• State
  • Locked Locked
  • Replies 2 replies
  • Subscribers 98 subscribers
  • Views 1397 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.

H264 Decoder issues

Mahesh Kulkarni
Prodigy 60 points

Hi All,

                We are analyzing TI’s H.264 decoder version 1.10.04.00 and we have come across some issues.

The scenarios and respective issues are listed below;-

 

1.       Passing cacheable input buffer to the VIDDEC2_process calls

                In this case the decoder fails for 2nd and subsequent process calls.

The error returned is ‘H264D_ERR_COPROCESSOR_LESS_MBS_DECODED’.

When we ensure that the a NON-cacheable CMEM buffer is passed to the H.264 decoder, the decoder doesn’t fail. (at the expense of performance)

 

2.       Multiple instances of the H.264 decoder

In this case, when we try to open two instances of H.264 decoder using the VIDDEC2_create call in two separate threads, one call returns error and the other call doesn’t return at all. This also hangs the DSP.

When we ensure that the two VIDDEC2_create calls are not done simultaneously with mutual exclusion, the two instances of the decoder are opened successfully. However this compels us to use the mutexes across all VISA calls to ensure that no VISA calls are done simultaneously.

Any inputs on resolving these issues will be of great help

Thanking in Advance

Mahesh K

  • 0
    TI__Intellectual 1790 points

    Mahesh,

    Currently we are verifying this issue, we will let you know once we found the root cause.

    Regards,

    Mahantesh

     

  • 0 in reply to Mahantesh S Madiwalar
    TI__Intellectual 1790 points

    Mahesh,

    Please find the following answers for your queries,

    1. Can you please provide the following details, 

        a. Does this buffer is read/written by only DMA and not by DSP.

        b. Make sure that the input buffer is aligned to 128 bytes and the size of the input buffer is multiple of 128 bytes.

    2. Can you please mention the following details,

        a. The type of error returned by the first instance?  

        b. L2SRAM scratch region used by the two instances and L2 configuration?

    You should be able to create two instances in  parallel without any issue. The problem you are facing might be either due to resource conflict or a timing issue which surfaces when more than one instances are run in parallel.

    Note: In the latest release version "01.10.05.03" few multi instance issues have been fixed (includes timing issues), it's better to download and use the latest library for multi instance use case.

    3. For erroneous case (I saw the third issue related to erroneous case in one of the mail)

    Can you please share the original and modified streams with us, as it's hard to tell about these issues for different error cases without proper analysis.

    Regards,

    Mahantesh