Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!sri-spam!sri-unix!quintus!ok
From: ok@quintus.UUCP (Richard A. O'Keefe)
Newsgroups: comp.arch
Subject: Re: Sticky overflow (was Shift and Subtract is NOT Multiply)
Message-ID: <724@cresswell.quintus.UUCP>
Date: 3 Mar 88 09:35:56 GMT
References: <7649@pur-ee.UUCP> <7276@sol.ARPA> <718@cresswell.quintus.UUCP> <7310@sol.ARPA>
Organization: Quintus Computer Systems, Mountain View, CA
Lines: 33

In article <7310@sol.ARPA>, crowl@cs.rochester.edu (Lawrence Crowl) writes:
> In article <718@cresswell.quintus.UUCP> ok@quintus.UUCP
> (Richard A. O'Keefe) writes:
> )Suppose there were two flavours of load/move instruction (one to clear the
> )overflow bit, one to leave it alone), and that integer arithmetic
> )instructions were to set the overflow bit, but not clear it.  Then this
> )argument against shift-and-subtract would go away.
> 
> You have missed the point.

I'm afraid Lawrence Crowl missed my point entirely.
I couldn't care less about multiplication by 7 or any other specific
operation.  I just used that (which was already being discussed) as
a lead-in.  Suppose I have a Pascal or ADA program which does
	K := (some hairy great complicated integer expression);
There seem to be three approaches in current use:
(1) have an integer overflow interrupt, which is enabled for Pascal
    and ADA, but disabled for C.
(2) have a non-sticky overflow bit, which has to be tested after each
    instruction which the compiler is unable to prove will not overflow.
    Again, Pascal and ADA would keep testing this flag, C would never test it.
(3) Ignore the problem and hope it will go away.
    Very popular for Pascal, normally illegal for ADA.

With a sticky-overflow bit,
(1) you don't have to switch modes when calling C from ADA or vice versa
(2) you only have to test just before stores and compares
(3) one 'trapv' per assignment or comparison is a sufficiently low cost
    that going to sea without a life jacket might seem less attractive.

In the Good Old Days I used to use a B6700, and since that I have never
been able to regard quiet overflow as a good idea.  Sticky overflow seems
like such an obvious idea (I stole it from IEEE floating point).  Why is
it not used, or is it?