Re: Imbedded computers

Don't get too enamored of AGP. It is over, no longer being designed into
motherboards. The serial/ parallel PCI-Express is WAY easier to design
for, so most new video cards are only available in PCI (relatively low
performance) and PCI Express. PCI Express uses (up to) eight high speed
serial lines in parallel. It is nearly impossible to lay out a 64 bit
parallel bus on a PC board without lethal time skews under 64 bit PCI. Thus
you mostly see 32 bit PCI cards so far, now jumping to PCI Express. (Has
anyone actually seen a true 64 bit PCI card, with the extra bus contacts
really hooked up? One that really works in 64 bit mode??)

It seems there are two practical ways to get a properly paced series of
commands to the CNC interface: either add a 'virus' that becomes a tiny
pre-emptive kernal which runs Windows as a task (see Mach 2/3) or add a
cheap external buffer to work off a standard I/O port (USB in the 21st
Century, right) deep enough to avoid the vagaries of the Windows dispatcher
timing.

This buffer can be as dumb as a PIC uProcessor plus buffers or as elaborate
as you like. Note that there are real benefits to placing the lowest
levels of driver processing 'down in the spine', like your bicycle skills
or the phase sequencing of a stepper. The hottest arguments seem to
revolve around details of this partitioning. Sort of like watching
churchmen splitting hairs over minor theological points - sheese! Just do
something! LOL

Jeff Thompson

>________________________________________________________________________
>
>Message: 16
> Date: Wed, 04 Jan 2006 16:36:12 -0600
> From: Codesuidae <codesuidae@...>
>Subject: Re: Imbedded computers
>
>Stephen Wille Padnos wrote:
>
> >The one thing that is different from most desktops is embedded video.
> >Not just "video on the motherboard", but "video using shared memory
> >access". This can be a problem for interrupt latency, and the CPU can't
> >necessarily even tell that anything bad is happening - it gets
> >effectively stopped while the video subsystem is using main memory
> >
> >
>This could be a problem with AGP specifications 1.0 and 2.0. AGP 3.0
>includes isochronous mode to address excessive delays caused by the
>video subsystem. This is specifically intended to address problems
>caused by AGP transactions taking excessive time away from
>time-sensitive applications. This mode might fix the problem for
>montion control applications as well.
>
>I'd be interested in seeing any data you have that documents problems in
>pulse generation caused by the AGP subsystem.
>
>Dave K
>