home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!decwrl!deccrl!news.crl.dec.com!rdg.dec.com!ryn.mro4.dec.com!ricks.enet.dec.com!leonard
- From: leonard@ricks.enet.dec.com ()
- Subject: Re: Translating ML (was Re: MVC and MVCL)
- Message-ID: <1992Jul27.170228.10737@ryn.mro4.dec.com>
- Lines: 17
- Sender: news@ryn.mro4.dec.com (USENET News System)
- Reply-To: leonard@ricks.enet.dec.com ()
- Organization: Digital Equipment Corporation
- References: <1992Jul27.002600.18548@nntpd.lkg.dec.com>
- Date: Mon, 27 Jul 1992 17:02:28 GMT
-
-
- As I remember, EX on the IBM S/370 is normally used to supply variable arguments
- for some instructions. A machine-language compiler should replace the EX not
- with a direct emulation of EX, but with a check that the executed instruction was the expected one (move string, for example), and code that either
- simulated the effects directly (doing a variable-length move) or calling the
- run-time simulator (for the unusual case where the instruction has actually been
- modified).
-
- It may be that self-modifying code is actually a problem for compilers of /370
- machine language, but the existence and frequent use of EX is not sufficient
- evidence.
-
- Tim Leonard
-
- I don't speak for Digital.
- In fact, Digital pays to have my proofs machine-checked.
-
-