[GRASS-dev] Re: [GRASS-user] problemsusing r.proj with large dataset

[Glynn Clements]

Jerry Nelson wrote:

> I did a cvs update around midday US CST on Wed (ie a couple of hours before
> I sent this email). I first ran r.proj. This time the last message I get is
> "Killed".

That suggests that the program is using too much RAM. If the kernel
has trouble obtaining enough physical RAM to operate reliably, it
chooses a process (typically one which is using a lot of memory but
isn't a "critical" system process) and kills it.

> With r.proj.seg I get the "File size limit exceeded" error.

Right.

The key difference between r.proj and r.proj.seg is that r.proj tries
to load the entire input map (or at least the portion which it
requires) into memory, whereas r.proj.seg copies the map to a file in
a convenient format for random access, essentially providing its own
form of virtual memory.

> In your email you said to change the Makefile for r.proj, then do a make in
> r.proj.seg. When I checked the Makefiles in both places, r.proj had what
> looked to be old stuff, with no large file support

r.proj uses memory rather than a file, so it doesn't need LFS. OTOH,
it requires large amounts of RAM instead.

> but r.proj.seg had exactly what you said to put in the r.proj
> makefile.

After my last email, I changed the CVS copy of the Makefile to the
version in my email.

> So I was wondering if you really meant to change the r.proj
> Makefile.

No; r.proj isn't relevant here. Once r.proj.seg has had some testing,
and has had any bugs ironed out, it will probably replace the existing
r.proj.

-- 
Glynn Clements <glynn at gclements.plus.com>