• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 98 subscribers
  • Views 1468 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Problems with Task Scheduling

Steve Kreyer
Intellectual 520 points
Other Parts Discussed in Thread: OMAP-L138

Hi,

 

I have a problem with the task scheduling model, and no idea how to solve it.

 

There are two tasks (lets call them T1 and T2), and one interrupt I. Task T1 does pend for a Semaphore and writes to SPI using the Stream_write function if the semaphore was posted. Task T2 does a complex calculation which can last for around 500 usecs til 4msecs depending on the size of some input data I have. Interrupt I is issued from an external source (each 5ms) and will post the semaphore that task T1 waits for. Because of some timing requirements of the Stream_write function (see please the problem under http://e2e.ti.com/support/embedded/bios/f/355/t/151706.aspx) task T1 does not necessarily wait on the semaphore when interrupt I posts it, thus task T1 will not be rescheduled (which is logic). The problem is now that if task T2 has a lot of input data to process, it seems that it slows down the reaction time of task T1 (even if it owns the required semaphore) in some way and this will lead to a system abort because it messes up the SPI communication mechanism. My question is if there is a way to cyclically interrupt the whole system (lets say each 500 usecs) and manually force a rescheduling of the tasks I have in the system. Task T1 already has the highest priority.

 

Here is the pseudo code of my system:

Void InterruptI(Void)

{

Semaphore_post(tx_event);

}

 

Void TaskT1(Void)

{

while (1)

{

Semaphore_pend(tx_event, BIOS_WAIT_FOREVER);

Stream_write(spiHandle, BIOS_WAIT_FOREVER); //interrupt I can also be triggered if Stream_write is currenty blocking => T1 does not wait on tx_event during post time

}

}

 

Void TaskT2(Void)

{

while(1)

{

Semaphore_pend(input_data_rx_event, BIOS_WAIT_FOREVER);

Do_heavy_computation(input_data);  //can last from 500usecs - 4ms

}

}

 

Btw for the future I've planned a task T3 which does even more complex calculations (processing times around 30-40ms). How can I ensure that I met the timing requirements of task T1 then?

 

Any help on this issue is very appreciated. If you need more info please let me  know. Btw Iam using  bios_6_32_05_54 and ipc_1_23_05_40 on an OMAP-L138 device.

 

Thanks in advance.

 

Kind Regards,

Steve

  • 0
    TI__Mastermind 30830 points

    Steve,

    If I understand you correctly, you'd like both tasks to be given CPU time in a time-slicing kind of way.

    While SYS/BIOS doesn't natively support this operation, it can be easily coerced into doing so.

    If you make both T1 and T2 tasks have the same priority, you can force a re-schedule of these two tasks (assuming they are both ready to run and higher priority than any other read tasks), by invoking the Task_yield() API from within a periodic interrupt.

    This will cause the two tasks to alternately pre-empt each other for the duration of the periodic interrupt.

    Alan

  • 0 in reply to Alan DeMars
    Intellectual 520 points

    Hi Alan,

     

    thanks for your answer. Yes this could solve the issue. The problem here is that I would need a relatively high resolution of that timer interrupt. Is it possible to run this interrupt lets say @500kHz without blocking the rest of the system? I only would need this interrupt running from the time I post the semaphore until Iam sure that I've reentered Stream_write. Is it possible to start such a timer within interrupt context, and stop it within the task before entering Stream_write?

     

    Thanks very much for your help.

     

    Steve