Xref: utzoo comp.compilers:1440 comp.lang.fortran:4003
Path: utzoo!attcan!uunet!ogicse!mintaka!spdcc!esegue!compilers-sender
From: eggert@twinsun.com (Paul Eggert)
Newsgroups: comp.compilers,comp.lang.fortran
Subject: Re: IEEE 754 vs Fortran arithmetic
Keywords: Fortran
Message-ID: <1990Oct25.010604.4796@twinsun.com>
Date: 25 Oct 90 01:06:04 GMT
References: <9010230628.AA22160@admin.ogi.edu> <1990Oct24.162529.20452@zoo.toronto.edu>
Sender: compilers-sender@esegue.segue.boston.ma.us
Reply-To: eggert@twinsun.com (Paul Eggert)
Organization: Twin Sun, Inc
Lines: 13
Approved: compilers@esegue.segue.boston.ma.us
Apparently-To: uunet!comp-compilers

>[Given that +0 = -0, it's not clear to me that the existence of -0 breaks
>anything.  Keep in mind that F77 is a permissive standard....  -John]

Not permissive enough, unfortunately: IEEE's -0.0 breaks F77's input/output.
ANSI X3.9-1978 says that -0.0 and +0.0 must print the same way.  IEEE 754-1985
says that -0.0 and +0.0 are distinct, and that printing any number and reading
it back in must yield the original number if the proper precision is used.
These provisions contradict each other.  An implementation cannot conform to
both standards without unnatural contortions.  I suspect most F77
implementations simply ignore the IEEE requirement here.
-- 
Send compilers articles to compilers@esegue.segue.boston.ma.us
{ima | spdcc | world}!esegue.  Meta-mail to compilers-request@esegue.