On Thu, 2004-07-22 at 17:56, Eric Valette wrote: > Philippe Gerum wrote: > > > Never put X in the loop; results cannot be reliable with it and it makes > > no sense to interpret them since the user-space X driver can do whatever > > it wants to with your hadware, especially when switching back and forth > > vt7. So I cannot comment these figures. > > May I make the same comment again, publicly this time : for me, one > reason of using a RTAI implementation is having the normal linux > environment including sophisticated X11/KDE, apps. Otherwyse, I would go > for RTEMS or eCos. > > So I think making benchmarks with X app running, is relevant (not > obviously the vt switch). Also, for embeded application, using the > framebuffer makes the X server behaves almost like any other linux > driver. Furthermore, the X server indeed performs IO but do not handle > interrupts directly. And if you use this argument, any user space > drivers would also be removed and this includes USB drivers, ...
To summarize my views here: o Yes, RTAI might work with X if your X driver does not fiddle with the same stuff that RTAI does, basically: on-boad/on-cpu timer or interrupt mask at CPU level (and a few other bus-related things I guess). Wrt to the latter, I guess that we both know why iopl() was implemented in the first place... o Indeed, I've experimented many crashes while transitioning back and forth X and non-X displays while a RTAI app was running. o Yes, frame buffer might help, but depending on your X-driver code, it might or might not be sufficient to have it enabled for the whole thing to work reliably. It's a matter of experimentation. o No, I cannot recommend using X blindly with RTAI, because I would have to know about the implementation of all pre-compiled drivers (which makes your USB argument irrelevant since USB support is compiled from kernel sources in any case, so it cannot remain in our way wrt interrupt handling). Any source for the NVidia ones btw? The above has nothing to do with the usefulness of RTAI as a RTOS, it's only a matter of tested configurations. -- Philippe.
