Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!cbatt!ihnp4!inuxc!pur-ee!uiucdcs!uiucdcsm!grunwald
From: grunwald@uiucdcsm.UUCP
Newsgroups: comp.arch
Subject: Re: Optimization vs. the programmer
Message-ID: <3300003@uiucdcsm>
Date: Thu, 9-Apr-87 22:35:00 EST
Article-I.D.: uiucdcsm.3300003
Posted: Thu Apr  9 22:35:00 1987
Date-Received: Sat, 11-Apr-87 23:38:25 EST
References: <479@danews.ATT.COM>
Lines: 16
Nf-ID: #R:danews.ATT.COM:479:uiucdcsm:3300003:000:888
Nf-From: uiucdcsm.cs.uiuc.edu!grunwald    Apr  9 21:35:00 1987


Strength reduction is replacement of operations such as "i = j * 8" by
"i = j << 3", or "i = j % 8" by "i = j & 0x3" (if j is unsigned).

As for source-to-source translation: not for me. A lot of the optimizations
(dead-code analysis, colored register allocation, etc) isn't at the level
of the source code.
	It is this class of optimization which kills statements such as
"n = n;". You could recognize ``useless'' statements like this & warn the
person about it, but I think it would be better to simply dis-able the
optimizations at that point and/or code that little-needed routine
in assembly code.
	Vector & concurrentization optimizations can usually be done in
a source-source fashion, but this stuff would just fill your source files
up with meaningless garbage which is impossible to edit/maintain. The
only ``source'' which it would be reasonable to convert it to is assembler.