Renesas / mbed-dsp-neon

Fork of mbed-dsp. CMSIS-DSP library of supporting NEON

Dependents:   mbed-os-example-cmsis_dsp_neon

Fork of mbed-dsp by mbed official

Information

Japanese version is available in lower part of this page.
このページの後半に日本語版が用意されています．

CMSIS-DSP of supporting NEON

What is this ?

A library for CMSIS-DSP of supporting NEON.
We supported the NEON to CMSIS-DSP Ver1.4.3(CMSIS V4.1) that ARM supplied, has achieved the processing speed improvement.
If you use the mbed-dsp library, you can use to replace this library.
CMSIS-DSP of supporting NEON is provied as a library.

Library Creation environment

CMSIS-DSP library of supporting NEON was created by the following environment.

  • Compiler
    ARMCC Version 5.03
  • Compile option switch[C Compiler]
   -DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm 
   --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp 
   --vectorize --asm
  • Compile option switch[Assembler]
   --cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access 
   --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp


Effects of NEON support

In the data which passes to each function, large size will be expected more effective than small size.
Also if the data is a multiple of 16, effect will be expected in every function in the CMSIS-DSP.


NEON対応CMSIS-DSP

概要

NEON対応したCMSIS-DSPのライブラリです。
ARM社提供のCMSIS-DSP Ver1.4.3(CMSIS V4.1)をターゲットにNEON対応を行ない、処理速度向上を実現しております。
mbed-dspライブラリを使用している場合は、本ライブラリに置き換えて使用することができます。
NEON対応したCMSIS-DSPはライブラリで提供します。

ライブラリ作成環境

NEON対応CMSIS-DSPライブラリは、以下の環境で作成しています。

  • コンパイラ
    ARMCC Version 5.03
  • コンパイルオプションスイッチ[C Compiler]
   -DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm 
   --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp 
   --vectorize --asm
  • コンパイルオプションスイッチ[Assembler]
   --cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access 
   --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp


NEON対応による効果について

CMSIS-DSP内の各関数へ渡すデータは、小さいサイズよりも大きいサイズの方が効果が見込めます。
また、16の倍数のデータであれば、CMSIS-DSP内のどの関数でも効果が見込めます。


cmsis_dsp/SupportFunctions/math_helper.c

Committer:
emilmont
Date:
2013-05-30
Revision:
2:da51fb522205
Parent:
1:fdd22bb7aa52
Child:
3:7a284390b0ce

File content as of revision 2:da51fb522205:

/* ----------------------------------------------------------------------
* Copyright (C) 2010 ARM Limited. All rights reserved.  
*  
* $Date:        29. November 2010  
* $Revision: 	V1.0.3  
*  
* Project: 	    CMSIS DSP Library 
*
* Title:	    math_helper.c
*
* Description:	Definition of all helper functions required.  
*  
* Target Processor: Cortex-M4/Cortex-M3
*  
* Version 1.0.3 2010/11/29 
*    Re-organized the CMSIS folders and updated documentation.  
*   
* Version 1.0.2 2010/11/11  
*    Documentation updated.   
*  
* Version 1.0.1 2010/10/05   
*    Production release and review comments incorporated.  
*  
* Version 1.0.0 2010/09/20   
*    Production release and review comments incorporated.  
*  
* Version 0.0.7  2010/06/10   
*    Misra-C changes done 
* -------------------------------------------------------------------- */

/* ----------------------------------------------------------------------
*		Include standard header files  
* -------------------------------------------------------------------- */
#include<math.h>

/* ----------------------------------------------------------------------
*		Include project header files  
* -------------------------------------------------------------------- */
#include "math_helper.h"

/** 
 * @brief  Caluclation of SNR
 * @param  float* 	Pointer to the reference buffer
 * @param  float*	Pointer to the test buffer
 * @param  uint32_t	total number of samples
 * @return float	SNR
 * The function Caluclates signal to noise ratio for the reference output 
 * and test output 
 */

float arm_snr_f32(float *pRef, float *pTest, uint32_t buffSize)
{
  float EnergySignal = 0.0, EnergyError = 0.0;
  uint32_t i;
  float SNR;
  int temp;
  int *test;

  for (i = 0; i < buffSize; i++)
    {
 	  /* Checking for a NAN value in pRef array */
	  test =   (int *)(&pRef[i]);
      temp =  *test;

	  if(temp == 0x7FC00000)
	  {
	  		return(0);
	  }

	  /* Checking for a NAN value in pTest array */
	  test =   (int *)(&pTest[i]);
      temp =  *test;

	  if(temp == 0x7FC00000)
	  {
	  		return(0);
	  }
      EnergySignal += pRef[i] * pRef[i];
      EnergyError += (pRef[i] - pTest[i]) * (pRef[i] - pTest[i]); 
    }

	/* Checking for a NAN value in EnergyError */
	test =   (int *)(&EnergyError);
    temp =  *test;

    if(temp == 0x7FC00000)
    {
  		return(0);
    }
	

  SNR = 10 * log10 (EnergySignal / EnergyError);

  return (SNR);

}


/** 
 * @brief  Provide guard bits for Input buffer
 * @param  q15_t* 	    Pointer to input buffer
 * @param  uint32_t 	blockSize
 * @param  uint32_t 	guard_bits
 * @return none
 * The function Provides the guard bits for the buffer 
 * to avoid overflow 
 */

void arm_provide_guard_bits_q15 (q15_t * input_buf, uint32_t blockSize,
                            uint32_t guard_bits)
{
  uint32_t i;

  for (i = 0; i < blockSize; i++)
    {
      input_buf[i] = input_buf[i] >> guard_bits;
    }
}

/** 
 * @brief  Converts float to fixed in q12.20 format
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point(q12.20) values 
 */

void arm_float_to_q12_20(float *pIn, q31_t * pOut, uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
	  /* 1048576.0f corresponds to pow(2, 20) */
      pOut[i] = (q31_t) (pIn[i] * 1048576.0f);

      pOut[i] += pIn[i] > 0 ? 0.5 : -0.5;

      if (pIn[i] == (float) 1.0)
        {
          pOut[i] = 0x000FFFFF;
        }
    }
}

/** 
 * @brief  Compare MATLAB Reference Output and ARM Test output
 * @param  q15_t* 	Pointer to Ref buffer
 * @param  q15_t* 	Pointer to Test buffer
 * @param  uint32_t 	number of samples in the buffer
 * @return none 
 */

uint32_t arm_compare_fixed_q15(q15_t *pIn, q15_t * pOut, uint32_t numSamples)
{
  uint32_t i; 
  int32_t diff, diffCrnt = 0;
  uint32_t maxDiff = 0;

  for (i = 0; i < numSamples; i++)
  {
  	diff = pIn[i] - pOut[i];
  	diffCrnt = (diff > 0) ? diff : -diff;

	if(diffCrnt > maxDiff)
	{
		maxDiff = diffCrnt;
	}	
  }

  return(maxDiff);
}

/** 
 * @brief  Compare MATLAB Reference Output and ARM Test output
 * @param  q31_t* 	Pointer to Ref buffer
 * @param  q31_t* 	Pointer to Test buffer
 * @param  uint32_t 	number of samples in the buffer
 * @return none 
 */

uint32_t arm_compare_fixed_q31(q31_t *pIn, q31_t * pOut, uint32_t numSamples)
{
  uint32_t i; 
  int32_t diff, diffCrnt = 0;
  uint32_t maxDiff = 0;

  for (i = 0; i < numSamples; i++)
  {
  	diff = pIn[i] - pOut[i];
  	diffCrnt = (diff > 0) ? diff : -diff;

	if(diffCrnt > maxDiff)
	{
		maxDiff = diffCrnt;
	}
  }

  return(maxDiff);
}

/** 
 * @brief  Provide guard bits for Input buffer
 * @param  q31_t* 	Pointer to input buffer
 * @param  uint32_t 	blockSize
 * @param  uint32_t 	guard_bits
 * @return none
 * The function Provides the guard bits for the buffer 
 * to avoid overflow 
 */

void arm_provide_guard_bits_q31 (q31_t * input_buf, 
								 uint32_t blockSize,
                                 uint32_t guard_bits)
{
  uint32_t i;

  for (i = 0; i < blockSize; i++)
    {
      input_buf[i] = input_buf[i] >> guard_bits;
    }
}

/** 
 * @brief  Provide guard bits for Input buffer
 * @param  q31_t* 	Pointer to input buffer
 * @param  uint32_t 	blockSize
 * @param  uint32_t 	guard_bits
 * @return none
 * The function Provides the guard bits for the buffer 
 * to avoid overflow 
 */

void arm_provide_guard_bits_q7 (q7_t * input_buf, 
								uint32_t blockSize,
                                uint32_t guard_bits)
{
  uint32_t i;

  for (i = 0; i < blockSize; i++)
    {
      input_buf[i] = input_buf[i] >> guard_bits;
    }
}



/** 
 * @brief  Caluclates number of guard bits 
 * @param  uint32_t 	number of additions
 * @return none
 * The function Caluclates the number of guard bits  
 * depending on the numtaps 
 */

uint32_t arm_calc_guard_bits (uint32_t num_adds)
{
  uint32_t i = 1, j = 0;

  if (num_adds == 1)
    {
      return (0);
    }

  while (i < num_adds)
    {
      i = i * 2;
      j++;
    }

  return (j);
}

/** 
 * @brief  Converts Q15 to floating-point
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 */

void arm_apply_guard_bits (float32_t * pIn, 
						   uint32_t numSamples, 
						   uint32_t guard_bits)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
      pIn[i] = pIn[i] * arm_calc_2pow(guard_bits);
    }
}

/** 
 * @brief  Calculates pow(2, numShifts)
 * @param  uint32_t 	number of shifts
 * @return pow(2, numShifts)
 */
uint32_t arm_calc_2pow(uint32_t numShifts)
{

  uint32_t i, val = 1;

  for (i = 0; i < numShifts; i++)
    {
      val = val * 2;
    }	

  return(val);
}



/** 
 * @brief  Converts float to fixed q14 
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point values 
 */

void arm_float_to_q14 (float *pIn, q15_t * pOut, 
                       uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
	  /* 16384.0f corresponds to pow(2, 14) */
      pOut[i] = (q15_t) (pIn[i] * 16384.0f);

      pOut[i] += pIn[i] > 0 ? 0.5 : -0.5;

      if (pIn[i] == (float) 2.0)
        {
          pOut[i] = 0x7FFF;
        }

    }

}

 
/** 
 * @brief  Converts float to fixed q30 format
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point values 
 */

void arm_float_to_q30 (float *pIn, q31_t * pOut, 
					   uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
	  /* 1073741824.0f corresponds to pow(2, 30) */
      pOut[i] = (q31_t) (pIn[i] * 1073741824.0f);

      pOut[i] += pIn[i] > 0 ? 0.5 : -0.5;

      if (pIn[i] == (float) 2.0)
        {
          pOut[i] = 0x7FFFFFFF;
        }
    }
}

/** 
 * @brief  Converts float to fixed q30 format
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point values 
 */

void arm_float_to_q29 (float *pIn, q31_t * pOut, 
					   uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
	  /* 1073741824.0f corresponds to pow(2, 30) */
      pOut[i] = (q31_t) (pIn[i] * 536870912.0f);

      pOut[i] += pIn[i] > 0 ? 0.5 : -0.5;

      if (pIn[i] == (float) 4.0)
        {
          pOut[i] = 0x7FFFFFFF;
        }
    }
}


/** 
 * @brief  Converts float to fixed q28 format
 * @param  uint32_t 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point values 
 */

void arm_float_to_q28 (float *pIn, q31_t * pOut, 
                       uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
	/* 268435456.0f corresponds to pow(2, 28) */
      pOut[i] = (q31_t) (pIn[i] * 268435456.0f);

      pOut[i] += pIn[i] > 0 ? 0.5 : -0.5;

      if (pIn[i] == (float) 8.0)
        {
          pOut[i] = 0x7FFFFFFF;
        }
    }
}

/** 
 * @brief  Clip the float values to +/- 1 
 * @param  pIn 	input buffer
 * @param  numSamples 	number of samples in the buffer
 * @return none
 * The function converts floating point values to fixed point values 
 */

void arm_clip_f32 (float *pIn, uint32_t numSamples)
{
  uint32_t i;

  for (i = 0; i < numSamples; i++)
    {
      if(pIn[i] > 1.0f)
	  {
	    pIn[i] = 1.0;
	  }
	  else if( pIn[i] < -1.0f)
	  {
	    pIn[i] = -1.0;
	  }
	       
    }
}