Run Time Stats failure in 7.0.2

ksoldavin wrote on Monday, November 21, 2011:

I am currently running version 7.0.2 on a dsPIC33. We recently upgraded from version 5.3.1 and aside from slightly more RAM and Flash usage everything seems OK. We have recently found that when we attempt to get the real time stats on the tasks we are getting strange results. When we request the stats the first time, everything seems to work fine. All subsequent requests for the stats returns 0% or <1% CPU usage for all tasks. This code worked perfectly under 5.3.1.

Any suggestions on where to look for the problem?

Thank you
Keith

edwards3 wrote on Monday, November 21, 2011:

What is the run time stats code? Can you post it?

ksoldavin wrote on Monday, November 21, 2011:

To start with, I have configGENERATE_RUN_TIME_STATS set to 1 in freeRTOSConfig.h.

Here is the code for setting up the timer, reading the timer, and the timer interrupt. The code that calculates the stats is in the FreeRTOS code and was not modified.

void vConfigureTimerForRunTimeStats( void )
{
	const unsigned long ulCompareMatch = ( configCPU_CLOCK_HZ / 8 ) / configTICK_RATE_HZ / 10;
	
	runTimeStatsTicker = 0;
	T4CON = 0;
	TMR4 = 0;
	PR4 = ( unsigned portSHORT ) ulCompareMatch;
	/* Setup timer 1 interrupt priority. */
	IPC6bits.T4IP = configKERNEL_INTERRUPT_PRIORITY;
	/* Clear the interrupt as a starting condition. */
	IFS1bits.T4IF = 0;
	/* Enable the interrupt. */
	IEC1bits.T4IE = 1;
	/* Setup the prescale value. */
	T4CONbits.TCKPS0 = 1;
	T4CONbits.TCKPS1 = 0;
	/* Start the timer. */
	T4CONbits.TON = 1;
} // vConfigureTimerForRunTimeStats
// ulGetTimerForRunTimeStats
unsigned long ulGetTimerForRunTimeStats( void )
{
	return( runTimeStatsTicker );
} // ulGetTimerForRunTimeStats
// ulGetTimerForRunTimeStats
void __attribute__((__interrupt__, auto_psv)) _T4Interrupt( void )
{
	/* Clear the timer interrupt. */
	IFS1bits.T4IF = 0;
	runTimeStatsTicker++;
} // _T4Interrupt

edwards3 wrote on Monday, November 21, 2011:

Looks ok, I think. Have you checked that _T4Interrupt is actually executing at the expected  frequency?

ksoldavin wrote on Tuesday, November 22, 2011:

Yes, That was one of the first things I checked. The interrupt is being triggered and the runTimeStatsTicker variable is being incremented.

ksoldavin wrote on Monday, November 28, 2011:

Anyone else have any suggestions? I have hit a wall.