Re: [RTnet-users] Round trip test

[Arno G. <arn...@gm...>]

> Mica Hallin schrieb:
> > Hi, I try to do some basic performance test of RTnet
> > 
> > What correspond exactly TDMA_CYCLE and TDMA_OFFSET ?
> > 
> 
> TDMA_CYCLE is the period length in microseconds of a basic cycle. The 
> master station sends a Start of Frame (SOF) packet at the beginning of a 
> new cycle. Each station has one time slot per cycle, its offset to the 
> SOF is defined by multiple of TDMA_OFFSET (master = 0*offset, 1. station 
> = 1*offset, ..., n. station = n*offset).
> 
> > here is a test with TDMA_CYCLE=500 and TDMA_OFFSET="50"
> > 
> > ip address 192.168.10.3=c0a80a03
> > 
> > Warniinterval = 500
> > ng: loading rt_cpacketsize = 58
> > Roundtrip = 504us (min: 504us, max: 504us)
> > Roundtrip = 23902us (min: 504us, max: 23902us)
> > Roundtrip = 502us (min: 502us, max: 23902us)
> > Roundtrip = 24100us (min: 502us, max: 24100us)
> > Roundtrip = 500us (min: 500us, max: 24100us)
> > Roundtrip = 500us (min: 500us, max: 24100us)
> > Roundtrip = 24099us (min: 500us, max: 24100us)
> > Roundtrip = 498us (min: 498us, max: 24100us)
> > Roundtrip = 26097us (min: 498us, max: 26097us)
> > Roundtrip = 496us (min: 496us, max: 26097us)
> > Roundtrip = 26097us (min: 496us, max: 26097us)
> > Roundtrip = 497us (min: 496us, max: 26097us)
> > 
> > it's no real-time
> > 
> 
> Well, it is real-time - as long as you can live with 26 ms latency.
> 
> Ok, I also don't think that this is acceptable. Are you sure that the 
> hardware latency of all involved systems does not have any quirks?

Hi,
I just want to give a nice example to this:
I'm using RTNET together with pcmcia_cs package on an embedded
system. Pcmcia_cs's Configure script is reading some configuration
from .config and generates an own config.h file, but does not worry
about CONFIG_RTHAL. After including asm/system.h the pcmcia_cs
modules are using the hard cli() and sti() macros and damage the
hard realtime ability of RTAI/RTNET.
I got an inacceptable jitter of about 1.3ms on a 2.4kHz UDP output.

Be shure if you are using driver packages outside of the kernel tree,
that they are behaving the right way regarding the usage of a RTHAL!

Arno


> 
> > When a try whit default setting
> > 
> > the round trip is = 7000µs but it lost 1µs evry seccond.
> > 
> 
> Are you measuring the round trip delay on the master or one some other 
> node? If it is not the master, you just quantified the drift between 
> both clocks.
> 
> As with the default parameter set you don't suffer from high jitters, I 
> assume that the scheduling may overrun with the faster setup. Try to 
> play a bit with TDMA_OFFSET, maybe also with TDMA_CYCLE. What are the 
> "naked" round trip times of your systems (i.e. without RTmac/TDMA)? This 
>   may give you some idea about the system load caused by RTnet. Typical 
> times on Pentium I systems (>= 133 MHz, PCI adapters, no other 
> busmastering DMA enabled) are around 100 us.
> 
> Jan
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system

-- 
+++ NEU bei GMX und erstmalig in Deutschland: TÜV-geprüfter Virenschutz +++
100% Virenerkennung nach Wildlist. Infos: http://www.gmx.net/virenschutz