This is a shot in the dark hope that somebody (Ville if anyone I would guess) has an idea why I would see the following error.
For the record, I am using directfb with the multi-app core from cvs as of today, as well as the fusion driver from cvs as of today with a 2.4.20ish kernel (matroxfb-vsync-irq, pre-empt and lowlat included as well as a few other goodies). I am also using 1.6.1.90 of the mjpeg_tools. My video hardware is a Matrox G400 Marvel which means it is a G400 with the MJPEG encoder on the card, so I can use mjpeg_tools to record from the TV input of the card. The problem I'm having is that while mplayer to tv-out normally consumes about 40-50% CPU, as soon as I start recording from the mjpeg encoder with lavrec, CPU usage climbs to >80%. Part of the strangeness is that it's not a case where the moment recording starts, the CPU usage of mplayer shoots up, or that as soon as I start mplayer while recording, it immediately goes up to >80% but rather it seems to climb to that point, getting worse over as little as a minute or two or as much as 10 or 20 minutes. One thing that that I suppose could cause this is if lavrec was disabling the vsync irq from the card. Is this possible with matrox fbdev driver? Or is it coded that the IRQ is always delivered and not disablable? If it's disableable, what sort of thing should I be looking for in lavrec (I recently upgraded my mjpeg_tools, but not the marvel driver. It's the same as it's always been) that would be disabling it? Any other ideas? b. -- My other computer is your Microsoft Windows server. Brian J. Murrell
signature.asc
Description: This is a digitally signed message part
