Peter Boyle 0baaddbe98 Pipeline mode commit on Aurora. 5+ TF/s on 16^3x32 per tile at 384 ...

nodes.
More concurrency/fine grained scheduling is possible.

2025-02-04 19:27:26 +00:00

..

bench1.pbs

Pipeline mode getting better -- 2 nodes @ 10TF/s per node on Aurora

2025-01-29 09:22:21 +00:00

bench2.pbs

Significantly better performance on Aurora without using pipeline mode

2025-01-30 16:36:46 +00:00

bench16.pbs

Pipeline mode commit on Aurora. 5+ TF/s on 16^3x32 per tile at 384

2025-02-04 19:27:26 +00:00

bench256.pbs

More benchmark scripts

2024-02-13 19:40:49 +00:00

bench512.pbs

More benchmark scripts

2024-02-13 19:40:49 +00:00

bench1024.pbs

More blasted shell variables

2024-03-06 00:03:59 +00:00

bench2048.pbs

Benchmark scripts

2024-02-13 19:47:02 +00:00

bench_scaling.pbs

Scaling benchmarks, verbosity and MPICH aware in acceleratorInit()

2024-02-13 19:48:03 +00:00

gpu_tile_compact4.sh

Aurora MPI standalone benchmake and options that work well

2024-02-06 16:28:40 +00:00

gpu_tile.sh

Pipeline mode commit on Aurora. 5+ TF/s on 16^3x32 per tile at 384

2025-02-04 19:27:26 +00:00