Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!shelby!neon!lucid.com!lnz From: lnz@lucid.com (Leonard N. Zubkoff) Newsgroups: comp.sys.apollo Subject: Re: Domain/Lisp on series 400 Summary: More Info on Series 400 Message-ID: <1991Feb5.180100.6800@lucid.com> Date: 5 Feb 91 18:01:00 GMT References: Distribution: comp Organization: Lucid, Inc. Lines: 30 Let me elaborate on this a little more on this for the curious. Domain/CommonLISP Version 4.01 will be supporting the Series 400 and DN5500 nodes, and will no longer support SR9.7. There are not massive problems with getting the Lisp to run on the new machines; there are really only two issues. The new machines have increased the segment size from 32k to 256k. Since the old segment size was a divisor of the internal segment size we use within Lisp for memory management (64k), mapping calls were always correctly aligned to a Domain/OS segment boundary. On the new machines, just as on PRISM, this is no longer the case, and appropriate modifications to the code have been made. The second problem is that Common Lisp is an interactive environment and creates machine code on the fly, either by dynamically loading it or by compilation in memory, and so the instruction and data caches must be resynchronized when this is done. The 68040 has caches which make addition of code to do this necessary; this was almost trivial to do, as we've done it before for many of our other implementations. Unfortunately, the above two modifications are not really patchable, as are most bugs people encounter, and so a new Lisp release is necessary. I encourage anyone who wants to *seriously* use Lisp on the Series 400 or DN5500 to contact their Apollo representative and request to be a beta site. You'll get a Lisp sooner, and we'll get better beta testing of the product! Leonard N. Zubkoff Lucid, Inc.