Tuukka Toivonen wrote:
On Fri, 26 Sep 2003, Johnny Strom wrote:
so my question is the bandwith enough on the PCI bus to start a capture
of four .jpeg files from four cards at the same time and then sleep for
one second after the capture?
Yes, jpeg compressed images don't use much bandwidth (if the compression is
performed in hardware before the image gets into main memory).
01:00.0 Multimedia video controller: Brooktree Corporation Bt878 Video
Capture (rev 11)
...but Bt878 can't do hardware compression. So I suppose the drivers
actually capture uncompressed and compress in software. Let's say it's PAL
resolution 768x576@25fps 4:2:0 = 16 MB/s, or 64 MB/s for four
simultaneous captures. PCI bandwidth (IIRC) is 133 MB/s, so in theory it
would work. In practice... it's difficult to say.
Yes that is right there is no hardware compression, that is done with
the streamer program.
start getting kernel oopses like the ones below, so could the tree
oopses below be becouse of a badblocks on the swap or memory related or
becouse of to much data on the pci bus at the same time?
Maybe bad blocks in the swap, if you suspect that, maybe bad memory. But
simply overloading PCI bandwidth definitely shouldn't cause oopses (instead
Bt8x8 drops data which would cause corrupted images).
I did try to cahnge the memory but I still get ooopses.
I would guess it's some hardware/software conflict/bug. Try different
kernel versions and compile it with different options.
Well I put back the old script for now and I turned off the swap becouse
I have found one badblock on it, efter testing this and if it works then
I will change to the new script that starts the capture in parallel and
without swap to see if I get oopses again.
I guess I could test 2.4.23 when it comes out as well.
Thanks a lot!
Cheers Johnny
