Uploaded image for project: 'Embedded Software & Tools'
  1. Embedded Software & Tools
  2. EXT_EP-10713

Long duration run of vision apps tidl object detection demo hangs at high usage of A72

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Urgent Urgent
    • Processor SDK - Jacinto
    • ADASVISION-5060
    • Hide
      PROCESSOR_SDK_08.00.01
      PROCESSOR_SDK_08.00.00
      Show
      PROCESSOR_SDK_08.00.01 PROCESSOR_SDK_08.00.00
    • PROCESSOR_SDK_08.02.00
    • j721e-evm
    • Hide
      1. Ask customers to have TI review board design
      2. Determine if sample has gone through full TMS production testing from die ID.

      Ask die ID using:
      devmem2 0x43000020 – for die id 0
      devmem2 0x43000024 – for die id 1
      devmem2 0x43000028 – for die id 2
      devmem2 0x4300002C – for die id 3

      Send die ID to Don Steiss.
      Show
      1. Ask customers to have TI review board design 2. Determine if sample has gone through full TMS production testing from die ID. Ask die ID using: devmem2 0x43000020 – for die id 0 devmem2 0x43000024 – for die id 1 devmem2 0x43000028 – for die id 2 devmem2 0x4300002C – for die id 3 Send die ID to Don Steiss.
    • Hide
      This issue is only seen on old engineering samples, and new samples which have latest screen testing done on it do not see this issue, so suspicion is that this is not a software issue, rather a hardware issue on units that fell through older screen tests, so this is now resolved for production units.
      Show
      This issue is only seen on old engineering samples, and new samples which have latest screen testing done on it do not see this issue, so suspicion is that this is not a software issue, rather a hardware issue on units that fell through older screen tests, so this is now resolved for production units.

      Customer has found that running run_app_tidl_od.sh while there is large load on A72 core causes the demo to hang and freeze during their 4 hour aging test. E2E thread linked below.

      We found an issue with idle task stack overflowing, so we disabled preemptive interrupts and increase idle task stack. 

      Attached are screenshots of the C7x registers and function that IEAR register is pointing to when the hang occurs.

      **New update 11/11/2021

      On my setup:

      • SE0_FAR register reports the address of 0x06485F400 which corresponds to the address pointed to by SE0_ADDR8 and SE0_ADDR9 when hang occurs
      • SE0_FSR register reports 0x01005 which documentation says is Internal Storage Parity Error
      • SE1_FAR and FSR registers report 0x0 which should mean that they are fine

      On customer setup:

      • SE*1*_FSR register reports 0x01005
      • SE1_FAR register reports 0x06482AA00 which does not correspond to any SE1 address at the time of error.
      • IEAR value was 0
      • IERR value was 0

      We have tried increasing processor voltage from 0.8V to 0.825V since parity error was seen during our voltage sweep test at low voltage. However, the error still persists.

      Attached screenshots of these registers. Additionally, attached a c7x register dump from customer

       

            syncuser TI User
            syncuser TI User
            Votes:
            1 Vote for this issue
            Watchers:
            4 Start watching this issue

              Created:
              Updated: