[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MIT-Scheme-devel] Cleaning weak hash tables in secondary GC daemon?

From: Matt Birkholz
Subject: Re: [MIT-Scheme-devel] Cleaning weak hash tables in secondary GC daemon?
Date: Wed, 21 Sep 2011 13:46:02 -0700

> From: Taylor R Campbell <address@hidden>
> Date: Wed, 21 Sep 2011 00:46:09 +0000
> [...]
> But maybe you really want to use a GC finalizer or a `primary' GC
> daemon.  What are you using secondary GC daemons for?

Nutt'n; it was ALL System.  My FIRST suspect was the NEW (my!) code,
but I could find no data explosion there.  I forget how I happened to
investigate trigger-secondary-gc-deamons!.  I perhaps thought there
was nothing in the system calling it (at a useful time).  Once I had
my leak plugged, I did not look back.

I have a demo going now withOUT my trigger.  It is GCing approx. every
10 minutes and is coming up with fewer and fewer free words each time
(sometimes down 200,000 words, sometimes just 20,000).

I don't think there is a bug here, just a caveat that an application
will want to trigger-secondary-gc-deamons! itself, occasionally, at
application-appropriate times.  The SYSTEM needs it, but won't do it
(without aborting some hapless thread).

> What I meant by `we don't have a scalable GC' is that our pauses are
> long (and unbounded above) and our throughput is low (and unbounded
> below).

I find the "long pauses" hardly noticeable in 10frame/sec-type
animations.  The "low throughput" is in terms of the instantaneous
rate of crunched numbers?

> This doesn't matter much with a 26-bit limit on the size of the
> heap, but it does matter with a 58-bit limit.

Because 58bits allows you to allocate a generation0 of unusual size?

reply via email to

[Prev in Thread] Current Thread [Next in Thread]