From owner-sc22wg5@open-std.org Wed Sep 22 20:12:27 2004 Return-Path: X-Original-To: sc22wg5-domo Delivered-To: sc22wg5-domo@ghz.klid.dk Received: by ghz.klid.dk (Postfix, from userid 521) id D158137607; Wed, 22 Sep 2004 20:12:27 +0200 (CET DST) X-Original-To: sc22wg5@open-std.org Delivered-To: sc22wg5@open-std.org Received: from dkuug.dk (ptah.dkuug.dk [195.215.30.66]) by ghz.klid.dk (Postfix) with ESMTP id 0A29337605 for ; Wed, 22 Sep 2004 20:12:26 +0200 (CET DST) Received: from purcell.obs-nice.fr (purcell.obs-nice.fr [192.54.174.72]) by dkuug.dk (8.12.10/8.9.2) with ESMTP id i8MI9ajH068139 for ; Wed, 22 Sep 2004 20:09:44 +0200 (CEST) (envelope-from Bernard.Pichon@obs-nice.fr) Received: from pcpichon (pcpichon.obs-nice.fr [192.54.175.45]) by purcell.obs-nice.fr (8.12.9/jtpda-5.4) with SMTP id i8MIAZgi295088 ; Wed, 22 Sep 2004 20:10:35 +0200 (CEST) Message-ID: <003c01c4a0cf$952f0e60$2daf36c0@pcpichon> From: "Bernard PICHON" To: Cc: "NOULLEZ, Alain" , Subject: Why (standard conforming OR platform implementation) Date: Wed, 22 Sep 2004 20:11:29 +0200 MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="----=_NextPart_000_0039_01C4A0E0.58964CA0" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1409 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1409 X-Spam-Score: 0 () Sender: owner-sc22wg5@open-std.org Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0039_01C4A0E0.58964CA0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit Dear Fortran ! (expands in : Dear WG5 members, dear Lionel) This program seems to me standard conforming. But I don't understand its (different) behavior between Alpha platforms (with True64) for which the result is OK (for me) and Windows/Intel platforms (with CVF V6.6B) with a (very) strange result (in my opinion). ---------------------------------------------------------- ---------------------------------------------------------- Program ANZBUG Integer, Parameter :: SP = KIND(1.) , DP = KIND(1.D0) Real(SP) :: sx, sy Real(DP) :: dres, dres1 ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! dres = 12345.0_SP + 0.123456_SP Write(*,"(G30.20)") dres dres1 = DBLE(12345.0_SP) + DBLE(0.123456_SP) Write(*,"(G30.20)") dres1 Write(*,*) dres - dres1 ! must be non zero and it is OK .... ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! sx = 12345.0_SP sy = 0.123456_SP dres = sx + sy Write(*,"(G30.20)") dres dres1 = DBLE(sx) + DBLE(sy) Write(*,"(G30.20)") dres1 Write(*,*) dres - dres1 ! must be ALSO non zero ! BUT it is OK on True64 & Alpha, ! AND NOT OK on Win & Intel (CVF 6.6B) ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! End Program ANZBUG ---------------------------------------------------------- ---------------------------------------------------------- Also attached, the program. Question 1 : What does the Standard really say about operations with litteral constants : dres = 12345.0_SP + 0.123456_SP OR variables : dres = sx + sy ????? Question 2 : If it is (only) a platform Windows/Intel/CVF problem, how to fix it ? Rmq: In fact, switching between SP and DP is very useful to test/improve accuracies in numerical algorithms. PS: For WG5 members only, a great thanks to Mike Metcalf for his Fortran resources over past years ! Bernard -- Dr. Bernard PICHON Département Cassiopée (UMR 6202 du CNRS) Observatoire de la Côte d'Azur (OCA) à Nice BP 4229 06304 Nice Cedex 04 Telephone : 04 92 00 31 83 Telecopie : 04 92 00 31 21 Sur la toile : http://www.obs-nice.fr/pichon/ ------=_NextPart_000_0039_01C4A0E0.58964CA0 Content-Type: application/octet-stream; name="anzbug.f90" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="anzbug.f90" Program ANZBUG Integer, Parameter :: SP = KIND(1.) , DP = KIND(1.D0) Real(SP) :: sx, sy Real(DP) :: dres, dres1 ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! dres = 12345.0_SP + 0.123456_SP Write(*,"(G30.20)") dres dres1 = DBLE(12345.0_SP) + DBLE(0.123456_SP) Write(*,"(G30.20)") dres1 Write(*,*) dres - dres1 ! must be non zero and it is OK .... ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! sx = 12345.0_SP sy = 0.123456_SP dres = sx + sy Write(*,"(G30.20)") dres dres1 = DBLE(sx) + DBLE(sy) Write(*,"(G30.20)") dres1 Write(*,*) dres - dres1 ! must be ALSO non zero ! BUT it is OK on True64 & Alpha, ! AND NOT OK on Win & Intel (CVF 6.6B) ! Write(*,*) Write(*,*) " ==================== " Write(*,*) ! End Program ANZBUG ------=_NextPart_000_0039_01C4A0E0.58964CA0--