[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gpsd-dev] RFC 2783
From: |
Gary E. Miller |
Subject: |
Re: [gpsd-dev] RFC 2783 |
Date: |
Tue, 10 Mar 2015 14:45:09 -0700 |
Yo Hal!
On Tue, 10 Mar 2015 13:54:17 -0700
Hal Murray <address@hidden> wrote:
> >> USB is polled. ...
> > No, USB 1.1 is scheduled, at least on the PL2303. 1024 slots. So
> > we see in practice about +/1 600 uSec on the GR-601W.
>
> We are saying the same thing, just using different words.
Mostly.
> The key idea is that the polling is using a periodic schedule. There
> is no interrupt mechanism. The closest equivalent of interrupt
> latency is polling interval.
Yup.
> > That may seem poor, but is way better than I get with just NTP over
> > my local GigE net.
>
> Most GigE drivers batch interrupts to reduce CPU load. What's your
> round trip time? Mine is typically 200 microseconds.
I'm 200 to 300 mSec. With software PTP I can get it down to 6 uSec.
> For the Raspberry Pi, the Ethernet is on USB so there is more fuzz
> there. I assume it's polled more frequently than 1 ms but I don't
> have numbers handy.
USB 2.0 10/100 Ethernet. Around 600 to 700 mSec. USB 2.0 micro frames
are 125 uSec. TBD how good a timeing I can get on 2.0.
RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97701
address@hidden Tel:+1(541)382-8588
Re: [gpsd-dev] RFC 2783, Hal Murray, 2015/03/10
Re: [gpsd-dev] RFC 2783, Eric S. Raymond, 2015/03/10