From mboxrd@z Thu Jan 1 00:00:00 1970 Received: (from majordomo@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id QAA03345; Tue, 3 Dec 2002 16:34:06 +0100 (MET) X-Authentication-Warning: pauillac.inria.fr: majordomo set sender to owner-caml-list@pauillac.inria.fr using -f Received: from concorde.inria.fr (concorde.inria.fr [192.93.2.39]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id QAA03472 for ; Tue, 3 Dec 2002 16:34:05 +0100 (MET) Received: from pauillac.inria.fr (pauillac.inria.fr [128.93.11.35]) by concorde.inria.fr (8.11.1/8.11.1) with ESMTP id gB3FXwX07374; Tue, 3 Dec 2002 16:33:58 +0100 (MET) Received: (from xleroy@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id QAA03473; Tue, 3 Dec 2002 16:33:57 +0100 (MET) Date: Tue, 3 Dec 2002 16:33:57 +0100 From: Xavier Leroy To: Kevin Patrick Murphy Cc: caml-list@inria.fr Subject: Re: [Caml-list] BigArray.Array2 vs Array.make_matrix Message-ID: <20021203163357.A100@pauillac.inria.fr> References: <3DE58BF1.936A93B4@ai.mit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 1.0i In-Reply-To: <3DE58BF1.936A93B4@ai.mit.edu>; from murphyk@ai.mit.edu on Wed, Nov 27, 2002 at 10:22:25PM -0500 Sender: owner-caml-list@pauillac.inria.fr Precedence: bulk > I was wondering what are the relative merits (in terms of speed, > portability and ease of use) between using arrays of arrays, and the > BigArray class. > > While web browsing, I saw a couple of people claim that BigArray is > signigicantly slower than the standard array class: > > http://caml.inria.fr/archives/200105/msg00217.html > claims BigArray is 3 times slower > > www.cs.cornell.edu/courses/cs612/2001sp/ projects/ocaml-arrays/OCaml.pd > claims BigArray is about 10 times slower There are earlier discussions of this issue in the archives of the caml-list. See for instance my reply: http://pauillac.inria.fr/bin/wilma_hiliter/caml-list/200207/msg00445.html In general, bigarrays are slower than regular arrays, although the slowdown depends a lot on how the code is written, e.g. ensuring that bigarrays are fully monomorphic improves performance quite a lot. The main purpose of bigarrays is not to speed up pure Caml code, but to facilitate interfacing with external numerical libraries. In some circumstances, bigarrays have a few added bonuses as well: compact representation of arrays of small floats, small integers, and complex numbers; support for dimensions >= 2^22 on 32-bit machines; and no-copying slicing and shrinking operations. But all this comes at some expense in speed for raw array accesses. > Plus an interface to some standard linear algebra code would be nice... You mean, like LACAML? http://www.ai.univie.ac.at/~markus/home/ocaml_sources.html - Xavier Leroy ------------------- To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/ Beginner's list: http://groups.yahoo.com/group/ocaml_beginners