V4L2 also provides for non-blocking streaming captures and reads by supporting select(), which will indicate the file descriptor has stuff to read either if a streaming buffer is available to be dequeued, or (if streaming is off), if a frame is available to be read with read(). This is something I really miss in V4L. I'm not sure if read() is working in bttv (it's not in the version I have - I use mmap()), but I believe it is in bttv2. Ben Justin Schoeman wrote: > Ragga Muffin wrote: > > > > Hello all, > > > > Could someone briefly explain (or point to up-to-date info) what is > > the status of the v4l and v4l2 apis, as well as the > > bttv/bttv2 driver ? > > > > Background: > > Last year I wrote a small and simple video capture to disk program > > using the v4l api. (bt848 board, NTSC) > > Now however, the app needs to be rewritten for *much* better > > performance (frame rate etc..) and configurability (capture depth and size > > etc..). > > > > Can someone recommend me a way to go api and driverwise ? > > > > Many TIA for any insights, > > > > Ragga > > 1) bttv has been around a lot longer, and is pretty heavily tested - it > is likely to have less problems than bttv2. > 2) bttv2 has completely rewritten capture core that gets rid of a lot of > historical cruft that bttv has/had (NOTE: I have not had an in depth > look at bttv for a long time. When I first started bttv2, bttv had a > number of very bad races and lockups). > 3) From a capture point of view, the v4l2 API has the following > advantages over v4l (in my opinion): > - Multi-buffering: You can choose how many capture buffers you want > (for short captures you can do a ramdisk type thing, where you allocate > enough buffers for your capture, capture to these buffers and write to > disk later). In normal use, this also allows you to absorb a lot of I/O > latency. > - Timestamped capture. Each buffer is timestamped and sequence > numbered, as it is captured. This allows you to detect drops, and also > synchronise audio capture. > 4) In terms of configurability, v4l and v4l2 are basically the same, > except that v4l2 allows you to do field capture as well. > > I'm not sure if there are any disadvantages (except for the fact that > bttv2 is relatively untested). > > Also, take note that I/O bandwidth can often be the determining factor > for frame rate. v4l2 does allow you to use a lot of buffers, so you can > get closer to the average throughput of your I/O device, but it cannot > (obviously) increase your I/O bandwidth. > > -justin > > _______________________________________________ > Video4linux-list mailing list > Video4linux-list@xxxxxxxxxx > https://listman.redhat.com/mailman/listinfo/video4linux-list
