Tests: Add CUDA tools tests for accelerated-python, cuda-cpp, and stdpar tutorials

tutorials/accelerated-python/brev/test.bash

@@ -4,25 +4,20 @@ START_TIME=$(date +%s.%N)
 
 nvidia-smi
 
-# Run regular package tests.
-echo "Running regular package tests..."
-pytest /accelerated-computing-hub/tutorials/accelerated-python/test/test_packages.py
-EXIT_CODE_PACKAGES=$?
+# Run tests.
+echo "Running tests..."
+pytest /accelerated-computing-hub/tutorials/accelerated-python/test/ \
+  --ignore=/accelerated-computing-hub/tutorials/accelerated-python/test/test_rapids.py
+EXIT_CODE_TESTS=$?
 
-# Run RAPIDS tests.
+# Run RAPIDS tests separately because they require a different virtual environment.
 echo ""
-echo "Running RAPIDS package tests in virtual environment..."
+echo "Running RAPIDS tests in virtual environment..."
 /opt/venvs/rapids/bin/pytest /accelerated-computing-hub/tutorials/accelerated-python/test/test_rapids.py
 EXIT_CODE_RAPIDS=$?
 
-# Test solution notebooks.
-echo ""
-echo "Running solution notebook tests..."
-pytest /accelerated-computing-hub/tutorials/accelerated-python/test/test_notebooks.py
-EXIT_CODE_NOTEBOOKS=$?
-
 # Overall exit code is non-zero if any test suite failed.
-EXIT_CODE=$((EXIT_CODE_PACKAGES || EXIT_CODE_RAPIDS || EXIT_CODE_NOTEBOOKS))
+EXIT_CODE=$((EXIT_CODE_TESTS || EXIT_CODE_RAPIDS))
 
 END_TIME=$(date +%s.%N)
 ELAPSED=$(awk "BEGIN {print $END_TIME - $START_TIME}")

tutorials/accelerated-python/test/test_cuda_tools.py

@@ -0,0 +1,105 @@
+"""
+Test that CUDA tools (nvcc, nsys, ncu) work correctly.
+
+Compiles a minimal CUDA C++ program using Thrust and verifies that both
+Nsight Systems and Nsight Compute can successfully profile it.
+"""
+
+import pytest
+import shutil
+import subprocess
+from pathlib import Path
+
+CUDA_PROGRAM = r"""
+#include <thrust/device_vector.h>
+#include <thrust/sequence.h>
+#include <thrust/reduce.h>
+#include <iostream>
+#include <cstdlib>
+
+int main() {
+    constexpr int n = 256;
+
+    thrust::device_vector<float> d(n);
+    thrust::sequence(d.begin(), d.end());
+
+    float sum = thrust::reduce(d.begin(), d.end());
+    float expected = n * (n - 1) / 2.0f;
+
+    if (sum != expected) {
+        std::cerr << "Mismatch: got " << sum << ", expected " << expected << std::endl;
+        return EXIT_FAILURE;
+    }
+
+    std::cout << "PASS" << std::endl;
+}
+"""
+
+
+@pytest.fixture(scope="module")
+def cuda_binary(tmp_path_factory):
+    """Compile a minimal CUDA C++ program and return the path to the binary."""
+    tmp_dir = tmp_path_factory.mktemp("nsight_test")
+    src_path = tmp_dir / "test_program.cu"
+    bin_path = tmp_dir / "test_program"
+
+    src_path.write_text(CUDA_PROGRAM)
+
+    nvcc_path = shutil.which("nvcc")
+    print(f"nvcc resolves to: {nvcc_path}")
+
+    result = subprocess.run(
+        ["nvcc", "-o", str(bin_path), str(src_path)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"nvcc compilation failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert bin_path.exists(), "Binary was not created"
+
+    return bin_path
+
+
+def test_cuda_binary_runs(cuda_binary):
+    """Verify the compiled CUDA binary runs successfully."""
+    result = subprocess.run(
+        [str(cuda_binary)],
+        capture_output=True, text=True, timeout=30
+    )
+    assert result.returncode == 0, \
+        f"CUDA binary failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert "PASS" in result.stdout
+
+
+def test_nsys_profile(cuda_binary, tmp_path):
+    """Test that nsys can profile the CUDA binary."""
+    nsys_path = shutil.which("nsys")
+    print(f"nsys resolves to: {nsys_path}")
+
+    report_path = tmp_path / "test_report.nsys-rep"
+
+    result = subprocess.run(
+        ["nsys", "profile",
+         "--force-overwrite=true",
+         "--output", str(report_path),
+         str(cuda_binary)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"nsys profile failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert report_path.exists(), "nsys report file was not created"
+
+
+def test_ncu_profile(cuda_binary):
+    """Test that ncu can profile the CUDA binary."""
+    ncu_path = shutil.which("ncu")
+    print(f"ncu resolves to: {ncu_path}")
+
+    result = subprocess.run(
+        ["ncu",
+         "--target-processes=all",
+         "--set=basic",
+         str(cuda_binary)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"ncu profile failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"

tutorials/accelerated-python/test/test_notebook.ipynb

@@ -0,0 +1,29 @@
+{
+  "cells": [
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from nsightful.notebook import is_interactive_notebook\n",
+        "\n",
+        "assert is_interactive_notebook() == False, \"nsightful interactive notebook check failed\""
+      ]
+    }
+  ],
+  "metadata": {
+    "accelerator": "GPU",
+    "colab": {
+      "gpuType": "T4",
+      "provenance": []
+    },
+    "kernelspec": {
+      "display_name": "Python 3 (ipykernel)",
+      "language": "python",
+      "name": "python3"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 5
+}

tutorials/accelerated-python/test/test_notebooks.py

@@ -11,15 +11,25 @@
 
 # Define the path to the notebooks directory
 NOTEBOOKS_DIR = Path(__file__).resolve().parent.parent / 'notebooks'
+TEST_DIR = Path(__file__).resolve().parent
 
 # Discover all solution notebooks (excluding checkpoint files)
 solution_notebooks = sorted([
     nb for nb in NOTEBOOKS_DIR.rglob('*SOLUTION*.ipynb')
     if '.ipynb_checkpoints' not in str(nb)
 ])
 
+# Discover test notebooks in the test directory
+test_notebooks = sorted([
+    nb for nb in TEST_DIR.glob('*.ipynb')
+    if '.ipynb_checkpoints' not in str(nb)
+])
+
+all_notebooks = solution_notebooks + test_notebooks
+
 # Create test IDs from notebook paths for better test output
-notebook_ids = [nb.relative_to(NOTEBOOKS_DIR).as_posix() for nb in solution_notebooks]
+notebook_ids = [nb.relative_to(NOTEBOOKS_DIR).as_posix() for nb in solution_notebooks] + \
+               [nb.name for nb in test_notebooks]
 
 
 def extract_cell_outputs(nb, cell_times=None):
@@ -66,7 +76,7 @@ def check_gpu_state():
         print(f"  GPU State check failed: {e}")
 
 
-@pytest.mark.parametrize('notebook_path', solution_notebooks, ids=notebook_ids)
+@pytest.mark.parametrize('notebook_path', all_notebooks, ids=notebook_ids)
 def test_solution_notebook_executes(notebook_path):
     """
     Test that a solution notebook executes without errors.

tutorials/cuda-cpp/brev/requirements.txt

@@ -8,3 +8,6 @@ jupyterlab-execute-time
 
 # NVIDIA devtools
 nsightful[notebook] @ git+https://github.com/brycelelbach/nsightful.git
+
+# Testing
+pytest

tutorials/cuda-cpp/brev/test.bash

@@ -1,3 +1,23 @@
 #! /bin/bash
 
+START_TIME=$(date +%s.%N)
+
 nvidia-smi
+
+# Run tests.
+echo "Running tests..."
+pytest /accelerated-computing-hub/tutorials/cuda-cpp/test/
+EXIT_CODE=$?
+
+END_TIME=$(date +%s.%N)
+ELAPSED=$(awk "BEGIN {print $END_TIME - $START_TIME}")
+
+echo ""
+awk -v elapsed="$ELAPSED" 'BEGIN {
+    hours = int(elapsed / 3600)
+    minutes = int((elapsed % 3600) / 60)
+    seconds = elapsed % 60
+    printf "Elapsed time: %dh %dm %.3fs\n", hours, minutes, seconds
+}'
+
+exit $EXIT_CODE

tutorials/cuda-cpp/test/pytest.ini

@@ -0,0 +1,2 @@
+[pytest]
+addopts = -v -s --durations=0 --durations-min=0.0

tutorials/cuda-cpp/test/test_cuda_tools.py

@@ -0,0 +1,105 @@
+"""
+Test that CUDA tools (nvcc, nsys, ncu) work correctly.
+
+Compiles a minimal CUDA C++ program using Thrust and verifies that both
+Nsight Systems and Nsight Compute can successfully profile it.
+"""
+
+import pytest
+import shutil
+import subprocess
+from pathlib import Path
+
+CUDA_PROGRAM = r"""
+#include <thrust/device_vector.h>
+#include <thrust/sequence.h>
+#include <thrust/reduce.h>
+#include <iostream>
+#include <cstdlib>
+
+int main() {
+    constexpr int n = 256;
+
+    thrust::device_vector<float> d(n);
+    thrust::sequence(d.begin(), d.end());
+
+    float sum = thrust::reduce(d.begin(), d.end());
+    float expected = n * (n - 1) / 2.0f;
+
+    if (sum != expected) {
+        std::cerr << "Mismatch: got " << sum << ", expected " << expected << std::endl;
+        return EXIT_FAILURE;
+    }
+
+    std::cout << "PASS" << std::endl;
+}
+"""
+
+
+@pytest.fixture(scope="module")
+def cuda_binary(tmp_path_factory):
+    """Compile a minimal CUDA C++ program and return the path to the binary."""
+    tmp_dir = tmp_path_factory.mktemp("nsight_test")
+    src_path = tmp_dir / "test_program.cu"
+    bin_path = tmp_dir / "test_program"
+
+    src_path.write_text(CUDA_PROGRAM)
+
+    nvcc_path = shutil.which("nvcc")
+    print(f"nvcc resolves to: {nvcc_path}")
+
+    result = subprocess.run(
+        ["nvcc", "-o", str(bin_path), str(src_path)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"nvcc compilation failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert bin_path.exists(), "Binary was not created"
+
+    return bin_path
+
+
+def test_cuda_binary_runs(cuda_binary):
+    """Verify the compiled CUDA binary runs successfully."""
+    result = subprocess.run(
+        [str(cuda_binary)],
+        capture_output=True, text=True, timeout=30
+    )
+    assert result.returncode == 0, \
+        f"CUDA binary failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert "PASS" in result.stdout
+
+
+def test_nsys_profile(cuda_binary, tmp_path):
+    """Test that nsys can profile the CUDA binary."""
+    nsys_path = shutil.which("nsys")
+    print(f"nsys resolves to: {nsys_path}")
+
+    report_path = tmp_path / "test_report.nsys-rep"
+
+    result = subprocess.run(
+        ["nsys", "profile",
+         "--force-overwrite=true",
+         "--output", str(report_path),
+         str(cuda_binary)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"nsys profile failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"
+    assert report_path.exists(), "nsys report file was not created"
+
+
+def test_ncu_profile(cuda_binary):
+    """Test that ncu can profile the CUDA binary."""
+    ncu_path = shutil.which("ncu")
+    print(f"ncu resolves to: {ncu_path}")
+
+    result = subprocess.run(
+        ["ncu",
+         "--target-processes=all",
+         "--set=basic",
+         str(cuda_binary)],
+        capture_output=True, text=True, timeout=120
+    )
+    assert result.returncode == 0, \
+        f"ncu profile failed:\nstdout: {result.stdout}\nstderr: {result.stderr}"

tutorials/stdpar/brev/requirements.txt

@@ -19,3 +19,6 @@ jupyterlab-execute-time
 
 # NVIDIA devtools
 nsightful[notebook] @ git+https://github.com/brycelelbach/nsightful.git
+
+# Testing
+pytest

tutorials/stdpar/brev/test.bash

@@ -1,3 +1,23 @@
 #! /bin/bash
 
+START_TIME=$(date +%s.%N)
+
 nvidia-smi
+
+# Run tests.
+echo "Running tests..."
+pytest /accelerated-computing-hub/tutorials/stdpar/test/
+EXIT_CODE=$?
+
+END_TIME=$(date +%s.%N)
+ELAPSED=$(awk "BEGIN {print $END_TIME - $START_TIME}")
+
+echo ""
+awk -v elapsed="$ELAPSED" 'BEGIN {
+    hours = int(elapsed / 3600)
+    minutes = int((elapsed % 3600) / 60)
+    seconds = elapsed % 60
+    printf "Elapsed time: %dh %dm %.3fs\n", hours, minutes, seconds
+}'
+
+exit $EXIT_CODE

tutorials/stdpar/test/pytest.ini

@@ -0,0 +1,2 @@
+[pytest]
+addopts = -v -s --durations=0 --durations-min=0.0