• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 98 subscribers
  • Views 360 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.

RTA Statistics for SWIs

Brad Baker
Prodigy 170 points


Using the clock feature I have manually measured the worse case execution time (instruction cycles) for a SWI with interrupts disabled and I get a substantially smaller number then I do when using RTA to do the measurement.

Does the RTA measurement account for the effect of interrupts on the calculated SWI duration (both HWI and SWI) ?

-- Brad

  • 0
    TI__Mastermind 30140 points

    Hi Brad,

    You are right about the RTA measurement including the overhead of DSP/BIOS, while the clock feature just counts cycles of raw execution. There are periodic timer interrupts, context switching, etc. that will increase the execution time required to execute the task.

    If you plan to use DSP/BIOS, then the RTA measurements provide the most accurate timings with regard to hitting deadlines, etc.

    Hope this helps,
    Mark

     

  • 0 in reply to Mark M
    Prodigy 170 points

    Mark Mckeown said:

    Hi Brad,

    You are right about the RTA measurement including the overhead of DSP/BIOS, while the clock feature just counts cycles of raw execution. There are periodic timer interrupts, context switching, etc. that will increase the execution time required to execute the task.

    If you plan to use DSP/BIOS, then the RTA measurements provide the most accurate timings with regard to hitting deadlines, etc.

    Hope this helps,
    Mark

    Mark,

    I understand the inclusion of overhead time by the OS, context switching, what-have-you.  Rather, my concern is that the time measured using RTA for a SWI from SWI entry to SWI exit does not measure the actual work time of the SWI , but rather includes the time consumed by some things that interrupt it, specifically HWIs.  My own timing instrumentation of SWI execution seems to indicate that the RTA timing of a SWI excludes the time consumed by those SWIs that preempt it but does not exclude the time consumed by those HWIs that preempt it.  I have not found documentation that describes the RTA calculation of SWI execution time in detail.

    Best regards,

    -- Brad

     

  • 0 in reply to Brad Baker
    Prodigy 170 points

    (I'm hoping I am wrong about this).  I ran a test that seems to confirm my earlier fears; the SWI execution time reported by the RTA tool does not take into account the time consumed by HWIs that interrupt it.  I presume that it accounts for other SWIs that interrupt it.  Regardless, I was really hoping to be able to use this for feature for RMA timing analysis.  It looks I'll have to use my own instrumentation code that does account for hardware and software interrupt effect on SWI execution time, but it is oh so piggish.

    -- Brad