Add graph generation tool

benchmark/gen_graph.py

@@ -0,0 +1,70 @@
+import subprocess
+import re
+import matplotlib.pyplot as plt
+import matplotlib.ticker as ticker
+from tqdm import tqdm
+import pickle
+import os
+
+# define the buffer sizes to test
+bufsizes = list(range(16, 4096, 8))
+
+# regex to extract cycles per byte for all ciphers
+cpb_pattern = re.compile(r"(AES-\d{3}-\w+\d? \w+ \(generic\)|AES-\d{3}-\w+\d? \w+ \(simd\)|Polyval hashing \(generic\)|Polyval hashing \(clmul\))\s+([\d\.]+)")
+
+# dictionary to store cycles per byte results for each cipher
+cpb_results = {}
+
+# Check if results exist
+if os.path.exists('cpb_results.pkl'):
+    # Load the results from disk
+    with open('cpb_results.pkl', 'rb') as f:
+        cpb_results = pickle.load(f)
+else:
+    # Run the benchmark and collect results
+    for bufsize in tqdm(bufsizes):
+        cmd = ["./build/host/cipherbench", "--bufsize={}".format(bufsize),
+               "--ntries=20", "HCTR2", "XTS"]
+        # execute the command and get the output
+        result = subprocess.run(cmd, stdout=subprocess.PIPE)
+        # decode the output and extract cycles per byte
+        output = result.stdout.decode()
+        cpb_matches = cpb_pattern.findall(output)
+
+        for cipher, cpb in cpb_matches:
+            if cipher not in cpb_results:
+                cpb_results[cipher] = []
+            cpb_results[cipher].append(float(cpb))
+
+    # Save the results
+    with open('cpb_results.pkl', 'wb') as f:
+        pickle.dump(cpb_results, f)
+
+included_ciphers = [
+    "AES-256-XTS encryption (simd)",
+    "AES-256-XTS decryption (simd)",
+    "AES-256-HCTR2 encryption (simd)",
+    "AES-256-HCTR2 decryption (simd)",
+]
+
+# plot the data
+for cipher, results in cpb_results.items():
+    if cipher in included_ciphers:
+        plt.plot(bufsizes, results, label=cipher)
+
+plt.title('Cipher Performance vs. Buffer Size')
+plt.xlabel('Buffer Size')
+plt.ylabel('Cycles Per Byte')
+plt.yscale('log')  # make y-axis logarithmic
+plt.grid(True)
+
+# set more granular y-axis labeling
+plt.gca().yaxis.set_minor_locator(ticker.LogLocator(subs='all'))
+plt.gca().yaxis.set_minor_formatter(ticker.FormatStrFormatter("%d"))
+plt.gca().yaxis.set_major_formatter(ticker.FormatStrFormatter("%d"))
+
+# place legend inside plot
+plt.legend(loc='best', fontsize='small')
+
+plt.show()
+