CMSIS DSP library
Dependents: KL25Z_FFT_Demo Hat_Board_v5_1 KL25Z_FFT_Demo_tony KL25Z_FFT_Demo_tony ... more
Fork of mbed-dsp by
cmsis_dsp/FilteringFunctions/arm_conv_partial_fast_q31.c@3:7a284390b0ce, 2013-11-08 (annotated)
- Committer:
- mbed_official
- Date:
- Fri Nov 08 13:45:10 2013 +0000
- Revision:
- 3:7a284390b0ce
- Parent:
- 2:da51fb522205
Synchronized with git revision e69956aba2f68a2a26ac26b051f8d349deaa1ce8
Who changed what in which revision?
User | Revision | Line number | New contents of line |
---|---|---|---|
emilmont | 1:fdd22bb7aa52 | 1 | /* ---------------------------------------------------------------------- |
mbed_official | 3:7a284390b0ce | 2 | * Copyright (C) 2010-2013 ARM Limited. All rights reserved. |
emilmont | 1:fdd22bb7aa52 | 3 | * |
mbed_official | 3:7a284390b0ce | 4 | * $Date: 17. January 2013 |
mbed_official | 3:7a284390b0ce | 5 | * $Revision: V1.4.1 |
emilmont | 1:fdd22bb7aa52 | 6 | * |
emilmont | 2:da51fb522205 | 7 | * Project: CMSIS DSP Library |
emilmont | 2:da51fb522205 | 8 | * Title: arm_conv_partial_fast_q31.c |
emilmont | 1:fdd22bb7aa52 | 9 | * |
emilmont | 2:da51fb522205 | 10 | * Description: Fast Q31 Partial convolution. |
emilmont | 1:fdd22bb7aa52 | 11 | * |
emilmont | 1:fdd22bb7aa52 | 12 | * Target Processor: Cortex-M4/Cortex-M3 |
emilmont | 1:fdd22bb7aa52 | 13 | * |
mbed_official | 3:7a284390b0ce | 14 | * Redistribution and use in source and binary forms, with or without |
mbed_official | 3:7a284390b0ce | 15 | * modification, are permitted provided that the following conditions |
mbed_official | 3:7a284390b0ce | 16 | * are met: |
mbed_official | 3:7a284390b0ce | 17 | * - Redistributions of source code must retain the above copyright |
mbed_official | 3:7a284390b0ce | 18 | * notice, this list of conditions and the following disclaimer. |
mbed_official | 3:7a284390b0ce | 19 | * - Redistributions in binary form must reproduce the above copyright |
mbed_official | 3:7a284390b0ce | 20 | * notice, this list of conditions and the following disclaimer in |
mbed_official | 3:7a284390b0ce | 21 | * the documentation and/or other materials provided with the |
mbed_official | 3:7a284390b0ce | 22 | * distribution. |
mbed_official | 3:7a284390b0ce | 23 | * - Neither the name of ARM LIMITED nor the names of its contributors |
mbed_official | 3:7a284390b0ce | 24 | * may be used to endorse or promote products derived from this |
mbed_official | 3:7a284390b0ce | 25 | * software without specific prior written permission. |
mbed_official | 3:7a284390b0ce | 26 | * |
mbed_official | 3:7a284390b0ce | 27 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
mbed_official | 3:7a284390b0ce | 28 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
mbed_official | 3:7a284390b0ce | 29 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
mbed_official | 3:7a284390b0ce | 30 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
mbed_official | 3:7a284390b0ce | 31 | * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
mbed_official | 3:7a284390b0ce | 32 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, |
mbed_official | 3:7a284390b0ce | 33 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
mbed_official | 3:7a284390b0ce | 34 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
mbed_official | 3:7a284390b0ce | 35 | * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
mbed_official | 3:7a284390b0ce | 36 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN |
mbed_official | 3:7a284390b0ce | 37 | * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
mbed_official | 3:7a284390b0ce | 38 | * POSSIBILITY OF SUCH DAMAGE. |
emilmont | 1:fdd22bb7aa52 | 39 | * -------------------------------------------------------------------- */ |
emilmont | 1:fdd22bb7aa52 | 40 | |
emilmont | 1:fdd22bb7aa52 | 41 | #include "arm_math.h" |
emilmont | 1:fdd22bb7aa52 | 42 | |
emilmont | 1:fdd22bb7aa52 | 43 | /** |
emilmont | 1:fdd22bb7aa52 | 44 | * @ingroup groupFilters |
emilmont | 1:fdd22bb7aa52 | 45 | */ |
emilmont | 1:fdd22bb7aa52 | 46 | |
emilmont | 1:fdd22bb7aa52 | 47 | /** |
emilmont | 1:fdd22bb7aa52 | 48 | * @addtogroup PartialConv |
emilmont | 1:fdd22bb7aa52 | 49 | * @{ |
emilmont | 1:fdd22bb7aa52 | 50 | */ |
emilmont | 1:fdd22bb7aa52 | 51 | |
emilmont | 1:fdd22bb7aa52 | 52 | /** |
emilmont | 1:fdd22bb7aa52 | 53 | * @brief Partial convolution of 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. |
emilmont | 1:fdd22bb7aa52 | 59 | * @param[in] firstIndex is the first output sample to start with. |
emilmont | 1:fdd22bb7aa52 | 60 | * @param[in] numPoints is the number of output points to be computed. |
emilmont | 1:fdd22bb7aa52 | 61 | * @return Returns either ARM_MATH_SUCCESS if the function completed correctly or ARM_MATH_ARGUMENT_ERROR if the requested subset is not in the range [0 srcALen+srcBLen-2]. |
emilmont | 1:fdd22bb7aa52 | 62 | * |
emilmont | 1:fdd22bb7aa52 | 63 | * \par |
emilmont | 1:fdd22bb7aa52 | 64 | * See <code>arm_conv_partial_q31()</code> for a slower implementation of this function which uses a 64-bit accumulator to provide higher precision. |
emilmont | 1:fdd22bb7aa52 | 65 | */ |
emilmont | 1:fdd22bb7aa52 | 66 | |
emilmont | 1:fdd22bb7aa52 | 67 | arm_status arm_conv_partial_fast_q31( |
emilmont | 1:fdd22bb7aa52 | 68 | q31_t * pSrcA, |
emilmont | 1:fdd22bb7aa52 | 69 | uint32_t srcALen, |
emilmont | 1:fdd22bb7aa52 | 70 | q31_t * pSrcB, |
emilmont | 1:fdd22bb7aa52 | 71 | uint32_t srcBLen, |
emilmont | 1:fdd22bb7aa52 | 72 | q31_t * pDst, |
emilmont | 1:fdd22bb7aa52 | 73 | uint32_t firstIndex, |
emilmont | 1:fdd22bb7aa52 | 74 | uint32_t numPoints) |
emilmont | 1:fdd22bb7aa52 | 75 | { |
emilmont | 1:fdd22bb7aa52 | 76 | q31_t *pIn1; /* inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 77 | q31_t *pIn2; /* inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 78 | q31_t *pOut = pDst; /* output pointer */ |
emilmont | 1:fdd22bb7aa52 | 79 | q31_t *px; /* Intermediate inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 80 | q31_t *py; /* Intermediate inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 81 | q31_t *pSrc1, *pSrc2; /* Intermediate pointers */ |
emilmont | 1:fdd22bb7aa52 | 82 | q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ |
emilmont | 1:fdd22bb7aa52 | 83 | q31_t x0, x1, x2, x3, c0; |
emilmont | 1:fdd22bb7aa52 | 84 | uint32_t j, k, count, check, blkCnt; |
emilmont | 1:fdd22bb7aa52 | 85 | int32_t blockSize1, blockSize2, blockSize3; /* loop counters */ |
emilmont | 1:fdd22bb7aa52 | 86 | arm_status status; /* status of Partial convolution */ |
emilmont | 1:fdd22bb7aa52 | 87 | |
emilmont | 1:fdd22bb7aa52 | 88 | |
emilmont | 1:fdd22bb7aa52 | 89 | /* Check for range of output samples to be calculated */ |
emilmont | 1:fdd22bb7aa52 | 90 | if((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u)))) |
emilmont | 1:fdd22bb7aa52 | 91 | { |
emilmont | 1:fdd22bb7aa52 | 92 | /* Set status as ARM_MATH_ARGUMENT_ERROR */ |
emilmont | 1:fdd22bb7aa52 | 93 | status = ARM_MATH_ARGUMENT_ERROR; |
emilmont | 1:fdd22bb7aa52 | 94 | } |
emilmont | 1:fdd22bb7aa52 | 95 | else |
emilmont | 1:fdd22bb7aa52 | 96 | { |
emilmont | 1:fdd22bb7aa52 | 97 | |
emilmont | 1:fdd22bb7aa52 | 98 | /* The algorithm implementation is based on the lengths of the inputs. */ |
emilmont | 1:fdd22bb7aa52 | 99 | /* srcB is always made to slide across srcA. */ |
emilmont | 1:fdd22bb7aa52 | 100 | /* So srcBLen is always considered as shorter or equal to srcALen */ |
emilmont | 1:fdd22bb7aa52 | 101 | if(srcALen >= srcBLen) |
emilmont | 1:fdd22bb7aa52 | 102 | { |
emilmont | 1:fdd22bb7aa52 | 103 | /* Initialization of inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 104 | pIn1 = pSrcA; |
emilmont | 1:fdd22bb7aa52 | 105 | |
emilmont | 1:fdd22bb7aa52 | 106 | /* Initialization of inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 107 | pIn2 = pSrcB; |
emilmont | 1:fdd22bb7aa52 | 108 | } |
emilmont | 1:fdd22bb7aa52 | 109 | else |
emilmont | 1:fdd22bb7aa52 | 110 | { |
emilmont | 1:fdd22bb7aa52 | 111 | /* Initialization of inputA pointer */ |
emilmont | 1:fdd22bb7aa52 | 112 | pIn1 = pSrcB; |
emilmont | 1:fdd22bb7aa52 | 113 | |
emilmont | 1:fdd22bb7aa52 | 114 | /* Initialization of inputB pointer */ |
emilmont | 1:fdd22bb7aa52 | 115 | pIn2 = pSrcA; |
emilmont | 1:fdd22bb7aa52 | 116 | |
emilmont | 1:fdd22bb7aa52 | 117 | /* srcBLen is always considered as shorter or equal to srcALen */ |
emilmont | 1:fdd22bb7aa52 | 118 | j = srcBLen; |
emilmont | 1:fdd22bb7aa52 | 119 | srcBLen = srcALen; |
emilmont | 1:fdd22bb7aa52 | 120 | srcALen = j; |
emilmont | 1:fdd22bb7aa52 | 121 | } |
emilmont | 1:fdd22bb7aa52 | 122 | |
emilmont | 1:fdd22bb7aa52 | 123 | /* Conditions to check which loopCounter holds |
emilmont | 1:fdd22bb7aa52 | 124 | * the first and last indices of the output samples to be calculated. */ |
emilmont | 1:fdd22bb7aa52 | 125 | check = firstIndex + numPoints; |
emilmont | 1:fdd22bb7aa52 | 126 | blockSize3 = ((int32_t) check - (int32_t) srcALen); |
emilmont | 1:fdd22bb7aa52 | 127 | blockSize3 = (blockSize3 > 0) ? blockSize3 : 0; |
emilmont | 1:fdd22bb7aa52 | 128 | blockSize1 = (((int32_t) srcBLen - 1) - (int32_t) firstIndex); |
emilmont | 1:fdd22bb7aa52 | 129 | blockSize1 = (blockSize1 > 0) ? ((check > (srcBLen - 1u)) ? blockSize1 : |
emilmont | 1:fdd22bb7aa52 | 130 | (int32_t) numPoints) : 0; |
emilmont | 1:fdd22bb7aa52 | 131 | blockSize2 = (int32_t) check - ((blockSize3 + blockSize1) + |
emilmont | 1:fdd22bb7aa52 | 132 | (int32_t) firstIndex); |
emilmont | 1:fdd22bb7aa52 | 133 | blockSize2 = (blockSize2 > 0) ? blockSize2 : 0; |
emilmont | 1:fdd22bb7aa52 | 134 | |
emilmont | 1:fdd22bb7aa52 | 135 | /* 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 | 136 | /* The function is internally |
emilmont | 1:fdd22bb7aa52 | 137 | * divided into three stages according to the number of multiplications that has to be |
emilmont | 1:fdd22bb7aa52 | 138 | * taken place between inputA samples and inputB samples. In the first stage of the |
emilmont | 1:fdd22bb7aa52 | 139 | * algorithm, the multiplications increase by one for every iteration. |
emilmont | 1:fdd22bb7aa52 | 140 | * In the second stage of the algorithm, srcBLen number of multiplications are done. |
emilmont | 1:fdd22bb7aa52 | 141 | * In the third stage of the algorithm, the multiplications decrease by one |
emilmont | 1:fdd22bb7aa52 | 142 | * for every iteration. */ |
emilmont | 1:fdd22bb7aa52 | 143 | |
emilmont | 1:fdd22bb7aa52 | 144 | /* Set the output pointer to point to the firstIndex |
emilmont | 1:fdd22bb7aa52 | 145 | * of the output sample to be calculated. */ |
emilmont | 1:fdd22bb7aa52 | 146 | pOut = pDst + firstIndex; |
emilmont | 1:fdd22bb7aa52 | 147 | |
emilmont | 1:fdd22bb7aa52 | 148 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 149 | * Initializations of stage1 |
emilmont | 1:fdd22bb7aa52 | 150 | * -------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 151 | |
emilmont | 1:fdd22bb7aa52 | 152 | /* sum = x[0] * y[0] |
emilmont | 1:fdd22bb7aa52 | 153 | * sum = x[0] * y[1] + x[1] * y[0] |
emilmont | 1:fdd22bb7aa52 | 154 | * .... |
emilmont | 1:fdd22bb7aa52 | 155 | * sum = x[0] * y[srcBlen - 1] + x[1] * y[srcBlen - 2] +...+ x[srcBLen - 1] * y[0] |
emilmont | 1:fdd22bb7aa52 | 156 | */ |
emilmont | 1:fdd22bb7aa52 | 157 | |
emilmont | 1:fdd22bb7aa52 | 158 | /* In this stage the MAC operations are increased by 1 for every iteration. |
emilmont | 1:fdd22bb7aa52 | 159 | The count variable holds the number of MAC operations performed. |
emilmont | 1:fdd22bb7aa52 | 160 | Since the partial convolution starts from firstIndex |
emilmont | 1:fdd22bb7aa52 | 161 | Number of Macs to be performed is firstIndex + 1 */ |
emilmont | 1:fdd22bb7aa52 | 162 | count = 1u + firstIndex; |
emilmont | 1:fdd22bb7aa52 | 163 | |
emilmont | 1:fdd22bb7aa52 | 164 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 165 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 166 | |
emilmont | 1:fdd22bb7aa52 | 167 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 168 | pSrc2 = pIn2 + firstIndex; |
emilmont | 1:fdd22bb7aa52 | 169 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 170 | |
emilmont | 1:fdd22bb7aa52 | 171 | /* ------------------------ |
emilmont | 1:fdd22bb7aa52 | 172 | * Stage1 process |
emilmont | 1:fdd22bb7aa52 | 173 | * ----------------------*/ |
emilmont | 1:fdd22bb7aa52 | 174 | |
emilmont | 1:fdd22bb7aa52 | 175 | /* The first loop starts here */ |
emilmont | 1:fdd22bb7aa52 | 176 | while(blockSize1 > 0) |
emilmont | 1:fdd22bb7aa52 | 177 | { |
emilmont | 1:fdd22bb7aa52 | 178 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 179 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 180 | |
emilmont | 1:fdd22bb7aa52 | 181 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 182 | k = count >> 2u; |
emilmont | 1:fdd22bb7aa52 | 183 | |
emilmont | 1:fdd22bb7aa52 | 184 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 185 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 186 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 187 | { |
emilmont | 1:fdd22bb7aa52 | 188 | /* x[0] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 189 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 190 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 191 | |
emilmont | 1:fdd22bb7aa52 | 192 | /* x[1] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 193 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 194 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 195 | |
emilmont | 1:fdd22bb7aa52 | 196 | /* x[2] * y[srcBLen - 3] */ |
emilmont | 1:fdd22bb7aa52 | 197 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 198 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 199 | |
emilmont | 1:fdd22bb7aa52 | 200 | /* x[3] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 201 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 202 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 203 | |
emilmont | 1:fdd22bb7aa52 | 204 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 205 | k--; |
emilmont | 1:fdd22bb7aa52 | 206 | } |
emilmont | 1:fdd22bb7aa52 | 207 | |
emilmont | 1:fdd22bb7aa52 | 208 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 209 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 210 | k = count % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 211 | |
emilmont | 1:fdd22bb7aa52 | 212 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 213 | { |
emilmont | 1:fdd22bb7aa52 | 214 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 215 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 216 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 217 | |
emilmont | 1:fdd22bb7aa52 | 218 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 219 | k--; |
emilmont | 1:fdd22bb7aa52 | 220 | } |
emilmont | 1:fdd22bb7aa52 | 221 | |
emilmont | 1:fdd22bb7aa52 | 222 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 223 | *pOut++ = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 224 | |
emilmont | 1:fdd22bb7aa52 | 225 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 226 | py = ++pSrc2; |
emilmont | 1:fdd22bb7aa52 | 227 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 228 | |
emilmont | 1:fdd22bb7aa52 | 229 | /* Increment the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 230 | count++; |
emilmont | 1:fdd22bb7aa52 | 231 | |
emilmont | 1:fdd22bb7aa52 | 232 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 233 | blockSize1--; |
emilmont | 1:fdd22bb7aa52 | 234 | } |
emilmont | 1:fdd22bb7aa52 | 235 | |
emilmont | 1:fdd22bb7aa52 | 236 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 237 | * Initializations of stage2 |
emilmont | 1:fdd22bb7aa52 | 238 | * ------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 239 | |
emilmont | 1:fdd22bb7aa52 | 240 | /* sum = x[0] * y[srcBLen-1] + x[1] * y[srcBLen-2] +...+ x[srcBLen-1] * y[0] |
emilmont | 1:fdd22bb7aa52 | 241 | * sum = x[1] * y[srcBLen-1] + x[2] * y[srcBLen-2] +...+ x[srcBLen] * y[0] |
emilmont | 1:fdd22bb7aa52 | 242 | * .... |
emilmont | 1:fdd22bb7aa52 | 243 | * sum = x[srcALen-srcBLen-2] * y[srcBLen-1] + x[srcALen] * y[srcBLen-2] +...+ x[srcALen-1] * y[0] |
emilmont | 1:fdd22bb7aa52 | 244 | */ |
emilmont | 1:fdd22bb7aa52 | 245 | |
emilmont | 1:fdd22bb7aa52 | 246 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 247 | px = pIn1; |
emilmont | 1:fdd22bb7aa52 | 248 | |
emilmont | 1:fdd22bb7aa52 | 249 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 250 | pSrc2 = pIn2 + (srcBLen - 1u); |
emilmont | 1:fdd22bb7aa52 | 251 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 252 | |
emilmont | 1:fdd22bb7aa52 | 253 | /* count is index by which the pointer pIn1 to be incremented */ |
emilmont | 1:fdd22bb7aa52 | 254 | count = 0u; |
emilmont | 1:fdd22bb7aa52 | 255 | |
emilmont | 1:fdd22bb7aa52 | 256 | /* ------------------- |
emilmont | 1:fdd22bb7aa52 | 257 | * Stage2 process |
emilmont | 1:fdd22bb7aa52 | 258 | * ------------------*/ |
emilmont | 1:fdd22bb7aa52 | 259 | |
emilmont | 1:fdd22bb7aa52 | 260 | /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed. |
emilmont | 1:fdd22bb7aa52 | 261 | * So, to loop unroll over blockSize2, |
emilmont | 1:fdd22bb7aa52 | 262 | * srcBLen should be greater than or equal to 4 */ |
emilmont | 1:fdd22bb7aa52 | 263 | if(srcBLen >= 4u) |
emilmont | 1:fdd22bb7aa52 | 264 | { |
emilmont | 1:fdd22bb7aa52 | 265 | /* Loop unroll over blockSize2 */ |
emilmont | 1:fdd22bb7aa52 | 266 | blkCnt = ((uint32_t) blockSize2 >> 2u); |
emilmont | 1:fdd22bb7aa52 | 267 | |
emilmont | 1:fdd22bb7aa52 | 268 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 269 | { |
emilmont | 1:fdd22bb7aa52 | 270 | /* Set all accumulators to zero */ |
emilmont | 1:fdd22bb7aa52 | 271 | acc0 = 0; |
emilmont | 1:fdd22bb7aa52 | 272 | acc1 = 0; |
emilmont | 1:fdd22bb7aa52 | 273 | acc2 = 0; |
emilmont | 1:fdd22bb7aa52 | 274 | acc3 = 0; |
emilmont | 1:fdd22bb7aa52 | 275 | |
emilmont | 1:fdd22bb7aa52 | 276 | /* read x[0], x[1], x[2] samples */ |
emilmont | 1:fdd22bb7aa52 | 277 | x0 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 278 | x1 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 279 | x2 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 280 | |
emilmont | 1:fdd22bb7aa52 | 281 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 282 | k = srcBLen >> 2u; |
emilmont | 1:fdd22bb7aa52 | 283 | |
emilmont | 1:fdd22bb7aa52 | 284 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 285 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 286 | do |
emilmont | 1:fdd22bb7aa52 | 287 | { |
emilmont | 1:fdd22bb7aa52 | 288 | /* Read y[srcBLen - 1] sample */ |
emilmont | 1:fdd22bb7aa52 | 289 | c0 = *(py--); |
emilmont | 1:fdd22bb7aa52 | 290 | |
emilmont | 1:fdd22bb7aa52 | 291 | /* Read x[3] sample */ |
emilmont | 1:fdd22bb7aa52 | 292 | x3 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 293 | |
emilmont | 1:fdd22bb7aa52 | 294 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 295 | /* acc0 += x[0] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 296 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 297 | |
emilmont | 1:fdd22bb7aa52 | 298 | /* acc1 += x[1] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 299 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 300 | |
emilmont | 1:fdd22bb7aa52 | 301 | /* acc2 += x[2] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 302 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 303 | |
emilmont | 1:fdd22bb7aa52 | 304 | /* acc3 += x[3] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 305 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 306 | |
emilmont | 1:fdd22bb7aa52 | 307 | /* Read y[srcBLen - 2] sample */ |
emilmont | 1:fdd22bb7aa52 | 308 | c0 = *(py--); |
emilmont | 1:fdd22bb7aa52 | 309 | |
emilmont | 1:fdd22bb7aa52 | 310 | /* Read x[4] sample */ |
emilmont | 1:fdd22bb7aa52 | 311 | x0 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 312 | |
emilmont | 1:fdd22bb7aa52 | 313 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 314 | /* acc0 += x[1] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 315 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 316 | /* acc1 += x[2] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 317 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 318 | /* acc2 += x[3] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 319 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 320 | /* acc3 += x[4] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 321 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 322 | |
emilmont | 1:fdd22bb7aa52 | 323 | /* Read y[srcBLen - 3] sample */ |
emilmont | 1:fdd22bb7aa52 | 324 | c0 = *(py--); |
emilmont | 1:fdd22bb7aa52 | 325 | |
emilmont | 1:fdd22bb7aa52 | 326 | /* Read x[5] sample */ |
emilmont | 1:fdd22bb7aa52 | 327 | x1 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 328 | |
emilmont | 1:fdd22bb7aa52 | 329 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 330 | /* acc0 += x[2] * y[srcBLen - 3] */ |
emilmont | 1:fdd22bb7aa52 | 331 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 332 | /* acc1 += x[3] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 333 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 334 | /* acc2 += x[4] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 335 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 336 | /* acc3 += x[5] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 337 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 338 | |
emilmont | 1:fdd22bb7aa52 | 339 | /* Read y[srcBLen - 4] sample */ |
emilmont | 1:fdd22bb7aa52 | 340 | c0 = *(py--); |
emilmont | 1:fdd22bb7aa52 | 341 | |
emilmont | 1:fdd22bb7aa52 | 342 | /* Read x[6] sample */ |
emilmont | 1:fdd22bb7aa52 | 343 | x2 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 344 | |
emilmont | 1:fdd22bb7aa52 | 345 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 346 | /* acc0 += x[3] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 347 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 348 | /* acc1 += x[4] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 349 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 350 | /* acc2 += x[5] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 351 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 352 | /* acc3 += x[6] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 353 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 354 | |
emilmont | 1:fdd22bb7aa52 | 355 | |
emilmont | 1:fdd22bb7aa52 | 356 | } while(--k); |
emilmont | 1:fdd22bb7aa52 | 357 | |
emilmont | 1:fdd22bb7aa52 | 358 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 359 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 360 | k = srcBLen % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 361 | |
emilmont | 1:fdd22bb7aa52 | 362 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 363 | { |
emilmont | 1:fdd22bb7aa52 | 364 | /* Read y[srcBLen - 5] sample */ |
emilmont | 1:fdd22bb7aa52 | 365 | c0 = *(py--); |
emilmont | 1:fdd22bb7aa52 | 366 | |
emilmont | 1:fdd22bb7aa52 | 367 | /* Read x[7] sample */ |
emilmont | 1:fdd22bb7aa52 | 368 | x3 = *(px++); |
emilmont | 1:fdd22bb7aa52 | 369 | |
emilmont | 1:fdd22bb7aa52 | 370 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 371 | /* acc0 += x[4] * y[srcBLen - 5] */ |
emilmont | 1:fdd22bb7aa52 | 372 | acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x0 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 373 | /* acc1 += x[5] * y[srcBLen - 5] */ |
emilmont | 1:fdd22bb7aa52 | 374 | acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x1 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 375 | /* acc2 += x[6] * y[srcBLen - 5] */ |
emilmont | 1:fdd22bb7aa52 | 376 | acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x2 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 377 | /* acc3 += x[7] * y[srcBLen - 5] */ |
emilmont | 1:fdd22bb7aa52 | 378 | acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); |
emilmont | 1:fdd22bb7aa52 | 379 | |
emilmont | 1:fdd22bb7aa52 | 380 | /* Reuse the present samples for the next MAC */ |
emilmont | 1:fdd22bb7aa52 | 381 | x0 = x1; |
emilmont | 1:fdd22bb7aa52 | 382 | x1 = x2; |
emilmont | 1:fdd22bb7aa52 | 383 | x2 = x3; |
emilmont | 1:fdd22bb7aa52 | 384 | |
emilmont | 1:fdd22bb7aa52 | 385 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 386 | k--; |
emilmont | 1:fdd22bb7aa52 | 387 | } |
emilmont | 1:fdd22bb7aa52 | 388 | |
emilmont | 1:fdd22bb7aa52 | 389 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 390 | *pOut++ = (q31_t) (acc0 << 1); |
emilmont | 1:fdd22bb7aa52 | 391 | *pOut++ = (q31_t) (acc1 << 1); |
emilmont | 1:fdd22bb7aa52 | 392 | *pOut++ = (q31_t) (acc2 << 1); |
emilmont | 1:fdd22bb7aa52 | 393 | *pOut++ = (q31_t) (acc3 << 1); |
emilmont | 1:fdd22bb7aa52 | 394 | |
emilmont | 1:fdd22bb7aa52 | 395 | /* Increment the pointer pIn1 index, count by 4 */ |
emilmont | 1:fdd22bb7aa52 | 396 | count += 4u; |
emilmont | 1:fdd22bb7aa52 | 397 | |
emilmont | 1:fdd22bb7aa52 | 398 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 399 | px = pIn1 + count; |
emilmont | 1:fdd22bb7aa52 | 400 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 401 | |
emilmont | 1:fdd22bb7aa52 | 402 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 403 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 404 | } |
emilmont | 1:fdd22bb7aa52 | 405 | |
emilmont | 1:fdd22bb7aa52 | 406 | /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here. |
emilmont | 1:fdd22bb7aa52 | 407 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 408 | blkCnt = (uint32_t) blockSize2 % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 409 | |
emilmont | 1:fdd22bb7aa52 | 410 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 411 | { |
emilmont | 1:fdd22bb7aa52 | 412 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 413 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 414 | |
emilmont | 1:fdd22bb7aa52 | 415 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 416 | k = srcBLen >> 2u; |
emilmont | 1:fdd22bb7aa52 | 417 | |
emilmont | 1:fdd22bb7aa52 | 418 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 419 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 420 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 421 | { |
emilmont | 1:fdd22bb7aa52 | 422 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 423 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 424 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 425 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 426 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 427 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 428 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 429 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 430 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 431 | |
emilmont | 1:fdd22bb7aa52 | 432 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 433 | k--; |
emilmont | 1:fdd22bb7aa52 | 434 | } |
emilmont | 1:fdd22bb7aa52 | 435 | |
emilmont | 1:fdd22bb7aa52 | 436 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 437 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 438 | k = srcBLen % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 439 | |
emilmont | 1:fdd22bb7aa52 | 440 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 441 | { |
emilmont | 1:fdd22bb7aa52 | 442 | /* Perform the multiply-accumulate */ |
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 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 451 | *pOut++ = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 452 | |
emilmont | 1:fdd22bb7aa52 | 453 | /* Increment the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 454 | count++; |
emilmont | 1:fdd22bb7aa52 | 455 | |
emilmont | 1:fdd22bb7aa52 | 456 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 457 | px = pIn1 + count; |
emilmont | 1:fdd22bb7aa52 | 458 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 459 | |
emilmont | 1:fdd22bb7aa52 | 460 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 461 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 462 | } |
emilmont | 1:fdd22bb7aa52 | 463 | } |
emilmont | 1:fdd22bb7aa52 | 464 | else |
emilmont | 1:fdd22bb7aa52 | 465 | { |
emilmont | 1:fdd22bb7aa52 | 466 | /* If the srcBLen is not a multiple of 4, |
emilmont | 1:fdd22bb7aa52 | 467 | * the blockSize2 loop cannot be unrolled by 4 */ |
emilmont | 1:fdd22bb7aa52 | 468 | blkCnt = (uint32_t) blockSize2; |
emilmont | 1:fdd22bb7aa52 | 469 | |
emilmont | 1:fdd22bb7aa52 | 470 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 471 | { |
emilmont | 1:fdd22bb7aa52 | 472 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 473 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 474 | |
emilmont | 1:fdd22bb7aa52 | 475 | /* srcBLen number of MACS should be performed */ |
emilmont | 1:fdd22bb7aa52 | 476 | k = srcBLen; |
emilmont | 1:fdd22bb7aa52 | 477 | |
emilmont | 1:fdd22bb7aa52 | 478 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 479 | { |
emilmont | 1:fdd22bb7aa52 | 480 | /* Perform the multiply-accumulate */ |
emilmont | 1:fdd22bb7aa52 | 481 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 482 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 483 | |
emilmont | 1:fdd22bb7aa52 | 484 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 485 | k--; |
emilmont | 1:fdd22bb7aa52 | 486 | } |
emilmont | 1:fdd22bb7aa52 | 487 | |
emilmont | 1:fdd22bb7aa52 | 488 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 489 | *pOut++ = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 490 | |
emilmont | 1:fdd22bb7aa52 | 491 | /* Increment the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 492 | count++; |
emilmont | 1:fdd22bb7aa52 | 493 | |
emilmont | 1:fdd22bb7aa52 | 494 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 495 | px = pIn1 + count; |
emilmont | 1:fdd22bb7aa52 | 496 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 497 | |
emilmont | 1:fdd22bb7aa52 | 498 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 499 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 500 | } |
emilmont | 1:fdd22bb7aa52 | 501 | } |
emilmont | 1:fdd22bb7aa52 | 502 | |
emilmont | 1:fdd22bb7aa52 | 503 | |
emilmont | 1:fdd22bb7aa52 | 504 | /* -------------------------- |
emilmont | 1:fdd22bb7aa52 | 505 | * Initializations of stage3 |
emilmont | 1:fdd22bb7aa52 | 506 | * -------------------------*/ |
emilmont | 1:fdd22bb7aa52 | 507 | |
emilmont | 1:fdd22bb7aa52 | 508 | /* sum += x[srcALen-srcBLen+1] * y[srcBLen-1] + x[srcALen-srcBLen+2] * y[srcBLen-2] +...+ x[srcALen-1] * y[1] |
emilmont | 1:fdd22bb7aa52 | 509 | * sum += x[srcALen-srcBLen+2] * y[srcBLen-1] + x[srcALen-srcBLen+3] * y[srcBLen-2] +...+ x[srcALen-1] * y[2] |
emilmont | 1:fdd22bb7aa52 | 510 | * .... |
emilmont | 1:fdd22bb7aa52 | 511 | * sum += x[srcALen-2] * y[srcBLen-1] + x[srcALen-1] * y[srcBLen-2] |
emilmont | 1:fdd22bb7aa52 | 512 | * sum += x[srcALen-1] * y[srcBLen-1] |
emilmont | 1:fdd22bb7aa52 | 513 | */ |
emilmont | 1:fdd22bb7aa52 | 514 | |
emilmont | 1:fdd22bb7aa52 | 515 | /* In this stage the MAC operations are decreased by 1 for every iteration. |
emilmont | 1:fdd22bb7aa52 | 516 | The count variable holds the number of MAC operations performed */ |
emilmont | 1:fdd22bb7aa52 | 517 | count = srcBLen - 1u; |
emilmont | 1:fdd22bb7aa52 | 518 | |
emilmont | 1:fdd22bb7aa52 | 519 | /* Working pointer of inputA */ |
emilmont | 1:fdd22bb7aa52 | 520 | pSrc1 = (pIn1 + srcALen) - (srcBLen - 1u); |
emilmont | 1:fdd22bb7aa52 | 521 | px = pSrc1; |
emilmont | 1:fdd22bb7aa52 | 522 | |
emilmont | 1:fdd22bb7aa52 | 523 | /* Working pointer of inputB */ |
emilmont | 1:fdd22bb7aa52 | 524 | pSrc2 = pIn2 + (srcBLen - 1u); |
emilmont | 1:fdd22bb7aa52 | 525 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 526 | |
emilmont | 1:fdd22bb7aa52 | 527 | /* ------------------- |
emilmont | 1:fdd22bb7aa52 | 528 | * Stage3 process |
emilmont | 1:fdd22bb7aa52 | 529 | * ------------------*/ |
emilmont | 1:fdd22bb7aa52 | 530 | |
emilmont | 1:fdd22bb7aa52 | 531 | while(blockSize3 > 0) |
emilmont | 1:fdd22bb7aa52 | 532 | { |
emilmont | 1:fdd22bb7aa52 | 533 | /* Accumulator is made zero for every iteration */ |
emilmont | 1:fdd22bb7aa52 | 534 | sum = 0; |
emilmont | 1:fdd22bb7aa52 | 535 | |
emilmont | 1:fdd22bb7aa52 | 536 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
emilmont | 1:fdd22bb7aa52 | 537 | k = count >> 2u; |
emilmont | 1:fdd22bb7aa52 | 538 | |
emilmont | 1:fdd22bb7aa52 | 539 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
emilmont | 1:fdd22bb7aa52 | 540 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 541 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 542 | { |
emilmont | 1:fdd22bb7aa52 | 543 | /* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */ |
emilmont | 1:fdd22bb7aa52 | 544 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 545 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 546 | |
emilmont | 1:fdd22bb7aa52 | 547 | /* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */ |
emilmont | 1:fdd22bb7aa52 | 548 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 549 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 550 | |
emilmont | 1:fdd22bb7aa52 | 551 | /* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */ |
emilmont | 1:fdd22bb7aa52 | 552 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 553 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 554 | |
emilmont | 1:fdd22bb7aa52 | 555 | /* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */ |
emilmont | 1:fdd22bb7aa52 | 556 | sum = (q31_t) ((((q63_t) sum << 32) + |
emilmont | 1:fdd22bb7aa52 | 557 | ((q63_t) * px++ * (*py--))) >> 32); |
emilmont | 1:fdd22bb7aa52 | 558 | |
emilmont | 1:fdd22bb7aa52 | 559 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 560 | k--; |
emilmont | 1:fdd22bb7aa52 | 561 | } |
emilmont | 1:fdd22bb7aa52 | 562 | |
emilmont | 1:fdd22bb7aa52 | 563 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
emilmont | 1:fdd22bb7aa52 | 564 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 565 | k = count % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 566 | |
emilmont | 1:fdd22bb7aa52 | 567 | while(k > 0u) |
emilmont | 1:fdd22bb7aa52 | 568 | { |
emilmont | 1:fdd22bb7aa52 | 569 | /* Perform the multiply-accumulates */ |
emilmont | 1:fdd22bb7aa52 | 570 | /* sum += x[srcALen-1] * y[srcBLen-1] */ |
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 | /* Store the result in the accumulator in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 579 | *pOut++ = sum << 1; |
emilmont | 1:fdd22bb7aa52 | 580 | |
emilmont | 1:fdd22bb7aa52 | 581 | /* Update the inputA and inputB pointers for next MAC calculation */ |
emilmont | 1:fdd22bb7aa52 | 582 | px = ++pSrc1; |
emilmont | 1:fdd22bb7aa52 | 583 | py = pSrc2; |
emilmont | 1:fdd22bb7aa52 | 584 | |
emilmont | 1:fdd22bb7aa52 | 585 | /* Decrement the MAC count */ |
emilmont | 1:fdd22bb7aa52 | 586 | count--; |
emilmont | 1:fdd22bb7aa52 | 587 | |
emilmont | 1:fdd22bb7aa52 | 588 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 589 | blockSize3--; |
emilmont | 1:fdd22bb7aa52 | 590 | |
emilmont | 1:fdd22bb7aa52 | 591 | } |
emilmont | 1:fdd22bb7aa52 | 592 | |
emilmont | 1:fdd22bb7aa52 | 593 | /* set status as ARM_MATH_SUCCESS */ |
emilmont | 1:fdd22bb7aa52 | 594 | status = ARM_MATH_SUCCESS; |
emilmont | 1:fdd22bb7aa52 | 595 | } |
emilmont | 1:fdd22bb7aa52 | 596 | |
emilmont | 1:fdd22bb7aa52 | 597 | /* Return to application */ |
emilmont | 1:fdd22bb7aa52 | 598 | return (status); |
emilmont | 1:fdd22bb7aa52 | 599 | |
emilmont | 1:fdd22bb7aa52 | 600 | } |
emilmont | 1:fdd22bb7aa52 | 601 | |
emilmont | 1:fdd22bb7aa52 | 602 | /** |
emilmont | 1:fdd22bb7aa52 | 603 | * @} end of PartialConv group |
emilmont | 1:fdd22bb7aa52 | 604 | */ |