Re: [ECOS] Link problem with i386-pc platform

[Alex Mathews <alexma at crosstor dot com>]

Jonathan Larmour wrote:
> 
> Alex Mathews wrote:
> >
> > Jonathan Larmour wrote:
> > >
> > > Fabrice Gautier wrote:
> > > > Before enabling tracing the downloaded size was about 350-400k, is it
> > > > possible that just enabling tracing  and asserts make the size more than the
> > > > maximum allowed size (which should be 640ko minus 64 or 128k I think) ?
> > >
> > > Yes.
> >
> > I've made changes to load above 1MB and have room upto 16MB.  I've
> > tested the change with the hello world example.  I still need to test
> > with larger programs.  I've made the below change to the memory layout.
> > Is there any reason why I cannot make the below change to extend the
> > length and begin above the 1M boundary for the i386 PC port?
> 
> Well, if you've tried it and it works, then probably not :-).  I'm curious
> as to why you load above 1M though - any particular reason to start that
> high?

We want to be running out of extended memory, because our product that
we will be porting to eCos will have an image size > 2MB.  So, running
in conventional memory is not an option.  We currently run our
application on top of VxWorks and QNX Neutrino on i386 PCs.

> The default is so low because we can't assume the amount of memory the PC
> has. The PC is particularly bad since it can vary so much, so the default
> assumption is quite low.

I understand this point, and this is something my company will have to
change because we only run on i386 PCs.  For the most part server level
PCs.  My company had a joint press release with Red Hat last week.  I'm
not sure if you're aware of the news.  The link is below.

http://www.redhat.com/about/2000/press_crosstor.html

Thanks.

-- 
Alex Mathews
CrosStor Software, Inc.
908-226-0100 ext 148
alexma@crosstor.com