Path: utzoo!attcan!uunet!ns-mx!iowasp.physics.uiowa.edu!maverick.ksu.ksu.edu!zaphod.mps.ohio-state.edu!rpi!bu.edu!purdue!mentor.cc.purdue.edu!l.cc.purdue.edu!cik
From: cik@l.cc.purdue.edu (Herman Rubin)
Newsgroups: comp.arch
Subject: Re: Compilers taking advantage of architectural enhancements
Summary: Ignorance should not determine the future
Message-ID: <2649@l.cc.purdue.edu>
Date: 17 Oct 90 12:58:51 GMT
References: <1990Oct9> <3300194@m.cs.uiuc.edu> <11922@ganymede.inmos.co.uk>
Organization: Purdue University Statistics Department
Lines: 46

In article <11922@ganymede.inmos.co.uk>, des@frogland.inmos.co.uk (David Shepherd) writes:
> In article <336@fjcp60.GOV> golds@fjcnet.GOV (Rich Goldschmidt) writes:

	[I have included the whole of Goldschmidt's article, so I
	can comment on that as well in this posting.]

|   Maybe this is naive or too futuristic, but is anyone working towards
|   methods for automatically generating a compiler based on the architecture
|   design?  It would seem that even before there is silicon, there is
|   enough info about a new CPU in the CAD system used for layout that the
|   features of special interest for generating a compiler are known.  To
|   the extent that generating a compiler is rule based (those tasks a
|   good CS grad can do?), why hasn't it been automated?  Or are there people
|   working on this now?  Chip designers might even take compilers into 
|   consideration in their designs :-).

> they already do! 4 years ago when the IMS T800 was being designed out of
> the 8 people involved in the fpu design 1 was writting standard function
> libraries and another was writing the compiler. This enabled the
> instructions to be tuned to take account of the code sequences that the
> compiler produced/needed. high performance now depends on both hardware
> and software technology so you shouldn't design one without considering
> the other!

If we limit languages and compilers to what is taught to CS graduate
students now, we can only freeze them at the present sorry state.  There
is so much more that a thinking human can do, and do better, faster, and
more clearly, than is provided for in the HLLs.  Much of this was even on
older computers.

The same holds for hardware design.  We are already in the vicious cycle
where constructs are not used because they are not in the language, and
they are not put in the hardware because they are not used, and so it goes.

It seems that people who can see the deficiencies in design even without
using it are too rare.  As has been stated,

	The reasonable man adjusts himself to his environment.  It is
	the unreasonable man who insists on adjusting the environment
	to himself.  Therefore, all progress depends on the 
	unreasonable man.

-- 
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907
Phone: (317)494-6054
hrubin@l.cc.purdue.edu (Internet, bitnet)	{purdue,pur-ee}!l.cc!cik(UUCP)