enable GPU tests

.buildkite/JuliaProject.toml

@@ -0,0 +1,14 @@
+[extras]
+HDF5_jll = "0234f1f7-429e-5d53-9886-15a909be8d59"
+MPIPreferences = "3da0fdf6-3ccc-4f1b-acd9-58baa6c99267"
+
+[preferences.HDF5_jll]
+libhdf5_path = "libhdf5"
+libhdf5_hl_path = "libhdf5_hl"
+
+[preferences.MPIPreferences]
+_format = "1.0"
+abi = "OpenMPI"
+binary = "system"
+libmpi = "libmpi"
+mpiexec = "mpiexec"

.buildkite/pipeline.yml

@@ -1,19 +1,21 @@
+
+agents:
+  queue: central
+  slurm_mem: 8G
+  modules: julia/1.8.5 openmpi/4.1.1 cuda/11.3 hdf5/1.12.1-ompi411
+
 env:
-  JULIA_VERSION: "1.8.3"
-  CUDA_VERSION: "11.2"
-  OPENMPI_VERSION: "4.1.1"
+  JULIA_LOAD_PATH: "${JULIA_LOAD_PATH}:${BUILDKITE_BUILD_CHECKOUT_PATH}/.buildkite"
+  JULIA_CUDA_USE_BINARYBUILDER: false
   OPENBLAS_NUM_THREADS: 1
-  CLIMATEMACHINE_SETTINGS_FIX_RNG_SEED: "true"
 
 steps:
   - label: "init cpu env"
     key: "init_cpu_env"
     command:
-      - echo "--- Configure MPI"
-      - julia -e 'using Pkg; Pkg.add("MPIPreferences"); using MPIPreferences; use_system_binary()'
-
       - echo "--- Instantiate project"
       - "julia --project -e 'using Pkg; Pkg.instantiate(;verbose=true); Pkg.precompile(;strict=true)'"
+      - "julia --project -e 'using CUDA; CUDA.precompile_runtime()'"
 
       - echo "--- Instantiate test"
       - "julia --project=test -e 'using Pkg; Pkg.develop(path=\".\"); Pkg.instantiate(;verbose=true); Pkg.precompile(;strict=true)'"
@@ -23,85 +25,38 @@ steps:
 
       - echo "--- Package status"
       - "julia --project -e 'using Pkg; Pkg.status()'"
-
     agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1
-
-  # - label: "init gpu env"
-  #   key: "init_gpu_env"
-  #   command:
-  #     - echo "--- Configure MPI"
-  #     - julia -e 'using Pkg; Pkg.add("MPIPreferences"); using MPIPreferences; use_system_binary()'
-
-  #     - echo "--- Instantiate project"
-  #     - "julia --project -e 'using Pkg; Pkg.instantiate(;verbose=true); Pkg.precompile(;strict=true)'"
-
-  #     - echo "--- Instantiate test"
-  #     - "julia --project=test -e 'using Pkg; Pkg.develop(path=\".\"); Pkg.instantiate(;verbose=true); Pkg.precompile()'"
-
-  #     - echo "--- Initialize CUDA runtime"
-  #     - "julia --project -e 'using CUDA; CUDA.precompile_runtime(); CUDA.versioninfo()'"
-
-  #     - echo "--- Package status"
-  #     - "julia --project -e 'using Pkg; Pkg.status()'"
-  #   agents:
-  #     config: gpu
-  #     queue: central
-  #     slurm_ntasks: 1
-  #     slurm_gres: "gpu:1"
+      slurm_cpus_per_task: 8
+    env:
+      JULIA_NUM_PRECOMPILE_TASKS: 8
 
   - wait
 
   - label: "CPU tests"
     command:
       - "julia --project=test --check-bounds=yes test/runtests.jl"
     artifact_paths: "output/*"
-    agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1
 
-  # - label: "GPU tests"
-  #   command:
-  #     - "julia --project=test --check-bounds=yes test/runtests.jl CuArray"
-  #   artifact_paths: "output/*"
-  #   agents:
-  #     config: gpu
-  #     queue: central
-  #     slurm_ntasks: 1
-  #     slurm_gres: "gpu:1"
+  - label: "GPU tests"
+    command:
+      - "julia --project=test --check-bounds=yes test/runtests.jl CuArray"
+    artifact_paths: "output/*"
+    agents:
+      slurm_gpus: 1
 
   - label: "Flame graph (1D diffusion)"
     command:
       - "julia --project=perf perf/flame.jl --job_id diffusion_1D"
     artifact_paths: "diffusion_1D/*"
-    agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1
 
   - label: "Jet graph (ode fun)"
     command:
       - "julia --project=perf perf/jet.jl --problem ode_fun"
-    agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1
 
   - label: "Jet (forward euler)"
     command:
       - "julia --project=perf perf/jet.jl --problem fe"
-    agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1
 
   - label: "Benchmark"
     command:
       - "julia --project=perf perf/benchmark.jl"
-    agents:
-      config: cpu
-      queue: central
-      slurm_ntasks: 1