home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!bnr.co.uk!uknet!pavo.csi.cam.ac.uk!ag129
- From: ag129@cus.cam.ac.uk (Alasdair Grant)
- Newsgroups: bit.listserv.ibm-main
- Subject: Re: Re: Is PL/S a dead language?
- Message-ID: <1992Dec29.134941.27874@infodev.cam.ac.uk>
- Date: 29 Dec 92 13:49:41 GMT
- References: <IBM-MAIN%92122814293609@RICEVM1.RICE.EDU>
- Sender: news@infodev.cam.ac.uk (USENET news)
- Organization: U of Cambridge, England
- Lines: 42
- Nntp-Posting-Host: bootes.cus.cam.ac.uk
-
- In article <IBM-MAIN%92122814293609@RICEVM1.RICE.EDU> Michael Stein <CSYSMAS@UCLAMVS.BITNET> writes:
- >Ha! Exactly backwards. PL/S allows writing a structure which
- >you *know* the mapping of, while C doesn't (the C mapping depends
-
- Good point. But PL/S only runs on one machine, so there are no endianness
- or word-size issues. It would be possible to say that _for any given CPU_
- the layout of C structures would be well-defined, but the C standard cannot
- do this. In fact (FLAME ON) IBM should have way back said what C compilers
- on MVS should do, how they should handle the record-oriented file system,
- what the ASCII-to-EBCDIC translation routines should be called, then we
- wouldn't have to put up with oddball compilers.
-
- >PL/S allows bare machine instructions (in addition to inline
- >assembly), ie: MVCL(R2,R4);
-
- Yes. I can't think why I implied it didn't; one of my peeves with PL/S
- is that programmers drop into machine instructions when they should use
- PL/S itself, e.g. why use MVCL instead of a PL/S assignment?
-
- >No, I've heard that at least one level of PL/S accepts
- >declarations of fields as needing CS and will generate the
- >required code. I've never seen any code which uses this though.
-
- Neither have I. What use is a language if programmers don't understand it?
-
- >-- CS and CDS use should be minimized by avoiding the updates
- >from asynchronous processes first. (Anyway, CS/CDS force a
- >pipeline drain and are slooooow compared to normal instructions).
-
- I've seen plenty of assembler code which uses CS for _all_ updates to
- certain variables, even when unnecessary (e.g. setting counters to zero).
- The whole point about high-level optimising compilers is that they should
- take this burden away from the programmer. Minimizing use of CS is
- something PL/S would be useful for, if it had the intelligence to
- understand multitasking issues. But it seems that most IBM programmers
- use PL/S as no more than one of those dreadful "structured macro" packages.
-
- Your point about SRBs etc. is a good one. But under an SRB it's important
- to have highly optimised code. I use MVS TCP/IP a lot and I think many of
- its problems (not just performance) are due to vast amounts of bloated PL/S
- in the system (IRB, SRB, PC-call etc.) code. Surely not all the assembler
- wizards at IBM are working on VM/CMS?
-