[MLton] self-compiled mlton on redhat 9

[Gerard Milmeister]

On Mon, 2003-11-03 at 06:51, Stephen Weeks wrote:

> The user+sys is in the right ballpark for your machine.  I don't
> understand why it is thrashing so badly.  By default the runtime
> shouldn't use more than 80% of RAM, which is more than enough to do a
> self compile on a 1GB machine.  Was anything else running
> simultaneously and holding on to memory?
top show SIZE ca 810M and RSS ca 650M, but %CPU is very low.
I noticed I had previously used mlton-20030711-1.i386.rpm
to bootstrap. I downloaded mlton-20030716-1.i386.rpm, and
it is just as bad as the self-compiled one. I tried
to limit -h addressspace 512M, and it started fine without
trashing, but along the way it couldn't allocate memory
anymore.
Could this be a problem of the VM in the kernel 2.4.20-20.9
shipped by RedHat?

> > The first patch replaces the -m flags for the gcc compiler with
> > -f flags. Maybe you should do this in the source distribution.
> 
> Yeah.  The situation is that older gcc's (e.g. 2.96) reject -f, while
> newer ones only warn about -m.  So the decision was made to stay with
> -m for now because I build the rpms on a RedHat 7.1 machine for
> maximum portability.  Maybe I will soon abandon the 7.x line and start
> building on 8.x with gcc 3.2, in which case I will apply the patch.
-- 
Gérard Milmeister
Tannenrauchstrasse 35
8038 Zürich
gemi@bluewin.ch