I'm also seeing the same type of behavior but it's kind of a quirky scenario. When I run my application using a NFS all runs fine and dandy, however when I load up a ramdisk (the one that come with dvevm) the loadmodules script fails (has a problem loading /dev/cmem) and then the application hangs. I was able to see that the application hangs when it calls TraceUtil_start, and if I remove it, then it'll hang when it tries to open the engine. I'm not quite sure where the problem area is, but I'm leaning towards how busybox on the ramdisk is handling the insmod command. Either that, or udev is not properly functioning? I'm running 2.6.10_mvl401-davinci kernel. Any ideas? (And I don't mean to do any sort of thread jack here)
BJ -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Sam Hague (shague) Sent: Thursday, January 31, 2008 8:45 AM To: [email protected] Subject: Codec Engine hangs when tracing is enabled Hi all, does anyone have insight into why the DSP/ARM would hang when the tracing collection thread is enabled? I have a codec engine algorithm developed by someone else. I have modified the decode demo to call this new algorithm after an h264 decode of a frame has been completed. The decodeCombo server has been modified to include this new algorithm Everything works fine if I do not call "TraceUtil_start(ENGINE_NAME)" to start the trace collection thread. In the decode.cfg file it does not matter what "TraceUtil.attrs = TraceUtil.SOCRATES_TRACING" is set to. Either SOCRATES_TRACING or FULL_TRACING. If "TraceUtil_start(ENGINE_NAME)" is called, then the decode demo app hangs when the VIDDEC_process(newAlg) all is made. The VIDDEC_process(h264) call works fine and then the decoded frame is passed into the VIDDEC_process(newAlg) call where it hangs. The VIDDEC_create(newAlg) also returns successfully. If "TraceUtil_start(ENGINE_NAME)" is not called then everything works fine. Also if the TraceUtil.attrs = TraceUtil.FULL_TRACING, the newAlg works fine and dumps the tracing messages to stdout. My thinking is there is probably something in the new algorithm that has not been implemented for supporting the trace collection. Or maybe some build option for the server. So, I am looking for ideas to poke the codec developer. Thanks, Sam _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
