Add script to generate combined plots from Google Benchmark JSON outputs

kratos/python_scripts/benchmark/generate_plot_google_benchmark.py

@@ -0,0 +1,89 @@
+import json
+import pandas as pd
+import matplotlib.pyplot as plt
+import argparse
+
+def main(filenames):
+    """
+    Main function to generate a combined plot from multiple Google Benchmark JSON outputs.
+
+    Args:
+        filenames (list of str): List of paths to the JSON files containing benchmark results.
+
+    The function performs the following steps:
+    1. Loads the JSON files specified in the filenames list.
+    2. Combines the 'benchmarks' section of all JSON data into a single pandas DataFrame.
+    3. Plots the 'cpu_time' and 'real_time' for each benchmark with bars grouped by input filename.
+
+    The resulting plot displays:
+    - X-axis: Benchmark names (with labels based on input filenames)
+    - Y-axis: Time in nanoseconds
+    - Bars grouped by CPU Time and Real Time for each file
+
+    NOTE: The JSON files must be generated using Google Benchmark with:
+    ./benchmark_name --benchmark_format=json --benchmark_out=filename.json
+    """
+    combined_data = []
+
+    # Process each JSON file
+    for filename in filenames:
+        with open(filename) as f:
+            data = json.load(f)
+
+        # Extract benchmark information and add filename as a source
+        benchmarks = data["benchmarks"]
+        for benchmark in benchmarks:
+            benchmark["source"] = filename
+        combined_data.extend(benchmarks)
+
+    # Convert combined data to a pandas DataFrame
+    benchmark_df = pd.DataFrame(combined_data)
+
+    # Plot the results
+    plt.figure(figsize=(12, 8))
+    bar_width = 0.35
+    x_labels = benchmark_df["name"].unique()
+    x = range(len(x_labels))
+    offset = 0
+
+    # Plot for each input file
+    for filename in filenames:
+        subset = benchmark_df[benchmark_df["source"] == filename]
+        plt.bar(
+            [pos + offset for pos in x],
+            subset["real_time"],
+            bar_width,
+            label=f"{filename} - Real Time",
+            alpha=0.7
+        )
+        plt.bar(
+            [pos + offset for pos in x],
+            subset["cpu_time"],
+            bar_width,
+            label=f"{filename} - CPU Time",
+            alpha=0.7
+        )
+        offset += bar_width
+
+    # Customize the plot
+    plt.title("Benchmark Performance Across Files")
+    plt.xlabel("Benchmark Name")
+    plt.ylabel("Time (ns)")
+    plt.xticks([pos + bar_width for pos in x], x_labels, rotation=45, ha="right")
+    plt.legend()
+    plt.tight_layout()
+
+    # Show the plot
+    plt.show()
+
+if __name__ == "__main__":
+    """
+    This script can be run from the command line to generate a combined plot from multiple Google Benchmark JSON outputs.
+    Example usage:
+    python generate_plot_google_benchmark.py --filenames filename1.json filename2.json
+    """
+    # Parse command line arguments
+    parser = argparse.ArgumentParser(description='Generate a combined plot from multiple Google Benchmark JSON results.')
+    parser.add_argument('--filenames', nargs='+', type=str, help='The list of JSON files containing benchmark results')
+    args = parser.parse_args()
+    main(args.filenames)