Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!whuxl!whuxlm!akgua!gatech!seismo!lll-crg!lll-lcc!qantel!hplabs!tektronix!tekcrl!tekchips!stevev
From: stevev@tekchips.UUCP
Newsgroups: net.arch
Subject: Re: Addressing modes
Message-ID: <120@tekchips.UUCP>
Date: Wed, 12-Mar-86 13:24:30 EST
Article-I.D.: tekchips.120
Posted: Wed Mar 12 13:24:30 1986
Date-Received: Mon, 17-Mar-86 03:26:37 EST
References: <946@garfield.UUCP> <1417@sdcsvax.UUCP>
Organization: Tektronix, Beaverton OR
Lines: 28

> You miss my point.  The optimisations described (in this case common 
> expression detection) are possible *regardless* of the target 
> machine's architecture.  Any good compiler should try to optimise 
> code like this.  However, once these "source level" and flow analysis 
> optimisations are finished, there should be only one reasonable way 
> of generating the machine code.  Otherwise, the poor compiler writer 
> has to do another set of optimisation analyses, all because of the 
> target machine's instruction set.  

Even though an the correctness of an optimization is machine-independent,
its effect on the quality of code is very machine-dependent.  The selection
of the ``best'' source-level optimizations will differ from machine to
machine, even for RISCs.  Consider common-subexpression elimination.
Sure, you can perform the optimization at source level, but for some
machines, it might be cheaper to redo the computation rather than to
save the result because saving the result requires a register, which could
register allocation poorer; in some cases, reperforming the computation
could even be free on RISC machines if it can be done in an instruction
that would otherwise be a delayed-branch-induced NOP.

The point is that architectural attributes such as *the number of available
registers* and *delayed branching* can effect which optimizations you should
(should not) do at the source-level.

		Steve Vegdahl
		Computer Research Lab.
		Tektronix, Inc.
		Beaverton, Oregon