Skip to content

Commit

Permalink
Merge pull request #930 from CliMA/gb/walltime
Browse files Browse the repository at this point in the history
Increase AMIP walltime
  • Loading branch information
Sbozzolo authored Aug 28, 2024
2 parents 3cac40f + 6931f19 commit 0e891dd
Show file tree
Hide file tree
Showing 3 changed files with 15 additions and 13 deletions.
7 changes: 3 additions & 4 deletions .buildkite/amip/pipeline.yml
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
agents:
queue: clima
slurm_time: 76:00:00
slurm_time: 96:00:00
modules: common

env:
Expand All @@ -9,7 +9,7 @@ env:
SLURM_KILL_BAD_EXIT: 1
CONFIG_PATH: "config/amip_configs"

timeout_in_minutes: 4560
timeout_in_minutes: 5760

steps:

Expand Down Expand Up @@ -46,7 +46,7 @@ steps:
- echo "--- Run simulation"
- "srun --cpu-bind=threads --cpus-per-task=4 julia --threads=3 --color=yes --project=experiments/ClimaEarth/ experiments/ClimaEarth/run_amip.jl --config_file $CONFIG_PATH/amip.yml --job_id amip"
artifact_paths: "experiments/ClimaEarth/output/amip/amip_artifacts/*"
timeout_in_minutes: 4320
timeout_in_minutes: 5760
env:
CLIMACOMMS_DEVICE: "CUDA"
agents:
Expand All @@ -70,4 +70,3 @@ steps:
find experiments/ClimaEarth/output/amip/amip_artifacts/ -type f -name 'bias*.png' -print0 | while IFS= read -r -d '' file; do
slack-upload -c "#coupler-report" -f "$$file" -m png -n "$$(basename "$$file" .png)" -x "$$(basename "$$file" .png)"
done
- SYPD="$(cat experiments/ClimaEarth/output/amip/amip_artifacts/sypd.txt)"; WALL="$(cat experiments/ClimaEarth/output/amip/amip_artifacts/walltime_per_atmos_step.txt)"; slack-message -c "#coupler-report" -x "SYPD: $SYPD Walltime per Atmos step: $WALL"
17 changes: 10 additions & 7 deletions experiments/ClimaEarth/run_amip.jl
Original file line number Diff line number Diff line change
Expand Up @@ -808,15 +808,18 @@ walltime_per_atmos_step = es.walltime / n_atmos_steps

## Save the SYPD and allocation information
if ClimaComms.iamroot(comms_ctx)
sypd_filename = joinpath(dir_paths.artifacts, "sypd.txt")
write(sypd_filename, "$sypd")
open(joinpath(dir_paths.artifacts, "sypd.txt"), "w") do sypd_filename
println(sypd_filename, "$sypd")
end

walltime_per_atmos_step_filename = joinpath(dir_paths.artifacts, "walltime_per_atmos_step.txt")
write(walltime_per_atmos_step_filename, "$(walltime_per_atmos_step)")
open(joinpath(dir_paths.artifacts, "walltime_per_atmos_step.txt"), "w") do walltime_per_atmos_step_filename
println(walltime_per_atmos_step_filename, "$(walltime_per_atmos_step)")
end

cpu_max_rss_GB = Utilities.show_memory_usage(comms_ctx)
cpu_max_rss_filename = joinpath(dir_paths.artifacts, "max_rss_cpu.txt")
write(cpu_max_rss_filename, cpu_max_rss_GB)
open(joinpath(dir_paths.artifacts, "max_rss_cpu.txt"), "w") do cpu_max_rss_filename
cpu_max_rss_GB = Utilities.show_memory_usage(comms_ctx)
println(cpu_max_rss_filename, cpu_max_rss_GB)
end
end

#=
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -71,11 +71,11 @@ sypd = CA.simulated_years_per_day(es)
comms_ctx = atmos_config.comms_ctx
if ClimaComms.iamroot(comms_ctx)
open(joinpath(output_dir, "sypd.txt"), "w") do sypd_filename
write(sypd_filename, "$sypd")
println(sypd_filename, "$sypd")
end

open(joinpath(output_dir, "max_rss_cpu.txt"), "w") do cpu_max_rss_filename
cpu_max_rss_GB = Utilities.show_memory_usage(comms_ctx)
write(cpu_max_rss_filename, cpu_max_rss_GB)
println(cpu_max_rss_filename, cpu_max_rss_GB)
end
end

0 comments on commit 0e891dd

Please sign in to comment.