204 lines
4.7 KiB
C
204 lines
4.7 KiB
C
/* ----------------------------------------------------------------------
|
|
* Project: CMSIS DSP Library
|
|
* Title: arm_mse_f16.c
|
|
* Description: Half floating point mean square error
|
|
*
|
|
* $Date: 05 April 2022
|
|
* $Revision: V1.10.0
|
|
*
|
|
* Target Processor: Cortex-M and Cortex-A cores
|
|
* -------------------------------------------------------------------- */
|
|
/*
|
|
* Copyright (C) 2010-2022 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 "dsp/statistics_functions_f16.h"
|
|
|
|
/**
|
|
@ingroup groupStats
|
|
*/
|
|
|
|
/**
|
|
@addtogroup MSE
|
|
@{
|
|
*/
|
|
|
|
/**
|
|
@brief Mean square error between two half floating point vectors.
|
|
@param[in] pSrcA points to the first input vector
|
|
@param[in] pSrcB points to the second input vector
|
|
@param[in] blockSize number of samples in input vector
|
|
@param[out] result mean square error
|
|
@return none
|
|
*/
|
|
|
|
#if !defined(ARM_MATH_AUTOVECTORIZE)
|
|
|
|
#if defined(ARM_MATH_MVE_FLOAT16)
|
|
#include "arm_helium_utils.h"
|
|
|
|
void arm_mse_f16(
|
|
const float16_t * pSrcA,
|
|
const float16_t * pSrcB,
|
|
uint32_t blockSize,
|
|
float16_t * result)
|
|
|
|
{
|
|
float16x8_t vecA, vecB;
|
|
float16x8_t vecSum;
|
|
uint32_t blkCnt;
|
|
_Float16 sum = 0.0f16;
|
|
vecSum = vdupq_n_f16(0.0f16);
|
|
|
|
blkCnt = (blockSize) >> 3;
|
|
while (blkCnt > 0U)
|
|
{
|
|
vecA = vld1q(pSrcA);
|
|
pSrcA += 8;
|
|
|
|
vecB = vld1q(pSrcB);
|
|
pSrcB += 8;
|
|
|
|
vecA = vsubq(vecA, vecB);
|
|
|
|
vecSum = vfmaq(vecSum, vecA, vecA);
|
|
/*
|
|
* Decrement the blockSize loop counter
|
|
*/
|
|
blkCnt --;
|
|
}
|
|
|
|
|
|
blkCnt = (blockSize) & 7;
|
|
if (blkCnt > 0U)
|
|
{
|
|
mve_pred16_t p0 = vctp16q(blkCnt);
|
|
vecA = vld1q(pSrcA);
|
|
vecB = vld1q(pSrcB);
|
|
|
|
vecA = vsubq(vecA, vecB);
|
|
vecSum = vfmaq_m(vecSum, vecA, vecA, p0);
|
|
}
|
|
|
|
sum = vecAddAcrossF16Mve(vecSum);
|
|
|
|
/* Store result in destination buffer */
|
|
*result = (_Float16)sum / (_Float16)blockSize;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
#endif /*#if !defined(ARM_MATH_AUTOVECTORIZE)*/
|
|
|
|
|
|
#if defined(ARM_FLOAT16_SUPPORTED)
|
|
|
|
#if (!defined(ARM_MATH_MVE_FLOAT16)) || defined(ARM_MATH_AUTOVECTORIZE)
|
|
|
|
|
|
|
|
void arm_mse_f16(
|
|
const float16_t * pSrcA,
|
|
const float16_t * pSrcB,
|
|
uint32_t blockSize,
|
|
float16_t * result)
|
|
|
|
{
|
|
uint32_t blkCnt; /* Loop counter */
|
|
_Float16 inA, inB;
|
|
_Float16 sum = 0.0f16; /* Temporary return variable */
|
|
#if defined (ARM_MATH_LOOPUNROLL)
|
|
blkCnt = (blockSize) >> 3;
|
|
|
|
|
|
while (blkCnt > 0U)
|
|
{
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
/* Decrement loop counter */
|
|
blkCnt--;
|
|
}
|
|
|
|
|
|
/* Loop unrolling: Compute remaining outputs */
|
|
blkCnt = (blockSize) & 7;
|
|
#else
|
|
/* Initialize blkCnt with number of samples */
|
|
blkCnt = blockSize;
|
|
#endif
|
|
while (blkCnt > 0U)
|
|
{
|
|
inA = *pSrcA++;
|
|
inB = *pSrcB++;
|
|
inA = (_Float16)inA - (_Float16)inB;
|
|
sum += (_Float16)inA * (_Float16)inA;
|
|
|
|
/* Decrement loop counter */
|
|
blkCnt--;
|
|
}
|
|
|
|
/* Store result in destination buffer */
|
|
*result = (_Float16)sum / (_Float16)blockSize;
|
|
}
|
|
|
|
#endif /* end of test for vector instruction availability */
|
|
|
|
#endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
|
|
/**
|
|
@} end of MSE group
|
|
*/
|