Re: SVC, BBQ, GEPOP problem

13 views
Skip to first unread message

Patrick Bellasi

unread,
Apr 17, 2013, 10:55:46 AM4/17/13
to Jens Brandenburg, Germain Haugou, Edoardo Paone, bosp-...@googlegroups.com
Hi Jens!

Running on the board I'm getting an error on the HOST side, maybe you know about:

NPM_init: Can't open directory '/vendor/npm' pointed to by NPM_HOME.
HOST[0]: ERROR: NPM init failed (-1)

Than the excution stops with a segfault.

In attachment the logcat and the application log.
Here also a video of the test: please check it for possible command issues...

No warninkg on the console side... not event at the BBS side...
It seems really an NPM_Init related problem...

Ciao,
Patrick



On Wed, Apr 17, 2013 at 4:31 PM, Jens Brandenburg <jens.bra...@hhi.fraunhofer.de> wrote:
Hi,

I have an update.zip for SDK version 159. For lack of better opportunity I have copied this file to numasaurus. You can find the file at:
/home/bjens/update_v159.zip

To start the test-application with my settings use:


adb shell "LD_PRELOAD=\"/data/bosp/lib/bbque/libbbque_rtlib.so /data/bosp/lib/libSvcDec.so\" /data/bosp/usr/bin/svcdec-test -b /data/bosp/lib/svcdec -i /data/bosp/usr/bin/2P-CGS-1-L1.264 -o /data/bosp/usr/bin/dec.yuv -f 1 -a 0"

Best regards,
Jens


On 04/17/2013 04:20 PM, Patrick Bellasi wrote:



On Wed, Apr 17, 2013 at 3:42 PM, Jens Brandenburg
<jens.brandenburg@hhi.fraunhofer.de <mailto:jens.brandenburg@hhi.fraunhofer.de>>

wrote:

    Hi,

    I don't think the issue is with BBQ. As explained below it also happens when
    running the SVC test application without BBQ. I have performed quite long
    tests (>2h) with HOST_TYPE=posix and FABRIC_TYPE=iss and there the issue did
    not appear. When running HOST_TYPE=arm and FABRIC_TYPE=iss, the application
    fails very fast (<16 frames decoded).

    I used --rt-trace=1000 to get these traces from NPM. When running with BBQ I
    did not use these settings.

    The message is prefixed with RT, so this might be the NPM part? When
    everything works fine I got:
    RT[PE(0,0)] : Called method notify on stub
    RT[PE(0,16)] : Received message 0x9fafb400 from 0x9f2200d0
    When the failure occurs the last message is:
    RT[PE(0,0)] : Called method notify on stub

    I assume BBQ requires HOST_TYPE=arm. Sorry for these problems but I have no
    idea howto circumvent. I hope Germain can provide a solution. But I think we
    should inform Christina and William quite soon.



I agree, let see if Germain could spot the issue by tomorrow.
Meanwhile, maybe we should check if the same issues happens when running on
physical board?
If the issue is QEmu/ARM related maybe we can circumnavigate it by switching to
board usage for this UC as well... what do you think?

Eventually we could try to have a test here in the board, just send us a link
with the update.zip ready to be deployed and run and the command to start the uc
via ADB once installed and BBQ up-and-running...


    Best regards,
    Jens


Ciao Patrick


    On 04/17/2013 03:27 PM, Patrick Bellasi wrote:




        On Wed, Apr 17, 2013 at 10:05 AM, Jens Brandenburg
        <jens.brandenburg@hhi.__fraunhofer.de
        <mailto:jens.brandenburg@hhi.fraunhofer.de>
        <mailto:jens.brandenburg@hhi.__fraunhofer.de

        <mailto:jens.brandenburg@hhi.fraunhofer.de>>>

        wrote:

             Hi,

             to simplify the use-case I tried the SVC decoder application
        without BBQ.
             There the same problem could be found. The execution stops after
        some time.
             Please find attached the log files. So I assume, it is an problem
        with the
             application, NPM or the simulator.

             p12run_log.txt generated by:
             p12run --platform=gepop --host=arm --fabric=iss --ocl-nb-cluster=0
             --rt-trace=1000 >


        Are you always running with --rt-trace=1000?
        In case, maybe by removing this option the stability could be improved?

             p12run_log.txt

             adb_log.txt generated by:
             adb shell "/data/test/host/svcdec_test_____app -b /data/test -i


             /data/test/2P-CGS-1-L1.264 -o /data/test/dec.yuv -a 0" > adb_log.txt

             As you can see at the end of p12run_log.txt. The last message is
        "Called
             method notify on stub".


        Do you have any clue on which code is generating that message?
        Application code, Fabric Code, fabric run-time?

        Ciao Patrick

        --
        #include <best/regards.h>

        Patrick Bellasi
        Post-Doc at Politecnico di Milano
        http://home.dei.polimi.it/__bellasi <http://home.dei.polimi.it/bellasi>



    --
    Mit freundlichen Grüßen / Best regards

    Jens Brandenburg

    Embedded Systems Group
    Image Processing Department
    jens.brandenburg@hhi.__fraunhofer.de <mailto:jens.brandenburg@hhi.fraunhofer.de>

    Tel +49 30 / 31002-660 <tel:%2B49%2030%20%2F%2031002-660>


    Fraunhofer Heinrich Hertz Institute
    Einsteinufer 37, D-10587 Berlin, Germany
    www.hhi.fraunhofer.de <http://www.hhi.fraunhofer.de>



    -----
    visit us at

    Girlsday 2013 / April 25 / Heinrich-Hertz-Institut, Berlin, Einsteinufer 37

    Laser World of Photonics 2013 / May 13-16 / Munich, Germany / booth 312 hall C1

    www.hhi.fraunhofer.de/events <http://www.hhi.fraunhofer.de/events>





--
#include <best/regards.h>

Patrick Bellasi
Post-Doc at Politecnico di Milano
http://home.dei.polimi.it/bellasi

--
Mit freundlichen Grüßen / Best regards

Jens Brandenburg

Embedded Systems Group
Image Processing Department
jens.brandenburg@hhi.fraunhofer.de

Tel +49 30 / 31002-660

Fraunhofer Heinrich Hertz Institute
Einsteinufer 37, D-10587 Berlin, Germany
www.hhi.fraunhofer.de

-----
visit us at

Girlsday 2013 / April 25 / Heinrich-Hertz-Institut, Berlin, Einsteinufer 37

Laser World of Photonics 2013 / May 13-16 / Munich, Germany / booth 312 hall C1

www.hhi.fraunhofer.de/events



--
#include <best/regards.h>

Patrick Bellasi
Post-Doc at Politecnico di Milano
http://home.dei.polimi.it/bellasi
logcat.txt
host.log

Germain HAUGOU

unread,
Apr 17, 2013, 5:39:46 PM4/17/13
to Patrick Bellasi, Jens Brandenburg, Edoardo Paone, bosp-...@googlegroups.com

Hi Jens, Patrick,

 

this issue will be difficult to fix for the timeframe we have, how many frames are we able to process before it happens ?

 

Bye,

Germain

        <jens.brandenburg@hhi.__fraunhofer.de
        <mailto:jens.bra...@hhi.fraunhofer.de>
        <mailto:jens.brandenburg@hhi.__fraunhofer.de


        <mailto:jens.bra...@hhi.fraunhofer.de>>>

    jens.brandenburg@hhi.__fraunhofer.de <mailto:jens.bra...@hhi.fraunhofer.de>



    Fraunhofer Heinrich Hertz Institute
    Einsteinufer 37, D-10587 Berlin, Germany

    www.hhi.fraunhofer.de <http://www.hhi.fraunhofer.de>




    -----
    visit us at

    Girlsday 2013 / April 25 / Heinrich-Hertz-Institut, Berlin, Einsteinufer 37

    Laser World of Photonics 2013 / May 13-16 / Munich, Germany / booth 312 hall C1

    www.hhi.fraunhofer.de/events <http://www.hhi.fraunhofer.de/events>






--
#include <best/regards.h>

Patrick Bellasi
Post-Doc at Politecnico di Milano
http://home.dei.polimi.it/bellasi


--
Mit freundlichen Grüßen / Best regards

Jens Brandenburg

Embedded Systems Group
Image Processing Department



Tel +49 30 / 31002-660

Fraunhofer Heinrich Hertz Institute
Einsteinufer 37, D-10587 Berlin, Germany
www.hhi.fraunhofer.de

-----
visit us at

Girlsday 2013 / April 25 / Heinrich-Hertz-Institut, Berlin, Einsteinufer 37

Laser World of Photonics 2013 / May 13-16 / Munich, Germany / booth 312 hall C1

www.hhi.fraunhofer.de/events

Reply all
Reply to author
Forward
0 new messages