Billy Biggs <vektor@xxxxxxxxxxxx> writes: > > On Mon, 4 Aug 2003, Trent Piepho wrote: > > > > > The waste of memory argument only works if you ignore user space. > > > If the driver doesn't provide multiple buffers, then the application > > > has to. The buffers are still there, they've just moved. > > > > You missed some messages? Yes, the buffers are still there, internal > > to the driver. They just aren't mmapped because the data in the > > internal buffers is in format which is not understood by any user > > application anyway (except my special test program). > > Oh great, you've written yet another driver that does format > conversion in the kernel? Thats common practice for compressed frames, because it is the only way to provide at least one of the standard v4l formats ... > When will these webcam authors (yourself included!) write a userspace > library for webcams that does all of these different format conversions > in userspace where they belong. At this point there are more drivers > being patched than the total sum of applications you would have to > (trivially!) patch to support your abstraction library. It isn't that trivial, there are also non-open source applications using the v4l API ... As first step we can export the native (compressed) format of the USB cams to userspace, additionally to one standard, uncompressed format. So applications can start moving uncompressing into userspace. v4l2 already has one vendor-specific format (V4L2_PIX_FMT_WNVA), I'll happily add more for the usb cams. A (single) library for uncompressing these would be needed too of course. Gerd -- sigfault
