On Tue, 2013-06-04 at 11:45 -0700, Linda Walsh wrote:

Roger Oberholtzer wrote:

...

...

...

Instead of the expected 80 msecs or so to do the compress and write, it takes, say, 4 or 5 seconds.

---

BTW, I forgot, have you (besides running iostat), run 'xlatencytop'?

That will show you what your longest system calls are taking.

@640x480 -- Are your jpegs that big I.e. what's the average size/jpeg?

1960x450 grayscale. The compression sizes are between 150K and 500K per image (x 50+ per second at 90 km/h).

Is the bottleneck in compression? If that's the case, overall, that's where your 10 cpu's can benefit you. Let the compression and writes to disk happen asynchronously in bckground serviced by 'N' workers.

Compression is not the bottleneck. It seems to be taking roughly 10-15 msecs per image.

At this point you say it takes 80ms to do the compress and write. You need to find out where things are spending their time. Is it 100% cpu bound (1 cpu@100%) for 30 minutes, then 100% disk-bound for 5 seconds? But if you wanted faster speeds, with 10 cpu's you could easily get 3-5x times that rate -- It sounds like you are doing everything in 1 execution thread right now?

The application is threaded for many parts (e.g., obtaining the images from numerous GigEVision cameras and dealing with control commands and providing status info. The compression is, at this time, in one thread. That is on our to-do list. Of course, the single thread helps us decide that there is no thread resource contention in the compression code.

-- Yours sincerely, Roger Oberholtzer OPQ Systems / Ramböll RST Office: Int +46 10-615 60 20 Mobile: Int +46 70-815 1696 roger.oberholtzer@ramboll.se ________________________________________ Ramböll Sverige AB Krukmakargatan 21 P.O. Box 17009 SE-104 62 Stockholm, Sweden www.rambollrst.se -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org