HPFF Conference Business Meeting Report

Houston, Texas, Jan. 31, 1995 11am
Record of Action: Mary Zosel

Executive Summary

HPFF meetings will continue in 1995/96 with the following charter.
High Performance Fortran Forum Charter 1995/1996

The objectives of the High Performance Fortran Forum in 1995/1996 are:

  1. To produce, by November 1996, a High Performance Fortran version 2.0 including additional features that broaden the applicability of HPF, and whose implementation can exploit successful research experience.
  2. To provide correction, clarification, and interpretation of the current language standard, HPF 1.1.
  3. To encourage high quality implementations.

Possible additional HPF features are listed at the end of this message.
Meeting plans for the first two meetings are:
May 11-12 and July 17-19 (note change!) in Denver, Colorado.

People interested in participation and detailed meeting plans should send email to chk@rice.edu requesting to be added to the email list for meeting participants

Record of Action

Following the contributed presentations at the January HPFF conference, Ken Kennedy led a one hour business meeting to discuss HPFF activities for the coming year.

Mary Zosel announced that a new feature will be added to the HPFF home page - a place to report HPF-related products. There have been a number of requests about how to advertise new HPF libraries, education courses, etc. Anyone with items to put on this page should send an email to hpf@erc.msstate.edu including a descriptive sentence and reference for more information (e.g. a pointer to another web page, or email address). As a reminder, the HPFF home page is here.

Ken led a discussion of activity that HPFF might undertake next. As a starting point, the list of features, presented earlier in talks by Rob Schreiber and Barbara Chapman, provided candidates for addition to the language. (The items are listed at the end of these minutes). David Presburg expressed the opinion that it is still too early to add to the language and more experience is needed. Larry Meadows expressed the opposite opinion, speaking from the vendor perspective: the features proposed were things needed and reasonable to implement. After additional discussion both pro and con, Ken proposed a straw poll:

Starting roughly in May, HPFF should start a one year activity, with the goal of a document completed, including public review, by SC96. We should also continue the CCI activities of HPFF94, possibly producing an intermediate HPF version 1.2 document as well as an HPF version 2 document.

This straw poll passed by a vote of 41 - 3 - 10.

The guidelines for the new HPFF activities are that we consider features for which there was demonstrated user need and successful research - implementation experience.

Meeting dates and location were discussed, selecting dates for the first two meetings, with the remaining dates to be determined later. Denver was proposed as the site for the first two meetings, with remaining meetings to be determined after experience with Denver. A vote about preferred meeting location confirmed this plan: 18 - 8 - 6 (only people likely to attend voting).

For the 1st meeting, May 11/12 is planned - a 1.5 day meeting instead of 2 days, although there may be some work planned for the afternoon of May 10.

For the second meeting, July 19-21 was selected --- BUT --- subsequent email discussion identified that this is in direct conflict with the PPoPP meeting. As of the time of these minutes, exact dates are unconfirmed, with July 17-19 is under consideration as the primary alternative. By the second meeting, we should have the schedule worked out for consideration of features for the HPF2 meetings.

At this time, flashing room lights were turned on and Andy Meltzer presented an LPF update, as is the HPFF tradition - including the following reports:
Reports on experience with LPF and motivating applications for the next version of LPF are much shorter than those for HPF. (There are none.)

LPF User comments were reviewed:

LPF Implementations: (Empty list)
Installed LPF sites - see list of HPF sites.
Performance slides: - (flat line at 0 flop)
[audience question about what size shallow-water test was used]
Performance per pc is right on target, but it scales linearly - that will be worked on in LPF2. Competition with HPF2 may heat up.
Meeting adjourned.

Summary of Features Proposed for HPF2 consideration:

(Disclaimer - this is not intended to prescribe an exact list for HPF2.)