From jls@uxb.liv.ac.uk Mon Feb 1 13:39:22 1993 Received: from ib.rl.ac.uk by dkuug.dk with SMTP id AA15941 (5.65c8/IDA-1.4.4j for ); Mon, 1 Feb 1993 14:50:10 +0100 Received: from mail.liv.ac.uk by ib.rl.ac.uk (IBM VM SMTP V2R1) with TCP; Mon, 01 Feb 93 13:39:30 GMT Received: from uxb.liverpool.ac.uk by mailhub.liverpool.ac.uk with SMTP (PP) id <03803-0@mailhub.liverpool.ac.uk>; Mon, 1 Feb 1993 13:39:28 +0000 From: "Dr.J.L.Schonfelder " Received: from uxb.liv.ac.uk (uxf.liv.ac.uk) by uxb.liv.ac.uk; Mon, 1 Feb 93 13:39:22 GMT Message-Id: <18572.9302011339@uxb.liv.ac.uk > Subject: HPF 1.0 Comments Chapter 1 To: hpff-comments@edu.rice.cs Date: Mon, 1 Feb 93 13:39:22 GMT Cc: SC22WG5@dkuug.dk (SC22/WG5 members), UKFORTRAN@edinburgh.ac.uk (UK Fortran Panel) X-Mailer: ELM [version 2.3 PL11] X-Charset: ASCII X-Char-Esc: 29 The following are my comments as a result of a weekend reading HPF 1.0 These are my opinions and have not been discussed with members of either ISO working group on Fortran (WG5) or the BSI Fortran panel. However I am including both groups on the mail circulation as I think it important that this proposal be debated within the formal Fortran standardisation bodies. Comment on HPF 1.0, Jan'93 Chapter 1, General Overview In general I believe the basic thinking behind the design of HPF to be sound and I would recommend that the formal standards bodies consider seriously the incorporation of much of HPF in the next revision of Fortran. However, I have a number of specific criticisms which I would like to put forward for consideration prior to the final adoption of HPF and the implementation of conforming processors. The major general feature I feel most unhappy about is the definition of an official subset. This is a thoroughly bad idea in principle and I am also very unhappy about the precise details of the subset. I will comment on the latter point on a message addressed to Chapter 9. In this note I will address the reasons for my opposition to the definition of any subset. I believe standards with subsets are a waste of time and effort for the committee defining and documenting them, and a serious dis-service to the potential user community of the standard. As a matter of empirical observation standard language subsets have seldom defined a truely useful base language. The Fortran 77 subset was a case in point. The subset language was never a very useful addition to Fortran 66 and Fortran 77 never became a really successful revision until the full language gained widespread implementation. In my opinion, as one involved in the design of Fortran 90 and now an active user of the language, the HPF subset is so seriously deficient that it would be highly unsatisfactory in practice. I would further tender the observation that subset definition usually has more to do with committee politics than with their utility to the end user. The usual effect of subsets is that a subset only implementation is rarely actually produced. Numerous incompatible subset-plus implementations appear, but because they have the right to claim conformance to the standard (subset in small print) the pressure to produce full language support is much reduced. Again the Fortran 77 experience is apposite. The user in this situation is left with the choice of using incompatible and therefore non-portable subset extensions or a linguistically impoverished subset only language. The subset in fact has bought the user nothing more than a delay in the availability of full implementations. Fortran 90 did not define a subset for the above reasons. The ISO committee involved in Fortran 90 design was throughout adamantly opposed to the definition of an official subset, and I believe this to remain the strongly held view of the vast majority of the continuing committee. This has not stopped suppliers coming to market with processors that support a subset of Fortran 90 features and from a users point of view a degree of commonality as useful as any officially defined subset is emerging. This has not however weakened the pressure for full language implementation. There is also ample demonstration that Fortran 90 is not too difficult to implement. One supplier, NAG, released a portable compiler for the whole language a month before the standard was actually published. This compiler was the work largely of one man and starting more or less from scratch it took approximately 18 months to produce the first releasable product. I would therefore recommend that the definition of the official subset be removed from the final document. -- Dr.J.L.Schonfelder Director, Computing Services Dept. University of Liverpool, UK Phone: +44(51)794 3716 FAX : +44(51)794 3759 email: jls@liv.ac.uk