From Miles.Ellis@etrc.ox.ac.uk Thu Jun 19 11:47:50 1997 Received: from oxmail4.ox.ac.uk (oxmail4.ox.ac.uk [163.1.2.33]) by dkuug.dk (8.6.12/8.6.12) with SMTP id LAA03071 for ; Thu, 19 Jun 1997 11:47:49 +0200 Received: from ermine.ox.ac.uk by oxmail4 with SMTP (PP) with ESMTP; Thu, 19 Jun 1997 10:47:41 +0100 Received: from [163.1.85.1] (com1.etrc.ox.ac.uk [163.1.85.1]) by ermine.ox.ac.uk (1.1/8.8.3) with ESMTP id KAA04104 for ; Thu, 19 Jun 1997 10:47:40 +0100 (BST) X-Sender: mellis@ermine.ox.ac.uk Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Thu, 19 Jun 1997 10:48:01 +0100 To: sc22wg5 From: Miles Ellis Subject: Further Floating Point Exceptions DTR Ballot During the recent ballot on the text of the Floating Point Exceptions DTR Malcolm Cohen pointed out an omission in the IEEE intrinsics covered. The revised document has already been issued for comment to the Exceptions Development Body, but it seems appropriate to also request a wider review from the whole of WG5. The revised document is on the WG5 server as N1281 in both pdf and PostScript forms, and the changes made to the earlier version (N1274) are as described in the following comments from John: "There are changes in the following places: 1/Foreword 2/1 and 5-6. 5 6 and elsewhere. Bullets added to lists. 9/ IEEE_SUPPORT_STANDARD 9/ IEEE_REM 13/-12 15.7 16/ IEEE_REM 18/9 19/-2 20/Result Characteristics (thrice) 21/ IEEE_REM 27/IEEE_SUPPORT_SQRT Malcolm was the only person to comment on the previous version and the changes are all in response: > Problem A: (Major) > ---------- > There seems to be a confusing duplication of functionality plus missing > functionality as follows: > > Missing functionality: > (4) no IEEE_REM I have added IEEE_REM as an extra function. > Problem B: (Minor) > ---------- > (a) The semantics of IEEE_SUPPORT_SQRT are buried in 15.7 "IEEE arithmetic". > I was lucky to spot them at all. 15.9.28 just says "supports" (without > mentioning the SQRT intrinsic), and 15.7 says that it means that the > SQRT intrinsic returns -0.0 for SQRT(-0.0). > > (b) Is the only requirement on SQRT for IEEE_SUPPORT_SQRT to return >.TRUE. the > return value for negative zero? What about the accuracy requirement in > the IEEE standard? You could just say that SQRT must return the value > specified by IEEE for the square root operation and give the SQRT(-0.0) > value as an example. I have changed the wording of 15.7 and the detailed description of IEEE_SUPPORT_SQRT slightly. > > Problem C: (Minor) > ---------- > IEEE_IS_FINITE, IEEE_IS_NAN, IEEE_IS_NEGATIVE, IEEE_IS_NORMAL are all > described as "Elemental function" but in each case the Result Characteristics > are "Default logical scalar". It should just be "Default logical". Done. > > Problem D: (More of an editorial quibble than a problem...) > ---------- > The document does not appear to follow the ISO/IEC guidelines - missing the > mandatory sections "Forward" and "Scope", no "Normative references" section > to list the F95 and IEEE 754 standards, "Rationale" section instead of an > "Introduction" section. Done. This has involved quite a lot of minor changes to the front material. In particular, pages 1 and 5 are quite different now. I also notice that we are not supposed to have lists withot bullets or dashes to mark each item. I have therefore added lots of bullets. > > Quibbles: > --------- > (i) page 7, second last line before heading of 2.7; "this standard" should be > "this Technical Report" since the TR is not a standard. Done. > (ii) The semantics of USE without INTRINSIC/NON_INTRINSIC are backward >from the > semantics of using an intrinsic procedure name without >INTRINSIC/EXTERNAL. I have not made this change since I believe it could alter the interpretation of a standard-conforming Fortran 95 program. . > (iv) typo "signal" for "set signalling", page 11, last sentence before the > Note. Done. > (v) I think it is worth noting in IEEE_LOGB that for normal numbers > IEEE_LOGB(X)==REAL(EXPONENT(X)-1,KIND(X)) Done. > (vi) "15.7 IEEE arithmetic" does not say whether +,-,* are required to be >done > to IEEE-specified accuracy. I have added a sentence to 15.7." It is my intention to formally agree to forward this document to JTC1 immediately after the Vienna meeting. If there are any further changes required they will be made in Vienna. Please review the document and indicate whether you believe that it should be submitted to JTC1 for final DTR approval ballot by returning the following ballot to me by noon GMT on Saturday 19th June 1997 ------------------------------ WG5 Ballot ---------------------------------- Please return this ballot to the Convenor NO LATER THAN NOON (GMT) on 27TH JUNE 1997. __________ I agree that N1281 should be forwarded for its final | | | DTR ballot in its present form | YES | NO | |_____|____| If NO, give changes required before it is balloted by JTC1: (signed) ................................ (date) .................... ----------------------------- End of WG5 Ballot ---------------------------- ------------------------------------------------------------ Dr Miles Ellis Director: Educational Technology Resources Centre University of Oxford, 37 Wellington Square, Oxford OX1 2JF, ENGLAND Telephone: +44 1865 270528 Fax: +44 1865 270527 Email: Miles.Ellis@etrc.ox.ac.uk WWW: http://www.etrc.ox.ac.uk/Personal/Miles/Miles_Ellis.html