I was testing for jitter although my code outputs the wake up latency. It is easier to see the jitter using the latency values. Jitter accuracy very important on real time systems. In QNX you have consistent 1 microsecond difference between sleeps and on Linux it was 9. At that time DragonFly BSD has good latency but jitter was up to 500 microseconds.
I tried it on Linux 2.6.22 and the latency is better similar to DragonFly BSD's. But the jitter is up to 206 microseconds. Maybe I'll try real time patches next time. On 9/26/07, Pablo Manalastas <[EMAIL PROTECTED]> wrote: > From the graph, it seems to me that for 0-600 sec wake > up latency, QNX has a delay of .001846-.001847 sec, > while for Linux for the same 0-600 sec wake up > latency, Linux has a delay of .001840-001849. I am > not sure how to interpret these graphs, but it looks > to me like QNX is quicker at "waking up", while Linux > has much greater variation in response. Is this > interpretation correct? And is the difference > significant for actual real-time behavior? > > Pablo Manalastas > --- Eduardo Tongson <[EMAIL PROTECTED]> wrote: > > > A bit OT. Two years ago I compared Linux, > > DragonFlyBSD, NetBSD, > > OpenBSD and QNX nanosleep() accuracy [1]. I'm not > > sure if there are > > any significant improvements with Linux nanosleep(). > > I didn't use real > > time patches on Linux because I wanted to test bare > > kernels. > > > > <http://leaf.dragonflybsd.org/~ed/misc/nanosleep/> > > > > _________________________________________________ > Philippine Linux Users' Group (PLUG) Mailing List > [email protected] (#PLUG @ irc.free.net.ph) > Read the Guidelines: http://linux.org.ph/lists > Searchable Archives: http://archives.free.net.ph > _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
