mem usage

[Stephen Weeks]

> Is anyone else finding that 85% of 512M is a little tight for a
> self-compile?  I know the increased program size is due to duplication of
> passes in CPS and SSA.

I have seen this as well, and some time ago had changed to fixed-heap
450m and had been getting by with that.  I haven't tried a
self-compile since the recent round of updates.  Hopefully we'll be
through this peak by the end of the week and can get rid of a lot of
the CPS stuff.

> The out-of-memory errors are really the run-time-system's faults since if  an
> mmap fails it should try to scale back the amount and live with it instead of
> dieing.  This still isn't a great solution since then you  are  running  with
> all  the memory you could get so all other processes are going to get nothing
> (more or less).
> 
> It looks like we should scale back a bit.  Maybe change the magic RAMSLOP  to
> 80%?

The out of memory errors in this case are not the runtime system's
fault.  We really are trying to have too much live data.  I don't want
to scale back RAMSLOP any more.