On Saturday 08 September 2007 11:49:53 Felix Fietkau wrote: > Tobias wrote: > > Hello, > > > > I'm trying to get some more performance out of the LinkSys NSLU2 to run > > my application. When I run it under nslu2-linux (slugosbe) I get enough > > performace. The CPU usage is < 30%. When I compile and run it under > > OpenWrt the CPU hits 99.5% and the application in unusable. I've been > > thinking of what could cause this. > > There's no way for us to tell without you giving some more info on the kind > of workload you're testing with.
Some additional information: The core of the application is the "select" system call. This is what drives it all so when sampling the audio device, select wait for an audio block to become ready from the audio device. Maybe it's something strange here. I have a test application which just open the audio device using my low level core libraries which, among other things, hide the select system call. It then sample audio and immediately plays it back. This application takes about 25% CPU. That's a lot for such a simple application! It does not do any audio processing. Just sampling, converting samples to floats and back to ints again and plays them back. I'm leaning towards that the problem must have something to do with the uClibc implementation. I might have done something stupid which does not show when using glibc. Is it possible to profile an application on the target system? Maybe I can identify what's taking so much time. / Tobias _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org http://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
