• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 99 subscribers
  • Views 273 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.

Keystone Maximum Power Draw Test

Mike JS
Intellectual 290 points

Is there any code available that is designed to maximize the current draw of the c66x Keystone core?

  • 0
    TI__Mastermind 30825 points

    Mike,

    It is highly recommend to do power analysis of your system using the Power Consumption spreadsheet, which should be available on the product page of the devices.

    This will give you the worst case power consumption numbers for the operating conditions and utilization of the device.

    I would not suggest getting power estimations from your system or an EVM for analysis of how much the Keystone DSP will draw as it will likely be a fair amount lower than worst case power consumption as nominal devices will be lower power than what's in the Power Consumption Spreadsheet.

    Best Regards,

    Chad

  • 0 in reply to Chad Courtney
    Intellectual 290 points

    Thanks for the reply!  Certainly the up-front worst case power consumption calculations are critical.  But measurements with the final system are important too to validate the original calculations, with the knowledge that the measurements are generally nominal given part variation.

    Finding the true absolute worst case is obviously less critical in this situation, but it's reasonable due diligence to set up the test so that each component is running in its highest power state where possible.

    If it turns out that the difference in power between a typical case and some tight maximum utilization instruction loop is known to be insignificant, then I can ignore it.  Otherwise it would be useful if one was available.  While they do not use every possible slot, the imgLib and dspLib algorithms seem like good candidates.  One of those should fit my purpose - but if there is a standard test I'd rather use that.

    Best Regards,

    Mike

  • 0 in reply to Mike JS
    TI__Mastermind 30825 points

    Mike,

    We don't have a worst case / critical power routine per se to provide.  I'd look at items your code usage that has the highest IPC (instructions per cycle) executed, as this is going to give the worst case power consumption on by the Core.  You may also want to drive data transfers in the background using EDMA traffic that's representative.

    Running task intensive items from the ImgLib would have pretty intensive code that should get you to realistic worst case utilization from the core perspective.

    Best Regards,

    Chad

  • 0 in reply to Chad Courtney
    TI__Genius 9065 points

    Hello,

    All Chad's point is valid.

    I just want to add: Linpack is a good candidate to run on all cores and make it as loaded as possible. You can download one example here for Keystone I device(C6678):

    http://www.pdc.kth.se/~noname/dist-ti-0.0.3-bin.tar.gz

    regards,

    David