MJPEG decoding problem in i.MX515

[ko-hey]

Hi all

My customer have a problem in i.MX515.

I checked the errata and patch but there are no errata which suit for this problem.

So please advise me how to avoid the symptoms.

They're decoding mjpeg using VPU, It seems that it decodes (converts) incorrectly.

Reproducibility is not very high.
According to the log, it seems that the data of Cr is garbled one bit.
The correct data 7F and the abnormal data is FF.

Would you like to teach how to avoid it ?

I think the informaiton to avoid the symptoms is missing.

So please tell me if there is missing information.

Ko-hey

[Yuri]

Hello,

 Note,  i.MX51 VPU supports JPEG baseline only.  Is Your image JPEG baseline ? 

Also, please check memory, using the stress test.

Have a great day,

Yuri

------------------------------------------------------------------------------

Note: If this post answers your question, please click the Correct

Answer button. Thank you!

[ko-hey]

Hello Yuri

Sorry for late reply.
I got test result which was tested with DDR Stress tester kit for the i.MX51 and i.MX53.
They testted in static DDR frequency mode (200MHz) and the log stopped following message.

348
t0: memcpy11 SSN test
t1: memcpy8 SSN test
t2: memcoy11 random pattern test
t3: byte-wise SSN test
t4: IRAM_to_DDRv2 test
t5: IRAM_to_DDRv1 test
t6: read noise walking ones and zeros test

349
t0: memcpy11 SSN test
t1: memcpy8 SSN test

I have some question about the relating to the above result.

Q1.

The static DDR frequency mode means it test many times at the static frequency.
Am I correct ?

Q2.

If Q1 is yes, from the above message, the test failed in memcoy11 random pattern test.
Am I correct ?

Q3.

If the test success, can user get a calibration value from the DDR Stress tester kit for the i.MX51 and i.MX53 whether user use i.MX51 or not ?

Q4.

When issued the message, it seems like i.MX51 is resetted.
Does the Stress tester kit include reset function when the test fail ?

Ko-hey

[Yuri]

Please look at my comments below.

1.

> The static DDR frequency mode means it test many times at the static frequency.
> Am I correct ?

  Yes.

2.

> If Q1 is yes, from the above message, the test failed in memcoy11 random pattern test.
> Am I correct ?

  I do not think so, since there are no errors messages. But looks like, the system (host or target) has 
been hanged.     

3.

> If the test success, can user get a calibration value from the DDR Stress tester kit for the i.MX51 and 
> i.MX53 whether user use i.MX51 or not ?

If the system hangs  with calibration and does not without it - hardly this is  reasonable to use the calibration values.

4.

> When issued the message, it seems like i.MX51 is resetted.
> Does the Stress tester kit include reset function when the test fail ?

  No -  just  error report is performed.

Regards,

Yuri.

[ko-hey]

Hi Yuri

Thank you for your comments.

> If the system hangs  with calibration and does not without it - hardly this is  reasonable to use the calibration values.

How about the test success ?

If the system doesn't hang, can we get the calibration value ?

Or do the tool just do the stress test ?

I don't have i.MX51 and i.MX53 board and can't see the case of success.

So please tell me what happens and gets when user success the test.

Ko-hey

[Yuri]

The test is end-less. If it works without hanging during long time - all is OK.
We can rely on memory initialization parameters (in init script), which are used for testing.   

Regards,

Yuri.

[ko-hey]

Do you have any recommendation number of times for testing ?

My customer tried it approximately 3,000 times.

Ko-hey

[Yuri]

> My customer tried it approximately 3,000 times.

this is OK.

[ko-hey]

They tested approximately 3,000 times and finally the message stopped as below.

t0: memcpy11 SSN test

t1: memcpy8 SSN test

According to the your answer, the DDR stress test was passed but the CPU hanged some reason.

I think we should check another part except for DDR.

What do you think about my understanding ?

Ko-hey

[ko-hey]

Hi Yuri

The customer did the incrementing DDR frequency mode too.

They tested three times and the test failed at 250MHz, 255MHz and 260MHz. (The start is 125MHz)

The operation freq is 200MHz in their system so I think it has enough margin.

From the both of result, I think the memory is OK.

What do you think ?

Please advice me which part do I check next.

Ko-hey

[Yuri]

Hello,

  Is the MJPEG decoding problem the only one - and there are no other

issues with the i.MX51 system (instability, hangs, say as for erratum

ENGcm12051 DPLL: Meta-stability Issue) ?

  If so - please create request.

https://community.nxp.com/docs/DOC-329745 

Regards,

Yuri.

[ko-hey]

Hello

Yes, there is no issue except for the MJPEG decoding problem.

I made a SR(Case#00101014).

Please follow it.

Ko-hey