February 3-5 2003 Cray-Fusion UltraScale Computing Workshop
Housing/travel information
Goals of Workshop
Fusion Computing Component
- Characterize models/codes in terms of physics objectives and model.
- Characterize performance extensions (including more cycles) needed
for making significant advances (i.e. the "Big-Science" problem).
- Identify code/code kernels/algorithm for analysis on a range of
platforms (IBMs/clusters/Crays/etc.). Perspective will be optimization
of algorithm/platform/software implementation combination as well
as performance.
- Develop plans for carrying out this task.
Cray Evaluation Component
- The science problems as in "2" above.
- Identify code/code kernels for evaluation on the Cray X1 and develop
a plan for carrying out that evaluation. (Much the same as "3" above.)
Summary of Workshop
Follow-up conference-call
on Feb 14 concerning Cray libraries
Follow-up conference-call
on Feb 28 about performance modeling
Agenda
In many cases there are overlapping physics areas, and some
coordination is needed in order to only present differences in
models/codes. Breakdown of times is left non-specific to allow
for flexibility in discussions.
There will be a dial-in conference call each day during the
workshop (allows up to 12 participants, including the meeting
room.) An email has been sent giving dialing instructions.
|
Monday, February 3
|
| 8:30-10:00 |
Refreshments |
| 10:00-10:15 |
Overview |
Lee Berry |
| Dial-in available 10:00-4:00 |
| 10:00-COB | Topical Discussions |
| | Extended MHD/Reconnection |
| | Overview | Steve Jardin (PPPLJ) |
| | NIMROD | Carl Sovinec (Wisconsin) |
| | M3D | Steve Jardin, Jin Chen (PPPLJ) |
| | CMRS unable to attend | Kai Germaschewski (Uni. Iowa) |
| 1:00-1:15 |
Welcome |
Jeff Nichols |
| | Radio Frequency-Plasma Interactions |
| | Overview | Don Batchelor (ORNL) |
| | TORIC/EMIR | Ed D'Azevedo (ORNL)
|
| | AORSA | Fred Jaeger (ORNL) |
|
Tuesday, February 4 |
|---|
| Dial-in available 9:00-4:00 |
| 9:00-10:30 | Tutorial | Trey White (ORNL) |
| | Atomic Physics |
| | Overview/Models | David Schultz (ORNL) |
| 10:30-2:00 | Topical Presentations Discussions (cont.) |
| | Micro Turbulence |
| | GTC | Stephane Ethier (PPPL) |
| | Lunch | |
| | GYRO | Jeff Candy (GA) |
| 2:00-3:00 | Cray Demo |
Nathan Wichmann (Cray) |
| | astro.pinfo.p | |
| | astro.pinfo.e | |
| | prof.statements | |
| | profline | |
| | prof.callstack | |
| | astro.out | |
| | LINPACK results | Nathan Wichmann (Cray) |
| |
Discussion of tasks for preparing
plans and summaries. There will likely be two or three working
groups with participation by Cray-knowledgeable folk in each.
Whatever is left in the day could be devoted to discussions
by the working groups.
|
|
Wednesday, February 5 |
|---|
| Dial-in available 9:00-12:00 |
| 9:30-9:45 |
Cray Evaluation |
Mark Fahey |
| 10:00-10:30 |
Fusion Ultra-Scale Computing Directions |
Lee Berry |
The presentations should include:
- Brief physics objectives of code.
- Description of physics model (suitable for computer scientists).
- What is the "Big-Science" problem? What is computational scale?
What are the anticipated bottlenecks?
- [The above can be sharply limited in the talks on particular models.
Points not included in the overviews may need to be addressed.]
- Description of numerical implementation, including needed libraries.
- Present limiting/controlling bottlenecks (size of data objects,
number of time steps, I/O bandwidth, etc).
- Information on scaling experiments on various systems that have
been carried out.
Needed Equipment:
- Net access (we would furnish hub hardware)
- LCD projector, screen
- Screen
- Conference Phone (They provide line, we provide phone). This
will be used for much of the M3D discussion, as well as other times
when necessary. To be useful, pdf copies of the presentations should
be provided ahead of time so that we can put them on a website (or
distribute them via an FTP site).
