Re: Re: Re: HARDINGE HNC RETROFIT

Hi Mark (comments mixed in)

>    From: "Mark" <markatswat@...>
> Subject: Re: Re: HARDINGE HNC RETROFIT
>
> --- In CAD_CAM_EDM_DRO@y..., Ray Henry <rehenry@u...> wrote:

<s>

> > My first question when asked about some "goofy" behavior on the
> > part of  an EMC is where'd you get it.
>
> I don't know if you want to hear the answer. ;) I got the BDI-TNG
> Install CD from you for $30.00

Good. That means it is a current version with some known
characteristics. I've been using TNG for my network and www stuff as
well as EMC testing for some time and it seems to be very stable for most
things. I have found that it doesn't always like me to start up more
than one emcsh at the same time because when I kill one of them, it kills
them all. I do this a lot while developing. I just let the EMC run on
one of the virtual terminals and start and stop my development stuff on
another. That is the only major problem that I've seen.

<s>>

> I was able to repeat the problem, both times the Backplotter was idle
> overnight, I think the Backplotter is the cause of the problem.

Yes this may well be true. There is no cap on the number of points that
the backplotter can create during one of it's runs and if there were
small differences in position while sitting still, it would pile those
points on top of each other until it consumed all available memory. Not
exactly a memory leak but a fault just the same. Is it possible that you
are running a stepper based system and have not set deadband so as to
idle the motors and positioning systems while the interpreter is idle.

I have started some plots here that took several hours of running and let
them complete while I did other things. They were still there when I got
back to them and they had not trashed the underlying software systems.
These make some big temporary files, often much larger than the RAM that
is available. Today, I ran Tub_Mold.ngc and wound up with emcsh owning
about 175000 when I show those things with kpm. After a couple of hours
idle it still showed the same value. It did take several seconds after
the kill command before it went away.

When I created backplotter, I gave no thought to the idea that it would
run for long periods of time. My only intent was to see the tool path
and this it does, but not well because there is no zoom or rotate of that
path after the plot is complete. (The newer version allows this but
looses some of the flexibility of the original.)

> I'll see if the problem happens without the Backplot. Hopefully I can
> narrow it down.

That will help, thanks.

> Mark
>
> PS - I noticed the kernels with the RT patches are setup for a single
> processor, am I correct in assuming the the RT kernels do not support
> SMP.

I've never tried SMP. I believe that some on the emc@... list have.
Seems like under rtlinux, mbuff and the shared memory things were
assigned to one of the processors and nonrealtime tasks were assigned to
the other. I have no clue what happens with the rtai that you use with
TNG.

Ray