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_correlate_fast_q31.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?
User | Revision | Line number | New 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_correlate_fast_q31.c |
emilmont | 1:fdd22bb7aa52 | 9 | * |
emilmont | 2:da51fb522205 | 10 | * Description: Fast Q31 Correlation. |
emilmont | 1:fdd22bb7aa52 | 11 | * |
emilmont | 1:fdd22bb7aa52 | 12 | * Target Processor: Cortex-M4/Cortex-M3 |
mbed_official | 3:7a284390b0ce | 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 Corr |
emilmont | 1:fdd22bb7aa52 | 49 | * @{ |
emilmont | 1:fdd22bb7aa52 | 50 | */ |
emilmont | 1:fdd22bb7aa52 | 51 | |
emilmont | 1:fdd22bb7aa52 | 52 | /** |
emilmont | 1:fdd22bb7aa52 | 53 | * @brief Correlation of Q31 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 2 * max(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 | * This function is optimized for speed at the expense of fixed-point precision and overflow protection. |
emilmont | 1:fdd22bb7aa52 | 66 | * The result of each 1.31 x 1.31 multiplication is truncated to 2.30 format. |
emilmont | 1:fdd22bb7aa52 | 67 | * These intermediate results are accumulated in a 32-bit register in 2.30 format. |
emilmont | 1:fdd22bb7aa52 | 68 | * Finally, the accumulator is saturated and converted to a 1.31 result. |
emilmont | 1:fdd22bb7aa52 | 69 | * |
emilmont | 1:fdd22bb7aa52 | 70 | * \par |
emilmont | 1:fdd22bb7aa52 | 71 | * The fast version has the same overflow behavior as the standard version but provides less precision since it discards the low 32 bits of each multiplication result. |
emilmont | 1:fdd22bb7aa52 | 72 | * In order to avoid overflows completely the input signals must be scaled down. |
emilmont | 1:fdd22bb7aa52 | 73 | * The input signals should be scaled down to avoid intermediate overflows. |
emilmont | 1:fdd22bb7aa52 | 74 | * Scale down one of the inputs by 1/min(srcALen, srcBLen)to avoid overflows since a |
emilmont | 1:fdd22bb7aa52 | 75 | * maximum of min(srcALen, srcBLen) number of additions is carried internally. |
emilmont | 1:fdd22bb7aa52 | 76 | * |
emilmont | 1:fdd22bb7aa52 | 77 | * \par |
emilmont | 1:fdd22bb7aa52 | 78 | * See <code>arm_correlate_q31()</code> for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. |
emilmont | 1:fdd22bb7aa52 | 79 | */ |
emilmont | 1:fdd22bb7aa52 | 80 | |
emilmont | 1:fdd22bb7aa52 | 81 | void arm_correlate_fast_q31( |
emilmont | 1:fdd22bb7aa52 | 82 | q31_t * pSrcA, |
emilmont | 1:fdd22bb7aa52 | 83 | uint32_t srcALen, |
emilmont | 1:fdd22bb7aa52 | 84 | q31_t * pSrcB, |
emilmont | 1:fdd22bb7aa52 | 85 | uint32_t srcBLen, |
emilmont | 1:fdd22bb7aa52 | 86 | q31_t * pDst) |
emilmont | 1:fdd22bb7aa52 | 87 | { |
emilmont | 1:fdd22bb7aa52 | 88 | q31_t *pIn1; /* inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 89 | q31_t *pIn2; /* inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 90 | q31_t *pOut = pDst; /* output pointer */ |
emilmont | 1:fdd22bb7aa52 | 91 | q31_t *px; /* Intermediate inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 92 | q31_t *py; /* Intermediate inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 93 | q31_t *pSrc1; /* Intermediate pointers */ |
emilmont | 1:fdd22bb7aa52 | 94 | q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ |
emilmont | 1:fdd22bb7aa52 | 95 | q31_t x0, x1, x2, x3, c0; /* temporary variables for holding input and coefficient values */ |
emilmont | 1:fdd22bb7aa52 | 96 | uint32_t j, k = 0u, count, blkCnt, outBlockSize, blockSize1, blockSize2, blockSize3; /* loop counter */ |
emilmont | 1:fdd22bb7aa52 | 97 | int32_t inc = 1; /* Destination address modifier */ |
emilmont | 1:fdd22bb7aa52 | 98 | |
emilmont | 1:fdd22bb7aa52 | 99 | |
emilmont | 1:fdd22bb7aa52 | 100 | /* The algorithm implementation is based on the lengths of the inputs. */ |
emilmont | 1:fdd22bb7aa52 | 101 | /* srcB is always made to slide across srcA. */ |
emilmont | 1:fdd22bb7aa52 | 102 | /* So srcBLen is always considered as shorter or equal to srcALen */ |
emilmont | 1:fdd22bb7aa52 | 103 | if(srcALen >= srcBLen) |
emilmont | 1:fdd22bb7aa52 | 104 | { |
emilmont | 1:fdd22bb7aa52 | 105 | /* Initialization of inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 106 | pIn1 = (pSrcA); |
emilmont | 1:fdd22bb7aa52 | 107 | |
emilmont | 1:fdd22bb7aa52 | 108 | /* Initialization of inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 109 | pIn2 = (pSrcB); |
emilmont | 1:fdd22bb7aa52 | 110 | |
emilmont | 1:fdd22bb7aa52 | 111 | /* Number of output samples is calculated */ |
emilmont | 1:fdd22bb7aa52 | 112 | outBlockSize = (2u * srcALen) - 1u; |
emilmont | 1:fdd22bb7aa52 | 113 | |
emilmont | 1:fdd22bb7aa52 | 114 | /* When srcALen > srcBLen, zero padding is done to srcB |
emilmont | 1:fdd22bb7aa52 | 115 | * to make their lengths equal. |
emilmont | 1:fdd22bb7aa52 | 116 | * Instead, (outBlockSize - (srcALen + srcBLen - 1)) |
emilmont | 1:fdd22bb7aa52 | 117 | * number of output samples are made zero */ |
emilmont | 1:fdd22bb7aa52 | 118 | j = outBlockSize - (srcALen + (srcBLen - 1u)); |
emilmont | 1:fdd22bb7aa52 | 119 | |
emilmont | 1:fdd22bb7aa52 | 120 | /* Updating the pointer position to non zero value */ |
emilmont | 1:fdd22bb7aa52 | 121 | pOut += j; |
emilmont | 1:fdd22bb7aa52 | 122 | |
emilmont | 1:fdd22bb7aa52 | 123 | } |
emilmont | 1:fdd22bb7aa52 | 124 | else |
emilmont | 1:fdd22bb7aa52 | 125 | { |
emilmont | 1:fdd22bb7aa52 | 126 | /* Initialization of inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 127 | pIn1 = (pSrcB); |
emilmont | 1:fdd22bb7aa52 | 128 | |
emilmont | 1:fdd22bb7aa52 | 129 | /* Initialization of inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 130 | pIn2 = (pSrcA); |
emilmont | 1:fdd22bb7aa52 | 131 | |
emilmont | 1:fdd22bb7aa52 | 132 | /* srcBLen is always considered as shorter or equal to srcALen */ |
emilmont | 1:fdd22bb7aa52 | 133 | j = srcBLen; |
emilmont | 1:fdd22bb7aa52 | 134 | srcBLen = srcALen; |
emilmont | 1:fdd22bb7aa52 | 135 | srcALen = j; |
emilmont | 1:fdd22bb7aa52 | 136 | |
emilmont | 1:fdd22bb7aa52 | 137 | /* CORR(x, y) = Reverse order(CORR(y, x)) */ |
emilmont | 1:fdd22bb7aa52 | 138 | /* Hence set the destination pointer to point to the last output sample */ |
emilmont | 1:fdd22bb7aa52 | 139 | pOut = pDst + ((srcALen + srcBLen) - 2u); |
emilmont | 1:fdd22bb7aa52 | 140 | |
emilmont | 1:fdd22bb7aa52 | 141 | /* Destination address modifier is set to -1 */ |
emilmont | 1:fdd22bb7aa52 | 142 | inc = -1; |
emilmont | 1:fdd22bb7aa52 | 143 | |
emilmont | 1:fdd22bb7aa52 | 144 | } |
emilmont | 1:fdd22bb7aa52 | 145 | |
emilmont | 1:fdd22bb7aa52 | 146 | /* The function is internally |
emilmont | 1:fdd22bb7aa52 | 147 | * divided into three parts according to the number of multiplications that has to be |
emilmont | 1:fdd22bb7aa52 | 148 | * taken place between inputA samples and inputB samples. In the first part of the |
emilmont | 1:fdd22bb7aa52 | 149 | * algorithm, the multiplications increase by one for every iteration. |
emilmont | 1:fdd22bb7aa52 | 150 | * In the second part of the algorithm, srcBLen number of multiplications are done. |
emilmont | 1:fdd22bb7aa52 | 151 | * In the third part of the algorithm, the multiplications decrease by one |
emilmont | 1:fdd22bb7aa52 | 152 | * for every iteration.*/ |
emilmont | 1:fdd22bb7aa52 | 153 | /* The algorithm is implemented in three stages. |
emilmont | 1:fdd22bb7aa52 | 154 | * The loop counters of each stage is initiated here. */ |
emilmont | 1:fdd22bb7aa52 | 155 | blockSize1 = srcBLen - 1u; |
emilmont | 1:fdd22bb7aa52 | 156 | blockSize2 = srcALen - (srcBLen - 1u); |
emilmont | 1:fdd22bb7aa52 | 157 | blockSize3 = blockSize1; |
emilmont | 1:fdd22bb7aa52 | 158 | |
emilmont | 1:fdd22bb7aa52 | 159 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 160 | * Initializations of stage1 |
emilmont | 1:fdd22bb7aa52 | 161 | * -------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 162 | |
emilmont | 1:fdd22bb7aa52 | 163 | /* sum = x[0] * y[srcBlen - 1] |
emilmont | 1:fdd22bb7aa52 | 164 | * sum = x[0] * y[srcBlen - 2] + x[1] * y[srcBlen - 1] |
emilmont | 1:fdd22bb7aa52 | 165 | * .... |
emilmont | 1:fdd22bb7aa52 | 166 | * sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen - 1] * y[srcBLen - 1] |
emilmont | 1:fdd22bb7aa52 | 167 | */ |
emilmont | 1:fdd22bb7aa52 | 168 | |
emilmont | 1:fdd22bb7aa52 | 169 | /* In this stage the MAC operations are increased by 1 for every iteration. |
emilmont | 1:fdd22bb7aa52 | 170 | The count variable holds the number of MAC operations performed */ |
emilmont | 1:fdd22bb7aa52 | 171 | count = 1u; |
emilmont | 1:fdd22bb7aa52 | 172 | |
emilmont | 1:fdd22bb7aa52 | 173 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 174 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 175 | |
emilmont | 1:fdd22bb7aa52 | 176 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 177 | pSrc1 = pIn2 + (srcBLen - 1u); |
emilmont | 1:fdd22bb7aa52 | 178 | py = pSrc1; |
emilmont | 1:fdd22bb7aa52 | 179 | |
emilmont | 1:fdd22bb7aa52 | 180 | /* ------------------------ |
emilmont | 1:fdd22bb7aa52 | 181 | * Stage1 process |
emilmont | 1:fdd22bb7aa52 | 182 | * ----------------------*/ |
emilmont | 1:fdd22bb7aa52 | 183 | |
emilmont | 1:fdd22bb7aa52 | 184 | /* The first stage starts here */ |
emilmont | 1:fdd22bb7aa52 | 185 | while(blockSize1 > 0u) |
emilmont | 1:fdd22bb7aa52 | 186 | { |
emilmont | 1:fdd22bb7aa52 | 187 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 188 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 189 | |
emilmont | 1:fdd22bb7aa52 | 190 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 191 | k = count >> 2; |
emilmont | 1:fdd22bb7aa52 | 192 | |
emilmont | 1:fdd22bb7aa52 | 193 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 194 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 195 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 196 | { |
emilmont | 1:fdd22bb7aa52 | 197 | /* x[0] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 198 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 199 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 200 | /* x[1] * y[srcBLen - 3] */ |
emilmont | 1:fdd22bb7aa52 | 201 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 202 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 203 | /* x[2] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 204 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 205 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 206 | /* x[3] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 207 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 208 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 209 | |
emilmont | 1:fdd22bb7aa52 | 210 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 211 | k--; |
emilmont | 1:fdd22bb7aa52 | 212 | } |
emilmont | 1:fdd22bb7aa52 | 213 | |
emilmont | 1:fdd22bb7aa52 | 214 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 215 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 216 | k = count % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 217 | |
emilmont | 1:fdd22bb7aa52 | 218 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 219 | { |
emilmont | 1:fdd22bb7aa52 | 220 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 221 | /* x[0] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 222 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 223 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 224 | |
emilmont | 1:fdd22bb7aa52 | 225 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 226 | k--; |
emilmont | 1:fdd22bb7aa52 | 227 | } |
emilmont | 1:fdd22bb7aa52 | 228 | |
emilmont | 1:fdd22bb7aa52 | 229 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 230 | *pOut = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 231 | /* Destination pointer is updated according to the address modifier, inc */ |
emilmont | 1:fdd22bb7aa52 | 232 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 233 | |
emilmont | 1:fdd22bb7aa52 | 234 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 235 | py = pSrc1 - count; |
emilmont | 1:fdd22bb7aa52 | 236 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 237 | |
emilmont | 1:fdd22bb7aa52 | 238 | /* Increment the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 239 | count++; |
emilmont | 1:fdd22bb7aa52 | 240 | |
emilmont | 1:fdd22bb7aa52 | 241 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 242 | blockSize1--; |
emilmont | 1:fdd22bb7aa52 | 243 | } |
emilmont | 1:fdd22bb7aa52 | 244 | |
emilmont | 1:fdd22bb7aa52 | 245 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 246 | * Initializations of stage2 |
emilmont | 1:fdd22bb7aa52 | 247 | * ------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 248 | |
emilmont | 1:fdd22bb7aa52 | 249 | /* sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen-1] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 250 | * sum = x[1] * y[0] + x[2] * y[1] +...+ x[srcBLen] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 251 | * .... |
emilmont | 1:fdd22bb7aa52 | 252 | * sum = x[srcALen-srcBLen-2] * y[0] + x[srcALen-srcBLen-1] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 253 | */ |
emilmont | 1:fdd22bb7aa52 | 254 | |
emilmont | 1:fdd22bb7aa52 | 255 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 256 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 257 | |
emilmont | 1:fdd22bb7aa52 | 258 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 259 | py = pIn2; |
emilmont | 1:fdd22bb7aa52 | 260 | |
emilmont | 1:fdd22bb7aa52 | 261 | /* count is index by which the pointer pIn1 to be incremented */ |
emilmont | 1:fdd22bb7aa52 | 262 | count = 0u; |
emilmont | 1:fdd22bb7aa52 | 263 | |
emilmont | 1:fdd22bb7aa52 | 264 | /* ------------------- |
emilmont | 1:fdd22bb7aa52 | 265 | * Stage2 process |
emilmont | 1:fdd22bb7aa52 | 266 | * ------------------*/ |
emilmont | 1:fdd22bb7aa52 | 267 | |
emilmont | 1:fdd22bb7aa52 | 268 | /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed. |
emilmont | 1:fdd22bb7aa52 | 269 | * So, to loop unroll over blockSize2, |
emilmont | 1:fdd22bb7aa52 | 270 | * srcBLen should be greater than or equal to 4 */ |
emilmont | 1:fdd22bb7aa52 | 271 | if(srcBLen >= 4u) |
emilmont | 1:fdd22bb7aa52 | 272 | { |
emilmont | 1:fdd22bb7aa52 | 273 | /* Loop unroll over blockSize2, by 4 */ |
emilmont | 1:fdd22bb7aa52 | 274 | blkCnt = blockSize2 >> 2u; |
emilmont | 1:fdd22bb7aa52 | 275 | |
emilmont | 1:fdd22bb7aa52 | 276 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 277 | { |
emilmont | 1:fdd22bb7aa52 | 278 | /* Set all accumulators to zero */ |
emilmont | 1:fdd22bb7aa52 | 279 | acc0 = 0; |
emilmont | 1:fdd22bb7aa52 | 280 | acc1 = 0; |
emilmont | 1:fdd22bb7aa52 | 281 | acc2 = 0; |
emilmont | 1:fdd22bb7aa52 | 282 | acc3 = 0; |
emilmont | 1:fdd22bb7aa52 | 283 | |
emilmont | 1:fdd22bb7aa52 | 284 | /* read x[0], x[1], x[2] samples */ |
emilmont | 1:fdd22bb7aa52 | 285 | x0 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 286 | x1 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 287 | x2 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 288 | |
emilmont | 1:fdd22bb7aa52 | 289 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 290 | k = srcBLen >> 2u; |
emilmont | 1:fdd22bb7aa52 | 291 | |
emilmont | 1:fdd22bb7aa52 | 292 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 293 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 294 | do |
emilmont | 1:fdd22bb7aa52 | 295 | { |
emilmont | 1:fdd22bb7aa52 | 296 | /* Read y[0] sample */ |
emilmont | 1:fdd22bb7aa52 | 297 | c0 = *(py++); |
emilmont | 1:fdd22bb7aa52 | 298 | |
emilmont | 1:fdd22bb7aa52 | 299 | /* Read x[3] sample */ |
emilmont | 1:fdd22bb7aa52 | 300 | x3 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 301 | |
emilmont | 1:fdd22bb7aa52 | 302 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 303 | /* acc0 += x[0] * y[0] */ |
emilmont | 1:fdd22bb7aa52 | 304 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 305 | /* acc1 += x[1] * y[0] */ |
emilmont | 1:fdd22bb7aa52 | 306 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 307 | /* acc2 += x[2] * y[0] */ |
emilmont | 1:fdd22bb7aa52 | 308 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 309 | /* acc3 += x[3] * y[0] */ |
emilmont | 1:fdd22bb7aa52 | 310 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 311 | |
emilmont | 1:fdd22bb7aa52 | 312 | /* Read y[1] sample */ |
emilmont | 1:fdd22bb7aa52 | 313 | c0 = *(py++); |
emilmont | 1:fdd22bb7aa52 | 314 | |
emilmont | 1:fdd22bb7aa52 | 315 | /* Read x[4] sample */ |
emilmont | 1:fdd22bb7aa52 | 316 | x0 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 317 | |
emilmont | 1:fdd22bb7aa52 | 318 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 319 | /* acc0 += x[1] * y[1] */ |
emilmont | 1:fdd22bb7aa52 | 320 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 321 | /* acc1 += x[2] * y[1] */ |
emilmont | 1:fdd22bb7aa52 | 322 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 323 | /* acc2 += x[3] * y[1] */ |
emilmont | 1:fdd22bb7aa52 | 324 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 325 | /* acc3 += x[4] * y[1] */ |
emilmont | 1:fdd22bb7aa52 | 326 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 327 | |
emilmont | 1:fdd22bb7aa52 | 328 | /* Read y[2] sample */ |
emilmont | 1:fdd22bb7aa52 | 329 | c0 = *(py++); |
emilmont | 1:fdd22bb7aa52 | 330 | |
emilmont | 1:fdd22bb7aa52 | 331 | /* Read x[5] sample */ |
emilmont | 1:fdd22bb7aa52 | 332 | x1 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 333 | |
emilmont | 1:fdd22bb7aa52 | 334 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 335 | /* acc0 += x[2] * y[2] */ |
emilmont | 1:fdd22bb7aa52 | 336 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 337 | /* acc1 += x[3] * y[2] */ |
emilmont | 1:fdd22bb7aa52 | 338 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 339 | /* acc2 += x[4] * y[2] */ |
emilmont | 1:fdd22bb7aa52 | 340 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 341 | /* acc3 += x[5] * y[2] */ |
emilmont | 1:fdd22bb7aa52 | 342 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 343 | |
emilmont | 1:fdd22bb7aa52 | 344 | /* Read y[3] sample */ |
emilmont | 1:fdd22bb7aa52 | 345 | c0 = *(py++); |
emilmont | 1:fdd22bb7aa52 | 346 | |
emilmont | 1:fdd22bb7aa52 | 347 | /* Read x[6] sample */ |
emilmont | 1:fdd22bb7aa52 | 348 | x2 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 349 | |
emilmont | 1:fdd22bb7aa52 | 350 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 351 | /* acc0 += x[3] * y[3] */ |
emilmont | 1:fdd22bb7aa52 | 352 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 353 | /* acc1 += x[4] * y[3] */ |
emilmont | 1:fdd22bb7aa52 | 354 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 355 | /* acc2 += x[5] * y[3] */ |
emilmont | 1:fdd22bb7aa52 | 356 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 357 | /* acc3 += x[6] * y[3] */ |
emilmont | 1:fdd22bb7aa52 | 358 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 359 | |
emilmont | 1:fdd22bb7aa52 | 360 | |
emilmont | 1:fdd22bb7aa52 | 361 | } while(--k); |
emilmont | 1:fdd22bb7aa52 | 362 | |
emilmont | 1:fdd22bb7aa52 | 363 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 364 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 365 | k = srcBLen % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 366 | |
emilmont | 1:fdd22bb7aa52 | 367 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 368 | { |
emilmont | 1:fdd22bb7aa52 | 369 | /* Read y[4] sample */ |
emilmont | 1:fdd22bb7aa52 | 370 | c0 = *(py++); |
emilmont | 1:fdd22bb7aa52 | 371 | |
emilmont | 1:fdd22bb7aa52 | 372 | /* Read x[7] sample */ |
emilmont | 1:fdd22bb7aa52 | 373 | x3 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 374 | |
emilmont | 1:fdd22bb7aa52 | 375 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 376 | /* acc0 += x[4] * y[4] */ |
emilmont | 1:fdd22bb7aa52 | 377 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 378 | /* acc1 += x[5] * y[4] */ |
emilmont | 1:fdd22bb7aa52 | 379 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 380 | /* acc2 += x[6] * y[4] */ |
emilmont | 1:fdd22bb7aa52 | 381 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 382 | /* acc3 += x[7] * y[4] */ |
emilmont | 1:fdd22bb7aa52 | 383 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 384 | |
emilmont | 1:fdd22bb7aa52 | 385 | /* Reuse the present samples for the next MAC */ |
emilmont | 1:fdd22bb7aa52 | 386 | x0 = x1; |
emilmont | 1:fdd22bb7aa52 | 387 | x1 = x2; |
emilmont | 1:fdd22bb7aa52 | 388 | x2 = x3; |
emilmont | 1:fdd22bb7aa52 | 389 | |
emilmont | 1:fdd22bb7aa52 | 390 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 391 | k--; |
emilmont | 1:fdd22bb7aa52 | 392 | } |
emilmont | 1:fdd22bb7aa52 | 393 | |
emilmont | 1:fdd22bb7aa52 | 394 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 395 | *pOut = (q31_t) (acc0 << 1); |
emilmont | 1:fdd22bb7aa52 | 396 | /* Destination pointer is updated according to the address modifier, inc */ |
emilmont | 1:fdd22bb7aa52 | 397 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 398 | |
emilmont | 1:fdd22bb7aa52 | 399 | *pOut = (q31_t) (acc1 << 1); |
emilmont | 1:fdd22bb7aa52 | 400 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 401 | |
emilmont | 1:fdd22bb7aa52 | 402 | *pOut = (q31_t) (acc2 << 1); |
emilmont | 1:fdd22bb7aa52 | 403 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 404 | |
emilmont | 1:fdd22bb7aa52 | 405 | *pOut = (q31_t) (acc3 << 1); |
emilmont | 1:fdd22bb7aa52 | 406 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 407 | |
emilmont | 1:fdd22bb7aa52 | 408 | /* Increment the pointer pIn1 index, count by 4 */ |
emilmont | 1:fdd22bb7aa52 | 409 | count += 4u; |
emilmont | 1:fdd22bb7aa52 | 410 | |
emilmont | 1:fdd22bb7aa52 | 411 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 412 | px = pIn1 + count; |
emilmont | 1:fdd22bb7aa52 | 413 | py = pIn2; |
emilmont | 1:fdd22bb7aa52 | 414 | |
emilmont | 1:fdd22bb7aa52 | 415 | |
emilmont | 1:fdd22bb7aa52 | 416 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 417 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 418 | } |
emilmont | 1:fdd22bb7aa52 | 419 | |
emilmont | 1:fdd22bb7aa52 | 420 | /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here. |
emilmont | 1:fdd22bb7aa52 | 421 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 422 | blkCnt = blockSize2 % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 423 | |
emilmont | 1:fdd22bb7aa52 | 424 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 425 | { |
emilmont | 1:fdd22bb7aa52 | 426 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 427 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 428 | |
emilmont | 1:fdd22bb7aa52 | 429 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 430 | k = srcBLen >> 2u; |
emilmont | 1:fdd22bb7aa52 | 431 | |
emilmont | 1:fdd22bb7aa52 | 432 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 433 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 434 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 435 | { |
emilmont | 1:fdd22bb7aa52 | 436 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 437 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 438 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 439 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 440 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 441 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 442 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 443 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 444 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 445 | |
emilmont | 1:fdd22bb7aa52 | 446 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 447 | k--; |
emilmont | 1:fdd22bb7aa52 | 448 | } |
emilmont | 1:fdd22bb7aa52 | 449 | |
emilmont | 1:fdd22bb7aa52 | 450 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 451 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 452 | k = srcBLen % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 453 | |
emilmont | 1:fdd22bb7aa52 | 454 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 455 | { |
emilmont | 1:fdd22bb7aa52 | 456 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 457 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 458 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 459 | |
emilmont | 1:fdd22bb7aa52 | 460 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 461 | k--; |
emilmont | 1:fdd22bb7aa52 | 462 | } |
emilmont | 1:fdd22bb7aa52 | 463 | |
emilmont | 1:fdd22bb7aa52 | 464 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 465 | *pOut = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 466 | /* Destination pointer is updated according to the address modifier, inc */ |
emilmont | 1:fdd22bb7aa52 | 467 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 468 | |
emilmont | 1:fdd22bb7aa52 | 469 | /* Increment the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 470 | count++; |
emilmont | 1:fdd22bb7aa52 | 471 | |
emilmont | 1:fdd22bb7aa52 | 472 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 473 | px = pIn1 + count; |
emilmont | 1:fdd22bb7aa52 | 474 | py = pIn2; |
emilmont | 1:fdd22bb7aa52 | 475 | |
emilmont | 1:fdd22bb7aa52 | 476 | |
emilmont | 1:fdd22bb7aa52 | 477 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 478 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 479 | } |
emilmont | 1:fdd22bb7aa52 | 480 | } |
emilmont | 1:fdd22bb7aa52 | 481 | else |
emilmont | 1:fdd22bb7aa52 | 482 | { |
emilmont | 1:fdd22bb7aa52 | 483 | /* If the srcBLen is not a multiple of 4, |
emilmont | 1:fdd22bb7aa52 | 484 | * the blockSize2 loop cannot be unrolled by 4 */ |
emilmont | 1:fdd22bb7aa52 | 485 | blkCnt = blockSize2; |
emilmont | 1:fdd22bb7aa52 | 486 | |
emilmont | 1:fdd22bb7aa52 | 487 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 488 | { |
emilmont | 1:fdd22bb7aa52 | 489 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 490 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 491 | |
emilmont | 1:fdd22bb7aa52 | 492 | /* Loop over srcBLen */ |
emilmont | 1:fdd22bb7aa52 | 493 | k = srcBLen; |
emilmont | 1:fdd22bb7aa52 | 494 | |
emilmont | 1:fdd22bb7aa52 | 495 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 496 | { |
emilmont | 1:fdd22bb7aa52 | 497 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 498 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 499 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 500 | |
emilmont | 1:fdd22bb7aa52 | 501 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 502 | k--; |
emilmont | 1:fdd22bb7aa52 | 503 | } |
emilmont | 1:fdd22bb7aa52 | 504 | |
emilmont | 1:fdd22bb7aa52 | 505 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 506 | *pOut = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 507 | /* Destination pointer is updated according to the address modifier, inc */ |
emilmont | 1:fdd22bb7aa52 | 508 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 509 | |
emilmont | 1:fdd22bb7aa52 | 510 | /* Increment the MAC count */ |
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 = pIn2; |
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 | |
emilmont | 1:fdd22bb7aa52 | 522 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 523 | * Initializations of stage3 |
emilmont | 1:fdd22bb7aa52 | 524 | * -------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 525 | |
emilmont | 1:fdd22bb7aa52 | 526 | /* sum += x[srcALen-srcBLen+1] * y[0] + x[srcALen-srcBLen+2] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 527 | * sum += x[srcALen-srcBLen+2] * y[0] + x[srcALen-srcBLen+3] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 528 | * .... |
emilmont | 1:fdd22bb7aa52 | 529 | * sum += x[srcALen-2] * y[0] + x[srcALen-1] * y[1] |
emilmont | 1:fdd22bb7aa52 | 530 | * sum += x[srcALen-1] * y[0] |
emilmont | 1:fdd22bb7aa52 | 531 | */ |
emilmont | 1:fdd22bb7aa52 | 532 | |
emilmont | 1:fdd22bb7aa52 | 533 | /* In this stage the MAC operations are decreased by 1 for every iteration. |
emilmont | 1:fdd22bb7aa52 | 534 | The count variable holds the number of MAC operations performed */ |
emilmont | 1:fdd22bb7aa52 | 535 | count = srcBLen - 1u; |
emilmont | 1:fdd22bb7aa52 | 536 | |
emilmont | 1:fdd22bb7aa52 | 537 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 538 | pSrc1 = ((pIn1 + srcALen) - srcBLen) + 1u; |
emilmont | 1:fdd22bb7aa52 | 539 | px = pSrc1; |
emilmont | 1:fdd22bb7aa52 | 540 | |
emilmont | 1:fdd22bb7aa52 | 541 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 542 | py = pIn2; |
emilmont | 1:fdd22bb7aa52 | 543 | |
emilmont | 1:fdd22bb7aa52 | 544 | /* ------------------- |
emilmont | 1:fdd22bb7aa52 | 545 | * Stage3 process |
emilmont | 1:fdd22bb7aa52 | 546 | * ------------------*/ |
emilmont | 1:fdd22bb7aa52 | 547 | |
emilmont | 1:fdd22bb7aa52 | 548 | while(blockSize3 > 0u) |
emilmont | 1:fdd22bb7aa52 | 549 | { |
emilmont | 1:fdd22bb7aa52 | 550 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 551 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 552 | |
emilmont | 1:fdd22bb7aa52 | 553 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 554 | k = count >> 2u; |
emilmont | 1:fdd22bb7aa52 | 555 | |
emilmont | 1:fdd22bb7aa52 | 556 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 557 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 558 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 559 | { |
emilmont | 1:fdd22bb7aa52 | 560 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 561 | /* sum += x[srcALen - srcBLen + 4] * y[3] */ |
emilmont | 1:fdd22bb7aa52 | 562 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 563 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 564 | /* sum += x[srcALen - srcBLen + 3] * y[2] */ |
emilmont | 1:fdd22bb7aa52 | 565 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 566 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 567 | /* sum += x[srcALen - srcBLen + 2] * y[1] */ |
emilmont | 1:fdd22bb7aa52 | 568 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 569 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 570 | /* sum += x[srcALen - srcBLen + 1] * y[0] */ |
emilmont | 1:fdd22bb7aa52 | 571 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 572 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 573 | |
emilmont | 1:fdd22bb7aa52 | 574 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 575 | k--; |
emilmont | 1:fdd22bb7aa52 | 576 | } |
emilmont | 1:fdd22bb7aa52 | 577 | |
emilmont | 1:fdd22bb7aa52 | 578 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 579 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 580 | k = count % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 581 | |
emilmont | 1:fdd22bb7aa52 | 582 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 583 | { |
emilmont | 1:fdd22bb7aa52 | 584 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 585 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 586 | ((q63_t) * px++ * (*py++))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 587 | |
emilmont | 1:fdd22bb7aa52 | 588 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 589 | k--; |
emilmont | 1:fdd22bb7aa52 | 590 | } |
emilmont | 1:fdd22bb7aa52 | 591 | |
emilmont | 1:fdd22bb7aa52 | 592 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 593 | *pOut = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 594 | /* Destination pointer is updated according to the address modifier, inc */ |
emilmont | 1:fdd22bb7aa52 | 595 | pOut += inc; |
emilmont | 1:fdd22bb7aa52 | 596 | |
emilmont | 1:fdd22bb7aa52 | 597 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 598 | px = ++pSrc1; |
emilmont | 1:fdd22bb7aa52 | 599 | py = pIn2; |
emilmont | 1:fdd22bb7aa52 | 600 | |
emilmont | 1:fdd22bb7aa52 | 601 | /* Decrement the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 602 | count--; |
emilmont | 1:fdd22bb7aa52 | 603 | |
emilmont | 1:fdd22bb7aa52 | 604 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 605 | blockSize3--; |
emilmont | 1:fdd22bb7aa52 | 606 | } |
emilmont | 1:fdd22bb7aa52 | 607 | |
emilmont | 1:fdd22bb7aa52 | 608 | } |
emilmont | 1:fdd22bb7aa52 | 609 | |
emilmont | 1:fdd22bb7aa52 | 610 | /** |
emilmont | 1:fdd22bb7aa52 | 611 | * @} end of Corr group |
emilmont | 1:fdd22bb7aa52 | 612 | */ |