mbed official
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 
mbed official
cmsis_dsp/FilteringFunctions/arm_correlate_q7.c@1:fdd22bb7aa52, 2012-11-28 (annotated)
- Committer:
- emilmont
- Date:
- Wed Nov 28 12:30:09 2012 +0000
- Revision:
- 1:fdd22bb7aa52
- Child:
- 2:da51fb522205
DSP library code
Who changed what in which revision?
| User | Revision | Line number | New contents of line |
|---|---|---|---|
| emilmont | 1:fdd22bb7aa52 | 1 | /* ---------------------------------------------------------------------- |
| emilmont | 1:fdd22bb7aa52 | 2 | * Copyright (C) 2010 ARM Limited. All rights reserved. |
| emilmont | 1:fdd22bb7aa52 | 3 | * |
| emilmont | 1:fdd22bb7aa52 | 4 | * $Date: 15. February 2012 |
| emilmont | 1:fdd22bb7aa52 | 5 | * $Revision: V1.1.0 |
| emilmont | 1:fdd22bb7aa52 | 6 | * |
| emilmont | 1:fdd22bb7aa52 | 7 | * Project: CMSIS DSP Library |
| emilmont | 1:fdd22bb7aa52 | 8 | * Title: arm_correlate_q7.c |
| emilmont | 1:fdd22bb7aa52 | 9 | * |
| emilmont | 1:fdd22bb7aa52 | 10 | * Description: Correlation of Q7 sequences. |
| emilmont | 1:fdd22bb7aa52 | 11 | * |
| emilmont | 1:fdd22bb7aa52 | 12 | * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 |
| emilmont | 1:fdd22bb7aa52 | 13 | * |
| emilmont | 1:fdd22bb7aa52 | 14 | * Version 1.1.0 2012/02/15 |
| emilmont | 1:fdd22bb7aa52 | 15 | * Updated with more optimizations, bug fixes and minor API changes. |
| emilmont | 1:fdd22bb7aa52 | 16 | * |
| emilmont | 1:fdd22bb7aa52 | 17 | * Version 1.0.11 2011/10/18 |
| emilmont | 1:fdd22bb7aa52 | 18 | * Bug Fix in conv, correlation, partial convolution. |
| emilmont | 1:fdd22bb7aa52 | 19 | * |
| emilmont | 1:fdd22bb7aa52 | 20 | * Version 1.0.10 2011/7/15 |
| emilmont | 1:fdd22bb7aa52 | 21 | * Big Endian support added and Merged M0 and M3/M4 Source code. |
| emilmont | 1:fdd22bb7aa52 | 22 | * |
| emilmont | 1:fdd22bb7aa52 | 23 | * Version 1.0.3 2010/11/29 |
| emilmont | 1:fdd22bb7aa52 | 24 | * Re-organized the CMSIS folders and updated documentation. |
| emilmont | 1:fdd22bb7aa52 | 25 | * |
| emilmont | 1:fdd22bb7aa52 | 26 | * Version 1.0.2 2010/11/11 |
| emilmont | 1:fdd22bb7aa52 | 27 | * Documentation updated. |
| emilmont | 1:fdd22bb7aa52 | 28 | * |
| emilmont | 1:fdd22bb7aa52 | 29 | * Version 1.0.1 2010/10/05 |
| emilmont | 1:fdd22bb7aa52 | 30 | * Production release and review comments incorporated. |
| emilmont | 1:fdd22bb7aa52 | 31 | * |
| emilmont | 1:fdd22bb7aa52 | 32 | * Version 1.0.0 2010/09/20 |
| emilmont | 1:fdd22bb7aa52 | 33 | * Production release and review comments incorporated |
| emilmont | 1:fdd22bb7aa52 | 34 | * |
| emilmont | 1:fdd22bb7aa52 | 35 | * Version 0.0.7 2010/06/10 |
| emilmont | 1:fdd22bb7aa52 | 36 | * Misra-C changes done |
| emilmont | 1:fdd22bb7aa52 | 37 | * |
| emilmont | 1:fdd22bb7aa52 | 38 | * -------------------------------------------------------------------- */ |
| emilmont | 1:fdd22bb7aa52 | 39 | |
| emilmont | 1:fdd22bb7aa52 | 40 | #include "arm_math.h" |
| emilmont | 1:fdd22bb7aa52 | 41 | |
| emilmont | 1:fdd22bb7aa52 | 42 | /** |
| emilmont | 1:fdd22bb7aa52 | 43 | * @ingroup groupFilters |
| emilmont | 1:fdd22bb7aa52 | 44 | */ |
| emilmont | 1:fdd22bb7aa52 | 45 | |
| emilmont | 1:fdd22bb7aa52 | 46 | /** |
| emilmont | 1:fdd22bb7aa52 | 47 | * @addtogroup Corr |
| emilmont | 1:fdd22bb7aa52 | 48 | * @{ |
| emilmont | 1:fdd22bb7aa52 | 49 | */ |
| emilmont | 1:fdd22bb7aa52 | 50 | |
| emilmont | 1:fdd22bb7aa52 | 51 | /** |
| emilmont | 1:fdd22bb7aa52 | 52 | * @brief Correlation of Q7 sequences. |
| emilmont | 1:fdd22bb7aa52 | 53 | * @param[in] *pSrcA points to the first input sequence. |
| emilmont | 1:fdd22bb7aa52 | 54 | * @param[in] srcALen length of the first input sequence. |
| emilmont | 1:fdd22bb7aa52 | 55 | * @param[in] *pSrcB points to the second input sequence. |
| emilmont | 1:fdd22bb7aa52 | 56 | * @param[in] srcBLen length of the second input sequence. |
| emilmont | 1:fdd22bb7aa52 | 57 | * @param[out] *pDst points to the location where the output result is written. Length 2 * max(srcALen, srcBLen) - 1. |
| emilmont | 1:fdd22bb7aa52 | 58 | * @return none. |
| emilmont | 1:fdd22bb7aa52 | 59 | * |
| emilmont | 1:fdd22bb7aa52 | 60 | * @details |
| emilmont | 1:fdd22bb7aa52 | 61 | * <b>Scaling and Overflow Behavior:</b> |
| emilmont | 1:fdd22bb7aa52 | 62 | * |
| emilmont | 1:fdd22bb7aa52 | 63 | * \par |
| emilmont | 1:fdd22bb7aa52 | 64 | * The function is implemented using a 32-bit internal accumulator. |
| emilmont | 1:fdd22bb7aa52 | 65 | * Both the inputs are represented in 1.7 format and multiplications yield a 2.14 result. |
| emilmont | 1:fdd22bb7aa52 | 66 | * The 2.14 intermediate results are accumulated in a 32-bit accumulator in 18.14 format. |
| emilmont | 1:fdd22bb7aa52 | 67 | * 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 | 68 | * The 18.14 result is then truncated to 18.7 format by discarding the low 7 bits and saturated to 1.7 format. |
| emilmont | 1:fdd22bb7aa52 | 69 | * |
| emilmont | 1:fdd22bb7aa52 | 70 | * \par |
| emilmont | 1:fdd22bb7aa52 | 71 | * Refer the function <code>arm_correlate_opt_q7()</code> for a faster implementation of this function. |
| emilmont | 1:fdd22bb7aa52 | 72 | * |
| emilmont | 1:fdd22bb7aa52 | 73 | */ |
| emilmont | 1:fdd22bb7aa52 | 74 | |
| emilmont | 1:fdd22bb7aa52 | 75 | void arm_correlate_q7( |
| emilmont | 1:fdd22bb7aa52 | 76 | q7_t * pSrcA, |
| emilmont | 1:fdd22bb7aa52 | 77 | uint32_t srcALen, |
| emilmont | 1:fdd22bb7aa52 | 78 | q7_t * pSrcB, |
| emilmont | 1:fdd22bb7aa52 | 79 | uint32_t srcBLen, |
| emilmont | 1:fdd22bb7aa52 | 80 | q7_t * pDst) |
| emilmont | 1:fdd22bb7aa52 | 81 | { |
| emilmont | 1:fdd22bb7aa52 | 82 | |
| emilmont | 1:fdd22bb7aa52 | 83 | |
| emilmont | 1:fdd22bb7aa52 | 84 | #ifndef ARM_MATH_CM0 |
| emilmont | 1:fdd22bb7aa52 | 85 | |
| emilmont | 1:fdd22bb7aa52 | 86 | /* Run the below code for Cortex-M4 and Cortex-M3 */ |
| emilmont | 1:fdd22bb7aa52 | 87 | |
| emilmont | 1:fdd22bb7aa52 | 88 | q7_t *pIn1; /* inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 89 | q7_t *pIn2; /* inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 90 | q7_t *pOut = pDst; /* output pointer */ |
| emilmont | 1:fdd22bb7aa52 | 91 | q7_t *px; /* Intermediate inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 92 | q7_t *py; /* Intermediate inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 93 | q7_t *pSrc1; /* Intermediate pointers */ |
| emilmont | 1:fdd22bb7aa52 | 94 | q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ |
| emilmont | 1:fdd22bb7aa52 | 95 | q31_t input1, input2; /* temporary variables */ |
| emilmont | 1:fdd22bb7aa52 | 96 | q15_t in1, in2; /* temporary variables */ |
| emilmont | 1:fdd22bb7aa52 | 97 | q7_t x0, x1, x2, x3, c0, c1; /* temporary variables for holding input and coefficient values */ |
| emilmont | 1:fdd22bb7aa52 | 98 | uint32_t j, k = 0u, count, blkCnt, outBlockSize, blockSize1, blockSize2, blockSize3; /* loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 99 | int32_t inc = 1; |
| emilmont | 1:fdd22bb7aa52 | 100 | |
| emilmont | 1:fdd22bb7aa52 | 101 | |
| emilmont | 1:fdd22bb7aa52 | 102 | /* The algorithm implementation is based on the lengths of the inputs. */ |
| emilmont | 1:fdd22bb7aa52 | 103 | /* srcB is always made to slide across srcA. */ |
| emilmont | 1:fdd22bb7aa52 | 104 | /* So srcBLen is always considered as shorter or equal to srcALen */ |
| emilmont | 1:fdd22bb7aa52 | 105 | /* But CORR(x, y) is reverse of CORR(y, x) */ |
| emilmont | 1:fdd22bb7aa52 | 106 | /* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 107 | /* and the destination pointer modifier, inc is set to -1 */ |
| emilmont | 1:fdd22bb7aa52 | 108 | /* If srcALen > srcBLen, zero pad has to be done to srcB to make the two inputs of same length */ |
| emilmont | 1:fdd22bb7aa52 | 109 | /* But to improve the performance, |
| emilmont | 1:fdd22bb7aa52 | 110 | * we include zeroes in the output instead of zero padding either of the the inputs*/ |
| emilmont | 1:fdd22bb7aa52 | 111 | /* If srcALen > srcBLen, |
| emilmont | 1:fdd22bb7aa52 | 112 | * (srcALen - srcBLen) zeroes has to included in the starting of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 113 | /* If srcALen < srcBLen, |
| emilmont | 1:fdd22bb7aa52 | 114 | * (srcALen - srcBLen) zeroes has to included in the ending of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 115 | if(srcALen >= srcBLen) |
| emilmont | 1:fdd22bb7aa52 | 116 | { |
| emilmont | 1:fdd22bb7aa52 | 117 | /* Initialization of inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 118 | pIn1 = (pSrcA); |
| emilmont | 1:fdd22bb7aa52 | 119 | |
| emilmont | 1:fdd22bb7aa52 | 120 | /* Initialization of inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 121 | pIn2 = (pSrcB); |
| emilmont | 1:fdd22bb7aa52 | 122 | |
| emilmont | 1:fdd22bb7aa52 | 123 | /* Number of output samples is calculated */ |
| emilmont | 1:fdd22bb7aa52 | 124 | outBlockSize = (2u * srcALen) - 1u; |
| emilmont | 1:fdd22bb7aa52 | 125 | |
| emilmont | 1:fdd22bb7aa52 | 126 | /* When srcALen > srcBLen, zero padding is done to srcB |
| emilmont | 1:fdd22bb7aa52 | 127 | * to make their lengths equal. |
| emilmont | 1:fdd22bb7aa52 | 128 | * Instead, (outBlockSize - (srcALen + srcBLen - 1)) |
| emilmont | 1:fdd22bb7aa52 | 129 | * number of output samples are made zero */ |
| emilmont | 1:fdd22bb7aa52 | 130 | j = outBlockSize - (srcALen + (srcBLen - 1u)); |
| emilmont | 1:fdd22bb7aa52 | 131 | |
| emilmont | 1:fdd22bb7aa52 | 132 | /* Updating the pointer position to non zero value */ |
| emilmont | 1:fdd22bb7aa52 | 133 | pOut += j; |
| emilmont | 1:fdd22bb7aa52 | 134 | |
| emilmont | 1:fdd22bb7aa52 | 135 | } |
| emilmont | 1:fdd22bb7aa52 | 136 | else |
| emilmont | 1:fdd22bb7aa52 | 137 | { |
| emilmont | 1:fdd22bb7aa52 | 138 | /* Initialization of inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 139 | pIn1 = (pSrcB); |
| emilmont | 1:fdd22bb7aa52 | 140 | |
| emilmont | 1:fdd22bb7aa52 | 141 | /* Initialization of inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 142 | pIn2 = (pSrcA); |
| emilmont | 1:fdd22bb7aa52 | 143 | |
| emilmont | 1:fdd22bb7aa52 | 144 | /* srcBLen is always considered as shorter or equal to srcALen */ |
| emilmont | 1:fdd22bb7aa52 | 145 | j = srcBLen; |
| emilmont | 1:fdd22bb7aa52 | 146 | srcBLen = srcALen; |
| emilmont | 1:fdd22bb7aa52 | 147 | srcALen = j; |
| emilmont | 1:fdd22bb7aa52 | 148 | |
| emilmont | 1:fdd22bb7aa52 | 149 | /* CORR(x, y) = Reverse order(CORR(y, x)) */ |
| emilmont | 1:fdd22bb7aa52 | 150 | /* Hence set the destination pointer to point to the last output sample */ |
| emilmont | 1:fdd22bb7aa52 | 151 | pOut = pDst + ((srcALen + srcBLen) - 2u); |
| emilmont | 1:fdd22bb7aa52 | 152 | |
| emilmont | 1:fdd22bb7aa52 | 153 | /* Destination address modifier is set to -1 */ |
| emilmont | 1:fdd22bb7aa52 | 154 | inc = -1; |
| emilmont | 1:fdd22bb7aa52 | 155 | |
| emilmont | 1:fdd22bb7aa52 | 156 | } |
| emilmont | 1:fdd22bb7aa52 | 157 | |
| emilmont | 1:fdd22bb7aa52 | 158 | /* The function is internally |
| emilmont | 1:fdd22bb7aa52 | 159 | * divided into three parts according to the number of multiplications that has to be |
| emilmont | 1:fdd22bb7aa52 | 160 | * taken place between inputA samples and inputB samples. In the first part of the |
| emilmont | 1:fdd22bb7aa52 | 161 | * algorithm, the multiplications increase by one for every iteration. |
| emilmont | 1:fdd22bb7aa52 | 162 | * In the second part of the algorithm, srcBLen number of multiplications are done. |
| emilmont | 1:fdd22bb7aa52 | 163 | * In the third part of the algorithm, the multiplications decrease by one |
| emilmont | 1:fdd22bb7aa52 | 164 | * for every iteration.*/ |
| emilmont | 1:fdd22bb7aa52 | 165 | /* The algorithm is implemented in three stages. |
| emilmont | 1:fdd22bb7aa52 | 166 | * The loop counters of each stage is initiated here. */ |
| emilmont | 1:fdd22bb7aa52 | 167 | blockSize1 = srcBLen - 1u; |
| emilmont | 1:fdd22bb7aa52 | 168 | blockSize2 = srcALen - (srcBLen - 1u); |
| emilmont | 1:fdd22bb7aa52 | 169 | blockSize3 = blockSize1; |
| emilmont | 1:fdd22bb7aa52 | 170 | |
| emilmont | 1:fdd22bb7aa52 | 171 | /* -------------------------- |
| emilmont | 1:fdd22bb7aa52 | 172 | * Initializations of stage1 |
| emilmont | 1:fdd22bb7aa52 | 173 | * -------------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 174 | |
| emilmont | 1:fdd22bb7aa52 | 175 | /* sum = x[0] * y[srcBlen - 1] |
| emilmont | 1:fdd22bb7aa52 | 176 | * sum = x[0] * y[srcBlen - 2] + x[1] * y[srcBlen - 1] |
| emilmont | 1:fdd22bb7aa52 | 177 | * .... |
| emilmont | 1:fdd22bb7aa52 | 178 | * sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen - 1] * y[srcBLen - 1] |
| emilmont | 1:fdd22bb7aa52 | 179 | */ |
| emilmont | 1:fdd22bb7aa52 | 180 | |
| emilmont | 1:fdd22bb7aa52 | 181 | /* In this stage the MAC operations are increased by 1 for every iteration. |
| emilmont | 1:fdd22bb7aa52 | 182 | The count variable holds the number of MAC operations performed */ |
| emilmont | 1:fdd22bb7aa52 | 183 | count = 1u; |
| emilmont | 1:fdd22bb7aa52 | 184 | |
| emilmont | 1:fdd22bb7aa52 | 185 | /* Working pointer of inputA */ |
| emilmont | 1:fdd22bb7aa52 | 186 | px = pIn1; |
| emilmont | 1:fdd22bb7aa52 | 187 | |
| emilmont | 1:fdd22bb7aa52 | 188 | /* Working pointer of inputB */ |
| emilmont | 1:fdd22bb7aa52 | 189 | pSrc1 = pIn2 + (srcBLen - 1u); |
| emilmont | 1:fdd22bb7aa52 | 190 | py = pSrc1; |
| emilmont | 1:fdd22bb7aa52 | 191 | |
| emilmont | 1:fdd22bb7aa52 | 192 | /* ------------------------ |
| emilmont | 1:fdd22bb7aa52 | 193 | * Stage1 process |
| emilmont | 1:fdd22bb7aa52 | 194 | * ----------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 195 | |
| emilmont | 1:fdd22bb7aa52 | 196 | /* The first stage starts here */ |
| emilmont | 1:fdd22bb7aa52 | 197 | while(blockSize1 > 0u) |
| emilmont | 1:fdd22bb7aa52 | 198 | { |
| emilmont | 1:fdd22bb7aa52 | 199 | /* Accumulator is made zero for every iteration */ |
| emilmont | 1:fdd22bb7aa52 | 200 | sum = 0; |
| emilmont | 1:fdd22bb7aa52 | 201 | |
| emilmont | 1:fdd22bb7aa52 | 202 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
| emilmont | 1:fdd22bb7aa52 | 203 | k = count >> 2; |
| emilmont | 1:fdd22bb7aa52 | 204 | |
| emilmont | 1:fdd22bb7aa52 | 205 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
| emilmont | 1:fdd22bb7aa52 | 206 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
| emilmont | 1:fdd22bb7aa52 | 207 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 208 | { |
| emilmont | 1:fdd22bb7aa52 | 209 | /* x[0] , x[1] */ |
| emilmont | 1:fdd22bb7aa52 | 210 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 211 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 212 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 213 | |
| emilmont | 1:fdd22bb7aa52 | 214 | /* y[srcBLen - 4] , y[srcBLen - 3] */ |
| emilmont | 1:fdd22bb7aa52 | 215 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 216 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 217 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 218 | |
| emilmont | 1:fdd22bb7aa52 | 219 | /* x[0] * y[srcBLen - 4] */ |
| emilmont | 1:fdd22bb7aa52 | 220 | /* x[1] * y[srcBLen - 3] */ |
| emilmont | 1:fdd22bb7aa52 | 221 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 222 | |
| emilmont | 1:fdd22bb7aa52 | 223 | /* x[2] , x[3] */ |
| emilmont | 1:fdd22bb7aa52 | 224 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 225 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 226 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 227 | |
| emilmont | 1:fdd22bb7aa52 | 228 | /* y[srcBLen - 2] , y[srcBLen - 1] */ |
| emilmont | 1:fdd22bb7aa52 | 229 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 230 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 231 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 232 | |
| emilmont | 1:fdd22bb7aa52 | 233 | /* x[2] * y[srcBLen - 2] */ |
| emilmont | 1:fdd22bb7aa52 | 234 | /* x[3] * y[srcBLen - 1] */ |
| emilmont | 1:fdd22bb7aa52 | 235 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 236 | |
| emilmont | 1:fdd22bb7aa52 | 237 | |
| emilmont | 1:fdd22bb7aa52 | 238 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 239 | k--; |
| emilmont | 1:fdd22bb7aa52 | 240 | } |
| emilmont | 1:fdd22bb7aa52 | 241 | |
| emilmont | 1:fdd22bb7aa52 | 242 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
| emilmont | 1:fdd22bb7aa52 | 243 | ** No loop unrolling is used. */ |
| emilmont | 1:fdd22bb7aa52 | 244 | k = count % 0x4u; |
| emilmont | 1:fdd22bb7aa52 | 245 | |
| emilmont | 1:fdd22bb7aa52 | 246 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 247 | { |
| emilmont | 1:fdd22bb7aa52 | 248 | /* Perform the multiply-accumulates */ |
| emilmont | 1:fdd22bb7aa52 | 249 | /* x[0] * y[srcBLen - 1] */ |
| emilmont | 1:fdd22bb7aa52 | 250 | sum += (q31_t) ((q15_t) * px++ * *py++); |
| emilmont | 1:fdd22bb7aa52 | 251 | |
| emilmont | 1:fdd22bb7aa52 | 252 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 253 | k--; |
| emilmont | 1:fdd22bb7aa52 | 254 | } |
| emilmont | 1:fdd22bb7aa52 | 255 | |
| emilmont | 1:fdd22bb7aa52 | 256 | /* Store the result in the accumulator in the destination buffer. */ |
| emilmont | 1:fdd22bb7aa52 | 257 | *pOut = (q7_t) (__SSAT(sum >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 258 | /* Destination pointer is updated according to the address modifier, inc */ |
| emilmont | 1:fdd22bb7aa52 | 259 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 260 | |
| emilmont | 1:fdd22bb7aa52 | 261 | /* Update the inputA and inputB pointers for next MAC calculation */ |
| emilmont | 1:fdd22bb7aa52 | 262 | py = pSrc1 - count; |
| emilmont | 1:fdd22bb7aa52 | 263 | px = pIn1; |
| emilmont | 1:fdd22bb7aa52 | 264 | |
| emilmont | 1:fdd22bb7aa52 | 265 | /* Increment the MAC count */ |
| emilmont | 1:fdd22bb7aa52 | 266 | count++; |
| emilmont | 1:fdd22bb7aa52 | 267 | |
| emilmont | 1:fdd22bb7aa52 | 268 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 269 | blockSize1--; |
| emilmont | 1:fdd22bb7aa52 | 270 | } |
| emilmont | 1:fdd22bb7aa52 | 271 | |
| emilmont | 1:fdd22bb7aa52 | 272 | /* -------------------------- |
| emilmont | 1:fdd22bb7aa52 | 273 | * Initializations of stage2 |
| emilmont | 1:fdd22bb7aa52 | 274 | * ------------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 275 | |
| emilmont | 1:fdd22bb7aa52 | 276 | /* sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen-1] * y[srcBLen-1] |
| emilmont | 1:fdd22bb7aa52 | 277 | * sum = x[1] * y[0] + x[2] * y[1] +...+ x[srcBLen] * y[srcBLen-1] |
| emilmont | 1:fdd22bb7aa52 | 278 | * .... |
| emilmont | 1:fdd22bb7aa52 | 279 | * sum = x[srcALen-srcBLen-2] * y[0] + x[srcALen-srcBLen-1] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
| emilmont | 1:fdd22bb7aa52 | 280 | */ |
| emilmont | 1:fdd22bb7aa52 | 281 | |
| emilmont | 1:fdd22bb7aa52 | 282 | /* Working pointer of inputA */ |
| emilmont | 1:fdd22bb7aa52 | 283 | px = pIn1; |
| emilmont | 1:fdd22bb7aa52 | 284 | |
| emilmont | 1:fdd22bb7aa52 | 285 | /* Working pointer of inputB */ |
| emilmont | 1:fdd22bb7aa52 | 286 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 287 | |
| emilmont | 1:fdd22bb7aa52 | 288 | /* count is index by which the pointer pIn1 to be incremented */ |
| emilmont | 1:fdd22bb7aa52 | 289 | count = 0u; |
| emilmont | 1:fdd22bb7aa52 | 290 | |
| emilmont | 1:fdd22bb7aa52 | 291 | /* ------------------- |
| emilmont | 1:fdd22bb7aa52 | 292 | * Stage2 process |
| emilmont | 1:fdd22bb7aa52 | 293 | * ------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 294 | |
| emilmont | 1:fdd22bb7aa52 | 295 | /* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed. |
| emilmont | 1:fdd22bb7aa52 | 296 | * So, to loop unroll over blockSize2, |
| emilmont | 1:fdd22bb7aa52 | 297 | * srcBLen should be greater than or equal to 4 */ |
| emilmont | 1:fdd22bb7aa52 | 298 | if(srcBLen >= 4u) |
| emilmont | 1:fdd22bb7aa52 | 299 | { |
| emilmont | 1:fdd22bb7aa52 | 300 | /* Loop unroll over blockSize2, by 4 */ |
| emilmont | 1:fdd22bb7aa52 | 301 | blkCnt = blockSize2 >> 2u; |
| emilmont | 1:fdd22bb7aa52 | 302 | |
| emilmont | 1:fdd22bb7aa52 | 303 | while(blkCnt > 0u) |
| emilmont | 1:fdd22bb7aa52 | 304 | { |
| emilmont | 1:fdd22bb7aa52 | 305 | /* Set all accumulators to zero */ |
| emilmont | 1:fdd22bb7aa52 | 306 | acc0 = 0; |
| emilmont | 1:fdd22bb7aa52 | 307 | acc1 = 0; |
| emilmont | 1:fdd22bb7aa52 | 308 | acc2 = 0; |
| emilmont | 1:fdd22bb7aa52 | 309 | acc3 = 0; |
| emilmont | 1:fdd22bb7aa52 | 310 | |
| emilmont | 1:fdd22bb7aa52 | 311 | /* read x[0], x[1], x[2] samples */ |
| emilmont | 1:fdd22bb7aa52 | 312 | x0 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 313 | x1 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 314 | x2 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 315 | |
| emilmont | 1:fdd22bb7aa52 | 316 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
| emilmont | 1:fdd22bb7aa52 | 317 | k = srcBLen >> 2u; |
| emilmont | 1:fdd22bb7aa52 | 318 | |
| emilmont | 1:fdd22bb7aa52 | 319 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
| emilmont | 1:fdd22bb7aa52 | 320 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
| emilmont | 1:fdd22bb7aa52 | 321 | do |
| emilmont | 1:fdd22bb7aa52 | 322 | { |
| emilmont | 1:fdd22bb7aa52 | 323 | /* Read y[0] sample */ |
| emilmont | 1:fdd22bb7aa52 | 324 | c0 = *py++; |
| emilmont | 1:fdd22bb7aa52 | 325 | /* Read y[1] sample */ |
| emilmont | 1:fdd22bb7aa52 | 326 | c1 = *py++; |
| emilmont | 1:fdd22bb7aa52 | 327 | |
| emilmont | 1:fdd22bb7aa52 | 328 | /* Read x[3] sample */ |
| emilmont | 1:fdd22bb7aa52 | 329 | x3 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 330 | |
| emilmont | 1:fdd22bb7aa52 | 331 | /* x[0] and x[1] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 332 | in1 = (q15_t) x0; |
| emilmont | 1:fdd22bb7aa52 | 333 | in2 = (q15_t) x1; |
| emilmont | 1:fdd22bb7aa52 | 334 | |
| emilmont | 1:fdd22bb7aa52 | 335 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 336 | |
| emilmont | 1:fdd22bb7aa52 | 337 | /* y[0] and y[1] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 338 | in1 = (q15_t) c0; |
| emilmont | 1:fdd22bb7aa52 | 339 | in2 = (q15_t) c1; |
| emilmont | 1:fdd22bb7aa52 | 340 | |
| emilmont | 1:fdd22bb7aa52 | 341 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 342 | |
| emilmont | 1:fdd22bb7aa52 | 343 | /* acc0 += x[0] * y[0] + x[1] * y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 344 | acc0 = __SMLAD(input1, input2, acc0); |
| emilmont | 1:fdd22bb7aa52 | 345 | |
| emilmont | 1:fdd22bb7aa52 | 346 | /* x[1] and x[2] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 347 | in1 = (q15_t) x1; |
| emilmont | 1:fdd22bb7aa52 | 348 | in2 = (q15_t) x2; |
| emilmont | 1:fdd22bb7aa52 | 349 | |
| emilmont | 1:fdd22bb7aa52 | 350 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 351 | |
| emilmont | 1:fdd22bb7aa52 | 352 | /* acc1 += x[1] * y[0] + x[2] * y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 353 | acc1 = __SMLAD(input1, input2, acc1); |
| emilmont | 1:fdd22bb7aa52 | 354 | |
| emilmont | 1:fdd22bb7aa52 | 355 | /* x[2] and x[3] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 356 | in1 = (q15_t) x2; |
| emilmont | 1:fdd22bb7aa52 | 357 | in2 = (q15_t) x3; |
| emilmont | 1:fdd22bb7aa52 | 358 | |
| emilmont | 1:fdd22bb7aa52 | 359 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 360 | |
| emilmont | 1:fdd22bb7aa52 | 361 | /* acc2 += x[2] * y[0] + x[3] * y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 362 | acc2 = __SMLAD(input1, input2, acc2); |
| emilmont | 1:fdd22bb7aa52 | 363 | |
| emilmont | 1:fdd22bb7aa52 | 364 | /* Read x[4] sample */ |
| emilmont | 1:fdd22bb7aa52 | 365 | x0 = *(px++); |
| emilmont | 1:fdd22bb7aa52 | 366 | |
| emilmont | 1:fdd22bb7aa52 | 367 | /* x[3] and x[4] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 368 | in1 = (q15_t) x3; |
| emilmont | 1:fdd22bb7aa52 | 369 | in2 = (q15_t) x0; |
| emilmont | 1:fdd22bb7aa52 | 370 | |
| emilmont | 1:fdd22bb7aa52 | 371 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 372 | |
| emilmont | 1:fdd22bb7aa52 | 373 | /* acc3 += x[3] * y[0] + x[4] * y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 374 | acc3 = __SMLAD(input1, input2, acc3); |
| emilmont | 1:fdd22bb7aa52 | 375 | |
| emilmont | 1:fdd22bb7aa52 | 376 | /* Read y[2] sample */ |
| emilmont | 1:fdd22bb7aa52 | 377 | c0 = *py++; |
| emilmont | 1:fdd22bb7aa52 | 378 | /* Read y[3] sample */ |
| emilmont | 1:fdd22bb7aa52 | 379 | c1 = *py++; |
| emilmont | 1:fdd22bb7aa52 | 380 | |
| emilmont | 1:fdd22bb7aa52 | 381 | /* Read x[5] sample */ |
| emilmont | 1:fdd22bb7aa52 | 382 | x1 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 383 | |
| emilmont | 1:fdd22bb7aa52 | 384 | /* x[2] and x[3] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 385 | in1 = (q15_t) x2; |
| emilmont | 1:fdd22bb7aa52 | 386 | in2 = (q15_t) x3; |
| emilmont | 1:fdd22bb7aa52 | 387 | |
| emilmont | 1:fdd22bb7aa52 | 388 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 389 | |
| emilmont | 1:fdd22bb7aa52 | 390 | /* y[2] and y[3] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 391 | in1 = (q15_t) c0; |
| emilmont | 1:fdd22bb7aa52 | 392 | in2 = (q15_t) c1; |
| emilmont | 1:fdd22bb7aa52 | 393 | |
| emilmont | 1:fdd22bb7aa52 | 394 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 395 | |
| emilmont | 1:fdd22bb7aa52 | 396 | /* acc0 += x[2] * y[2] + x[3] * y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 397 | acc0 = __SMLAD(input1, input2, acc0); |
| emilmont | 1:fdd22bb7aa52 | 398 | |
| emilmont | 1:fdd22bb7aa52 | 399 | /* x[3] and x[4] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 400 | in1 = (q15_t) x3; |
| emilmont | 1:fdd22bb7aa52 | 401 | in2 = (q15_t) x0; |
| emilmont | 1:fdd22bb7aa52 | 402 | |
| emilmont | 1:fdd22bb7aa52 | 403 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 404 | |
| emilmont | 1:fdd22bb7aa52 | 405 | /* acc1 += x[3] * y[2] + x[4] * y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 406 | acc1 = __SMLAD(input1, input2, acc1); |
| emilmont | 1:fdd22bb7aa52 | 407 | |
| emilmont | 1:fdd22bb7aa52 | 408 | /* x[4] and x[5] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 409 | in1 = (q15_t) x0; |
| emilmont | 1:fdd22bb7aa52 | 410 | in2 = (q15_t) x1; |
| emilmont | 1:fdd22bb7aa52 | 411 | |
| emilmont | 1:fdd22bb7aa52 | 412 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 413 | |
| emilmont | 1:fdd22bb7aa52 | 414 | /* acc2 += x[4] * y[2] + x[5] * y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 415 | acc2 = __SMLAD(input1, input2, acc2); |
| emilmont | 1:fdd22bb7aa52 | 416 | |
| emilmont | 1:fdd22bb7aa52 | 417 | /* Read x[6] sample */ |
| emilmont | 1:fdd22bb7aa52 | 418 | x2 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 419 | |
| emilmont | 1:fdd22bb7aa52 | 420 | /* x[5] and x[6] are packed */ |
| emilmont | 1:fdd22bb7aa52 | 421 | in1 = (q15_t) x1; |
| emilmont | 1:fdd22bb7aa52 | 422 | in2 = (q15_t) x2; |
| emilmont | 1:fdd22bb7aa52 | 423 | |
| emilmont | 1:fdd22bb7aa52 | 424 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 425 | |
| emilmont | 1:fdd22bb7aa52 | 426 | /* acc3 += x[5] * y[2] + x[6] * y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 427 | acc3 = __SMLAD(input1, input2, acc3); |
| emilmont | 1:fdd22bb7aa52 | 428 | |
| emilmont | 1:fdd22bb7aa52 | 429 | } while(--k); |
| emilmont | 1:fdd22bb7aa52 | 430 | |
| emilmont | 1:fdd22bb7aa52 | 431 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
| emilmont | 1:fdd22bb7aa52 | 432 | ** No loop unrolling is used. */ |
| emilmont | 1:fdd22bb7aa52 | 433 | k = srcBLen % 0x4u; |
| emilmont | 1:fdd22bb7aa52 | 434 | |
| emilmont | 1:fdd22bb7aa52 | 435 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 436 | { |
| emilmont | 1:fdd22bb7aa52 | 437 | /* Read y[4] sample */ |
| emilmont | 1:fdd22bb7aa52 | 438 | c0 = *py++; |
| emilmont | 1:fdd22bb7aa52 | 439 | |
| emilmont | 1:fdd22bb7aa52 | 440 | /* Read x[7] sample */ |
| emilmont | 1:fdd22bb7aa52 | 441 | x3 = *px++; |
| emilmont | 1:fdd22bb7aa52 | 442 | |
| emilmont | 1:fdd22bb7aa52 | 443 | /* Perform the multiply-accumulates */ |
| emilmont | 1:fdd22bb7aa52 | 444 | /* acc0 += x[4] * y[4] */ |
| emilmont | 1:fdd22bb7aa52 | 445 | acc0 += ((q15_t) x0 * c0); |
| emilmont | 1:fdd22bb7aa52 | 446 | /* acc1 += x[5] * y[4] */ |
| emilmont | 1:fdd22bb7aa52 | 447 | acc1 += ((q15_t) x1 * c0); |
| emilmont | 1:fdd22bb7aa52 | 448 | /* acc2 += x[6] * y[4] */ |
| emilmont | 1:fdd22bb7aa52 | 449 | acc2 += ((q15_t) x2 * c0); |
| emilmont | 1:fdd22bb7aa52 | 450 | /* acc3 += x[7] * y[4] */ |
| emilmont | 1:fdd22bb7aa52 | 451 | acc3 += ((q15_t) x3 * c0); |
| emilmont | 1:fdd22bb7aa52 | 452 | |
| emilmont | 1:fdd22bb7aa52 | 453 | /* Reuse the present samples for the next MAC */ |
| emilmont | 1:fdd22bb7aa52 | 454 | x0 = x1; |
| emilmont | 1:fdd22bb7aa52 | 455 | x1 = x2; |
| emilmont | 1:fdd22bb7aa52 | 456 | x2 = x3; |
| emilmont | 1:fdd22bb7aa52 | 457 | |
| emilmont | 1:fdd22bb7aa52 | 458 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 459 | k--; |
| emilmont | 1:fdd22bb7aa52 | 460 | } |
| emilmont | 1:fdd22bb7aa52 | 461 | |
| emilmont | 1:fdd22bb7aa52 | 462 | /* Store the result in the accumulator in the destination buffer. */ |
| emilmont | 1:fdd22bb7aa52 | 463 | *pOut = (q7_t) (__SSAT(acc0 >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 464 | /* Destination pointer is updated according to the address modifier, inc */ |
| emilmont | 1:fdd22bb7aa52 | 465 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 466 | |
| emilmont | 1:fdd22bb7aa52 | 467 | *pOut = (q7_t) (__SSAT(acc1 >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 468 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 469 | |
| emilmont | 1:fdd22bb7aa52 | 470 | *pOut = (q7_t) (__SSAT(acc2 >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 471 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 472 | |
| emilmont | 1:fdd22bb7aa52 | 473 | *pOut = (q7_t) (__SSAT(acc3 >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 474 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 475 | |
| emilmont | 1:fdd22bb7aa52 | 476 | count += 4u; |
| emilmont | 1:fdd22bb7aa52 | 477 | /* Update the inputA and inputB pointers for next MAC calculation */ |
| emilmont | 1:fdd22bb7aa52 | 478 | px = pIn1 + count; |
| emilmont | 1:fdd22bb7aa52 | 479 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 480 | |
| emilmont | 1:fdd22bb7aa52 | 481 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 482 | blkCnt--; |
| emilmont | 1:fdd22bb7aa52 | 483 | } |
| emilmont | 1:fdd22bb7aa52 | 484 | |
| emilmont | 1:fdd22bb7aa52 | 485 | /* If the blockSize2 is not a multiple of 4, compute any remaining output samples here. |
| emilmont | 1:fdd22bb7aa52 | 486 | ** No loop unrolling is used. */ |
| emilmont | 1:fdd22bb7aa52 | 487 | blkCnt = blockSize2 % 0x4u; |
| emilmont | 1:fdd22bb7aa52 | 488 | |
| emilmont | 1:fdd22bb7aa52 | 489 | while(blkCnt > 0u) |
| emilmont | 1:fdd22bb7aa52 | 490 | { |
| emilmont | 1:fdd22bb7aa52 | 491 | /* Accumulator is made zero for every iteration */ |
| emilmont | 1:fdd22bb7aa52 | 492 | sum = 0; |
| emilmont | 1:fdd22bb7aa52 | 493 | |
| emilmont | 1:fdd22bb7aa52 | 494 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
| emilmont | 1:fdd22bb7aa52 | 495 | k = srcBLen >> 2u; |
| emilmont | 1:fdd22bb7aa52 | 496 | |
| emilmont | 1:fdd22bb7aa52 | 497 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
| emilmont | 1:fdd22bb7aa52 | 498 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
| emilmont | 1:fdd22bb7aa52 | 499 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 500 | { |
| emilmont | 1:fdd22bb7aa52 | 501 | /* Reading two inputs of SrcA buffer and packing */ |
| emilmont | 1:fdd22bb7aa52 | 502 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 503 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 504 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 505 | |
| emilmont | 1:fdd22bb7aa52 | 506 | /* Reading two inputs of SrcB buffer and packing */ |
| emilmont | 1:fdd22bb7aa52 | 507 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 508 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 509 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 510 | |
| emilmont | 1:fdd22bb7aa52 | 511 | /* Perform the multiply-accumulates */ |
| emilmont | 1:fdd22bb7aa52 | 512 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 513 | |
| emilmont | 1:fdd22bb7aa52 | 514 | /* Reading two inputs of SrcA buffer and packing */ |
| emilmont | 1:fdd22bb7aa52 | 515 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 516 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 517 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 518 | |
| emilmont | 1:fdd22bb7aa52 | 519 | /* Reading two inputs of SrcB buffer and packing */ |
| emilmont | 1:fdd22bb7aa52 | 520 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 521 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 522 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 523 | |
| emilmont | 1:fdd22bb7aa52 | 524 | /* Perform the multiply-accumulates */ |
| emilmont | 1:fdd22bb7aa52 | 525 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 526 | |
| emilmont | 1:fdd22bb7aa52 | 527 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 528 | k--; |
| emilmont | 1:fdd22bb7aa52 | 529 | } |
| emilmont | 1:fdd22bb7aa52 | 530 | |
| emilmont | 1:fdd22bb7aa52 | 531 | /* If the srcBLen is not a multiple of 4, compute any remaining MACs here. |
| emilmont | 1:fdd22bb7aa52 | 532 | ** No loop unrolling is used. */ |
| emilmont | 1:fdd22bb7aa52 | 533 | k = srcBLen % 0x4u; |
| emilmont | 1:fdd22bb7aa52 | 534 | |
| emilmont | 1:fdd22bb7aa52 | 535 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 536 | { |
| emilmont | 1:fdd22bb7aa52 | 537 | /* Perform the multiply-accumulates */ |
| 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 >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 546 | /* Destination pointer is updated according to the address modifier, inc */ |
| emilmont | 1:fdd22bb7aa52 | 547 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 548 | |
| emilmont | 1:fdd22bb7aa52 | 549 | /* Increment the pointer pIn1 index, count by 1 */ |
| emilmont | 1:fdd22bb7aa52 | 550 | count++; |
| emilmont | 1:fdd22bb7aa52 | 551 | |
| emilmont | 1:fdd22bb7aa52 | 552 | /* Update the inputA and inputB pointers for next MAC calculation */ |
| emilmont | 1:fdd22bb7aa52 | 553 | px = pIn1 + count; |
| emilmont | 1:fdd22bb7aa52 | 554 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 555 | |
| emilmont | 1:fdd22bb7aa52 | 556 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 557 | blkCnt--; |
| emilmont | 1:fdd22bb7aa52 | 558 | } |
| emilmont | 1:fdd22bb7aa52 | 559 | } |
| emilmont | 1:fdd22bb7aa52 | 560 | else |
| emilmont | 1:fdd22bb7aa52 | 561 | { |
| emilmont | 1:fdd22bb7aa52 | 562 | /* If the srcBLen is not a multiple of 4, |
| emilmont | 1:fdd22bb7aa52 | 563 | * the blockSize2 loop cannot be unrolled by 4 */ |
| emilmont | 1:fdd22bb7aa52 | 564 | blkCnt = blockSize2; |
| emilmont | 1:fdd22bb7aa52 | 565 | |
| emilmont | 1:fdd22bb7aa52 | 566 | while(blkCnt > 0u) |
| emilmont | 1:fdd22bb7aa52 | 567 | { |
| emilmont | 1:fdd22bb7aa52 | 568 | /* Accumulator is made zero for every iteration */ |
| emilmont | 1:fdd22bb7aa52 | 569 | sum = 0; |
| emilmont | 1:fdd22bb7aa52 | 570 | |
| emilmont | 1:fdd22bb7aa52 | 571 | /* Loop over srcBLen */ |
| emilmont | 1:fdd22bb7aa52 | 572 | k = srcBLen; |
| emilmont | 1:fdd22bb7aa52 | 573 | |
| emilmont | 1:fdd22bb7aa52 | 574 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 575 | { |
| emilmont | 1:fdd22bb7aa52 | 576 | /* Perform the multiply-accumulate */ |
| emilmont | 1:fdd22bb7aa52 | 577 | sum += ((q15_t) * px++ * *py++); |
| emilmont | 1:fdd22bb7aa52 | 578 | |
| emilmont | 1:fdd22bb7aa52 | 579 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 580 | k--; |
| emilmont | 1:fdd22bb7aa52 | 581 | } |
| emilmont | 1:fdd22bb7aa52 | 582 | |
| emilmont | 1:fdd22bb7aa52 | 583 | /* Store the result in the accumulator in the destination buffer. */ |
| emilmont | 1:fdd22bb7aa52 | 584 | *pOut = (q7_t) (__SSAT(sum >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 585 | /* Destination pointer is updated according to the address modifier, inc */ |
| emilmont | 1:fdd22bb7aa52 | 586 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 587 | |
| emilmont | 1:fdd22bb7aa52 | 588 | /* Increment the MAC count */ |
| emilmont | 1:fdd22bb7aa52 | 589 | count++; |
| emilmont | 1:fdd22bb7aa52 | 590 | |
| emilmont | 1:fdd22bb7aa52 | 591 | /* Update the inputA and inputB pointers for next MAC calculation */ |
| emilmont | 1:fdd22bb7aa52 | 592 | px = pIn1 + count; |
| emilmont | 1:fdd22bb7aa52 | 593 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 594 | |
| emilmont | 1:fdd22bb7aa52 | 595 | |
| emilmont | 1:fdd22bb7aa52 | 596 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 597 | blkCnt--; |
| emilmont | 1:fdd22bb7aa52 | 598 | } |
| emilmont | 1:fdd22bb7aa52 | 599 | } |
| emilmont | 1:fdd22bb7aa52 | 600 | |
| emilmont | 1:fdd22bb7aa52 | 601 | /* -------------------------- |
| emilmont | 1:fdd22bb7aa52 | 602 | * Initializations of stage3 |
| emilmont | 1:fdd22bb7aa52 | 603 | * -------------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 604 | |
| emilmont | 1:fdd22bb7aa52 | 605 | /* sum += x[srcALen-srcBLen+1] * y[0] + x[srcALen-srcBLen+2] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
| emilmont | 1:fdd22bb7aa52 | 606 | * sum += x[srcALen-srcBLen+2] * y[0] + x[srcALen-srcBLen+3] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] |
| emilmont | 1:fdd22bb7aa52 | 607 | * .... |
| emilmont | 1:fdd22bb7aa52 | 608 | * sum += x[srcALen-2] * y[0] + x[srcALen-1] * y[1] |
| emilmont | 1:fdd22bb7aa52 | 609 | * sum += x[srcALen-1] * y[0] |
| emilmont | 1:fdd22bb7aa52 | 610 | */ |
| emilmont | 1:fdd22bb7aa52 | 611 | |
| emilmont | 1:fdd22bb7aa52 | 612 | /* In this stage the MAC operations are decreased by 1 for every iteration. |
| emilmont | 1:fdd22bb7aa52 | 613 | The count variable holds the number of MAC operations performed */ |
| emilmont | 1:fdd22bb7aa52 | 614 | count = srcBLen - 1u; |
| emilmont | 1:fdd22bb7aa52 | 615 | |
| emilmont | 1:fdd22bb7aa52 | 616 | /* Working pointer of inputA */ |
| emilmont | 1:fdd22bb7aa52 | 617 | pSrc1 = pIn1 + (srcALen - (srcBLen - 1u)); |
| emilmont | 1:fdd22bb7aa52 | 618 | px = pSrc1; |
| emilmont | 1:fdd22bb7aa52 | 619 | |
| emilmont | 1:fdd22bb7aa52 | 620 | /* Working pointer of inputB */ |
| emilmont | 1:fdd22bb7aa52 | 621 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 622 | |
| emilmont | 1:fdd22bb7aa52 | 623 | /* ------------------- |
| emilmont | 1:fdd22bb7aa52 | 624 | * Stage3 process |
| emilmont | 1:fdd22bb7aa52 | 625 | * ------------------*/ |
| emilmont | 1:fdd22bb7aa52 | 626 | |
| emilmont | 1:fdd22bb7aa52 | 627 | while(blockSize3 > 0u) |
| emilmont | 1:fdd22bb7aa52 | 628 | { |
| emilmont | 1:fdd22bb7aa52 | 629 | /* Accumulator is made zero for every iteration */ |
| emilmont | 1:fdd22bb7aa52 | 630 | sum = 0; |
| emilmont | 1:fdd22bb7aa52 | 631 | |
| emilmont | 1:fdd22bb7aa52 | 632 | /* Apply loop unrolling and compute 4 MACs simultaneously. */ |
| emilmont | 1:fdd22bb7aa52 | 633 | k = count >> 2u; |
| emilmont | 1:fdd22bb7aa52 | 634 | |
| emilmont | 1:fdd22bb7aa52 | 635 | /* First part of the processing with loop unrolling. Compute 4 MACs at a time. |
| emilmont | 1:fdd22bb7aa52 | 636 | ** a second loop below computes MACs for the remaining 1 to 3 samples. */ |
| emilmont | 1:fdd22bb7aa52 | 637 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 638 | { |
| emilmont | 1:fdd22bb7aa52 | 639 | /* x[srcALen - srcBLen + 1] , x[srcALen - srcBLen + 2] */ |
| emilmont | 1:fdd22bb7aa52 | 640 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 641 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 642 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 643 | |
| emilmont | 1:fdd22bb7aa52 | 644 | /* y[0] , y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 645 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 646 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 647 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 648 | |
| emilmont | 1:fdd22bb7aa52 | 649 | /* sum += x[srcALen - srcBLen + 1] * y[0] */ |
| emilmont | 1:fdd22bb7aa52 | 650 | /* sum += x[srcALen - srcBLen + 2] * y[1] */ |
| emilmont | 1:fdd22bb7aa52 | 651 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 652 | |
| emilmont | 1:fdd22bb7aa52 | 653 | /* x[srcALen - srcBLen + 3] , x[srcALen - srcBLen + 4] */ |
| emilmont | 1:fdd22bb7aa52 | 654 | in1 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 655 | in2 = (q15_t) * px++; |
| emilmont | 1:fdd22bb7aa52 | 656 | input1 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 657 | |
| emilmont | 1:fdd22bb7aa52 | 658 | /* y[2] , y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 659 | in1 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 660 | in2 = (q15_t) * py++; |
| emilmont | 1:fdd22bb7aa52 | 661 | input2 = ((q31_t) in1 & 0x0000FFFF) | ((q31_t) in2 << 16); |
| emilmont | 1:fdd22bb7aa52 | 662 | |
| emilmont | 1:fdd22bb7aa52 | 663 | /* sum += x[srcALen - srcBLen + 3] * y[2] */ |
| emilmont | 1:fdd22bb7aa52 | 664 | /* sum += x[srcALen - srcBLen + 4] * y[3] */ |
| emilmont | 1:fdd22bb7aa52 | 665 | sum = __SMLAD(input1, input2, sum); |
| emilmont | 1:fdd22bb7aa52 | 666 | |
| emilmont | 1:fdd22bb7aa52 | 667 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 668 | k--; |
| emilmont | 1:fdd22bb7aa52 | 669 | } |
| emilmont | 1:fdd22bb7aa52 | 670 | |
| emilmont | 1:fdd22bb7aa52 | 671 | /* If the count is not a multiple of 4, compute any remaining MACs here. |
| emilmont | 1:fdd22bb7aa52 | 672 | ** No loop unrolling is used. */ |
| emilmont | 1:fdd22bb7aa52 | 673 | k = count % 0x4u; |
| emilmont | 1:fdd22bb7aa52 | 674 | |
| emilmont | 1:fdd22bb7aa52 | 675 | while(k > 0u) |
| emilmont | 1:fdd22bb7aa52 | 676 | { |
| emilmont | 1:fdd22bb7aa52 | 677 | /* Perform the multiply-accumulates */ |
| emilmont | 1:fdd22bb7aa52 | 678 | sum += ((q15_t) * px++ * *py++); |
| emilmont | 1:fdd22bb7aa52 | 679 | |
| emilmont | 1:fdd22bb7aa52 | 680 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 681 | k--; |
| emilmont | 1:fdd22bb7aa52 | 682 | } |
| emilmont | 1:fdd22bb7aa52 | 683 | |
| emilmont | 1:fdd22bb7aa52 | 684 | /* Store the result in the accumulator in the destination buffer. */ |
| emilmont | 1:fdd22bb7aa52 | 685 | *pOut = (q7_t) (__SSAT(sum >> 7, 8)); |
| emilmont | 1:fdd22bb7aa52 | 686 | /* Destination pointer is updated according to the address modifier, inc */ |
| emilmont | 1:fdd22bb7aa52 | 687 | pOut += inc; |
| emilmont | 1:fdd22bb7aa52 | 688 | |
| emilmont | 1:fdd22bb7aa52 | 689 | /* Update the inputA and inputB pointers for next MAC calculation */ |
| emilmont | 1:fdd22bb7aa52 | 690 | px = ++pSrc1; |
| emilmont | 1:fdd22bb7aa52 | 691 | py = pIn2; |
| emilmont | 1:fdd22bb7aa52 | 692 | |
| emilmont | 1:fdd22bb7aa52 | 693 | /* Decrement the MAC count */ |
| emilmont | 1:fdd22bb7aa52 | 694 | count--; |
| emilmont | 1:fdd22bb7aa52 | 695 | |
| emilmont | 1:fdd22bb7aa52 | 696 | /* Decrement the loop counter */ |
| emilmont | 1:fdd22bb7aa52 | 697 | blockSize3--; |
| emilmont | 1:fdd22bb7aa52 | 698 | } |
| emilmont | 1:fdd22bb7aa52 | 699 | |
| emilmont | 1:fdd22bb7aa52 | 700 | #else |
| emilmont | 1:fdd22bb7aa52 | 701 | |
| emilmont | 1:fdd22bb7aa52 | 702 | /* Run the below code for Cortex-M0 */ |
| emilmont | 1:fdd22bb7aa52 | 703 | |
| emilmont | 1:fdd22bb7aa52 | 704 | q7_t *pIn1 = pSrcA; /* inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 705 | q7_t *pIn2 = pSrcB + (srcBLen - 1u); /* inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 706 | q31_t sum; /* Accumulator */ |
| emilmont | 1:fdd22bb7aa52 | 707 | uint32_t i = 0u, j; /* loop counters */ |
| emilmont | 1:fdd22bb7aa52 | 708 | uint32_t inv = 0u; /* Reverse order flag */ |
| emilmont | 1:fdd22bb7aa52 | 709 | uint32_t tot = 0u; /* Length */ |
| emilmont | 1:fdd22bb7aa52 | 710 | |
| emilmont | 1:fdd22bb7aa52 | 711 | /* The algorithm implementation is based on the lengths of the inputs. */ |
| emilmont | 1:fdd22bb7aa52 | 712 | /* srcB is always made to slide across srcA. */ |
| emilmont | 1:fdd22bb7aa52 | 713 | /* So srcBLen is always considered as shorter or equal to srcALen */ |
| emilmont | 1:fdd22bb7aa52 | 714 | /* But CORR(x, y) is reverse of CORR(y, x) */ |
| emilmont | 1:fdd22bb7aa52 | 715 | /* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 716 | /* and a varaible, inv is set to 1 */ |
| emilmont | 1:fdd22bb7aa52 | 717 | /* If lengths are not equal then zero pad has to be done to make the two |
| emilmont | 1:fdd22bb7aa52 | 718 | * inputs of same length. But to improve the performance, we include zeroes |
| emilmont | 1:fdd22bb7aa52 | 719 | * in the output instead of zero padding either of the the inputs*/ |
| emilmont | 1:fdd22bb7aa52 | 720 | /* If srcALen > srcBLen, (srcALen - srcBLen) zeroes has to included in the |
| emilmont | 1:fdd22bb7aa52 | 721 | * starting of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 722 | /* If srcALen < srcBLen, (srcALen - srcBLen) zeroes has to included in the |
| emilmont | 1:fdd22bb7aa52 | 723 | * ending of the output buffer */ |
| emilmont | 1:fdd22bb7aa52 | 724 | /* Once the zero padding is done the remaining of the output is calcualted |
| emilmont | 1:fdd22bb7aa52 | 725 | * using convolution but with the shorter signal time shifted. */ |
| emilmont | 1:fdd22bb7aa52 | 726 | |
| emilmont | 1:fdd22bb7aa52 | 727 | /* Calculate the length of the remaining sequence */ |
| emilmont | 1:fdd22bb7aa52 | 728 | tot = ((srcALen + srcBLen) - 2u); |
| emilmont | 1:fdd22bb7aa52 | 729 | |
| emilmont | 1:fdd22bb7aa52 | 730 | if(srcALen > srcBLen) |
| emilmont | 1:fdd22bb7aa52 | 731 | { |
| emilmont | 1:fdd22bb7aa52 | 732 | /* Calculating the number of zeros to be padded to the output */ |
| emilmont | 1:fdd22bb7aa52 | 733 | j = srcALen - srcBLen; |
| emilmont | 1:fdd22bb7aa52 | 734 | |
| emilmont | 1:fdd22bb7aa52 | 735 | /* Initialise the pointer after zero padding */ |
| emilmont | 1:fdd22bb7aa52 | 736 | pDst += j; |
| emilmont | 1:fdd22bb7aa52 | 737 | } |
| emilmont | 1:fdd22bb7aa52 | 738 | |
| emilmont | 1:fdd22bb7aa52 | 739 | else if(srcALen < srcBLen) |
| emilmont | 1:fdd22bb7aa52 | 740 | { |
| emilmont | 1:fdd22bb7aa52 | 741 | /* Initialization to inputB pointer */ |
| emilmont | 1:fdd22bb7aa52 | 742 | pIn1 = pSrcB; |
| emilmont | 1:fdd22bb7aa52 | 743 | |
| emilmont | 1:fdd22bb7aa52 | 744 | /* Initialization to the end of inputA pointer */ |
| emilmont | 1:fdd22bb7aa52 | 745 | pIn2 = pSrcA + (srcALen - 1u); |
| emilmont | 1:fdd22bb7aa52 | 746 | |
| emilmont | 1:fdd22bb7aa52 | 747 | /* Initialisation of the pointer after zero padding */ |
| emilmont | 1:fdd22bb7aa52 | 748 | pDst = pDst + tot; |
| emilmont | 1:fdd22bb7aa52 | 749 | |
| emilmont | 1:fdd22bb7aa52 | 750 | /* Swapping the lengths */ |
| emilmont | 1:fdd22bb7aa52 | 751 | j = srcALen; |
| emilmont | 1:fdd22bb7aa52 | 752 | srcALen = srcBLen; |
| emilmont | 1:fdd22bb7aa52 | 753 | srcBLen = j; |
| emilmont | 1:fdd22bb7aa52 | 754 | |
| emilmont | 1:fdd22bb7aa52 | 755 | /* Setting the reverse flag */ |
| emilmont | 1:fdd22bb7aa52 | 756 | inv = 1; |
| emilmont | 1:fdd22bb7aa52 | 757 | |
| emilmont | 1:fdd22bb7aa52 | 758 | } |
| emilmont | 1:fdd22bb7aa52 | 759 | |
| emilmont | 1:fdd22bb7aa52 | 760 | /* Loop to calculate convolution for output length number of times */ |
| emilmont | 1:fdd22bb7aa52 | 761 | for (i = 0u; i <= tot; i++) |
| emilmont | 1:fdd22bb7aa52 | 762 | { |
| emilmont | 1:fdd22bb7aa52 | 763 | /* Initialize sum with zero to carry on MAC operations */ |
| emilmont | 1:fdd22bb7aa52 | 764 | sum = 0; |
| emilmont | 1:fdd22bb7aa52 | 765 | |
| emilmont | 1:fdd22bb7aa52 | 766 | /* Loop to perform MAC operations according to convolution equation */ |
| emilmont | 1:fdd22bb7aa52 | 767 | for (j = 0u; j <= i; j++) |
| emilmont | 1:fdd22bb7aa52 | 768 | { |
| emilmont | 1:fdd22bb7aa52 | 769 | /* Check the array limitations */ |
| emilmont | 1:fdd22bb7aa52 | 770 | if((((i - j) < srcBLen) && (j < srcALen))) |
| emilmont | 1:fdd22bb7aa52 | 771 | { |
| emilmont | 1:fdd22bb7aa52 | 772 | /* z[i] += x[i-j] * y[j] */ |
| emilmont | 1:fdd22bb7aa52 | 773 | sum += ((q15_t) pIn1[j] * pIn2[-((int32_t) i - j)]); |
| emilmont | 1:fdd22bb7aa52 | 774 | } |
| emilmont | 1:fdd22bb7aa52 | 775 | } |
| emilmont | 1:fdd22bb7aa52 | 776 | /* Store the output in the destination buffer */ |
| emilmont | 1:fdd22bb7aa52 | 777 | if(inv == 1) |
| emilmont | 1:fdd22bb7aa52 | 778 | *pDst-- = (q7_t) __SSAT((sum >> 7u), 8u); |
| emilmont | 1:fdd22bb7aa52 | 779 | else |
| emilmont | 1:fdd22bb7aa52 | 780 | *pDst++ = (q7_t) __SSAT((sum >> 7u), 8u); |
| emilmont | 1:fdd22bb7aa52 | 781 | } |
| emilmont | 1:fdd22bb7aa52 | 782 | |
| emilmont | 1:fdd22bb7aa52 | 783 | #endif /* #ifndef ARM_MATH_CM0 */ |
| emilmont | 1:fdd22bb7aa52 | 784 | |
| emilmont | 1:fdd22bb7aa52 | 785 | } |
| emilmont | 1:fdd22bb7aa52 | 786 | |
| emilmont | 1:fdd22bb7aa52 | 787 | /** |
| emilmont | 1:fdd22bb7aa52 | 788 | * @} end of Corr group |
| emilmont | 1:fdd22bb7aa52 | 789 | */ |