Ilyes Gouta wrote: > > Hi Erik, > > Congratulations for the achievement! I'd like to ask you a couple of > questions: > > 1. What was the bug? Does it affect the m5603c-mt9v011 branch too? It was a silly bug related to that I use three linked lists acting as FIFO queues. The applications queue up buffers using the QBUF ioctl on the inqueue buffer. When a frame has been fully received it is moved over to the post processing queue. When the frame is fully post processed (color is recovered, frame is resized if necessary) it is moved to the out queue.
The bug was that I accessed the first frame in the post processing queue without removing it and moved the frame from the post processing queue when the post processing was finished (which may take some time). This lead to a race condition where it was possible for the next scheduled frame post processing to access the same frame leading to a lockup. The correct solution was to remove the frame from the post processing queue the moment post processing started. > 2. I intend to apply your mechanism, i.e doing the bayer decode in > another workqueue. Is it really worth the trouble? How does it impact > the performance of the driver? I have very good results and haven't noticed any means of a slow down. Right now I'm using the system workqueue which is shared with other drivers. It's possible when the system is loaded that this may lead to jittery display. I haven't investigated it further though. I also don't know if this soft-scaling is frowned-upon by the larger v4l community. But as we don't have any datasheets this will have to do for now. I belive you could basically copy and paste my code if you want to. Regards, Erik > > Regards, > Ilyes Gouta. > > Erik Andrén wrote: >> I'm proud to say that we're finally going somewhere with the m5602- >> ov9650-2 branch of the driver. Today I finally squashed a major bug >> causing a lot of instability and hindered amsn, cheese and others to >> work. >> >> Please test the m5602-ov9650-2 branch and see how it works out for >> you, report any issues to the list. >> >> Known issues are: >> * Some programs convert the image into a YUV format, this causes the >> image to get very dark and the colors get disorted. I'm currently >> investigating this. >> >> * The v4l2 controls may still be unstable and badly calibrated >> >> Regards, >> Erik Andrén >> >> ------------------------------------------------------------------------- >> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference >> Don't miss this year's exciting event. There's still time to save >> $100. Use priority code J8TL2D2. >> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone >> >> >> _______________________________________________ >> M560x-driver-devel mailing list >> [email protected] >> https://lists.sourceforge.net/lists/listinfo/m560x-driver-devel > ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ M560x-driver-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/m560x-driver-devel
