From cse.psu.edu!9fans-outgoing-owner Sat Oct 28 13:02:16 1995
Received: from colossus.cse.psu.edu ([130.203.1.2]) by cannon.ecf.toronto.edu with SMTP id <1016>; Sat, 28 Oct 1995 13:02:13 -0400
Received: by colossus.cse.psu.edu id <78453>; Sat, 28 Oct 1995 12:51:16 -0400
Received: from plan9.att.com ([192.20.225.252]) by colossus.cse.psu.edu with SMTP id <78454>; Sat, 28 Oct 1995 12:50:50 -0400
From:	philw@plan9.att.com
To:	9fans@cse.psu.edu
Date:	Sat, 28 Oct 1995 12:32:02 -0400
Subject: Re: alef fp bug on 386
Message-Id: <95Oct28.125050edt.78454@colossus.cse.psu.edu>
Sender: owner-9fans@cse.psu.edu
Precedence: bulk
Reply-To: 9fans@cse.psu.edu
Status: RO

Compiler bug. This fixes it:

philw-thresher% diff inst.c /n/bootes/sys/src/alef/8
498c498
< 		a = AFCOMD;
---
> 		a = AFCOMF;
502c502
< 		a = AFCOMDP;
---
> 		a = AFCOMFP;

phil


From 9fans@cse.psu.edu Wed Nov  1 16:43:27 EST 1995
Article: 734 of comp.os.plan9
Xref: cannon.ecf comp.os.plan9:734
Newsgroups: comp.os.plan9
Path: cannon.ecf!utnut!cs.utexas.edu!usc!howland.reston.ans.net!EU.net!uknet!strath-cs!dcl-cs!bath.ac.uk!ccsis
From: dhog@plan9.CS.SU.oz.AU
Subject: alef fp bug on 386
Message-ID: <95Oct28.065316edt.78430@colossus.cse.psu.edu>
Sender: ccsis@bath.ac.uk (Icarus Sparry)
Reply-To: 9fans@cse.psu.edu
Organization: Plan 9 mailing list
Date: Sat, 28 Oct 1995 02:40:54 GMT
Approved: plan9mod@bath.ac.uk
Lines: 31
Status: RO

The following alef program prints the wrong answer on a 386 architecture
machine (actually a pentium...).

----------------
#include <alef.h>

void
main()
{
	float x, y;

	x = 3.0;
	y = 5.0;
	if (x > y)
		print("%f > %f\n", x, y);
	else
		print("%f <= %f\n", x, y);
}
----------------
Here's the output on a pentium:

	term% 8.out
	3.000000 > 5.000000

and on a mips:

	cpu% v.out
	3.000000 <= 5.000000

(that's more like it!).  Does anyone know why?