From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-help-return-45928-listarch-gcc-help=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 11803 invoked by alias); 13 May 2011 18:12:10 -0000
Received: (qmail 11793 invoked by uid 22791); 13 May 2011 18:12:09 -0000
X-SWARE-Spam-Status: No, hits=-1.8 required=5.0	tests=AWL,BAYES_00,TW_TM,T_RP_MATCHES_RCVD
X-Spam-Check-By: sourceware.org
Received: from mail.agmk.net (HELO mail.agmk.net) (91.192.224.71)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Fri, 13 May 2011 18:11:55 +0000
Received: from localhost (unknown [91.192.224.71])	by mail.agmk.net (Postfix) with ESMTP id 5828B203005E	for <gcc-help@gcc.gnu.org>; Fri, 13 May 2011 20:11:50 +0200 (CEST)
Received: from mail.agmk.net ([91.192.224.71])	by localhost (agmk.net [91.192.224.71]) (amavisd-new, port 10024)	with ESMTP id W6VzbA28PVnF for <gcc-help@gcc.gnu.org>;	Fri, 13 May 2011 20:11:37 +0200 (CEST)
Received: from vmx.localnet (unknown [89.79.206.92])	(using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits))	(No client certificate requested)	by mail.agmk.net (Postfix) with ESMTP id C0E8B203005D	for <gcc-help@gcc.gnu.org>; Fri, 13 May 2011 20:11:37 +0200 (CEST)
From: =?utf-8?q?Pawe=C5=82_Sikora?= <pluto@agmk.net>
To: gcc-help@gcc.gnu.org
Subject: question about equivalent x87/x64-64 fpu code...
Date: Fri, 13 May 2011 18:12:00 -0000
User-Agent: KMail/1.13.7 (Linux/2.6.37.6-2; KDE/4.6.3; x86_64; ; )
MIME-Version: 1.0
Content-Type: Text/Plain;  charset="utf-8"
Content-Transfer-Encoding: 8bit
Message-Id: <201105132011.34453.pluto@agmk.net>
X-IsSubscribed: yes
Mailing-List: contact gcc-help-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-help.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-help/>
List-Post: <mailto:gcc-help@gcc.gnu.org>
List-Help: <mailto:gcc-help-help@gcc.gnu.org>
Sender: gcc-help-owner@gcc.gnu.org
X-SW-Source: 2011-05/txt/msg00182.txt.bz2

Hi,

i'm using a 3rd-party engine http://glaros.dtc.umn.edu/gkhome/metis/metis/overview
for partitioning some complex data. it worked fine for years until today (may 13)...

observations:
- the 32-bit metis build produces nice and balanced partitons.
- the 64-bit metis build produces bad and unbalanced partitons.

the metis' engine uses arrays of integers on the public interface and internally
some float-based and unsafe in terms of precison (x<y and x==y) operations.

so, i've built/tested following metis variants:

1). -m32 -march=pentium4 -O1                         - works fine.
2). -m32 -march=pentium4 -O1 -mfpmath=sse            - works fine.
3). -m64 -march=x86-64 -O1                           - bad/unbalanced partitions.
4). -m64 -march=x86-64 -O1 -mfpmath=387              - bad/unbalanced partitions.

at this point i've expected wrong results (< 80-bit precision) from variants 2/3
and good results from variants 1/4 but the real world differs.

next, i've isolated a one place in sources with float x<y stmt and changed it
to (x-y)<0.00001. with such change both native 1/3 variants give nice/equivalent results.

so, where is the problem? is the variants 1/4 really equivalent?

BR,
Paweł.