John Kasunich wrote: >Stephen Wille Padnos wrote: > > >>I have done reasonably extensive testing on an Intel Core 2 Duo CPU, and >>I found that the latencies weren't improved much simply by using an >>isolated CPU. What did help (a lot!) was making the Linux-managed core >>do a lot of nothing. Running the bash script `while true ; do echo >>"nothing" > /dev/null ; done` improved matters a lot. Some additional >>trimming of loaded modules, combined with using ext2 instead of ext3 >>(kjournald made a blip every 5 seconds) made that machine get 200-400ns >>average latencies, with the highest spikes still in the 2 us range. >> >> >It didn't occur to me at the time, but I bet I know why the "do a lot of >nothing" task helped. I bet the Linux idle task puts the CPU into a >HALT or some other low power state. Even though the RT stuff is running >on the other core, there might be some side effects of that state that >hurt latency. I could certainly see how it would help on a single core >system - if the CPU halts, then the latency will include whatever time >is needed to get it out of that halt state, which might be non-trivial. > > Though that is possible, I got the same results using the kernel "no-hlt" parameter (or however it's spelled), which is supposed to prevent that.
- Steve ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
