pigweed / third_party / github / STMicroelectronics / cmsis_core / be0e91fb4ba19a2ee427ba8b6b75b43eb02b2810 / . / DSP / Source / FilteringFunctions / arm_lms_norm_q15.c

/* ---------------------------------------------------------------------- | |

* Project: CMSIS DSP Library | |

* Title: arm_lms_norm_q15.c | |

* Description: Processing function for Q15 normalized LMS filter | |

* | |

* $Date: 18. March 2019 | |

* $Revision: V1.6.0 | |

* | |

* Target Processor: Cortex-M cores | |

* -------------------------------------------------------------------- */ | |

/* | |

* Copyright (C) 2010-2019 ARM Limited or its affiliates. All rights reserved. | |

* | |

* SPDX-License-Identifier: Apache-2.0 | |

* | |

* Licensed under the Apache License, Version 2.0 (the License); you may | |

* not use this file except in compliance with the License. | |

* You may obtain a copy of the License at | |

* | |

* www.apache.org/licenses/LICENSE-2.0 | |

* | |

* Unless required by applicable law or agreed to in writing, software | |

* distributed under the License is distributed on an AS IS BASIS, WITHOUT | |

* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |

* See the License for the specific language governing permissions and | |

* limitations under the License. | |

*/ | |

#include "arm_math.h" | |

/** | |

@ingroup groupFilters | |

*/ | |

/** | |

@addtogroup LMS_NORM | |

@{ | |

*/ | |

/** | |

@brief Processing function for Q15 normalized LMS filter. | |

@param[in] S points to an instance of the Q15 normalized LMS filter structure | |

@param[in] pSrc points to the block of input data | |

@param[in] pRef points to the block of reference data | |

@param[out] pOut points to the block of output data | |

@param[out] pErr points to the block of error data | |

@param[in] blockSize number of samples to process | |

@return none | |

@par Scaling and Overflow Behavior | |

The function is implemented using a 64-bit internal accumulator. | |

Both coefficients and state variables are represented in 1.15 format and | |

multiplications yield a 2.30 result. The 2.30 intermediate results are | |

accumulated in a 64-bit accumulator in 34.30 format. | |

There is no risk of internal overflow with this approach and the full | |

precision of intermediate multiplications is preserved. After all additions | |

have been performed, the accumulator is truncated to 34.15 format by | |

discarding low 15 bits. Lastly, the accumulator is saturated to yield a | |

result in 1.15 format. | |

@par | |

In this filter, filter coefficients are updated for each sample and the | |

updation of filter cofficients are saturted. | |

*/ | |

void arm_lms_norm_q15( | |

arm_lms_norm_instance_q15 * S, | |

const q15_t * pSrc, | |

q15_t * pRef, | |

q15_t * pOut, | |

q15_t * pErr, | |

uint32_t blockSize) | |

{ | |

q15_t *pState = S->pState; /* State pointer */ | |

q15_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */ | |

q15_t *pStateCurnt; /* Points to the current sample of the state */ | |

q15_t *px, *pb; /* Temporary pointers for state and coefficient buffers */ | |

q15_t mu = S->mu; /* Adaptive factor */ | |

uint32_t numTaps = S->numTaps; /* Number of filter coefficients in the filter */ | |

uint32_t tapCnt, blkCnt; /* Loop counters */ | |

q63_t acc; /* Accumulator */ | |

q31_t energy; /* Energy of the input */ | |

q15_t e = 0, d = 0; /* Error, reference data sample */ | |

q15_t w = 0, in; /* Weight factor and state */ | |

q15_t x0; /* Temporary variable to hold input sample */ | |

q15_t errorXmu, oneByEnergy; /* Temporary variables to store error and mu product and reciprocal of energy */ | |

q15_t postShift; /* Post shift to be applied to weight after reciprocal calculation */ | |

q31_t coef; /* Temporary variable for coefficient */ | |

q31_t acc_l, acc_h; /* Temporary input */ | |

int32_t lShift = (15 - (int32_t) S->postShift); /* Post shift */ | |

int32_t uShift = (32 - lShift); | |

energy = S->energy; | |

x0 = S->x0; | |

/* S->pState points to buffer which contains previous frame (numTaps - 1) samples */ | |

/* pStateCurnt points to the location where the new input data should be written */ | |

pStateCurnt = &(S->pState[(numTaps - 1U)]); | |

/* initialise loop count */ | |

blkCnt = blockSize; | |

while (blkCnt > 0U) | |

{ | |

/* Copy the new input sample into the state buffer */ | |

*pStateCurnt++ = *pSrc; | |

/* Initialize pState pointer */ | |

px = pState; | |

/* Initialize coefficient pointer */ | |

pb = pCoeffs; | |

/* Read the sample from input buffer */ | |

in = *pSrc++; | |

/* Update the energy calculation */ | |

energy -= (((q31_t) x0 * (x0)) >> 15); | |

energy += (((q31_t) in * (in)) >> 15); | |

/* Set the accumulator to zero */ | |

acc = 0; | |

#if defined (ARM_MATH_LOOPUNROLL) | |

/* Loop unrolling: Compute 4 taps at a time. */ | |

tapCnt = numTaps >> 2U; | |

while (tapCnt > 0U) | |

{ | |

/* Perform the multiply-accumulate */ | |

/* acc += b[N] * x[n-N] + b[N-1] * x[n-N-1] */ | |

acc = __SMLALD(read_q15x2_ia (&px), read_q15x2_ia (&pb), acc); | |

acc = __SMLALD(read_q15x2_ia (&px), read_q15x2_ia (&pb), acc); | |

/* Decrement loop counter */ | |

tapCnt--; | |

} | |

/* Loop unrolling: Compute remaining taps */ | |

tapCnt = numTaps % 0x4U; | |

#else | |

/* Initialize tapCnt with number of samples */ | |

tapCnt = numTaps; | |

#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ | |

while (tapCnt > 0U) | |

{ | |

/* Perform the multiply-accumulate */ | |

acc += (q63_t) (((q31_t) (*px++) * (*pb++))); | |

/* Decrement the loop counter */ | |

tapCnt--; | |

} | |

/* Calc lower part of acc */ | |

acc_l = acc & 0xffffffff; | |

/* Calc upper part of acc */ | |

acc_h = (acc >> 32) & 0xffffffff; | |

/* Apply shift for lower part of acc and upper part of acc */ | |

acc = (uint32_t) acc_l >> lShift | acc_h << uShift; | |

/* Converting the result to 1.15 format and saturate the output */ | |

acc = __SSAT(acc, 16U); | |

/* Store the result from accumulator into the destination buffer. */ | |

*pOut++ = (q15_t) acc; | |

/* Compute and store error */ | |

d = *pRef++; | |

e = d - (q15_t) acc; | |

*pErr++ = e; | |

/* Calculation of 1/energy */ | |

postShift = arm_recip_q15((q15_t) energy + DELTA_Q15, &oneByEnergy, S->recipTable); | |

/* Calculation of e * mu value */ | |

errorXmu = (q15_t) (((q31_t) e * mu) >> 15); | |

/* Calculation of (e * mu) * (1/energy) value */ | |

acc = (((q31_t) errorXmu * oneByEnergy) >> (15 - postShift)); | |

/* Weighting factor for the normalized version */ | |

w = (q15_t) __SSAT((q31_t) acc, 16); | |

/* Initialize pState pointer */ | |

px = pState; | |

/* Initialize coefficient pointer */ | |

pb = pCoeffs; | |

#if defined (ARM_MATH_LOOPUNROLL) | |

/* Loop unrolling: Compute 4 taps at a time. */ | |

tapCnt = numTaps >> 2U; | |

/* Update filter coefficients */ | |

while (tapCnt > 0U) | |

{ | |

coef = (q31_t) *pb + (((q31_t) w * (*px++)) >> 15); | |

*pb++ = (q15_t) __SSAT(coef, 16); | |

coef = (q31_t) *pb + (((q31_t) w * (*px++)) >> 15); | |

*pb++ = (q15_t) __SSAT(coef, 16); | |

coef = (q31_t) *pb + (((q31_t) w * (*px++)) >> 15); | |

*pb++ = (q15_t) __SSAT(coef, 16); | |

coef = (q31_t) *pb + (((q31_t) w * (*px++)) >> 15); | |

*pb++ = (q15_t) __SSAT(coef, 16); | |

/* Decrement loop counter */ | |

tapCnt--; | |

} | |

/* Loop unrolling: Compute remaining taps */ | |

tapCnt = numTaps % 0x4U; | |

#else | |

/* Initialize tapCnt with number of samples */ | |

tapCnt = numTaps; | |

#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ | |

while (tapCnt > 0U) | |

{ | |

/* Perform the multiply-accumulate */ | |

coef = (q31_t) *pb + (((q31_t) w * (*px++)) >> 15); | |

*pb++ = (q15_t) __SSAT(coef, 16); | |

/* Decrement loop counter */ | |

tapCnt--; | |

} | |

x0 = *pState; | |

/* Advance state pointer by 1 for the next sample */ | |

pState = pState + 1; | |

/* Decrement loop counter */ | |

blkCnt--; | |

} | |

/* Save energy and x0 values for the next frame */ | |

S->energy = (q15_t) energy; | |

S->x0 = x0; | |

/* Processing is complete. | |

Now copy the last numTaps - 1 samples to the start of the state buffer. | |

This prepares the state buffer for the next function call. */ | |

/* Points to the start of the pState buffer */ | |

pStateCurnt = S->pState; | |

/* copy data */ | |

#if defined (ARM_MATH_LOOPUNROLL) | |

/* Loop unrolling: Compute 4 taps at a time. */ | |

tapCnt = (numTaps - 1U) >> 2U; | |

while (tapCnt > 0U) | |

{ | |

write_q15x2_ia (&pStateCurnt, read_q15x2_ia (&pState)); | |

write_q15x2_ia (&pStateCurnt, read_q15x2_ia (&pState)); | |

/* Decrement loop counter */ | |

tapCnt--; | |

} | |

/* Loop unrolling: Compute remaining taps */ | |

tapCnt = (numTaps - 1U) % 0x4U; | |

#else | |

/* Initialize tapCnt with number of samples */ | |

tapCnt = (numTaps - 1U); | |

#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ | |

while (tapCnt > 0U) | |

{ | |

*pStateCurnt++ = *pState++; | |

/* Decrement loop counter */ | |

tapCnt--; | |

} | |

} | |

/** | |

@} end of LMS_NORM group | |

*/ |