diff options
Diffstat (limited to 'Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c')
-rw-r--r-- | Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c | 193 |
1 files changed, 93 insertions, 100 deletions
diff --git a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c index ce3e33473..e87eddc73 100644 --- a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c +++ b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_conv_fast_q31.c @@ -3,13 +3,13 @@ * Title: arm_conv_fast_q31.c * Description: Fast Q31 Convolution * - * $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 * @@ -29,57 +29,54 @@ #include "arm_math.h" /** - * @ingroup groupFilters + @ingroup groupFilters */ /** - * @addtogroup Conv - * @{ + @addtogroup Conv + @{ */ /** - * @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 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. - * Scale down the inputs by log2(min(srcALen, srcBLen)) (log2 is read as log to the base 2) times to avoid overflows, - * as maximum of min(srcALen, srcBLen) number of additions are carried internally. - * - * \par - * See <code>arm_conv_q31()</code> for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. + @brief Convolution 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 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. + Scale down the inputs by log2(min(srcALen, srcBLen)) (log2 is read as log to the base 2) times to avoid overflows, + as maximum of min(srcALen, srcBLen) number of additions are carried internally. + @remark + Refer to \ref arm_conv_q31() for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. */ void arm_conv_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, *pSrc2; /* Intermediate pointers */ - q31_t sum, acc0, acc1, acc2, acc3; /* Accumulator */ - q31_t x0, x1, x2, x3, c0; /* Temporary variables to hold state and coefficient values */ - uint32_t j, k, count, blkCnt, blockSize1, blockSize2, blockSize3; /* loop counter */ + 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, *pSrc2; /* Intermediate pointers */ + q31_t sum, acc0, acc1, acc2, acc3; /* Accumulators */ + q31_t x0, x1, x2, x3, c0; /* Temporary variables to hold state and coefficient values */ + uint32_t blockSize1, blockSize2, blockSize3; /* Loop counters */ + uint32_t j, k, count, blkCnt; /* Loop counters */ /* The algorithm implementation is based on the lengths of the inputs. */ /* srcB is always made to slide across srcA. */ @@ -161,21 +158,21 @@ void arm_conv_fast_q31( { /* x[0] * y[srcBLen - 1] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* x[1] * y[srcBLen - 2] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* x[2] * y[srcBLen - 3] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* x[3] * y[srcBLen - 4] */ 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--; } @@ -187,9 +184,9 @@ void arm_conv_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--; } @@ -200,10 +197,10 @@ void arm_conv_fast_q31( py = pIn2 + count; px = pIn1; - /* Increment the MAC count */ + /* Increment MAC count */ count++; - /* Decrement the loop counter */ + /* Decrement loop counter */ blockSize1--; } @@ -248,9 +245,9 @@ void arm_conv_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; @@ -260,29 +257,25 @@ void arm_conv_fast_q31( do { /* Read y[srcBLen - 1] sample */ - c0 = *(py--); - + c0 = *py--; /* Read x[3] sample */ - x3 = *(px++); + x3 = *px++; - /* Perform the multiply-accumulates */ + /* Perform the multiply-accumulate */ /* acc0 += x[0] * y[srcBLen - 1] */ acc0 = (q31_t) ((((q63_t) acc0 << 32) + ((q63_t) x0 * c0)) >> 32); - /* acc1 += x[1] * y[srcBLen - 1] */ acc1 = (q31_t) ((((q63_t) acc1 << 32) + ((q63_t) x1 * c0)) >> 32); - /* acc2 += x[2] * y[srcBLen - 1] */ acc2 = (q31_t) ((((q63_t) acc2 << 32) + ((q63_t) x2 * c0)) >> 32); - /* acc3 += x[3] * y[srcBLen - 1] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x3 * c0)) >> 32); - /* Read y[srcBLen - 2] sample */ - c0 = *(py--); + /* Read y[srcBLen - 2] sample */ + c0 = *py--; /* Read x[4] sample */ - x0 = *(px++); + x0 = *px++; /* Perform the multiply-accumulate */ /* acc0 += x[1] * y[srcBLen - 2] */ @@ -294,11 +287,11 @@ void arm_conv_fast_q31( /* acc3 += x[4] * y[srcBLen - 2] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x0 * c0)) >> 32); - /* Read y[srcBLen - 3] sample */ - c0 = *(py--); + /* Read y[srcBLen - 3] sample */ + c0 = *py--; /* Read x[5] sample */ - x1 = *(px++); + x1 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[2] * y[srcBLen - 3] */ @@ -310,11 +303,11 @@ void arm_conv_fast_q31( /* acc3 += x[5] * y[srcBLen - 3] */ acc3 = (q31_t) ((((q63_t) acc3 << 32) + ((q63_t) x1 * c0)) >> 32); - /* Read y[srcBLen - 4] sample */ - c0 = *(py--); + /* Read y[srcBLen - 4] sample */ + c0 = *py--; /* Read x[6] sample */ - x2 = *(px++); + x2 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[3] * y[srcBLen - 4] */ @@ -336,10 +329,9 @@ void arm_conv_fast_q31( while (k > 0U) { /* Read y[srcBLen - 5] sample */ - c0 = *(py--); - + c0 = *py--; /* Read x[7] sample */ - x3 = *(px++); + x3 = *px++; /* Perform the multiply-accumulates */ /* acc0 += x[4] * y[srcBLen - 5] */ @@ -356,11 +348,11 @@ void arm_conv_fast_q31( x1 = x2; x2 = x3; - /* Decrement the loop counter */ + /* Decrement loop counter */ k--; } - /* Store the results in the accumulators in the destination buffer. */ + /* Store the result in the accumulator in the destination buffer. */ *pOut++ = (q31_t) (acc0 << 1); *pOut++ = (q31_t) (acc1 << 1); *pOut++ = (q31_t) (acc2 << 1); @@ -373,7 +365,7 @@ void arm_conv_fast_q31( px = pIn1 + count; py = pSrc2; - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } @@ -395,15 +387,15 @@ void arm_conv_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--; } @@ -415,23 +407,23 @@ void arm_conv_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--; } /* Store the result in the accumulator in the destination buffer. */ *pOut++ = sum << 1; - /* Increment the MAC count */ + /* Increment MAC count */ count++; /* Update the inputA and inputB pointers for next MAC calculation */ px = pIn1 + count; py = pSrc2; - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } } @@ -453,23 +445,23 @@ void arm_conv_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--; } /* Store the result in the accumulator in the destination buffer. */ *pOut++ = sum << 1; - /* Increment the MAC count */ + /* Increment MAC count */ count++; /* Update the inputA and inputB pointers for next MAC calculation */ px = pIn1 + count; py = pSrc2; - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } } @@ -513,23 +505,24 @@ void arm_conv_fast_q31( ** a second loop below computes MACs for the remaining 1 to 3 samples. */ while (k > 0U) { + /* Perform the multiply-accumulate */ /* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */ sum = (q31_t) ((((q63_t) sum << 32) + - ((q63_t) * px++ * (*py--))) >> 32); + ((q63_t) *px++ * (*py--))) >> 32); /* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */ 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--; } @@ -541,9 +534,9 @@ void arm_conv_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--; } @@ -554,12 +547,12 @@ void arm_conv_fast_q31( px = ++pSrc1; py = pSrc2; - /* Decrement the loop counter */ + /* Decrement loop counter */ blockSize3--; } } /** - * @} end of Conv group + @} end of Conv group */ |