|
|
|
In response to the high priority placed on enhancing the performance of SciDAC applications on
Leadership Class systems within the Department of Energys Office of Science (DOE SC), the Performance Engineering
Research Institute (PERI) will be working with selected applications to improve their performance on these systems.
The purpose of this questionnaire is to determine the performance needs and goals of the application projects.
Our assistance will require access to the full source code, and the relevance of that assistance will be enhanced
if we can apply our efforts to the current version.
This online survey allows data to be entered and can be updated at any time, e.g. when more complete information becomes available.
Thank you for your time.
Instructions: This survey should take no more than 30 minutes to complete for a person familiar with your application code base.
For multiple choice questions, please complete each question that best fits your code project.
If no choice is a fit, explain the answer that best describes your code project using the Other field.
If you need further clarification on any question, please email the PERI team.
| Survey Summary Statistics |
| Number of entries: |
30 |
| Number of codes represented: |
29 |
| Number of projects represented: |
16 |
|
1. Project Contact Information
|
|
|
2. Performance Issues
|
a.
What performance is required, or what is your performance goal, and why?
|
b. What is your current performance and how does this depend on the computing platform?
(please provide any information you have available -- e.g., performance measurements, published performance results).
|
c. Do you know how to get from the current performance to the required performance?
|
d. What is the criticality of achieving your performance goals (e.g, needed by the end of SciDAC2, needed yesterday)?
|
e. Do you want our help?
|
f. Can you devote the necessary resources to collaboration with the PERI team?
|
3. Code Characteristics and Structure |
a. Which solution methods do you employ (e.g., adaptive mesh refinement, fast fourier transforms)?
Other solution methods, please specify.
|
b. Which of your solution methods are known performance bottlenecks?
|
c. Select the different computer languages used by your project?
Other languages, please specify.
Comments.
|
d. What is the size of your executable code ( number of lines, excluding libraries )?
|
e. Please provide a breakdown of the code types your project uses by entering percentages for each code type.
|
f. On average how frequently is your code released?
|
g. How do you foresee the future of this code?
|
h. Please select all library dependencies?
Other library dependencies, please specify.
|
i. How do you implement parallelism in your code project?
Other implementations, please specify.
|
j. If you use domain decomposition, please describe how?
|
k. Please list the computer platforms that your code runs on using the table below?
[+]
|
l. Do you use checkpoint restarts?
|
4. Review |
a. Additional Comments.
|
|