On 5/19/07, Hans Verkuil <[EMAIL PROTECTED]> wrote: > Hi all, > > I would appreciate it if people could start testing the latest ivtv on > the 0.10 branch, available here: > > http://ivtvdriver.org/viewcvs/ivtv/branches/0.10.tar.gz?view=tar > > The current driver has a problem with copying data from the MPEG card to > memory that are done using PIO instead of DMA (for good reasons). > Unfortunately PIO accesses are slow, so too much time is spent inside > the interrupt handler. This can lead to missing clock ticks and > problems with remotes (key hits that are missed). > > The patch I've committed moves these PIO copies to a separate handler > outside the interrupt handler, so this should solve these problems. > > However, I did not have much time to test it thoroughly, so I'd very > much appreciate it if people could start testing this and report > whether or not any new problems were introduced. > > This patch is also queued for the 2.6.22 kernel, by the way. > > If this change solves the problems that I've mentioned above, then I'll > make a new 0.10.3 release in two weeks time.
The good news is that the remote seems much more responsive. The bad news is that NTSC CC support appears to be broken. Assuming that it was related to VBI processing, I used test/vbi to dump the data from each of my cards (a PVR-150 and a PVR-350). I noticed something that I hope is useful: without anything else running, captions are largely intelligible, with only a few characters missing or changed here and there. If I attempt to watch "live" TV in MythTV, i.e. recording and playing back at the same time, the higher CPU load seems to cause more lost or changed characters, which is also mirrored by the vbi utility. For reference, I'm running a Coppermine P-III @ 800 MHz w/133 MHz FSB. --scott _______________________________________________ ivtv-devel mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-devel
