From owner-sc22wg5@open-std.org Thu Oct 27 11:58:51 2005 Return-Path: X-Original-To: sc22wg5-domo2 Delivered-To: sc22wg5-domo2@open-std.org Received: by open-std.org (Postfix, from userid 521) id 4725512AE9; Thu, 27 Oct 2005 11:58:51 +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 open-std.org (Postfix) with ESMTP id B0EB711473 for ; Thu, 27 Oct 2005 11:58:44 +0200 (CET DST) Received: from oin.rl.ac.uk (oin.rl.ac.uk [130.246.135.200]) by dkuug.dk (8.12.10/8.9.2) with ESMTP id j9R9wWZb052000 for ; Thu, 27 Oct 2005 11:58:37 +0200 (CEST) (envelope-from j.k.reid@rl.ac.uk) X-RAL-MFrom: X-RAL-Connect: Received: from jkr.cse.rl.ac.uk (jkr.cse.rl.ac.uk [130.246.9.202]) by oin.rl.ac.uk (8.12.11/8.12.11) with ESMTP id j9R91DSe023363 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 27 Oct 2005 10:01:13 +0100 Received: from [127.0.0.1] (localhost.localdomain [127.0.0.1]) by jkr.cse.rl.ac.uk (8.12.10/8.12.8) with ESMTP id j9R93O89014103; Thu, 27 Oct 2005 10:03:24 +0100 Message-ID: <436097DC.5000408@rl.ac.uk> Date: Thu, 27 Oct 2005 10:03:24 +0100 From: John Reid Reply-To: j.k.reid@rl.ac.uk Organization: Rutherford Appleton Laboratory User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.10) Gecko/20050811 Fedora/1.7.10-1.1.1.legacy X-Accept-Language: en-us, en MIME-Version: 1.0 To: WG5 Cc: "Stevens, Charles C" Subject: Re: SC22 meeting Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-CCLRC-SPAM-report: -4.9 : BAYES_00 X-Scanned-By: MIMEDefang 2.39 X-Spam-Score: 0 () Sender: owner-sc22wg5@open-std.org Precedence: bulk WG5, In my message of 7 October, I said > The other resolutions that affect us are > > 1. Resolution 05-08: Revision of IEEE 754 Standard for Floating-Point > Arithmetic > > We are asked to provide our input regarding requirements for our decimal > floating-point work. The Cobol convener gave me to understand that the > current problems are associated with choosing the base: 10 or 1000. > Cobol very firmly wants 10. I said that we had no strong views on this > and would probably support whatever the standard finally says. Does > anyone disagree? I have looked into this further and find that there is no suggestion of the use of any base other than 2 or 10. The deadlock is over how to represent decimal floating-point numbers. Some members of the IEEE committee use the term 'base 1000' for encodings that group the digits into sets of 3 (which can be represented efficiently in 10 bits). I was misled and in turn misled you. Sorry! The principle contenders are 1. Densely-Packed Decimal (DPD). Values 1 bit for the manitissa sign. 1 1 1 5 bits for the 'combination'. 5 5 5 w bits for the exponent. 6 8 12 10J bits for 3J manitissa digits. 20 50 110 Totals 32 64 128 This allows the following for the numbers of mantissa digits and the exponent ranges: digits emax emin 7 96 -95 16 384 -383 34 6144 -6143 The key is that the 5 bits of the combination are enough to hold one decimal digit (part of the mantissa) and one base-3 digit (part of the exponent). 2. Binary Interchange Decimal (BID) The numbers of mantissa digits and the exponent ranges are as for DPD. The essential difference is that the mantissa is held as a binary number. This needs 10J+4 bits. If the leading digit is zero, it is held directly in 10J+3 bits. If the leading digit is one, the second digit is zero and the trailing 10J+1 bits are held directly. The advantage of BID is that is it easier to simulate with software and floating-point arithmetic runs faster. The disadvantage is for actions that depend on the digits themselves, such as shifting or extraction of digit groups. It is DPD that the Cobol people favour. It has also been proposed that decimal arithmetic be removed altogether. I think that would be a shame. We all use decimal arithmetic for our thinking and hand calculations. It would be great if our computers did, too - for example, no rounding would be needed for formatted output. Does anyone disagree with this position? Of course, the Cobol people strongly favour retention of decimal arithmetic in the new standard. Cheers, John.