From mboxrd@z Thu Jan 1 00:00:00 1970 Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id KAA25696 for caml-red; Mon, 15 Jan 2001 10:56:12 +0100 (MET) 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 DAA17105 for ; Mon, 15 Jan 2001 03:01:55 +0100 (MET) Received: from elbereth.pgh.arsdigita.com ([63.124.128.66]) by concorde.inria.fr (8.11.1/8.10.0) with ESMTP id f0F21s525694; Mon, 15 Jan 2001 03:01:54 +0100 (MET) Received: by elbereth.pgh.arsdigita.com (Postfix, from userid 1000) id 0987636B92; Sun, 14 Jan 2001 21:09:27 -0500 (EST) To: Xavier Leroy Cc: Dave Berry , Norman Ramsey , caml-list@inria.fr Subject: Re: questions about costs of nativeint vs int References: <3145774E67D8D111BE6E00C0DF418B663AF87C@nt.kal.com> <20010114121329.A562@pauillac.inria.fr> From: John Prevost In-Reply-To: <20010114121329.A562@pauillac.inria.fr> User-Agent: Gnus/5.090001 (Oort Gnus v0.01) Emacs/21.0.91 Message-ID: <873del5zzi.fsf@elbereth.pgh.arsdigita.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: 14 Jan 2001 18:09:27 -0800 Sender: weis@pauillac.inria.fr >>>>> "xl" == Xavier Leroy writes: xl> Ocaml performs some amount of local (intra-function) unboxing xl> on large integers (nativeint, int32, int64) as well as on xl> floats. Version 3.00 missed some opportunities for local xl> unboxing of large integers, but the next release will improve xl> this, bringing integer unboxing to the level of float xl> unboxing. However, large integers as well as floats remain xl> boxed across functions and inside data structures (with a few xl> special cases for arrays and big arrays). You say that this doesn't happen across function-call boundaries. Am I right in assuming that function-call boundaries that are smoothed away by inlining are an exception to this? What about tail-recursion boundaries? John.