Hi, Noel.
Thanks for the quick reply.
Ah. If the GC needs 2X actual that would explain 2 GB --> 1 GB usable.
That's a shame but I understand.
Since I posted, my git bisect suggests it wasn't my code that changed,
instead there was some change from PLT 4 to Racket. But next I need to
install PLT 4 again to confirm.
FWIW, if you check the RAM graph in performance monitor on Windows you see
this:
______ kablooey
/| /| / \/
/ | / |/
/ |/
/\/\/\/\/
Yeah I'll keep my day job rather than pursue a career in ASCII art. What
I'm trying to convey is a gradually rising sawtooth ("normal"). Then it
enters a second phase of long slopes up to 1 GB followed by a reclaim
drop-off, then marching up again but of course recovering less each time.
And finally the third phase is this frantic flatline. During which lots of
little ~30 msec GCs (according to DrRacket log) alternating with long, 5000
msec ones. Fourth phase is the poor DrRacket GUI gets sluggish and then
kablooey.
I'll re-install last PLT 4 version and see if it behaves differently as I
expect or if there's more to the story.
Greg
On Mon, Jul 12, 2010 at 4:19 PM, Noel Welsh <noelwe...@gmail.com> wrote:
> The copying collector needs about as much space to copy into as it
> allocates. Hence Racket keeps about 1GB free of the 2GB it is
> allocated. (This was certainly the old behaviour, but I thought it was
> made a bit less conservative so perhaps something else is going on.)
>
> HTH,
> N.
>
> On Mon, Jul 12, 2010 at 8:43 PM, Greg Hendershott
> <greghendersh...@gmail.com> wrote:
> > Regardless, the DrRacket process never seems to exceed about 1,069,999
> > bytes (Windows 7 64-bit).
> >
> > My understanding is that 32 bit processes on Win64 should get 2 GB (I
> > have 4 GB physical). So why is it maxing out at just under 1 GB?
> >...
>
_________________________________________________
For list-related administrative tasks:
http://lists.racket-lang.org/listinfo/users
