[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sbcl-help
Subject:    Re: [Sbcl-help] Maxima and sbcl
From:       Christophe Rhodes <csr21 () cantab ! net>
Date:       2007-06-30 10:37:51
Message-ID: 87d4zdafsw.fsf () cantab ! net
[Download RAW message or body]

C.Frangos <cfrangos@telkomsa.net> writes:

> I would be very grateful for any assistance with the following:
>
> (1) I installed the file sbcl-1.0.6-x86-64-linux-binary.tar.bz2 on a suse 
> 10.1, 64 bit system (an amd64). I am getting the following error:
>
> cfrangos@omega:~/gnu> sbcl
> mmap: Cannot allocate memory
> ensure_space: failed to validate 8589869056 bytes at 0x1000000000
> (hint: Try "ulimit -a"; maybe you should increase memory limits.)
> cfrangos@omega:~/gnu> ulimit -a
> virtual memory          (kbytes, -v) 5009920

You are enforcing a virtual memory limit for each process on your
system, such that each process should be able to allocate no more than
5GB of address space.  SBCL at startup wishes to reserve 8GB of
address space for its heap, which your limits are not allowing it to
do.  This is what the error message was trying to tell you, I think.
If you are unable or unwilling to raise your memory limit, you should
be able to request a smaller dynamic space using the
--dynamic-space-size command line parameter documented in the manual
page and in the help text from the --help command line option.

> (2) I installed the file sbcl-1.0.6-x86-linux-binary.tar.bz2 sbcl on
> a suse 9.1 32 bit system, recompiled maxima 5.12 using this
> sbcl. However, the running times for my representative program are
> much better on the old maxima 5.9.1, etc:
>
> P3, Celeron suse 9.1, Maxima 5.12, sbcl: 2862 sec.
> P3, Celeron suse 9.1, Maxima 5.9.0, clisp: 300 sec.
> P3,  suse 7.2, Maxima 5.9.1, cmucl: 21 sec.
> amd64 suse 10.1, Maxima 5.12, clisp: 600 sec.

Without knowing anything about your representative program, nor about
your setup and what you did to create the various versions of the
operating systems, lisps and other binaries, it's not possible to
comment meaningfully on the numbers you've quoted here.  To assess
which implementations work better for your needs, I suggest that you
work towards running the same body of code on the same platform, take
measurements on a program which is similar to those you need (ideally,
the _same_ as the programs you need to run), and if there is still a
question, to use profilers to find hotspots in the various
implementations' executions of those programs.

Cheers,

Christophe


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Sbcl-help mailing list
Sbcl-help@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-help
[prev in list] [next in list] [prev in thread] [next in thread]