Status of CCSM on BG/L. Effort now focuses on BG/P and CCSM4alpha

CCSM3 concurrent

  • CCSM with full models (single executable)
    • build supported on frost and suny
    • run completed T31_gx3v5 compset B on 32 procs with output.
  • CCSM with dead models (single executable)
  • CCSM with data models (single executable)
    • not attempted

      CCSM4 (sequential)

  • with dead models
    • Using 32 procs and all models with identical resolution
      • FV2x2.5: runs
      • T62: decomposition error
      • T85: runs but "end of main loop" message doesn't appear in output
      • FV1x1125: as for T85.
      • T170: as for T85.
  • with full models
    • high res simulation (0.5 atmosphere and 0.1 ocean) runs on BG/L at SUNY but has no output and crashes after a couple of days (sea ice error)

      Component models (as standalone)

  • POP2: Compiles and runs. JohnD has version with some mods for BlueGene.
  • CLM:
  • CSIM:
  • CAM:
    • Current Work
      • Latest CAM compiles and runs on Argonne BlueGene
      • Eulerian T42 ran ok up to 64 proc limit both co/vn
        • Eulerian T85 ran ok up to 128 proc limit co; vn has memory problems
        • FV 1x1.25: CO runs on 32,64,128,256, and 480
        • FV 1x1.25: all VN runs (8-128) fails. Allocation errors.
    • Old Work
      • Sidd Gosh Notes Used 3.3.31. T31 and T42 ran with little mdifications in co and vn mode. 1x1.25 ran fine in co mode.
        • Sidd's and Yu Heng's parallel I/O mods will not be ported. Will wait for PIO instead.
      • pilgrim memory fixed as of 3.3.35.

Results

CCSM with dead models:

cpl_map_mod problems a history of what we had to do with cpl_map_mod

  • No labels