Part Number: PROCESSOR-SDK-J721E
Many a times, there is a need to move a function/API to a specific section of the memory map, like internal memory for CPU performance optimization or memory optimization. This article explains how to do it in SYSBIOS.
Part Number: AM5718
Looking at the posts in the forum, I found the following related post in the MSP low-power microcontroller forum:
https://e2e.ti.com/support/microcontrollers/msp-low-power-microcontrollers-group/msp430/f/msp-low-power-microcontroller-forum/843943/msp432p4111-sysbios-clock-function-swi-appears-to-be-blocking-all-tasks-from-running
The issue identified appears to be the same: the value of TicksToService is an impossible large number, close to 2^32 and the code gets stuck in the same loop, located in the function Clock_workFuncDynamic of the kernel module Clock.c;
while (ticksToService >= distance) {
serviceTick = serviceTick + distance;
ticksToService -= distance;
distance = Clock_walkQueueDynamic(TRUE, serviceTick);
}
In my case, the system works correctly for the first 3m37s of real time measured with a chronometer. Then it starts triggering the SWI function continuously. It should trigger it at 2ms intervals. The value of tickPeriod is 100 (us). So, to get a SWI execution each 2ms, I'm using 20 as the value of clockParams.period. The number of counts per tick is 1919 as used in the Timer_setNextTick function in the timer module.
3m37s or 217 seconds equates approximately to 4164230000 counts of the timer. Close to 2^32. If I count the number of executions of the SWI before it starts triggering continuously, I get 111889. So, 111889 executions made each 20 ticks of 1919 timer counts, yields 4294299820 counts. Now it is closer still to 2^32. To that count 2 extra milliseconds should be added, accounting for the first timeout. In the end, I am convinced that the overflow causes the problem to appear.
The observed behavior is repeatable. The value of ticksToService is obtained by the following code:
nowTick = Clock_TimerProxy_getCurrentTick(Clock_module->timer, TRUE);
...
/* determine first tick expiration to service (the anticipated next tick) */
serviceTick = Clock_module->nextScheduledTick;
ticksToService = nowTick - serviceTick;
I think that during the first 217 seconds of operation (approx) nowTick is larger than serviceTick and the resulting value is positive and close to zero. However, when the timer overflows, the subtraction results in a number close to 2^32, as nowTick is close to zero, while serviceTick is close to 2^32.
Following the suggestion from Alan De Mars in the referred post, I changed tickMode from TickMode_DYNAMIC (the default) to TickMode_PERIODIC and the problem disappeared. Everything works as expected.
I am starting this thread with the hope that anyone encountering this problem might avoid it faster than myself.
Part Number: PROCESSOR-SDK-AM57X
Hi Team,
Can you guide me step by step for modifying DDR/RAM configurations on AM57x.
Part Number: TDA4VM
Can eMMC boot mode, be used to boot to QNX using Uboot/SPL for TDA4VM?
Part Number: TDA4VM
When running QNX on A72, how can CPU usage and memory usage be viewed?
Part Number: AM6442
Hi
There's a question about RESET circuit. Please answer to question.
1. Is there any problem that is expected when simplified with POWER-ON and Cold RESET as above circuit?
Thanks.
Part Number: TDA4VM
How to build PDK libraries and example applications, how to link them to custom applications and how to run them using CCS or through bootloader (either SD card or OSPI) on Jacinto, AM65x and AM57x class of devices ?
Part Number: TDA4VM
Part Number: TDA4VM
Hi,
How to configure ENET LLD and Ethernet Firmware for MAC 2 MAC (phy-less) connection ? Includes CPSW and ICSSG.
Regards
Vineet
Part Number: DRA821U
How do I change the frequency of A72 and Main R5F core in DRA821/DRA829 on both RTOS/SBL and Linux/SPL bootflow ? and disable one A72 core in Uboot ?
Regards
Vineet
