caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
From: Sam Steingold <sds@gnu.org>
To: caml-list@inria.fr
Subject: Re: compiling large file hogs RAM and takes a long time.
Date: Thu, 07 Jun 2007 11:52:03 -0400	[thread overview]
Message-ID: <466829A3.2090508@gnu.org> (raw)
In-Reply-To: <4666E11F.6000308@podval.org>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Sam Steingold wrote:
> I wrote a parser generator for a tick data file.
> The generated OCaml file is ~1Mb and contains ~120 variant type
> definitions (each with 2 to ~100 variants) plus one polymorphic variant
> type.
> When I tried to compile it (3.09.3), it took almost 10 minutes and
> consumed ~500MB RAM (Firefox and Thunderbird were killed by the kernel
> to make space for ocaml).
> I run Linux 2.6.18.8 on a dual CPU Pentium D 2.80GHz with 1GB of RAM.
> (in 32-bit mode)
> 
> So, is this something to be expected? 10 min / 500MB seems like a little
> bit excessive to me, given that the file is really very simple conceptually.

update: the main problem is a polymorphic variant type with 3050
variants and a function that returns its instances.
chunking the type, i.e., replacing
type foo=[|`A|`B|`C|`D]
with
type foo_1=[|`A|`B]
type foo_2=[|`C|`D]
type foo=[|foo_1|foo_2]
(but with ~55 variants in ~55 types) did not help.
Removing the type definition did solve the RAM problem - according to
top(1), ocamlopt.opt was taking ~80MB, and it compiled the file in about
10 minutes.

Any chance there is some quadratic code in polymorphic variant type
processing?!

Sam.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGaCmjPp1Qsf2qnMcRAn52AJ4vKgALmumuKvzgo3aiG4LaavmKJACeKfWn
glb7ZDcjbMt/lIWYWMmPF5M=
=xr5y
-----END PGP SIGNATURE-----


  parent reply	other threads:[~2007-06-07 15:52 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-06-06 16:30 Sam Steingold
2007-06-06 16:51 ` [Caml-list] " skaller
2007-06-06 17:05   ` Sam Steingold
2007-06-07 15:52 ` Sam Steingold [this message]
2007-06-08  1:02   ` [Caml-list] " Jacques Garrigue
2007-06-08  1:51     ` skaller
2007-06-08  2:26       ` Yaron Minsky
2007-06-08  9:05       ` Thomas Fischbacher
2007-06-08  9:35         ` skaller
2007-06-08  9:55           ` Thomas Fischbacher
2007-06-08 13:39       ` Sam Steingold
2007-06-08 12:30     ` [Caml-list] " Jacques Garrigue
2007-06-15 15:41       ` Sam Steingold
2007-06-15 18:56         ` [Caml-list] " Jon Harrop
2007-06-15 20:06           ` Sam Steingold
2007-07-09 20:22 ` large parametrized polymorphic variant type combinations take forever to compile Sam Steingold
2007-07-09 22:45   ` Sam Steingold
2007-07-09 23:37   ` [Caml-list] " Jacques Garrigue
2007-07-10  7:09     ` Christophe Raffalli
2007-07-10  7:31       ` Jacques Garrigue
2007-07-10 14:16     ` Sam Steingold
2007-07-10 16:49     ` Sam Steingold
     [not found]     ` <46938BDA.1090605@podval.org>
2007-07-11  0:10       ` [Caml-list] " Jacques Garrigue
2007-07-11  1:19         ` Jon Harrop
2007-07-11  2:23           ` Jacques GARRIGUE
2007-07-11 13:12         ` Sam Steingold
2007-07-11 19:17           ` [Caml-list] " Jon Harrop
2007-07-10  3:34   ` [Caml-list] " skaller
2007-07-10 13:27     ` Sam Steingold

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=466829A3.2090508@gnu.org \
    --to=sds@gnu.org \
    --cc=caml-list@inria.fr \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).