Project 36: Hardware Features Sophia-Antipolis 1985: ====================== IEEE Conference on Computer Arithmetic (Hull) --------------------------------------------- The next IEEE Conference on Computer Arithmetic is to be held at Lake Como (near Milan) in 1987. Co-chairpersons are L. Dadda (Milan) and M.J. Irwin (Penn State). At the last conference (in Urbana, June 4-6, 1985) it was decided to broaden the scope for the next conference by encouraging the participation of users, and it was hoped that WG 2.5 might help in organizing the next one. It was also decided to change the format from having only submitted papers to having invited papers as well, and possibly tutorials. The Working Group was enthusiastic about this proposal, and the Group decided to hold its 1987 meeting immediately after the IEEE meeting, at or near the same place, provided satisfactory dates can be arranged. Brian Ford would be host, and someone from the Joint Research Centre at Ispra would probably be in charge of local arrangements for the WG 2.5 meeting. The question of whether WG 2.5 could "co-sponsor" the conference was discussed, Hull agreed to phone Mason, the chairman of TC2, to find out if we could use this term, or if we should only "cooperate with". A subcommittee consisting of Hull, Cody, Kulisch, Paul and Smith considered further details and reported back to the Group. It was suggested that two tutorials might be useful, one on numerical analysis and the implied hardware requirements, and another on the design issues related to numerical needs (including issues related to exception facilities, parallelism, and possibly elementary functions in hardware). It was agreed that Hull should coordinate contributions of the Group. Other persons were suggested as members of the program committee and possible speakers. Argonne 1986: ============= IEEE Conference on Computer Arithmetic (ARITH 8) ------------------------------------------------ Documents: IFIP/WG 2.5 (Argonne-10) 1310, 1 page, IFIP/WG 2.5 (Argonne-14) 1314, 1 page. B. Cody summarized his phone conversation with T. Hull. There will be only a few invited speakers at the ARITH 8 conference. T. Hull suggested that Kahan ought to be one of them. T. Hull should be urged to press this matter further (Reid). The omission of the mention of the IFIP WG 2.5 on the announcements for the conference will be corrected. The Working Group was explicitly invited to submit papers and encourage others to do it (Ford). A show of hands indicated that majority of the WG 2.5 expects to attend both the conference and the WG 2.5 meeting (12 to 3). Members of the Group who intend to submit a paper should contact B. Ford. Como 1987: ========== ARITH 8 Hull reported that at the meeting of the WG 2.5 two years ago in Sophia-Antipolis, it was agreed to work in cooperation with the IEEE group on Computer Arithmetic, to help plan their next Symposium in Como in May 1987, and in particular to promote the inclusion of more coverage of material of common interest to users of computer arithmetic and designers of computer arithmetic. Hull was asked to provide liaison, and to suggest tutorials, invited talks, and more time for discussion after talks. The IEEE head office wouldn't approve tutorials, and invited talks were ruled out because of lack of money. The discussions idea was ignored. However, three WG 2.5 members (Cody, Hull, Kulisch) were put on the programme committee. Steffanelli and Irwin handled almost everything, and did so very efficiently. Both attended a meeting of about half the program committee in Los Angeles in December, 1986. Eighty-six papers had been received and refereed - and 36 were accepted. Proceedings of ARITH 8 may be ordered from IEEE: catalog number 87CH2419-0, ISBN 0-8186-077-2. ARITH 9 is to take place in Santa Monica (Ca, USA) in September 1989, with Earl Schwartzlander and Milos Ercegovac as co-chairmen. The old program committee wants to continue working "in cooperation with WG 2.5" and felt we should try again for invited speakers (such as Cody and/or Moler, and some hardware person(s)). Perhaps schedule 3 days. Hull said he believes that they should push for more discussion time. IFIP/WG 2.5 thanked T. Hull for his efforts and involvement with ARITH 8. IFIP/WG 2.5 approved further cooperation with ARITH 9. J. Reid will write to R.E.A Mason for IFIP approval. T. Hull will continue as liaison with ARITH 9. B. Ford will look into the possibility of a speaker from the Diamond project for ARITH 9. All members of WG 2.5 are asked to think about and send potential speaker names to T. Hull. A general discussion about the ARITH 8 conference developed. It was noted that there was considerable diversity in the topics covered (Stetter), and that it appeared that often the flow of information was unidirectional, lacking in feedback from the users. It was suggested that tutorial talks should be from the customers rather than vendors. It was noted that, in general, there was not enough time for complete discussions of papers, and that the input to users and hardware designers was insufficient. However, on the whole the conference was very interesting and useful and should definitely be supported by WG 2.5. (Gentleman, Delves, Ford, Rice, Reid, Hull, Dekker) Stanford 1988: ============== ARITH 9 ------- T. Hull, reported on the preparations for the ARITH 9 meeting. ARITH 9 will take place in Santa Monica (Ca, USA) in September 1989, with A. Avizienis (UCLA) as general chair, and E. Swartzlander (TRW) and M. Ercegovac (UCLA) as program co-chairmen. The conference will be organized in cooperation with WG 2.5. He said that there will be no invited talks. Brief discussion followed (Feldman. Ford). Jerusalem 1990: =============== ARITH 10 -------- Document: IFIP/WG 2.5 (Jerusalem-12) 1712, 2 pages. Cody has informed the group that he has accepted invitation to join the program committee for ARITH 10. Washington 2004: ================ Math libraries for the AMD64 ---------------------------- The implications for numerical software of the introduction of the 64 bit processor (the AMD64) were identified by John Reid. The implications for the NAG library, including the redesign of critical components and the exhaustive testing and performance modeling were discussed. Tools for making this task less tedious were discussed and some sample performance results were presented.