home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.arch:8334 comp.lang.misc:2702
- Path: sparky!uunet!ferkel.ucsb.edu!taco!gatech!swrinde!mips!spool.mu.edu!news.cs.indiana.edu!noose.ecn.purdue.edu!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
- From: hrubin@pop.stat.purdue.edu (Herman Rubin)
- Newsgroups: comp.arch,comp.lang.misc
- Subject: Re: MVC and MVCL (was Re: RISC Mainframe)
- Message-ID: <55124@mentor.cc.purdue.edu>
- Date: 26 Jul 92 18:17:41 GMT
- References: <1992Jul19.212901.8857@bcars64a.bnr.ca> <GLEW.92Jul20084651@pdx007.intel.com> <1992Jul20.154604.12548@bcars64a.bnr.ca>
- Sender: news@mentor.cc.purdue.edu
- Followup-To: comp.arch
- Organization: Purdue University Statistics Department
- Lines: 45
-
- In article <1992Jul20.154604.12548@bcars64a.bnr.ca> schow@bqneh3.bnr.ca (Stanley T.H. Chow) writes:
- >In article <GLEW.92Jul20084651@pdx007.intel.com> glew@pdx007.intel.com (Andy Glew) writes:
-
- >> >[Henry Spencer]
- >> >Of course, a RISC bigot like me :-) would say that this means your machine
- >> >is unbalanced: the hardware has useful capabilities that the software is
- >> >unable to exploit in a general-purpose way. A straight byte-for-byte copy
- >> >is not the only thing one might want to do at top speed...
-
- >> [Stanley Chow]
- >> Ah, I see you belong to the rewrite-code-for-each-new-machine school
- >> of programming. Otherwise, how do you write optimal loops that depend
- >> on the cache line transfer size?
-
- >>Playing both sides, it is rather easy to create programs to find
- >>and generate the optimal loops automatically.
-
- >Clearly then, you belong to the compile-code-for-each-different-machine
- >school of administration. I think I will start a disk drive company to
- >supply your disk requirements :-)
-
- Until there is a language which is powerful enough that the compiler
- can look at all the manifold ways to do what should be done, it is
- necessary to even write code for each different machine. Certainly,
- a compiler should make a different translation of HLL code into
- machine instructions for each different machine.
-
- Now how should the code be written? If the language can say, copy
- this block of memory to there, one can hope that the compiler can
- do things in a reasonably optimal manner. If the language cannot,
- then not only should the code be compiled separately, but it should
- be written separately.
-
- None of the common current HLLs is anywhere near powerful enough to do
- this, and other even essentially machine independent constructs, in
- such a way that reasonable optimizing can be done. That is, unless
- you allow the compiler to do such rash things as to say that "this is
- what the user MEANT to do." And if the user does something slightly
- different in between, the resulting code does not even accomplish the
- design goals, however inefficiently.
- --
- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
- Phone: (317)494-6054
- hrubin@pop.stat.purdue.edu (Internet, bitnet)
- {purdue,pur-ee}!pop.stat!hrubin(UUCP)
-