-
Notifications
You must be signed in to change notification settings - Fork 23
Summit
Matteo Turilli edited this page Jan 25, 2019
·
19 revisions
Legenda:
- #CPU/pilot = number of CPU per pilot
- #GPU/pilot = number of GPU per pilot
- #Conc. CU/pilot = number of concurrent compute units per pilot
- #Gen./pilot = number of generations per pilot
- #core/CU = number of cores per compute unit
- #GPU/CU = number of GPUs per compute unit
- ram/CU = amount of ram per compute unit
- CU = Compute Unit
- - = Not yet tested
#CPU/pilot | #GPU/pilot | #Conc. CU/pilot | #Gen./pilot | #core/CU | #GPU/CU | #ram/CU |
---|---|---|---|---|---|---|
4 * 1024 | - | 1024 | 4 | 1 | - | - |
Max | Reason | Ticket | |
---|---|---|---|
#CPU/pilot | - | ||
#GPU/pilot | - | ||
#Conc. CU/pilot | 1024 | # of available processes on the mom node | yes |
#Generations/pilot | - | ||
#Core/CU | - | ||
#GPU/CU | - | ||
#ram/CU | - | ||
error rate | 1% | jsrun | yes |
- HPC Workflows: Seisflow
- Arvind: Hyperspace
A collection of all the documentation relevant to the porting.
- BSUB+jsrun: https://www.olcf.ornl.gov/for-users/system-user-guides/summit/running-jobs/
- jsrun Presentation (David Solt – IBM): slides, recording
- LSF/jsrun examples: https://code.ornl.gov/t4p/Hello_jsrun