AM2612: Unnecessary critical sections in xTaskGetTickCount

Manuel Philippin

Part Number: AM2612

Hi team,

I noticed that xTaskGetTickCount (in task.c) implements a critical section to protect the xTickCount variable:

/* Critical section required if running on a 16 bit processor. */
portTICK_TYPE_ENTER_CRITICAL();
{
xTicks = xTickCount;
}
portTICK_TYPE_EXIT_CRITICAL();

Since xTickCount is a 32-bit variable and the Cortex-R5 is a 32-bit architecture, these locks appear unnecessary.

I noted that portTICK_TYPE_IS_ATOMIC=1 is defined in portmacro.h for the M4 to disable these locks, but this definition seems to be missing for the R5 in my current SDK version.

This is causing measurable latency in real-time applications.

Could you please confirm if this is an intended behavior or an oversight in the SDK?

Best regards
Manuel

4 days ago

0 Shaunak Deshpande 4 days ago

TI__Mastermind 25832 points

Hi Manuel,

Let me take it back to the SDK team and confirm. Btw, since you mention this causes latency increase in real-time applications, do we have any customer benchmarks or any data that can help me show the impact to SDK team.

Regards,
Shaunak

0 Manuel Philippin 17 hours ago in reply to Shaunak Deshpande

TI__Intellectual 1895 points

Hi Shaunak,

any feedback from the development team here?

Best regards
Manuel

0 Shaunak Deshpande 9 hours ago

TI__Mastermind 25832 points

Hi Manuel,

I confirmed with our team, this is a bug. I have raised a JIRA for this: jira.itg.ti.com/.../MCUSDK-15966

We can use portTICK_TYPE_IS_ATOMIC=1 for ARM R5F port as well.

Regards,
Shaunak

Arm-based microcontrollers

Arm-based microcontrollers forum

AM2612: Unnecessary critical sections in xTaskGetTickCount