task1

task1/scripts/run_cuda.slurm

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+#SBATCH --job-name=task1-cuda
+#SBATCH --partition=tornado-k40
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --time=00:20:00
+#SBATCH --output=results/%x-%j.out
+#SBATCH --error=results/%x-%j.err
+
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
+cd "$ROOT_DIR"
+
+mkdir -p results bin
+
+./scripts/build.sh
+
+echo "===== account info ====="
+whoami
+hostname
+date
+
+echo
+echo "===== slurm info ====="
+echo "SLURM_JOB_ID=${SLURM_JOB_ID:-unknown}"
+echo "SLURM_JOB_NAME=${SLURM_JOB_NAME:-unknown}"
+echo "SLURM_JOB_PARTITION=${SLURM_JOB_PARTITION:-unknown}"
+echo "SLURM_JOB_NUM_NODES=${SLURM_JOB_NUM_NODES:-unknown}"
+echo "SLURM_NODELIST=${SLURM_NODELIST:-unknown}"
+echo "CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-unset}"
+scontrol show job "${SLURM_JOB_ID}" || true
+
+echo
+echo "===== node config ====="
+lscpu | sed -n '1,20p'
+if [ -n "${SLURMD_NODENAME:-}" ]; then
+    scontrol show node "${SLURMD_NODENAME}" || true
+fi
+nvidia-smi -L || true
+nvidia-smi || true
+
+echo
+echo "===== benchmark ====="
+./bin/linpack_cuda \
+    --start 1000 \
+    --step 500 \
+    --count 6 \
+    --eps 1e-6 \
+    --max-iters 15000 \
+    --threads 256 \
+    --repeat 3 \
+    --warmup 1 \
+    --csv "results/task1-cuda-${SLURM_JOB_ID}.csv"