1 / 5

“Dream Big”: Mid-to-Long Term Issues for SWIM…

“Dream Big”: Mid-to-Long Term Issues for SWIM…. Jeeeeeeeeeembo Kohl (Randy’s Evil More Handsome Twin) Oak Ridge National Laboratory Wednesday, September 27, 2006. What CS Needs from Developers Later… (or vice versa… ). Need for parallel I/O in the fast and slow campaigns

perry
Download Presentation

“Dream Big”: Mid-to-Long Term Issues for SWIM…

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. “Dream Big”:Mid-to-Long Term Issues for SWIM… Jeeeeeeeeeembo Kohl (Randy’s Evil More Handsome Twin) Oak Ridge National Laboratory Wednesday, September 27, 2006 DBB

  2. What CS Needs from Developers Later…(or vice versa… ) • Need for parallel I/O in the fast and slow campaigns • Do we want or need ability to run components concurrently? • For computational components • For utilities like graphics generators or live validations • Transitioning from separate executables for fast MHD to single or smaller number of executables for slow MHD. • CCA technology, or • MCT, or • the FACETS approach, or ....? • Handling multiple parallel codes within single batch queue wait • Well-designed and thorough regression tests. DBB

  3. What CS Needs from Developers Later…(or vice versa… ) • Parallel Coupling Infrastructure (PCI) • “Real” Coupling of Independent Parallel Codes… • Ultimate Goal at the end of our Evolutionary Coupling Approach… • Start with File-based I/O • Evolve to some level of In-Core Memory Swapping/Multiplexing • Direct Parallel Messaging ~ e.g. MPI mayhem… • Full Parallel Coupling ~ e.g. MCT… • As we progress through this evolution: • The ISP will run faster and faster…  • The ISP interfaces will be increasingly generalized…  • Your codes will have to change more and more… :-o • Good News: • The CS guys already have some preliminary PCI technology… • Bad News: • The Developers are already complaining how slow File-based I/O is… DBB

  4. What CS Needs from Developers Later…(or vice versa… ) • May Your Wishes Come True Here: • ??? DBB

  5. ZTABLE.TXT Data file data category string.nc (netcdf) Code string.ps Library An Example of Something Handy for CS Geeks (“Dudes”): AORSA-CQL3D data flow (SWIM) RF specific data Plasma state data RF specific data Plasma state data aorsa2d.in, geqdsk cql3dinput, geqdsk eqdsk_setup grfont.dat eqdsk.out pgplot out_cql3d.coef AORSA (mpi) CQL3D (mpi) FFTpack BLAS scalapack BLACS aorsa2d.cps out_edotj (plots) (plots) (power absorbed) DBB

More Related