Repo PIs and Managers: Make a Request for Cori KNL allocation
May 26, 2017 by Richard Gerber
To: Principal Investigators and Repository Managers:
Charging will begin on July 1, 2017 on Cori’s Intel Xeon Phi ( or “KNL,” which is short for the internal Intel code name “Knight’s Landing”) nodes, at which time the current free usage period will end. Any codes run on Cori KNL after June 30 will be charged against your repository's allocation. In June DOE allocation managers will distribute additional time to repositories for use on KNL nodes for the period July 1, 2017 to January 8, 2018. You do not need to submit a new ERCAP request for time on KNL, but DOE program managers need to know your requests and judge your applications’ ability to run well on these nodes.
Please follow the links below to request a KNL allocation from your allocation manager and give him/her information about your ability to use KNL nodes efficiently. Please note that codes that have not been optimized for KNL will not run well (or maybe not at all) on those nodes. If you are not yet familiar with issues surrounding running on KNL, please refer to the bottom two links below.
The addition of the KNL nodes on Cori greatly increases NERSC’s computational capability and capacity and offers new opportunities for you to conduct transformational science. We hope you can take full advantage of the system. If you have questions about, or need assistance, using the Cori KNL nodes, please contact us at firstname.lastname@example.org.
[Note 1: This email does not apply to ALCC, Director’s Reserve, or Startup projects.]
[Note 2: To those who have already completed this form, thank you!]
[Note 3: Projects in HEP (High Energy Physics) should have already filled out a separate form as requested earlier by L. Chatterjee. HEP projects do not need to fill out the form linked below.]
KNL allocation request and info form (submit one per repo)
For your information:
Repo usage report during Cori KNL free usage period
Codes known to run on Cori KNL nodes
Cori Intel Xeon Phi Nodes
Application Porting and Performance