Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!mips!winchester!mash
From: mash@mips.COM (John Mashey)
Newsgroups: comp.arch
Subject: Re: Compiler Costs
Message-ID: <40052@mips.mips.COM>
Date: 11 Jul 90 10:50:33 GMT
Sender: news@mips.COM
Reply-To: mash@mips.COM (John Mashey)
Organization: MIPS Computer Systems, Inc.
Lines: 68

Since this discussion pops up here every 6-9 months, maybe it's worth
a shot at some generalizations, to put some structure around the
flood of anecdotal examples and counter-examples. 
First MY anecdote: I wrote all of the C string library in assembler for
the R2000 a long while back, because you ALWAYS do that when you do a new
UNIX.  THEN, after looking at the generated code from the standard C functions,
we threw out most of the assmebly code, because the compiler was good
enough that it wasn't worth QAing the assembly code.

Generalizations: the following factors will encourage one to use
assembly code (left) or compiled code (right).  I think most of the
anecdotes so far fit at least one of these.  The interesting questions are:
1) Are there more good genralizations?
2) What kind of experiments can be done to quantify these more specifically?

ASSEMBLER			COMPILER
1. Old, non-optimizing compiler	Modern, global optimizer, at least
				In scientific world, vectorizer/parallizer
					if appropriate.
2. Architecture/language mismatch	Architecture/language well matched

	2a. Architecture weird
	2b. Language not expressive enough


3. One architecture target, forever	Portability

	(as in some embedded)	(as in UNIX commands)

4. Absolute speed requirement	Premium on development cost, maintenance

5. Absolute space requirement	Space costs, but no hard, low limit

6. Programming time is cheap	Programming time is not cheap

As usual, in a domain-of-discourse-confusion type of discussion,
much of the arguments would disappear if people would preface their
opinion by specifiying which domain they're living in (i.e., which of
the attributes above apply to what they do.)
ALso, it would probably encourage light, and reduce heat, if people
would avoid statements of the form
	I haven't seen it so, it must not exist
in favor of
	In domain X, it doesn't seem to exist.  Does it elsewhere?

In this light:
	a) People with RISCs, good optimizers, doing general-purpose
	code, using C or FORTRAN seldom write assembler code.
	b) People with CISCs are more likely to go to assembler,
	especially when using pcc-style compilers. Likewise, people
	using machineswith vector units, special graphics operations,
	unusual pipelines or some parallel architectures.
	c) People that need operations not easily expressed in
	popular languages go to assembler more often.
	d) Embedded control folks are more likely to use assembler.
	etc.

Of course, it is the job of computer system designers to make systems
for which the assembler/compiler tradeoff increasingly favors compilers,
and the higher the level the better (i.e., do it in shell or awk ...
or 4GLs, or whatever, if possible), but this doesn't mean
that there are not sometimes good reasons for doing it in assembler
some of the time.....  Let's just hope that we can make it less often.
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	 mash@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash 
DDD:  	408-524-7015, 524-8253 or (main number) 408-720-1700
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086