Association for Computing Machinery Gordon Bell Prize
Teams of up to 12 individuals may apply for the award. Nominations will be evaluated on the basis of the following considerations:
- Evidence of important algorithmic and/or implementation innovations
- Clear improvement over the previous state-of-the-art
- Performance is not dependent on an architecture that is specialized or cannot be replicated
- Detailed performance measurements demonstrate the submission’s claims in terms of scalability (strong as well as weak scaling), time to solution, and efficiency in using bottleneck resources (such as memory size or bandwidth, communications bandwidth, I/O), as well as peak performance.
- Achievement is generalizable, in the sense that other scientists can learn and benefit from the innovations
Although solving an important scientific or engineering challenge is important to demonstrate/justify the work, scientific outcomes alone are not sufficient for this prize.
April 15, 2018 - End of Day, Anywhere on Earth (AoE), UTC -12 hr.
Nominations for the Gordon Bell Prize must explain the innovations, detail the performance levels achieved on one or more real-world applications, and articulate the implications of the approach for the broader HPC community. Prepare materials in the format specified below, and submit them using the online nomination form.
- Name, address, phone number, and email address of nominator (person making the submission).
- List of up to 12 names (including the nominator, if appropriate) to be cited as awardees, along with the institutional affiliation for each.
- Suggested citation if the nomination is selected. This is normally the same as the title under which the nomination statement will be published in the SC proceedings.
- Nomination (PDF not exceeding 11 pages in length, following typical technical paper page standards: 11pt font, single spaced text, fitting within 7.5” x 10” text area). Each nomination must provide the information necessary for judging and comparison, but also be suitable for inclusion in the SC conference proceedings, organized into the following sections. Note that the materials should explain the contribution in terms understandable to a non-specialist. Only nominations meeting all requirements will be considered.
- Optional but desirable: Up to 3 endorsement letters in PDF. Letters should be provided by independent individuals or institutions from the discipline of the candidates' real-world application, describing how the candidates' result could potentially produce key scientific advances in that discipline. The nominator should collect the letters and bundle them for submission.
- (150 word max)
- Justification for ACM Gordon Bell Prize
- (50 word max)
- indicate what implementation or performance “high watermark” was achieved (rather than the science that was enabled)
- Performance Attributes
- (use a table listing each attribute title and value in a separate row)
- Category of achievement (1+ of: scalability, time-to-solution, peak performance)
- Type of method used (1 of: explicit, implicit, both explicit and implicit, semi-implicit, n/a)
- Results reported on the basis of (1 of: whole application including I/O; whole application except I/O; kernel only; other [specify])
- Precision reported (1 of: single precision, double precision, mixed precision)
- System scale (1 of: results measured on full-scale system, projected from results of smaller system, other[specify])
- Measurement mechanism (1 of: timers, FLOP count, static analysis tool, performance modeling, other [specify] )
- Overview of the Problem
- description of the problem and its importance, in terms understandable to a non-specialist (1 p max)
- Current State of the Art
- quantitative discussion of current SoA, with emphasis on performance-related aspects (1 p max)
- Innovations Realized
- what the innovations are and how they were achieved (2 pp max)
- How Performance Was Measured
- (Note that preference is given to performance actually measured [not projected], based on the entire application [including I/O] and with uniform precision. Explain in detail if any portion of total runtime was not included in the measurements, if and where different precisions were used, or any attributes listed in Section 3 as “other”).
- what application(s) was used to measure performance (1 p max)
- system and environment where performance was measured (1 p max)
- Performance Results
- include scalability (weak and strong), time to solution, efficiency (of bottleneck resources), and peak performance(2 pp max)
- implications for future systems and applications (1 p max)
- References (1 p max)
The committee will select up to 6 nominations as “finalists” made on the basis of empirical performance measurements submitted (not extrapolated performance predictions). Each nominator should be prepared to reproduce both performance and scientific results.
Finalist submission will be published in the SC proceedings and presented at the SC conference. Prior to publication, finalists will be allowed to correct typographic errors, but no other changes to the text will be permitted prior to publication unless specifically requested by the committee.
Finalists will be given the opportunity to provide updated results in early August, prior to the final award decision.