From z Tue Dec 05 00:05:53 2000 Received: from maguro.epr.com ([198.3.162.27]) by exchange.epr.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2650.21) id WNKA3DYG; Mon, 4 Dec 2000 15:04:02 -0800 Received: from magrathea.epr.com (firewall-user@magrathea.epr.com [198.3.160.1]) by maguro.epr.com (8.9.3/8.9.3) with ESMTP id PAA09952 for ; Mon, 4 Dec 2000 15:07:56 -0800 (PST) Received: (from uucp@localhost) by magrathea.epr.com (8.9.3/8.7.3) id PAA10442 for ; Mon, 4 Dec 2000 15:07:55 -0800 (PST) Received: from nodnsquery(199.222.69.4) by magrathea.epr.com via smap (V5.5) id xma009955; Mon, 4 Dec 00 15:07:03 -0800 Received: from clairv.com (IDENT:root@pc246.edi.nunet.net [199.249.165.246]) by mail.acm.org (8.9.3/8.9.3) with ESMTP id SAA30690 for ; Mon, 4 Dec 2000 18:06:57 -0500 Received: from magrathea.epr.com (firewall-user@magrathea.epr.com [198.3.163.3]) by clairv.com (8.9.3/8.9.3) with ESMTP id RAA01563 for ; Mon, 4 Dec 2000 17:06:57 -0600 Received: (from uucp@localhost) by magrathea.epr.com (8.9.3/8.7.3) id PAA09837 for ; Mon, 4 Dec 2000 15:06:55 -0800 (PST) Received: from unknown(198.3.162.27) by magrathea.epr.com via smap (V5.5) id xma009282; Mon, 4 Dec 00 15:06:07 -0800 Received: from eponym.epr.com (eponym.epr.com [10.1.1.123]) by maguro.epr.com (8.9.3/8.9.3) with ESMTP id PAA09879 for ; Mon, 4 Dec 2000 15:06:06 -0800 (PST) Received: (from sweeks@localhost) by eponym.epr.com (8.9.3/8.9.3) id PAA24373; Mon, 4 Dec 2000 15:05:53 -0800 From: "Stephen Weeks" MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <14892.9041.816128.676359@eponym.epr.com> Date: Mon, 4 Dec 2000 15:05:53 -0800 (PST) To: MLton@sourcelight.com Reply-to: MLton@sourcelight.com Subject: Re: CWS paper In-Reply-To: <200012042132.PAA30467@syzygy.clairv.com> References: <200012042132.PAA30467@syzygy.clairv.com> X-Mailer: VM 6.71 under 20.4 "Emerald" XEmacs Lucid X-Spam-Checker-Version: SpamAssassin 2.60 (1.212-2003-09-23-exp) on eponym.sweeks.com X-Spam-Level: X-Spam-Status: No, hits=0.0 required=3.0 tests=AWL autolearn=no version=2.60 > Is the contification strategy you are working on different than the one I > suggested (where the continuation of a function f ends with return-from-f)? > That seems pretty much optimal to me. I guess it could also end with > tail-call-to-g. The one you suggested (as I understood it) failed on the example that I sent out on Nov 22. > fun f () = ... h () ... g () ... > fun g () = ... h () ... f () ... > fun h () = ... > ... K (f ()) ... > > All calls are tail calls. It seems to me that the extended analysis will not > be able to contify h because it will be called with two different > continuations: "return to f" and "return to g". On the other hand, with the > old analysis, the set of continuations for f, g, and h would be {K}, and all > would be contified. Matthew sent a proposed fix to the continuation based analysis to handle this problem. Intuitively, it made sense, but I haven't quite figured out how to make an analysis of it. > A completely different possibility is to observe that the main thing being > saved (in turning a procedure into a continuation) is construction of a > fram. Thus you could turn procedures which are only called with a few > continuations into continuations which take a flag argument to indicate where > they should go when they are done. I.e., it really is a form of closure > representation in the world of full CPS. This could work. I think the problem is that the CPS IL ties together the notions of environment and stack. For example, there is no way to keep around one environment while doing a nontail recursion that refers to that environment.