Re: [Caml-list] Strange Behaviour with Hashtables and Marshalling

Re: [Caml-list] Strange Behaviour with Hashtables and Marshalling

[Damien Doligez]

>From: Amit Dubey <adubey@hopper.math.uwaterloo.ca>

>    (1) If I'm doing some processing, then I save a few relevant data
>structures to disk with Marshal, terminate, then restart, reload
>the data and then continue processing, I use significantly less memory than
>if I process everything without stopping.  I'd imagine with garbage
>collection, this wouldn't be a problem,

It looks like you have a fragmentation problem.  Try Gc.compact ().

-- Damien
-------------------
Bug reports: http://caml.inria.fr/bin/caml-bugs  FAQ: http://caml.inria.fr/FAQ/
To unsubscribe, mail caml-list-request@inria.fr  Archives: http://caml.inria.fr

[Caml-list] Strange Behaviour with Hashtables and Marshalling

[Amit Dubey]

    Hi Everyone,

    I've recetly have two odd things happen to me in O'Caml, and I 
was wondering if anyone else has come across the same behaviour --

    (1) If I'm doing some processing, then I save a few relevant data
structures to disk with Marshal, terminate, then restart, reload
the data and then continue processing, I use significantly less memory than
if I process everything without stopping.  I'd imagine with garbage
collection, this wouldn't be a problem, yet I have come across it with both
the x86 and sparc native code versions (haven't tried bytecodes or any
other platform).  Forcing a GC collection doesn't seem to help.

    (2) If I'm saving really really big hashtables, which are part of
an enumerated type that can also contain closures ie: 

module MyHash = Hashtbl.Make(...)
type t = Table of float MyHash.t | Junk of foo | Func of (float -> float)

    After marshalling and reloading an instance of Table(s), some elements
in the table have mysteriously changed.  This isn't a problem if I change
the type of t to:

type t = Table of float MyHash.t | Junk of foo

    Again, I think I noticed this on both the sparc and x86 versions.

    I didn't find out exactly what was going on (deadlines!), but
I was just wondering if anyone might have ideas as to what the
problem might be.

    Thanks,
    Amit Dubey

-------------------
Bug reports: http://caml.inria.fr/bin/caml-bugs  FAQ: http://caml.inria.fr/FAQ/
To unsubscribe, mail caml-list-request@inria.fr  Archives: http://caml.inria.fr