RM46L852: Calculated value using FPU was corrupted, the root cause seems to be ISR

Hi Yamai,

Floating point calculations are performed on a separate register bank inside the FPU. If both the main program and ISR use the FPU, extra context saving and restoring are required to ensure that the ISR does not corrupt the data used by the main thread. 

Hello Wang san,

Thanks for your help.

ISR does some  a little work (see below) without FPU calculation, I think. Main work by ISR is deferred.

So I thought I do not need to take care about FPU things.

void etpwmNotification(etpwmBASE_t *node)
{
/* enter user code between the USER CODE BEGIN and USER CODE END. */
/* USER CODE BEGIN (44) */
if(node == etpwmREG2)
{
uletpwmInterruptHandler();
}
/* USER CODE END */
}

uint32_t uletpwmInterruptHandler( void )
{
BaseType_t xHigherPriorityTaskWoken;

xHigherPriorityTaskWoken = pdFALSE;

xSemaphoreGiveFromISR( xBinarySemaphore, &xHigherPriorityTaskWoken );

portYIELD_FROM_ISR( xHigherPriorityTaskWoken );

return 0;

}

But floating calculation corrupted if interrupt is enabled. 

So uletpwmInterruptHandler may have floating point calculation.

I have several questions regarding this.

1) Do you have some sample code to preserve FPU registers ?

2) os_port.c in RTOS says following, so could I use no FPU with RTOS ?

   Do we have any alternative to use FPU if "NO" ?

  if( ( ( uint32_t ) pxCode & 0x01UL ) != 0x00 )

  {

  /* The task will start in thumb mode. */

  *pxTopOfStack |= portTHUMB_MODE_BIT;

  }

  pxTopOfStack--;

  /* The last thing on the stack is the tasks ulUsingFPU value, which by

  default is set to indicate that the stack frame does not include FPU

  registers. */

  *pxTopOfStack = pdFALSE;

  return pxTopOfStack;

Thanks again.

Yamai

Hi Wang san,

I found several hints to resolve my issue, but I am not sure if they are enough or not .

1) https://e2e.ti.com/support/microcontrollers/arm-based-microcontrollers-group/arm-based-microcontrollers/f/arm-based-microcontrollers-forum/700651/rm46l430-using-floating-point-registers-in-irq-and-fiq-rm46

I put described some codes, in accordance with its contents.

2) https://www.freertos.org/Using-FreeRTOS-on-Cortex-A-Embedded-Processors.html

I put vPortTaskUsesFPU() once on task entry, before entering the loop that implements its functionality. 

In above URL, portTASK_USES_FLOATING_POINT() is shown instead of vPortTaskUsesFPU(), I think, we should use vPortTaskUsesFPU() for RM46.

After I applied what are given in above URL, my code seems correctly to be working.

Please let me know if it is enough or not.

Thanks for your cooperation.

Yamai

Hi Yamai,

#1 is one solution. Regarding #2, I haven't tried it before. 

Here is another useful link:

https://e2e.ti.com/support/microcontrollers/arm-based-microcontrollers-group/arm-based-microcontrollers/f/arm-based-microcontrollers-forum/376083/how-to-use-float-in-irq-with-freertos

Hello Wang san,

Thanks for giving me another useful URL. I also will consider it.

RM46x 16/32-Bit RISC Flash Microcontroller Technical Reference Manual, I referred, is not enough to understand FPU, anyway.

Do you have a document regarding floating point unit in RM46 ?

Hopefully, you have it.

Yamai

Hi Yamai,

The FPU is not included in RM46 TRM. It's ARM IP and in ARM TRM: ARM DDI0460D

https://developer.arm.com/documentation/ddi0460/c/FPU-Programmers-Model

Wang san,

Greetings.

Thanks, I will check it out in detail, later.

Yamai