Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-lcc!styx!ames!necntc!linus!philabs!micomvax!musocs!mcgill-vision!mouse
From: mouse@mcgill-vision.UUCP (der Mouse)
Newsgroups: comp.arch
Subject: Re: Optimization vs. the programmer
Message-ID: <745@mcgill-vision.UUCP>
Date: Sun, 26-Apr-87 15:03:09 EDT
Article-I.D.: mcgill-v.745
Posted: Sun Apr 26 15:03:09 1987
Date-Received: Sun, 3-May-87 18:39:57 EDT
References: <479@danews.ATT.COM> <3300003@uiucdcsm> <919@aw.sei.cmu.edu.sei.cmu.edu>
Organization: McGill University, Montreal
Lines: 19

In article <919@aw.sei.cmu.edu.sei.cmu.edu>, firth@sei.cmu.edu (Robert Firth) writes:
> In article <3300003@uiucdcsm> grunwald@uiucdcsm.cs.uiuc.edu writes:
>> 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).

> [second example should read 7, not 3].  And in the first, the
> replacement is invalid if the expression overflows.

Why?  j*8 will overflow if and only if j<<3 overflows....the only way I
can see a problem is if overflow checking is turned on and either (a)
multiply overflows trigger it but shift overflows don't or (b) shift
overflows trigger it but multiply overflows don't.  How many machines
are there that trigger overflow exceptions on the one sort of overflow
but not on the other?  (Yes, a VAX does too trap on a shift overflow,
or at least the Architecture Reference Manual says it's supposed to).

					der Mouse

				(mouse@mcgill-vision.uucp)