diff --git a/samples/genbenchmarkscripts/CMakeLists.txt b/samples/genbenchmarkscripts/CMakeLists.txt
new file mode 100644
index 0000000..6115fde
--- /dev/null
+++ b/samples/genbenchmarkscripts/CMakeLists.txt
@@ -0,0 +1,8 @@
+add_executable(genbenchmarkscripts main.c)
+
+add_custom_command(
+  TARGET genbenchmarkscripts POST_BUILD
+  COMMAND genbenchmarkscripts
+  WORKING_DIRECTORY ${PROJECT_BINARY_DIR}
+  COMMENT "Generating benchmark scripts"
+)
diff --git a/samples/genbenchmarkscripts/main.c b/samples/genbenchmarkscripts/main.c
new file mode 100644
index 0000000..f9d5506
--- /dev/null
+++ b/samples/genbenchmarkscripts/main.c
@@ -0,0 +1,46 @@
+#include <assert.h>
+#include <math.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+int
+main(void)
+{
+    const int max_nprocs = 128;
+    for (int nprocs = 1; nprocs <= max_nprocs; nprocs *= 2) {
+        char filename[4096];
+        sprintf(filename, "benchmark_%d.sh", nprocs);
+
+        FILE* fp = fopen(filename, "w");
+        assert(fp);
+
+        // Boilerplate
+        fprintf(fp, "#!/bin/bash\n");
+        fprintf(fp, "#BATCH --job-name=astaroth\n");
+        fprintf(fp, "#SBATCH --account=project_2000403\n");
+        fprintf(fp, "#SBATCH --time=00:14:59\n");
+        fprintf(fp, "#SBATCH --mem=24000\n");
+        fprintf(fp, "#SBATCH --partition=gputest\n");
+
+        // nprocs, nodes, gpus
+        const int max_gpus_per_node = 4;
+        const int gpus_per_node     = nprocs < max_gpus_per_node ? nprocs : max_gpus_per_node;
+        const int nodes             = (int)ceil((double)nprocs / max_gpus_per_node);
+        fprintf(fp, "#SBATCH --gres=gpu:v100:%d\n", gpus_per_node);
+        fprintf(fp, "#SBATCH -n %d\n", nprocs);
+        fprintf(fp, "#SBATCH -N %d\n", nodes);
+
+        // Modules
+        fprintf(fp, "module load gcc/8.3.0 cuda/10.1.168 cmake hpcx-mpi/2.5.0-cuda nccl\n");
+        fprintf(fp, "export UCX_MEMTYPE_CACHE=n\n");
+
+        // Profile and run
+        fprintf(fp, "mkdir -p profile_%d\n", nprocs);
+        fprintf(fp, "srun nvprof --annotate-mpi openmpi -o profile_%d/%%p.nvprof ./benchmark\n",
+                nprocs);
+
+        fclose(fp);
+    }
+
+    return EXIT_SUCCESS;
+}
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 152f811..e1d25a1 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -1,11 +1,14 @@
+find_package(CUDAToolkit)
+
 ## Astaroth Core
 add_library(astaroth_core STATIC device.cc node.cc astaroth.cc)
-target_link_libraries(astaroth_core astaroth_utils astaroth_kernels cudart)
+target_link_libraries(astaroth_core astaroth_utils astaroth_kernels CUDA::cudart_static)
 
 ## Options
 if (MPI_ENABLED)
     find_package(MPI)
-    target_link_libraries(astaroth_core MPI::MPI_CXX)
+    find_package(OpenMP)
+    target_link_libraries(astaroth_core MPI::MPI_CXX OpenMP::OpenMP_CXX)
 endif()
 
 if (MULTIGPU_ENABLED)