diff options
Diffstat (limited to 'Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c')
-rw-r--r-- | Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c | 237 |
1 files changed, 119 insertions, 118 deletions
diff --git a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c index 4a006aa89..a5840b733 100644 --- a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c +++ b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_correlate_fast_q31.c @@ -3,13 +3,13 @@ * Title: arm_correlate_fast_q31.c * Description: Fast Q31 Correlation * - * $Date: 27. January 2017 - * $Revision: V.1.5.1 + * $Date: 18. March 2019 + * $Revision: V1.6.0 * * Target Processor: Cortex-M cores * -------------------------------------------------------------------- */ /* - * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved. + * Copyright (C) 2010-2019 ARM Limited or its affiliates. All rights reserved. * * SPDX-License-Identifier: Apache-2.0 * @@ -29,61 +29,58 @@ #include "arm_math.h" /** - * @ingroup groupFilters + @ingroup groupFilters */ /** - * @addtogroup Corr - * @{ + @addtogroup Corr + @{ */ /** - * @brief Correlation of Q31 sequences (fast version) for Cortex-M3 and Cortex-M4. - * @param[in] *pSrcA points to the first input sequence. - * @param[in] srcALen length of the first input sequence. - * @param[in] *pSrcB points to the second input sequence. - * @param[in] srcBLen length of the second input sequence. - * @param[out] *pDst points to the location where the output result is written. Length 2 * max(srcALen, srcBLen) - 1. - * @return none. - * - * @details - * <b>Scaling and Overflow Behavior:</b> - * - * \par - * This function is optimized for speed at the expense of fixed-point precision and overflow protection. - * The result of each 1.31 x 1.31 multiplication is truncated to 2.30 format. - * These intermediate results are accumulated in a 32-bit register in 2.30 format. - * Finally, the accumulator is saturated and converted to a 1.31 result. - * - * \par - * 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. - * In order to avoid overflows completely the input signals must be scaled down. - * The input signals should be scaled down to avoid intermediate overflows. - * Scale down one of the inputs by 1/min(srcALen, srcBLen)to avoid overflows since a - * maximum of min(srcALen, srcBLen) number of additions is carried internally. - * - * \par - * See <code>arm_correlate_q31()</code> for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. + @brief Correlation of Q31 sequences (fast version). + @param[in] pSrcA points to the first input sequence + @param[in] srcALen length of the first input sequence + @param[in] pSrcB points to the second input sequence + @param[in] srcBLen length of the second input sequence + @param[out] pDst points to the location where the output result is written. Length 2 * max(srcALen, srcBLen) - 1. + @return none + + @par Scaling and Overflow Behavior + This function is optimized for speed at the expense of fixed-point precision and overflow protection. + The result of each 1.31 x 1.31 multiplication is truncated to 2.30 format. + These intermediate results are accumulated in a 32-bit register in 2.30 format. + Finally, the accumulator is saturated and converted to a 1.31 result. + @par + 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. + In order to avoid overflows completely the input signals must be scaled down. + The input signals should be scaled down to avoid intermediate overflows. + Scale down one of the inputs by 1/min(srcALen, srcBLen)to avoid overflows since a + maximum of min(srcALen, srcBLen) number of additions is carried internally. + + @remark + Refer to \ref arm_correlate_q31() for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. */ void arm_correlate_fast_q31( - q31_t * pSrcA, - uint32_t srcALen, - q31_t * pSrcB, - uint32_t srcBLen, - q31_t * pDst) + const q31_t * pSrcA, + uint32_t srcALen, + const q31_t * pSrcB, + uint32_t srcBLen, + q31_t * pDst) { - q31_t *pIn1; /* inputA pointer */ - q31_t *pIn2; /* inputB pointer */ - q31_t *pOut = pDst; /* output pointer */ - q31_t *px; /* Intermediate inputA pointer */ - q31_t *py; /* Intermediate inputB pointer */ - q31_t *pSrc1; /* Intermediate pointers */ - q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ - q31_t x0, x1, x2, x3, c0; /* temporary variables for holding input and coefficient values */ - uint32_t j, k = 0U, count, blkCnt, outBlockSize, blockSize1, blockSize2, blockSize3; /* loop counter */ - int32_t inc = 1; /* Destination address modifier */ - + const q31_t *pIn1; /* InputA pointer */ + const q31_t *pIn2; /* InputB pointer */ + q31_t *pOut = pDst; /* Output pointer */ + const q31_t *px; /* Intermediate inputA pointer */ + const q31_t *py; /* Intermediate inputB pointer */ + const q31_t *pSrc1; /* Intermediate pointers */ + q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ + q31_t x0, x1, x2, x3, c0; /* Temporary variables for holding input and coefficient values */ + uint32_t blockSize1, blockSize2, blockSize3; /* Loop counters */ + uint32_t j, k, count, blkCnt; /* Loop counters */ + uint32_t outBlockSize; + int32_t inc = 1; /* Destination address modifier */ /* The algorithm implementation is based on the lengths of the inputs. */ /* srcB is always made to slide across srcA. */ @@ -91,10 +88,10 @@ void arm_correlate_fast_q31( if (srcALen >= srcBLen) { /* Initialization of inputA pointer */ - pIn1 = (pSrcA); + pIn1 = pSrcA; /* Initialization of inputB pointer */ - pIn2 = (pSrcB); + pIn2 = pSrcB; /* Number of output samples is calculated */ outBlockSize = (2U * srcALen) - 1U; @@ -112,10 +109,10 @@ void arm_correlate_fast_q31( else { /* Initialization of inputA pointer */ - pIn1 = (pSrcB); + pIn1 = pSrcB; /* Initialization of inputB pointer */ - pIn2 = (pSrcA); + pIn2 = pSrcA; /* srcBLen is always considered as shorter or equal to srcALen */ j = srcBLen; @@ -132,14 +129,15 @@ void arm_correlate_fast_q31( } /* The function is internally - * divided into three parts according to the number of multiplications that has to be - * taken place between inputA samples and inputB samples. In the first part of the + * divided into three stages according to the number of multiplications that has to be + * taken place between inputA samples and inputB samples. In the first stage of the * algorithm, the multiplications increase by one for every iteration. - * In the second part of the algorithm, srcBLen number of multiplications are done. - * In the third part of the algorithm, the multiplications decrease by one - * for every iteration.*/ + * In the second stage of the algorithm, srcBLen number of multiplications are done. + * In the third stage of the algorithm, the multiplications decrease by one + * for every iteration. */ + /* The algorithm is implemented in three stages. - * The loop counters of each stage is initiated here. */ + The loop counters of each stage is initiated here. */ blockSize1 = srcBLen - 1U; blockSize2 = srcALen - (srcBLen - 1U); blockSize3 = blockSize1; @@ -176,7 +174,7 @@ void arm_correlate_fast_q31( sum = 0; /* Apply loop unrolling and compute 4 MACs simultaneously. */ - k = count >> 2; + k = count >> 2U; /* First part of the processing with loop unrolling. Compute 4 MACs at a time. ** a second loop below computes MACs for the remaining 1 to 3 samples. */ @@ -184,18 +182,21 @@ void arm_correlate_fast_q31( { /* x[0] * y[srcBLen - 4] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* x[1] * y[srcBLen - 3] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* x[2] * y[srcBLen - 2] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* x[3] * y[srcBLen - 1] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -205,12 +206,12 @@ void arm_correlate_fast_q31( while (k > 0U) { - /* Perform the multiply-accumulates */ + /* Perform the multiply-accumulate */ /* x[0] * y[srcBLen - 1] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -223,10 +224,10 @@ void arm_correlate_fast_q31( py = pSrc1 - count; px = pIn1; - /* Increment the MAC count */ + /* Increment MAC count */ count++; - /* Decrement the loop counter */ + /* Decrement loop counter */ blockSize1--; } @@ -270,9 +271,9 @@ void arm_correlate_fast_q31( acc3 = 0; /* read x[0], x[1], x[2] samples */ - x0 = *(px++); - x1 = *(px++); - x2 = *(px++); + x0 = *px++; + x1 = *px++; + x2 = *px++; /* Apply loop unrolling and compute 4 MACs simultaneously. */ k = srcBLen >> 2U; @@ -282,10 +283,9 @@ void arm_correlate_fast_q31( do { /* Read y[0] sample */ - c0 = *(py++); - + c0 = *py++; /* Read x[3] sample */ - x3 = *(px++); + x3 = *px++; /* Perform the multiply-accumulate */ /* acc0 += x[0] * y[0] */ @@ -297,13 +297,13 @@ void arm_correlate_fast_q31( /* acc3 += x[3] * y[0] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); - /* Read y[1] sample */ - c0 = *(py++); + /* Read y[1] sample */ + c0 = *py++; /* Read x[4] sample */ - x0 = *(px++); + x0 = *px++; - /* Perform the multiply-accumulates */ + /* Perform the multiply-accumulate */ /* acc0 += x[1] * y[1] */ acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x1 * c0)) >> 32); /* acc1 += x[2] * y[1] */ @@ -313,11 +313,11 @@ void arm_correlate_fast_q31( /* acc3 += x[4] * y[1] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x0 * c0)) >> 32); - /* Read y[2] sample */ - c0 = *(py++); + /* Read y[2] sample */ + c0 = *py++; /* Read x[5] sample */ - x1 = *(px++); + x1 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[2] * y[2] */ @@ -329,11 +329,11 @@ void arm_correlate_fast_q31( /* acc3 += x[5] * y[2] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x1 * c0)) >> 32); - /* Read y[3] sample */ - c0 = *(py++); + /* Read y[3] sample */ + c0 = *py++; /* Read x[6] sample */ - x2 = *(px++); + x2 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[3] * y[3] */ @@ -355,10 +355,9 @@ void arm_correlate_fast_q31( while (k > 0U) { /* Read y[4] sample */ - c0 = *(py++); - + c0 = *py++; /* Read x[7] sample */ - x3 = *(px++); + x3 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[4] * y[4] */ @@ -375,7 +374,7 @@ void arm_correlate_fast_q31( x1 = x2; x2 = x3; - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -400,8 +399,7 @@ void arm_correlate_fast_q31( px = pIn1 + count; py = pIn2; - - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } @@ -423,15 +421,15 @@ void arm_correlate_fast_q31( { /* Perform the multiply-accumulates */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -443,9 +441,9 @@ void arm_correlate_fast_q31( { /* Perform the multiply-accumulate */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -454,15 +452,14 @@ void arm_correlate_fast_q31( /* Destination pointer is updated according to the address modifier, inc */ pOut += inc; - /* Increment the MAC count */ + /* Increment MAC count */ count++; /* Update the inputA and inputB pointers for next MAC calculation */ px = pIn1 + count; py = pIn2; - - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } } @@ -477,16 +474,16 @@ void arm_correlate_fast_q31( /* Accumulator is made zero for every iteration */ sum = 0; - /* Loop over srcBLen */ + /* srcBLen number of MACS should be performed */ k = srcBLen; while (k > 0U) { /* Perform the multiply-accumulate */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -495,18 +492,19 @@ void arm_correlate_fast_q31( /* Destination pointer is updated according to the address modifier, inc */ pOut += inc; - /* Increment the MAC count */ + /* Increment MAC count */ count++; /* Update the inputA and inputB pointers for next MAC calculation */ px = pIn1 + count; py = pIn2; - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } } + /* -------------------------- * Initializations of stage3 * -------------------------*/ @@ -545,21 +543,24 @@ void arm_correlate_fast_q31( ** a second loop below computes MACs for the remaining 1 to 3 samples. */ while (k > 0U) { - /* Perform the multiply-accumulates */ + /* Perform the multiply-accumulate */ /* sum += x[srcALen - srcBLen + 4] * y[3] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* sum += x[srcALen - srcBLen + 3] * y[2] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* sum += x[srcALen - srcBLen + 2] * y[1] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); + /* sum += x[srcALen - srcBLen + 1] * y[0] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -569,11 +570,11 @@ void arm_correlate_fast_q31( while (k > 0U) { - /* Perform the multiply-accumulates */ + /* Perform the multiply-accumulate */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py++))) >> 32); + ((q63_t) *px++ * (*py++))) >> 32); - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } @@ -586,15 +587,15 @@ void arm_correlate_fast_q31( px = ++pSrc1; py = pIn2; - /* Decrement the MAC count */ + /* Decrement MAC count */ count--; - /* Decrement the loop counter */ + /* Decrement loop counter */ blockSize3--; } } /** - * @} end of Corr group + @} end of Corr group */ |