emh203 0:3d9c67d97d6f 365 b = *(pIn2 + 3);
emh203 0:3d9c67d97d6f 366 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 367 y2 = __PKHBT(a, b, 16);
emh203 0:3d9c67d97d6f 368 #else
emh203 0:3d9c67d97d6f 369 y2 = __PKHBT(b, a, 16);
emh203 0:3d9c67d97d6f 370 #endif
emh203 0:3d9c67d97d6f 371
emh203 0:3d9c67d97d6f 372 acc0 = __SMLAD(x1, y1, acc0);
emh203 0:3d9c67d97d6f 373
emh203 0:3d9c67d97d6f 374 acc2 = __SMLAD(x2, y1, acc2);
emh203 0:3d9c67d97d6f 375
emh203 0:3d9c67d97d6f 376 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 377 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 378 #else
emh203 0:3d9c67d97d6f 379 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 380 #endif
emh203 0:3d9c67d97d6f 381
emh203 0:3d9c67d97d6f 382 acc1 = __SMLADX(x3, y1, acc1);
emh203 0:3d9c67d97d6f 383
emh203 0:3d9c67d97d6f 384 a = *pScr1;
emh203 0:3d9c67d97d6f 385 b = *(pScr1 + 1);
emh203 0:3d9c67d97d6f 386
emh203 0:3d9c67d97d6f 387 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 388 x1 = __PKHBT(a, b, 16);
emh203 0:3d9c67d97d6f 389 #else
emh203 0:3d9c67d97d6f 390 x1 = __PKHBT(b, a, 16);
emh203 0:3d9c67d97d6f 391 #endif
emh203 0:3d9c67d97d6f 392
emh203 0:3d9c67d97d6f 393 acc0 = __SMLAD(x2, y2, acc0);
emh203 0:3d9c67d97d6f 394
emh203 0:3d9c67d97d6f 395 acc2 = __SMLAD(x1, y2, acc2);
emh203 0:3d9c67d97d6f 396
emh203 0:3d9c67d97d6f 397 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 398 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 399 #else
emh203 0:3d9c67d97d6f 400 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 401 #endif
emh203 0:3d9c67d97d6f 402
emh203 0:3d9c67d97d6f 403 acc3 = __SMLADX(x3, y1, acc3);
emh203 0:3d9c67d97d6f 404
emh203 0:3d9c67d97d6f 405 acc1 = __SMLADX(x3, y2, acc1);
emh203 0:3d9c67d97d6f 406
emh203 0:3d9c67d97d6f 407 a = *(pScr1 + 2);
emh203 0:3d9c67d97d6f 408 b = *(pScr1 + 3);
emh203 0:3d9c67d97d6f 409
emh203 0:3d9c67d97d6f 410 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 411 x2 = __PKHBT(a, b, 16);
emh203 0:3d9c67d97d6f 412 #else
emh203 0:3d9c67d97d6f 413 x2 = __PKHBT(b, a, 16);
emh203 0:3d9c67d97d6f 414 #endif
emh203 0:3d9c67d97d6f 415
emh203 0:3d9c67d97d6f 416 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 417 x3 = __PKHBT(x2, x1, 0);
emh203 0:3d9c67d97d6f 418 #else
emh203 0:3d9c67d97d6f 419 x3 = __PKHBT(x1, x2, 0);
emh203 0:3d9c67d97d6f 420 #endif
emh203 0:3d9c67d97d6f 421
emh203 0:3d9c67d97d6f 422 acc3 = __SMLADX(x3, y2, acc3);
emh203 0:3d9c67d97d6f 423
emh203 0:3d9c67d97d6f 424 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emh203 0:3d9c67d97d6f 425
emh203 0:3d9c67d97d6f 426 /* update scratch pointers */
emh203 0:3d9c67d97d6f 427 pIn2 += 4u;
emh203 0:3d9c67d97d6f 428 pScr1 += 4u;
emh203 0:3d9c67d97d6f 429
emh203 0:3d9c67d97d6f 430
emh203 0:3d9c67d97d6f 431 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 432 tapCnt--;
emh203 0:3d9c67d97d6f 433 }
emh203 0:3d9c67d97d6f 434
emh203 0:3d9c67d97d6f 435 /* Update scratch pointer for remaining samples of smaller length sequence */
emh203 0:3d9c67d97d6f 436 pScr1 -= 4u;
emh203 0:3d9c67d97d6f 437
emh203 0:3d9c67d97d6f 438 /* apply same above for remaining samples of smaller length sequence */
emh203 0:3d9c67d97d6f 439 tapCnt = (srcBLen) & 3u;
emh203 0:3d9c67d97d6f 440
emh203 0:3d9c67d97d6f 441 while(tapCnt > 0u)
emh203 0:3d9c67d97d6f 442 {
emh203 0:3d9c67d97d6f 443
emh203 0:3d9c67d97d6f 444 /* accumlate the results */
emh203 0:3d9c67d97d6f 445 acc0 += (*pScr1++ * *pIn2);
emh203 0:3d9c67d97d6f 446 acc1 += (*pScr1++ * *pIn2);
emh203 0:3d9c67d97d6f 447 acc2 += (*pScr1++ * *pIn2);
emh203 0:3d9c67d97d6f 448 acc3 += (*pScr1++ * *pIn2++);
emh203 0:3d9c67d97d6f 449
emh203 0:3d9c67d97d6f 450 pScr1 -= 3u;
emh203 0:3d9c67d97d6f 451
emh203 0:3d9c67d97d6f 452 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 453 tapCnt--;
emh203 0:3d9c67d97d6f 454 }
emh203 0:3d9c67d97d6f 455
emh203 0:3d9c67d97d6f 456 blkCnt--;
emh203 0:3d9c67d97d6f 457
emh203 0:3d9c67d97d6f 458
emh203 0:3d9c67d97d6f 459 /* Store the results in the accumulators in the destination buffer. */
emh203 0:3d9c67d97d6f 460
emh203 0:3d9c67d97d6f 461 #ifndef ARM_MATH_BIG_ENDIAN
emh203 0:3d9c67d97d6f 462
emh203 0:3d9c67d97d6f 463 *__SIMD32(pOut)++ =
emh203 0:3d9c67d97d6f 464 __PKHBT(__SSAT((acc0 >> 15), 16), __SSAT((acc1 >> 15), 16), 16);
emh203 0:3d9c67d97d6f 465
emh203 0:3d9c67d97d6f 466 *__SIMD32(pOut)++ =
emh203 0:3d9c67d97d6f 467 __PKHBT(__SSAT((acc2 >> 15), 16), __SSAT((acc3 >> 15), 16), 16);
emh203 0:3d9c67d97d6f 468
emh203 0:3d9c67d97d6f 469
emh203 0:3d9c67d97d6f 470 #else
emh203 0:3d9c67d97d6f 471
emh203 0:3d9c67d97d6f 472 *__SIMD32(pOut)++ =
emh203 0:3d9c67d97d6f 473 __PKHBT(__SSAT((acc1 >> 15), 16), __SSAT((acc0 >> 15), 16), 16);
emh203 0:3d9c67d97d6f 474
emh203 0:3d9c67d97d6f 475 *__SIMD32(pOut)++ =
emh203 0:3d9c67d97d6f 476 __PKHBT(__SSAT((acc3 >> 15), 16), __SSAT((acc2 >> 15), 16), 16);
emh203 0:3d9c67d97d6f 477
emh203 0:3d9c67d97d6f 478
emh203 0:3d9c67d97d6f 479
emh203 0:3d9c67d97d6f 480 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emh203 0:3d9c67d97d6f 481
emh203 0:3d9c67d97d6f 482 /* Initialization of inputB pointer */
emh203 0:3d9c67d97d6f 483 pIn2 = py;
emh203 0:3d9c67d97d6f 484
emh203 0:3d9c67d97d6f 485 pScratch1 += 4u;
emh203 0:3d9c67d97d6f 486
emh203 0:3d9c67d97d6f 487 }
emh203 0:3d9c67d97d6f 488
emh203 0:3d9c67d97d6f 489
emh203 0:3d9c67d97d6f 490 blkCnt = (srcALen + srcBLen - 1u) & 0x3;
emh203 0:3d9c67d97d6f 491
emh203 0:3d9c67d97d6f 492 /* Calculate convolution for remaining samples of Bigger length sequence */
emh203 0:3d9c67d97d6f 493 while(blkCnt > 0)
emh203 0:3d9c67d97d6f 494 {
emh203 0:3d9c67d97d6f 495 /* Initialze temporary scratch pointer as scratch1 */
emh203 0:3d9c67d97d6f 496 pScr1 = pScratch1;
emh203 0:3d9c67d97d6f 497
emh203 0:3d9c67d97d6f 498 /* Clear Accumlators */
emh203 0:3d9c67d97d6f 499 acc0 = 0;
emh203 0:3d9c67d97d6f 500
emh203 0:3d9c67d97d6f 501 tapCnt = (srcBLen) >> 1u;
emh203 0:3d9c67d97d6f 502
emh203 0:3d9c67d97d6f 503 while(tapCnt > 0u)
emh203 0:3d9c67d97d6f 504 {
emh203 0:3d9c67d97d6f 505
emh203 0:3d9c67d97d6f 506 acc0 += (*pScr1++ * *pIn2++);
emh203 0:3d9c67d97d6f 507 acc0 += (*pScr1++ * *pIn2++);
emh203 0:3d9c67d97d6f 508
emh203 0:3d9c67d97d6f 509 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 510 tapCnt--;
emh203 0:3d9c67d97d6f 511 }
emh203 0:3d9c67d97d6f 512
emh203 0:3d9c67d97d6f 513 tapCnt = (srcBLen) & 1u;
emh203 0:3d9c67d97d6f 514
emh203 0:3d9c67d97d6f 515 /* apply same above for remaining samples of smaller length sequence */
emh203 0:3d9c67d97d6f 516 while(tapCnt > 0u)
emh203 0:3d9c67d97d6f 517 {
emh203 0:3d9c67d97d6f 518
emh203 0:3d9c67d97d6f 519 /* accumlate the results */
emh203 0:3d9c67d97d6f 520 acc0 += (*pScr1++ * *pIn2++);
emh203 0:3d9c67d97d6f 521
emh203 0:3d9c67d97d6f 522 /* Decrement the loop counter */
emh203 0:3d9c67d97d6f 523 tapCnt--;
emh203 0:3d9c67d97d6f 524 }
emh203 0:3d9c67d97d6f 525
emh203 0:3d9c67d97d6f 526 blkCnt--;
emh203 0:3d9c67d97d6f 527
emh203 0:3d9c67d97d6f 528 /* The result is in 2.30 format. Convert to 1.15 with saturation.
emh203 0:3d9c67d97d6f 529 ** Then store the output in the destination buffer. */
emh203 0:3d9c67d97d6f 530 *pOut++ = (q15_t) (__SSAT((acc0 >> 15), 16));
emh203 0:3d9c67d97d6f 531
emh203 0:3d9c67d97d6f 532 /* Initialization of inputB pointer */
emh203 0:3d9c67d97d6f 533 pIn2 = py;
emh203 0:3d9c67d97d6f 534
emh203 0:3d9c67d97d6f 535 pScratch1 += 1u;
emh203 0:3d9c67d97d6f 536
emh203 0:3d9c67d97d6f 537 }
emh203 0:3d9c67d97d6f 538
emh203 0:3d9c67d97d6f 539 }
emh203 0:3d9c67d97d6f 540
emh203 0:3d9c67d97d6f 541 /**
emh203 0:3d9c67d97d6f 542 * @} end of Conv group
emh203 0:3d9c67d97d6f 543 */