how to use FPU in STM32F407

[Chopra.1]

I am using STM32F407 discovery board which is ARM cortex M-4 based , which comes with FPU.

I need to perform some Floating point number and exponent calculations. for that i need to use to FPU to speed up the process.

I have have written all the peripheral drivers needed for my application myself ( i haven't use the cube HAL layer or cmsis layer code for that )

how can i use the FPU of processor to perform the calculation ? Do i have to write code for that or do i have to add some library ?

i tried to enable the HW based FPU in the setting but when i perform the float / double calculation the code got stuck infinite loop, i think it is happing because the code to perform floating calculations is missing or FPU is not being used?

also can someone tell me if i have to perform software based floating point calculation.

which library i have to add to my code.

software i am using : STM32 Cube IDE

Thank you

kamal chopra

[TDK]

At compile-time, there are flags which are used to tell the compiler the chip has an FPU and to create hardware FPU instructions rather than software emulation. This is a compiler-time choice and not something that can be changed at run-time.

 

Enabling the FPU at run-time is also required, but is a separate step from this. In order to use hardware FPU, it needs to be compiled as such and the FPU needs to be enabled during startup.

[Chopra.1]

Thanks but ,

I have tried doing this but the result is same ,

(i have attacted some img for referance )

 

 

this is the first statement in my program here FPU calculation is required ( shown in above photo)

	double accelX_FT_val , accelY_FT_val ,accelZ_FT_val ;
	if ( accel_XTestdata!= 0 )
	{
		//FT[Xg] = 25*131*1.046^(XG_TEST - 1 )
		accelX_FT_val = pow(3425.65 , ( accel_XTestdata - 1)) ;
 
	}else

when this statement runs a hardware error occurs and program goes to infinite loop

 

Can you help with what i am missing

also .

can you elaborate this Enabling the FPU at run-time is also required, but is a separate step from this. In order to use hardware FPU, it needs to be compiled as such and the FPU needs to be enabled during startup.

thank you

[TDK]

You're using the 64-bit "double" type, which is not supported by the FPU on the STM32F4. Only the 32-bit "float" is.

In the infinite loop, look at the value of the VECTACTIVE field in SCB_ICSR to see what interrupt the chip is in.

[Chopra.1]

thanks,

I have changed variable from double to float

then also fault occurs,

As you said,

i have checked , VECTACTIVE field in SCB_ICSR  : VECTACTIVE - 0x3 { so it means hard fault occurs }

also i have checked , these bits :

SCB->CFSR-> FORCED - 0x1 (it is telling me hard fault is taken because of bus fault , usage fault , memory management fault )

SCB->CFSR->NOCP - 0x1 (it tells a attempt to execute coprocessor instruction )

also if have tried some other things

// i tried to run these instruction separately for testing  purpose
 
accelX_FT_val = pow (2,5); // works fine 
 
accelX_FT_val = pow(1 , ( accel_XTestdata - 1)) ;// works fine 
 
accelX_FT_val = pow(1.046 , ( accel_XTestdata - 1)) ; // hard fault ouccrs 
accelX_FT_val = 25*131*accelX_FT_val; // hard fault ouccrs 
accelX_FT_val = pow(3425 , ( accel_XTestdata - 1)) ;  // hard fault ouccrs 
accelX_FT_val = pow(3425.65 , ( accel_XTestdata - 1)) ; // hard fault ouccrs 

as this : accelX_FT_val = pow (2,5); // works fine  so math.h library has been added no problem in that case

accelX_FT_val = pow(1.046 , ( accel_XTestdata - 1)) ; // hard fault occurs  { as it is floating point calculation, could it be due to the reason FPU is disabled ??? }

accelX_FT_val = 25*131*accelX_FT_val; // hard fault ouccrs  ( if i calculated this explicitly result is 8.9245738115672328584839563027024e+45 )

accelX_FT_val = pow(3425 , ( accel_XTestdata - 1)) ; // hard fault ouccrs  ( if i calculate this explicitly result is 8.9245738115672328584839563027024e+45 )

In these two cases hard fault occurs due to over flow of variables ???

[Tesla DeLorean]

The FPU on the CM4F is single precision only, perhaps powf() ?

FPU typically needs to be enabled with code in SystemInit() or before

void SystemInit(void)
{
  /* FPU settings ------------------------------------------------------------*/
  #if (__FPU_PRESENT == 1) && (__FPU_USED == 1)
    SCB->CPACR |= ((3UL << 10*2)|(3UL << 11*2));  /* set CP10 and CP11 Full Access */
  #endif
...