Bozitaai wrote: > Dear Jan, > > Sorry for asking stupid question and re-post the question > > Most of time the UDP packets can be sent to the receiver with precision > time, but I > observed the kernel sometimes occurs LONG DELAY, about 20 ~ 60 milliseconds. > The results are not repeatable. > > Busy-loop is replaced the nanosleep function, which is used to control the > time now. > > You think this problem can be solved by RTnet? Actually, any method can
The problem is twofold: reliable low-jitter timing (that's what an RT-extended Kernel + libs are for) and deterministic packet emission (that's what RTnet is for). > capture the message about the status of kernel during the LONG DELAY period? Yes, we have that Ipipe latency tracer available for RT-extension that make use of this patch (Xenomai and RTAI). We are heavily using the tracer for spotting issues under Xenomai, but it should also work with some patches under RTAI (at least I once made it work a few months ago). The tracer is available from the download section on www.adeos.org. Documentation is sparse unfortunately. Search the adeos mailing lists for related posts. Jan PS: I put the list into the CC again. Please use "reply to all".
signature.asc
Description: OpenPGP digital signature
