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_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_q7.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Convolution of Q7 sequences.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
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.
emilmont 1:fdd22bb7aa52 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 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. Length srcALen+srcBLen-1.
emilmont 1:fdd22bb7aa52 59 * @return none.
emilmont 1:fdd22bb7aa52 60 *
emilmont 1:fdd22bb7aa52 61 * @details
emilmont 1:fdd22bb7aa52 62 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 63 *
emilmont 1:fdd22bb7aa52 64 * \par
emilmont 1:fdd22bb7aa52 65 * The function is implemented using a 32-bit internal accumulator.
emilmont 1:fdd22bb7aa52 66 * Both the inputs are represented in 1.7 format and multiplications yield a 2.14 result.
emilmont 1:fdd22bb7aa52 67 * The 2.14 intermediate results are accumulated in a 32-bit accumulator in 18.14 format.
emilmont 1:fdd22bb7aa52 68 * This approach provides 17 guard bits and there is no risk of overflow as long as <code>max(srcALen, srcBLen)<131072</code>.
emilmont 1:fdd22bb7aa52 69 * The 18.14 result is then truncated to 18.7 format by discarding the low 7 bits and then saturated to 1.7 format.
emilmont 1:fdd22bb7aa52 70 *
emilmont 1:fdd22bb7aa52 71 * \par
emilmont 1:fdd22bb7aa52 72 * Refer the function <code>arm_conv_opt_q7()</code> for a faster implementation of this function.
emilmont 1:fdd22bb7aa52 73 *
emilmont 1:fdd22bb7aa52 74 */
emilmont 1:fdd22bb7aa52 75
emilmont 1:fdd22bb7aa52 76 void arm_conv_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 {
emilmont 1:fdd22bb7aa52 83
emilmont 1:fdd22bb7aa52 84
mbed_official 3:7a284390b0ce 85 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 86
emilmont 1:fdd22bb7aa52 87 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 88
emilmont 1:fdd22bb7aa52 89 q7_t *pIn1; /* inputA pointer */
emilmont 1:fdd22bb7aa52 90 q7_t *pIn2; /* inputB pointer */
emilmont 1:fdd22bb7aa52 91 q7_t *pOut = pDst; /* output pointer */
emilmont 1:fdd22bb7aa52 92 q7_t *px; /* Intermediate inputA pointer */
emilmont 1:fdd22bb7aa52 93 q7_t *py; /* Intermediate inputB pointer */
emilmont 1:fdd22bb7aa52 94 q7_t *pSrc1, *pSrc2; /* Intermediate pointers */
emilmont 1:fdd22bb7aa52 95 q7_t x0, x1, x2, x3, c0, c1; /* Temporary variables to hold state and coefficient values */
emilmont 1:fdd22bb7aa52 96 q31_t sum, acc0, acc1, acc2, acc3; /* Accumulator */
emilmont 1:fdd22bb7aa52 97 q31_t input1, input2; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 98 q15_t in1, in2; /* Temporary input variables */
emilmont 1:fdd22bb7aa52 99 uint32_t j, k, count, blkCnt, blockSize1, blockSize2, blockSize3; /* loop counter */
emilmont 1:fdd22bb7aa52 100
emilmont 1:fdd22bb7aa52 101 /* The algorithm implementation is based on the lengths of the inputs. */
emilmont 1:fdd22bb7aa52 102 /* srcB is always made to slide across srcA. */
emilmont 1:fdd22bb7aa52 103 /* So srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 104 if(srcALen >= srcBLen)
emilmont 1:fdd22bb7aa52 105 {
emilmont 1:fdd22bb7aa52 106 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 107 pIn1 = pSrcA;
emilmont 1:fdd22bb7aa52 108
emilmont 1:fdd22bb7aa52 109 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 110 pIn2 = pSrcB;
emilmont 1:fdd22bb7aa52 111 }
emilmont 1:fdd22bb7aa52 112 else
emilmont 1:fdd22bb7aa52 113 {
emilmont 1:fdd22bb7aa52 114 /* Initialization of inputA pointer */
emilmont 1:fdd22bb7aa52 115 pIn1 = pSrcB;
emilmont 1:fdd22bb7aa52 116
emilmont 1:fdd22bb7aa52 117 /* Initialization of inputB pointer */
emilmont 1:fdd22bb7aa52 118 pIn2 = pSrcA;
emilmont 1:fdd22bb7aa52 119
emilmont 1:fdd22bb7aa52 120 /* srcBLen is always considered as shorter or equal to srcALen */
emilmont 1:fdd22bb7aa52 121 j = srcBLen;
emilmont 1:fdd22bb7aa52 122 srcBLen = srcALen;
emilmont 1:fdd22bb7aa52 123 srcALen = j;
emilmont 1:fdd22bb7aa52 124 }
emilmont 1:fdd22bb7aa52 125
emilmont 1:fdd22bb7aa52 126 /* conv(x,y) at n = x[n] * y[0] + x[n-1] * y[1] + x[n-2] * y[2] + ...+ x[n-N+1] * y[N -1] */
emilmont 1:fdd22bb7aa52 127 /* The function is internally
emilmont 1:fdd22bb7aa52 128 * divided into three stages according to the number of multiplications that has to be
emilmont 1:fdd22bb7aa52 129 * taken place between inputA samples and inputB samples. In the first stage of the
emilmont 1:fdd22bb7aa52 130 * algorithm, the multiplications increase by one for every iteration.
emilmont 1:fdd22bb7aa52 131 * In the second stage of the algorithm, srcBLen number of multiplications are done.
emilmont 1:fdd22bb7aa52 132 * In the third stage of the algorithm, the multiplications decrease by one
emilmont 1:fdd22bb7aa52 133 * for every iteration. */
emilmont 1:fdd22bb7aa52 134
emilmont 1:fdd22bb7aa52 135 /* The algorithm is implemented in three stages.
emilmont 1:fdd22bb7aa52 136 The loop counters of each stage is initiated here. */
emilmont 1:fdd22bb7aa52 137 blockSize1 = srcBLen - 1u;
emilmont 1:fdd22bb7aa52 138 blockSize2 = (srcALen - srcBLen) + 1u;
emilmont 1:fdd22bb7aa52 139 blockSize3 = blockSize1;
emilmont 1:fdd22bb7aa52 140
emilmont 1:fdd22bb7aa52 141 /* --------------------------
emilmont 1:fdd22bb7aa52 142 * Initializations of stage1
emilmont 1:fdd22bb7aa52 143 * -------------------------*/
emilmont 1:fdd22bb7aa52 144
emilmont 1:fdd22bb7aa52 145 /* sum = x[0] * y[0]
emilmont 1:fdd22bb7aa52 146 * sum = x[0] * y[1] + x[1] * y[0]
emilmont 1:fdd22bb7aa52 147 * ....
emilmont 1:fdd22bb7aa52 148 * sum = x[0] * y[srcBlen - 1] + x[1] * y[srcBlen - 2] +...+ x[srcBLen - 1] * y[0]
emilmont 1:fdd22bb7aa52 149 */
emilmont 1:fdd22bb7aa52 150
emilmont 1:fdd22bb7aa52 151 /* In this stage the MAC operations are increased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 152 The count variable holds the number of MAC operations performed */
emilmont 1:fdd22bb7aa52 153 count = 1u;
emilmont 1:fdd22bb7aa52 154
emilmont 1:fdd22bb7aa52 155 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 156 px = pIn1;
emilmont 1:fdd22bb7aa52 157
emilmont 1:fdd22bb7aa52 158 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 159 py = pIn2;
emilmont 1:fdd22bb7aa52 160
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* ------------------------
emilmont 1:fdd22bb7aa52 163 * Stage1 process
emilmont 1:fdd22bb7aa52 164 * ----------------------*/
emilmont 1:fdd22bb7aa52 165
emilmont 1:fdd22bb7aa52 166 /* The first stage starts here */
emilmont 1:fdd22bb7aa52 167 while(blockSize1 > 0u)
emilmont 1:fdd22bb7aa52 168 {
emilmont 1:fdd22bb7aa52 169 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 170 sum = 0;
emilmont 1:fdd22bb7aa52 171
emilmont 1:fdd22bb7aa52 172 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 173 k = count >> 2u;
emilmont 1:fdd22bb7aa52 174
emilmont 1:fdd22bb7aa52 175 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 176 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 177 while(k > 0u)
emilmont 1:fdd22bb7aa52 178 {
emilmont 1:fdd22bb7aa52 179 /* x[0] , x[1] */
emilmont 1:fdd22bb7aa52 180 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 181 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 182 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 183
emilmont 1:fdd22bb7aa52 184 /* y[srcBLen - 1] , y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 185 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 186 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 187 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 188
emilmont 1:fdd22bb7aa52 189 /* x[0] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 190 /* x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 191 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 /* x[2] , x[3] */
emilmont 1:fdd22bb7aa52 194 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 195 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 196 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 197
emilmont 1:fdd22bb7aa52 198 /* y[srcBLen - 3] , y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 199 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 200 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 201 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 /* x[2] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 204 /* x[3] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 205 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 206
emilmont 1:fdd22bb7aa52 207 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 208 k--;
emilmont 1:fdd22bb7aa52 209 }
emilmont 1:fdd22bb7aa52 210
emilmont 1:fdd22bb7aa52 211 /* If the count is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 212 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 213 k = count % 0x4u;
emilmont 1:fdd22bb7aa52 214
emilmont 1:fdd22bb7aa52 215 while(k > 0u)
emilmont 1:fdd22bb7aa52 216 {
emilmont 1:fdd22bb7aa52 217 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 218 sum += ((q15_t) * px++ * *py--);
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 221 k--;
emilmont 1:fdd22bb7aa52 222 }
emilmont 1:fdd22bb7aa52 223
emilmont 1:fdd22bb7aa52 224 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 225 *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 228 py = pIn2 + count;
emilmont 1:fdd22bb7aa52 229 px = pIn1;
emilmont 1:fdd22bb7aa52 230
emilmont 1:fdd22bb7aa52 231 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 232 count++;
emilmont 1:fdd22bb7aa52 233
emilmont 1:fdd22bb7aa52 234 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 235 blockSize1--;
emilmont 1:fdd22bb7aa52 236 }
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 /* --------------------------
emilmont 1:fdd22bb7aa52 239 * Initializations of stage2
emilmont 1:fdd22bb7aa52 240 * ------------------------*/
emilmont 1:fdd22bb7aa52 241
emilmont 1:fdd22bb7aa52 242 /* sum = x[0] * y[srcBLen-1] + x[1] * y[srcBLen-2] +...+ x[srcBLen-1] * y[0]
emilmont 1:fdd22bb7aa52 243 * sum = x[1] * y[srcBLen-1] + x[2] * y[srcBLen-2] +...+ x[srcBLen] * y[0]
emilmont 1:fdd22bb7aa52 244 * ....
emilmont 1:fdd22bb7aa52 245 * sum = x[srcALen-srcBLen-2] * y[srcBLen-1] + x[srcALen] * y[srcBLen-2] +...+ x[srcALen-1] * y[0]
emilmont 1:fdd22bb7aa52 246 */
emilmont 1:fdd22bb7aa52 247
emilmont 1:fdd22bb7aa52 248 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 249 px = pIn1;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 252 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 253 py = pSrc2;
emilmont 1:fdd22bb7aa52 254
emilmont 1:fdd22bb7aa52 255 /* count is index by which the pointer pIn1 to be incremented */
emilmont 1:fdd22bb7aa52 256 count = 0u;
emilmont 1:fdd22bb7aa52 257
emilmont 1:fdd22bb7aa52 258 /* -------------------
emilmont 1:fdd22bb7aa52 259 * Stage2 process
emilmont 1:fdd22bb7aa52 260 * ------------------*/
emilmont 1:fdd22bb7aa52 261
emilmont 1:fdd22bb7aa52 262 /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed.
emilmont 1:fdd22bb7aa52 263 * So, to loop unroll over blockSize2,
emilmont 1:fdd22bb7aa52 264 * srcBLen should be greater than or equal to 4 */
emilmont 1:fdd22bb7aa52 265 if(srcBLen >= 4u)
emilmont 1:fdd22bb7aa52 266 {
emilmont 1:fdd22bb7aa52 267 /* Loop unroll over blockSize2, by 4 */
emilmont 1:fdd22bb7aa52 268 blkCnt = blockSize2 >> 2u;
emilmont 1:fdd22bb7aa52 269
emilmont 1:fdd22bb7aa52 270 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 271 {
emilmont 1:fdd22bb7aa52 272 /* Set all accumulators to zero */
emilmont 1:fdd22bb7aa52 273 acc0 = 0;
emilmont 1:fdd22bb7aa52 274 acc1 = 0;
emilmont 1:fdd22bb7aa52 275 acc2 = 0;
emilmont 1:fdd22bb7aa52 276 acc3 = 0;
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 /* read x[0], x[1], x[2] samples */
emilmont 1:fdd22bb7aa52 279 x0 = *(px++);
emilmont 1:fdd22bb7aa52 280 x1 = *(px++);
emilmont 1:fdd22bb7aa52 281 x2 = *(px++);
emilmont 1:fdd22bb7aa52 282
emilmont 1:fdd22bb7aa52 283 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 284 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 285
emilmont 1:fdd22bb7aa52 286 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 287 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 288 do
emilmont 1:fdd22bb7aa52 289 {
emilmont 1:fdd22bb7aa52 290 /* Read y[srcBLen - 1] sample */
emilmont 1:fdd22bb7aa52 291 c0 = *(py--);
emilmont 1:fdd22bb7aa52 292 /* Read y[srcBLen - 2] sample */
emilmont 1:fdd22bb7aa52 293 c1 = *(py--);
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 /* Read x[3] sample */
emilmont 1:fdd22bb7aa52 296 x3 = *(px++);
emilmont 1:fdd22bb7aa52 297
emilmont 1:fdd22bb7aa52 298 /* x[0] and x[1] are packed */
emilmont 1:fdd22bb7aa52 299 in1 = (q15_t) x0;
emilmont 1:fdd22bb7aa52 300 in2 = (q15_t) x1;
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 303
emilmont 1:fdd22bb7aa52 304 /* y[srcBLen - 1] and y[srcBLen - 2] are packed */
emilmont 1:fdd22bb7aa52 305 in1 = (q15_t) c0;
emilmont 1:fdd22bb7aa52 306 in2 = (q15_t) c1;
emilmont 1:fdd22bb7aa52 307
emilmont 1:fdd22bb7aa52 308 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 309
emilmont 1:fdd22bb7aa52 310 /* acc0 += x[0] * y[srcBLen - 1] + x[1] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 311 acc0 = __SMLAD(input1, input2, acc0);
emilmont 1:fdd22bb7aa52 312
emilmont 1:fdd22bb7aa52 313 /* x[1] and x[2] are packed */
emilmont 1:fdd22bb7aa52 314 in1 = (q15_t) x1;
emilmont 1:fdd22bb7aa52 315 in2 = (q15_t) x2;
emilmont 1:fdd22bb7aa52 316
emilmont 1:fdd22bb7aa52 317 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 318
emilmont 1:fdd22bb7aa52 319 /* acc1 += x[1] * y[srcBLen - 1] + x[2] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 320 acc1 = __SMLAD(input1, input2, acc1);
emilmont 1:fdd22bb7aa52 321
emilmont 1:fdd22bb7aa52 322 /* x[2] and x[3] are packed */
emilmont 1:fdd22bb7aa52 323 in1 = (q15_t) x2;
emilmont 1:fdd22bb7aa52 324 in2 = (q15_t) x3;
emilmont 1:fdd22bb7aa52 325
emilmont 1:fdd22bb7aa52 326 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 327
emilmont 1:fdd22bb7aa52 328 /* acc2 += x[2] * y[srcBLen - 1] + x[3] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 329 acc2 = __SMLAD(input1, input2, acc2);
emilmont 1:fdd22bb7aa52 330
emilmont 1:fdd22bb7aa52 331 /* Read x[4] sample */
emilmont 1:fdd22bb7aa52 332 x0 = *(px++);
emilmont 1:fdd22bb7aa52 333
emilmont 1:fdd22bb7aa52 334 /* x[3] and x[4] are packed */
emilmont 1:fdd22bb7aa52 335 in1 = (q15_t) x3;
emilmont 1:fdd22bb7aa52 336 in2 = (q15_t) x0;
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 339
emilmont 1:fdd22bb7aa52 340 /* acc3 += x[3] * y[srcBLen - 1] + x[4] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 341 acc3 = __SMLAD(input1, input2, acc3);
emilmont 1:fdd22bb7aa52 342
emilmont 1:fdd22bb7aa52 343 /* Read y[srcBLen - 3] sample */
emilmont 1:fdd22bb7aa52 344 c0 = *(py--);
emilmont 1:fdd22bb7aa52 345 /* Read y[srcBLen - 4] sample */
emilmont 1:fdd22bb7aa52 346 c1 = *(py--);
emilmont 1:fdd22bb7aa52 347
emilmont 1:fdd22bb7aa52 348 /* Read x[5] sample */
emilmont 1:fdd22bb7aa52 349 x1 = *(px++);
emilmont 1:fdd22bb7aa52 350
emilmont 1:fdd22bb7aa52 351 /* x[2] and x[3] are packed */
emilmont 1:fdd22bb7aa52 352 in1 = (q15_t) x2;
emilmont 1:fdd22bb7aa52 353 in2 = (q15_t) x3;
emilmont 1:fdd22bb7aa52 354
emilmont 1:fdd22bb7aa52 355 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 356
emilmont 1:fdd22bb7aa52 357 /* y[srcBLen - 3] and y[srcBLen - 4] are packed */
emilmont 1:fdd22bb7aa52 358 in1 = (q15_t) c0;
emilmont 1:fdd22bb7aa52 359 in2 = (q15_t) c1;
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 362
emilmont 1:fdd22bb7aa52 363 /* acc0 += x[2] * y[srcBLen - 3] + x[3] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 364 acc0 = __SMLAD(input1, input2, acc0);
emilmont 1:fdd22bb7aa52 365
emilmont 1:fdd22bb7aa52 366 /* x[3] and x[4] are packed */
emilmont 1:fdd22bb7aa52 367 in1 = (q15_t) x3;
emilmont 1:fdd22bb7aa52 368 in2 = (q15_t) x0;
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 /* acc1 += x[3] * y[srcBLen - 3] + x[4] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 373 acc1 = __SMLAD(input1, input2, acc1);
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 /* x[4] and x[5] are packed */
emilmont 1:fdd22bb7aa52 376 in1 = (q15_t) x0;
emilmont 1:fdd22bb7aa52 377 in2 = (q15_t) x1;
emilmont 1:fdd22bb7aa52 378
emilmont 1:fdd22bb7aa52 379 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 380
emilmont 1:fdd22bb7aa52 381 /* acc2 += x[4] * y[srcBLen - 3] + x[5] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 382 acc2 = __SMLAD(input1, input2, acc2);
emilmont 1:fdd22bb7aa52 383
emilmont 1:fdd22bb7aa52 384 /* Read x[6] sample */
emilmont 1:fdd22bb7aa52 385 x2 = *(px++);
emilmont 1:fdd22bb7aa52 386
emilmont 1:fdd22bb7aa52 387 /* x[5] and x[6] are packed */
emilmont 1:fdd22bb7aa52 388 in1 = (q15_t) x1;
emilmont 1:fdd22bb7aa52 389 in2 = (q15_t) x2;
emilmont 1:fdd22bb7aa52 390
emilmont 1:fdd22bb7aa52 391 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 392
emilmont 1:fdd22bb7aa52 393 /* acc3 += x[5] * y[srcBLen - 3] + x[6] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 394 acc3 = __SMLAD(input1, input2, acc3);
emilmont 1:fdd22bb7aa52 395
emilmont 1:fdd22bb7aa52 396 } while(--k);
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 399 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 400 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402 while(k > 0u)
emilmont 1:fdd22bb7aa52 403 {
emilmont 1:fdd22bb7aa52 404 /* Read y[srcBLen - 5] sample */
emilmont 1:fdd22bb7aa52 405 c0 = *(py--);
emilmont 1:fdd22bb7aa52 406
emilmont 1:fdd22bb7aa52 407 /* Read x[7] sample */
emilmont 1:fdd22bb7aa52 408 x3 = *(px++);
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 411 /* acc0 += x[4] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 412 acc0 += ((q15_t) x0 * c0);
emilmont 1:fdd22bb7aa52 413 /* acc1 += x[5] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 414 acc1 += ((q15_t) x1 * c0);
emilmont 1:fdd22bb7aa52 415 /* acc2 += x[6] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 416 acc2 += ((q15_t) x2 * c0);
emilmont 1:fdd22bb7aa52 417 /* acc3 += x[7] * y[srcBLen - 5] */
emilmont 1:fdd22bb7aa52 418 acc3 += ((q15_t) x3 * c0);
emilmont 1:fdd22bb7aa52 419
emilmont 1:fdd22bb7aa52 420 /* Reuse the present samples for the next MAC */
emilmont 1:fdd22bb7aa52 421 x0 = x1;
emilmont 1:fdd22bb7aa52 422 x1 = x2;
emilmont 1:fdd22bb7aa52 423 x2 = x3;
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 426 k--;
emilmont 1:fdd22bb7aa52 427 }
emilmont 1:fdd22bb7aa52 428
emilmont 1:fdd22bb7aa52 429
emilmont 1:fdd22bb7aa52 430 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 431 *pOut++ = (q7_t) (__SSAT(acc0 >> 7u, 8));
emilmont 1:fdd22bb7aa52 432 *pOut++ = (q7_t) (__SSAT(acc1 >> 7u, 8));
emilmont 1:fdd22bb7aa52 433 *pOut++ = (q7_t) (__SSAT(acc2 >> 7u, 8));
emilmont 1:fdd22bb7aa52 434 *pOut++ = (q7_t) (__SSAT(acc3 >> 7u, 8));
emilmont 1:fdd22bb7aa52 435
emilmont 1:fdd22bb7aa52 436 /* Increment the pointer pIn1 index, count by 4 */
emilmont 1:fdd22bb7aa52 437 count += 4u;
emilmont 1:fdd22bb7aa52 438
emilmont 1:fdd22bb7aa52 439 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 440 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 441 py = pSrc2;
emilmont 1:fdd22bb7aa52 442
emilmont 1:fdd22bb7aa52 443 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 444 blkCnt--;
emilmont 1:fdd22bb7aa52 445 }
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here.
emilmont 1:fdd22bb7aa52 448 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 449 blkCnt = blockSize2 % 0x4u;
emilmont 1:fdd22bb7aa52 450
emilmont 1:fdd22bb7aa52 451 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 452 {
emilmont 1:fdd22bb7aa52 453 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 454 sum = 0;
emilmont 1:fdd22bb7aa52 455
emilmont 1:fdd22bb7aa52 456 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 457 k = srcBLen >> 2u;
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 460 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 461 while(k > 0u)
emilmont 1:fdd22bb7aa52 462 {
emilmont 1:fdd22bb7aa52 463
emilmont 1:fdd22bb7aa52 464 /* Reading two inputs of SrcA buffer and packing */
emilmont 1:fdd22bb7aa52 465 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 466 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 467 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 468
emilmont 1:fdd22bb7aa52 469 /* Reading two inputs of SrcB buffer and packing */
emilmont 1:fdd22bb7aa52 470 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 471 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 472 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 473
emilmont 1:fdd22bb7aa52 474 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 475 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 476
emilmont 1:fdd22bb7aa52 477 /* Reading two inputs of SrcA buffer and packing */
emilmont 1:fdd22bb7aa52 478 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 479 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 480 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* Reading two inputs of SrcB buffer and packing */
emilmont 1:fdd22bb7aa52 483 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 484 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 485 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 486
emilmont 1:fdd22bb7aa52 487 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 488 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 489
emilmont 1:fdd22bb7aa52 490 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 491 k--;
emilmont 1:fdd22bb7aa52 492 }
emilmont 1:fdd22bb7aa52 493
emilmont 1:fdd22bb7aa52 494 /* If the srcBLen is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 495 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 496 k = srcBLen % 0x4u;
emilmont 1:fdd22bb7aa52 497
emilmont 1:fdd22bb7aa52 498 while(k > 0u)
emilmont 1:fdd22bb7aa52 499 {
emilmont 1:fdd22bb7aa52 500 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 501 sum += ((q15_t) * px++ * *py--);
emilmont 1:fdd22bb7aa52 502
emilmont 1:fdd22bb7aa52 503 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 504 k--;
emilmont 1:fdd22bb7aa52 505 }
emilmont 1:fdd22bb7aa52 506
emilmont 1:fdd22bb7aa52 507 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 508 *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));
emilmont 1:fdd22bb7aa52 509
emilmont 1:fdd22bb7aa52 510 /* Increment the pointer pIn1 index, count by 1 */
emilmont 1:fdd22bb7aa52 511 count++;
emilmont 1:fdd22bb7aa52 512
emilmont 1:fdd22bb7aa52 513 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 514 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 515 py = pSrc2;
emilmont 1:fdd22bb7aa52 516
emilmont 1:fdd22bb7aa52 517 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 518 blkCnt--;
emilmont 1:fdd22bb7aa52 519 }
emilmont 1:fdd22bb7aa52 520 }
emilmont 1:fdd22bb7aa52 521 else
emilmont 1:fdd22bb7aa52 522 {
emilmont 1:fdd22bb7aa52 523 /* If the srcBLen is not a multiple of 4,
emilmont 1:fdd22bb7aa52 524 * the blockSize2 loop cannot be unrolled by 4 */
emilmont 1:fdd22bb7aa52 525 blkCnt = blockSize2;
emilmont 1:fdd22bb7aa52 526
emilmont 1:fdd22bb7aa52 527 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 528 {
emilmont 1:fdd22bb7aa52 529 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 530 sum = 0;
emilmont 1:fdd22bb7aa52 531
emilmont 1:fdd22bb7aa52 532 /* srcBLen number of MACS should be performed */
emilmont 1:fdd22bb7aa52 533 k = srcBLen;
emilmont 1:fdd22bb7aa52 534
emilmont 1:fdd22bb7aa52 535 while(k > 0u)
emilmont 1:fdd22bb7aa52 536 {
emilmont 1:fdd22bb7aa52 537 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 538 sum += ((q15_t) * px++ * *py--);
emilmont 1:fdd22bb7aa52 539
emilmont 1:fdd22bb7aa52 540 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 541 k--;
emilmont 1:fdd22bb7aa52 542 }
emilmont 1:fdd22bb7aa52 543
emilmont 1:fdd22bb7aa52 544 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 545 *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));
emilmont 1:fdd22bb7aa52 546
emilmont 1:fdd22bb7aa52 547 /* Increment the MAC count */
emilmont 1:fdd22bb7aa52 548 count++;
emilmont 1:fdd22bb7aa52 549
emilmont 1:fdd22bb7aa52 550 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 551 px = pIn1 + count;
emilmont 1:fdd22bb7aa52 552 py = pSrc2;
emilmont 1:fdd22bb7aa52 553
emilmont 1:fdd22bb7aa52 554 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 555 blkCnt--;
emilmont 1:fdd22bb7aa52 556 }
emilmont 1:fdd22bb7aa52 557 }
emilmont 1:fdd22bb7aa52 558
emilmont 1:fdd22bb7aa52 559
emilmont 1:fdd22bb7aa52 560 /* --------------------------
emilmont 1:fdd22bb7aa52 561 * Initializations of stage3
emilmont 1:fdd22bb7aa52 562 * -------------------------*/
emilmont 1:fdd22bb7aa52 563
emilmont 1:fdd22bb7aa52 564 /* sum += x[srcALen-srcBLen+1] * y[srcBLen-1] + x[srcALen-srcBLen+2] * y[srcBLen-2] +...+ x[srcALen-1] * y[1]
emilmont 1:fdd22bb7aa52 565 * sum += x[srcALen-srcBLen+2] * y[srcBLen-1] + x[srcALen-srcBLen+3] * y[srcBLen-2] +...+ x[srcALen-1] * y[2]
emilmont 1:fdd22bb7aa52 566 * ....
emilmont 1:fdd22bb7aa52 567 * sum += x[srcALen-2] * y[srcBLen-1] + x[srcALen-1] * y[srcBLen-2]
emilmont 1:fdd22bb7aa52 568 * sum += x[srcALen-1] * y[srcBLen-1]
emilmont 1:fdd22bb7aa52 569 */
emilmont 1:fdd22bb7aa52 570
emilmont 1:fdd22bb7aa52 571 /* In this stage the MAC operations are decreased by 1 for every iteration.
emilmont 1:fdd22bb7aa52 572 The blockSize3 variable holds the number of MAC operations performed */
emilmont 1:fdd22bb7aa52 573
emilmont 1:fdd22bb7aa52 574 /* Working pointer of inputA */
emilmont 1:fdd22bb7aa52 575 pSrc1 = pIn1 + (srcALen - (srcBLen - 1u));
emilmont 1:fdd22bb7aa52 576 px = pSrc1;
emilmont 1:fdd22bb7aa52 577
emilmont 1:fdd22bb7aa52 578 /* Working pointer of inputB */
emilmont 1:fdd22bb7aa52 579 pSrc2 = pIn2 + (srcBLen - 1u);
emilmont 1:fdd22bb7aa52 580 py = pSrc2;
emilmont 1:fdd22bb7aa52 581
emilmont 1:fdd22bb7aa52 582 /* -------------------
emilmont 1:fdd22bb7aa52 583 * Stage3 process
emilmont 1:fdd22bb7aa52 584 * ------------------*/
emilmont 1:fdd22bb7aa52 585
emilmont 1:fdd22bb7aa52 586 while(blockSize3 > 0u)
emilmont 1:fdd22bb7aa52 587 {
emilmont 1:fdd22bb7aa52 588 /* Accumulator is made zero for every iteration */
emilmont 1:fdd22bb7aa52 589 sum = 0;
emilmont 1:fdd22bb7aa52 590
emilmont 1:fdd22bb7aa52 591 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 592 k = blockSize3 >> 2u;
emilmont 1:fdd22bb7aa52 593
emilmont 1:fdd22bb7aa52 594 /* First part of the processing with loop unrolling. Compute 4 MACs at a time.
emilmont 1:fdd22bb7aa52 595 ** a second loop below computes MACs for the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 596 while(k > 0u)
emilmont 1:fdd22bb7aa52 597 {
emilmont 1:fdd22bb7aa52 598 /* Reading two inputs, x[srcALen - srcBLen + 1] and x[srcALen - srcBLen + 2] of SrcA buffer and packing */
emilmont 1:fdd22bb7aa52 599 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 600 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 601 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 602
emilmont 1:fdd22bb7aa52 603 /* Reading two inputs, y[srcBLen - 1] and y[srcBLen - 2] of SrcB buffer and packing */
emilmont 1:fdd22bb7aa52 604 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 605 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 606 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 607
emilmont 1:fdd22bb7aa52 608 /* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */
emilmont 1:fdd22bb7aa52 609 /* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */
emilmont 1:fdd22bb7aa52 610 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 611
emilmont 1:fdd22bb7aa52 612 /* Reading two inputs, x[srcALen - srcBLen + 3] and x[srcALen - srcBLen + 4] of SrcA buffer and packing */
emilmont 1:fdd22bb7aa52 613 in1 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 614 in2 = (q15_t) * px++;
emilmont 1:fdd22bb7aa52 615 input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 616
emilmont 1:fdd22bb7aa52 617 /* Reading two inputs, y[srcBLen - 3] and y[srcBLen - 4] of SrcB buffer and packing */
emilmont 1:fdd22bb7aa52 618 in1 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 619 in2 = (q15_t) * py--;
emilmont 1:fdd22bb7aa52 620 input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16u);
emilmont 1:fdd22bb7aa52 621
emilmont 1:fdd22bb7aa52 622 /* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */
emilmont 1:fdd22bb7aa52 623 /* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */
emilmont 1:fdd22bb7aa52 624 sum = __SMLAD(input1, input2, sum);
emilmont 1:fdd22bb7aa52 625
emilmont 1:fdd22bb7aa52 626 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 627 k--;
emilmont 1:fdd22bb7aa52 628 }
emilmont 1:fdd22bb7aa52 629
emilmont 1:fdd22bb7aa52 630 /* If the blockSize3 is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 631 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 632 k = blockSize3 % 0x4u;
emilmont 1:fdd22bb7aa52 633
emilmont 1:fdd22bb7aa52 634 while(k > 0u)
emilmont 1:fdd22bb7aa52 635 {
emilmont 1:fdd22bb7aa52 636 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 637 sum += ((q15_t) * px++ * *py--);
emilmont 1:fdd22bb7aa52 638
emilmont 1:fdd22bb7aa52 639 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 640 k--;
emilmont 1:fdd22bb7aa52 641 }
emilmont 1:fdd22bb7aa52 642
emilmont 1:fdd22bb7aa52 643 /* Store the result in the accumulator in the destination buffer. */
emilmont 1:fdd22bb7aa52 644 *pOut++ = (q7_t) (__SSAT(sum >> 7u, 8));
emilmont 1:fdd22bb7aa52 645
emilmont 1:fdd22bb7aa52 646 /* Update the inputA and inputB pointers for next MAC calculation */
emilmont 1:fdd22bb7aa52 647 px = ++pSrc1;
emilmont 1:fdd22bb7aa52 648 py = pSrc2;
emilmont 1:fdd22bb7aa52 649
emilmont 1:fdd22bb7aa52 650 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 651 blockSize3--;
emilmont 1:fdd22bb7aa52 652 }
emilmont 1:fdd22bb7aa52 653
emilmont 1:fdd22bb7aa52 654 #else
emilmont 1:fdd22bb7aa52 655
emilmont 1:fdd22bb7aa52 656 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 657
emilmont 1:fdd22bb7aa52 658 q7_t *pIn1 = pSrcA; /* input pointer */
emilmont 1:fdd22bb7aa52 659 q7_t *pIn2 = pSrcB; /* coefficient pointer */
emilmont 1:fdd22bb7aa52 660 q31_t sum; /* Accumulator */
emilmont 1:fdd22bb7aa52 661 uint32_t i, j; /* loop counter */
emilmont 1:fdd22bb7aa52 662
emilmont 1:fdd22bb7aa52 663 /* Loop to calculate output of convolution for output length number of times */
emilmont 1:fdd22bb7aa52 664 for (i = 0; i < (srcALen + srcBLen - 1); i++)
emilmont 1:fdd22bb7aa52 665 {
emilmont 1:fdd22bb7aa52 666 /* Initialize sum with zero to carry on MAC operations */
emilmont 1:fdd22bb7aa52 667 sum = 0;
emilmont 1:fdd22bb7aa52 668
emilmont 1:fdd22bb7aa52 669 /* Loop to perform MAC operations according to convolution equation */
emilmont 1:fdd22bb7aa52 670 for (j = 0; j <= i; j++)
emilmont 1:fdd22bb7aa52 671 {
emilmont 1:fdd22bb7aa52 672 /* Check the array limitations */
emilmont 1:fdd22bb7aa52 673 if(((i - j) < srcBLen) && (j < srcALen))
emilmont 1:fdd22bb7aa52 674 {
emilmont 1:fdd22bb7aa52 675 /* z[i] += x[i-j] * y[j] */
emilmont 1:fdd22bb7aa52 676 sum += (q15_t) pIn1[j] * (pIn2[i - j]);
emilmont 1:fdd22bb7aa52 677 }
emilmont 1:fdd22bb7aa52 678 }
emilmont 1:fdd22bb7aa52 679
emilmont 1:fdd22bb7aa52 680 /* Store the output in the destination buffer */
emilmont 1:fdd22bb7aa52 681 pDst[i] = (q7_t) __SSAT((sum >> 7u), 8u);
emilmont 1:fdd22bb7aa52 682 }
emilmont 1:fdd22bb7aa52 683
mbed_official 3:7a284390b0ce 684 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 685
emilmont 1:fdd22bb7aa52 686 }
emilmont 1:fdd22bb7aa52 687
emilmont 1:fdd22bb7aa52 688 /**
emilmont 1:fdd22bb7aa52 689 * @} end of Conv group
emilmont 1:fdd22bb7aa52 690 */