format

Author: usx95

clang/test/Analysis/lifetime_safety/benchmark.py

@@ -9,9 +9,32 @@
 from scipy.optimize import curve_fit
 from scipy.stats import t
 
+
 def generate_cpp_cycle_test(n: int) -> str:
     """
     Generates a C++ code snippet with a specified number of pointers in a cycle.
+    Example:
+        struct MyObj { int id; ~MyObj() {} };
+
+        void long_cycle_4(bool condition) {
+            MyObj v1{1};
+            MyObj v2{1};
+            MyObj v3{1};
+            MyObj v4{1};
+
+            MyObj* p1 = &v1;
+            MyObj* p2 = &v2;
+            MyObj* p3 = &v3;
+            MyObj* p4 = &v4;
+
+            while (condition) {
+                MyObj* temp = p1;
+                p1 = p2;
+                p2 = p3;
+                p3 = p4;
+                p4 = temp;
+            }
+        }
     """
     if n <= 0:
         return "// Number of variables must be positive."
@@ -34,9 +57,22 @@ def generate_cpp_cycle_test(n: int) -> str:
     cpp_code += f"\nint main() {{ long_cycle_{n}(false); return 0; }}\n"
     return cpp_code
 
+
 def generate_cpp_merge_test(n: int) -> str:
     """
     Generates a C++ code snippet with N independent conditional assignments.
+    Example:
+        struct MyObj { int id; ~MyObj() {} };
+
+        void conditional_merges_4(bool condition) {
+            MyObj v1, v2, v3, v4;
+            MyObj *p1 = nullptr, *p2 = nullptr, *p3 = nullptr, *p4 = nullptr;
+
+            if(condition) { p1 = &v1; }
+            if(condition) { p2 = &v2; }
+            if(condition) { p3 = &v3; }
+            if(condition) { p4 = &v4; }
+        }
     """
     if n <= 0:
         return "// Number of variables must be positive."
@@ -55,6 +91,7 @@ def generate_cpp_merge_test(n: int) -> str:
     cpp_code += f"\nint main() {{ conditional_merges_{n}(false); return 0; }}\n"
     return cpp_code
 
+
 def analyze_trace_file(trace_path: str) -> tuple[float, float]:
     """
     Parses the -ftime-trace JSON output to find durations.
@@ -65,29 +102,32 @@ def analyze_trace_file(trace_path: str) -> tuple[float, float]:
     lifetime_duration = 0.0
     total_duration = 0.0
     try:
-        with open(trace_path, 'r') as f:
+        with open(trace_path, "r") as f:
             trace_data = json.load(f)
-            for event in trace_data.get('traceEvents', []):
-                if event.get('name') == 'LifetimeAnalysis':
-                    lifetime_duration += float(event.get('dur', 0))
-                if event.get('name') == 'ExecuteCompiler':
-                    total_duration += float(event.get('dur', 0))
+            for event in trace_data.get("traceEvents", []):
+                if event.get("name") == "LifetimeSafetyAnalysis":
+                    lifetime_duration += float(event.get("dur", 0))
+                if event.get("name") == "ExecuteCompiler":
+                    total_duration += float(event.get("dur", 0))
 
     except (IOError, json.JSONDecodeError) as e:
         print(f"Error reading or parsing trace file {trace_path}: {e}", file=sys.stderr)
         return 0.0, 0.0
     return lifetime_duration, total_duration
 
+
 def power_law(n, c, k):
     """Represents the power law function: y = c * n^k"""
     return c * np.power(n, k)
 
+
 def human_readable_time(ms: float) -> str:
     """Converts milliseconds to a human-readable string (ms or s)."""
     if ms >= 1000:
         return f"{ms / 1000:.2f} s"
     return f"{ms:.2f} ms"
 
+
 def generate_markdown_report(results: dict) -> str:
     """Generates a Markdown-formatted report from the benchmark results."""
     report = []
@@ -97,7 +137,7 @@ def generate_markdown_report(results: dict) -> str:
     report.append("\n---\n")
 
     for test_type, data in results.items():
-        title = 'Pointer Cycle in Loop' if test_type == 'cycle' else 'CFG Merges'
+        title = "Pointer Cycle in Loop" if test_type == "cycle" else "CFG Merges"
         report.append(f"## Test Case: {title}")
         report.append("")
 
@@ -106,9 +146,9 @@ def generate_markdown_report(results: dict) -> str:
         report.append("|:----|--------------:|-----------------:|")
 
         # Table rows
-        n_data = np.array(data['n'])
-        analysis_data = np.array(data['lifetime_ms'])
-        total_data = np.array(data['total_ms'])
+        n_data = np.array(data["n"])
+        analysis_data = np.array(data["lifetime_ms"])
+        total_data = np.array(data["total_ms"])
         for i in range(len(n_data)):
             analysis_str = human_readable_time(analysis_data[i])
             total_str = human_readable_time(total_data[i])
@@ -119,28 +159,36 @@ def generate_markdown_report(results: dict) -> str:
         # Complexity analysis
         report.append(f"**Complexity Analysis:**")
         try:
-            popt, pcov = curve_fit(power_law, n_data, analysis_data, p0=[0, 2], maxfev=5000)
+            popt, pcov = curve_fit(
+                power_law, n_data, analysis_data, p0=[0, 2], maxfev=5000
+            )
             _, k = popt
-            
+
             # R-squared calculation
             residuals = analysis_data - power_law(n_data, *popt)
             ss_res = np.sum(residuals**2)
-            ss_tot = np.sum((analysis_data - np.mean(analysis_data))**2)
+            ss_tot = np.sum((analysis_data - np.mean(analysis_data)) ** 2)
             r_squared = 1 - (ss_res / ss_tot)
-            
+
             # Confidence Interval for k
             alpha = 0.05  # 95% confidence
-            dof = max(0, len(n_data) - len(popt)) # degrees of freedom
-            t_val = t.ppf(1.0 - alpha / 2., dof)
+            dof = max(0, len(n_data) - len(popt))  # degrees of freedom
+            t_val = t.ppf(1.0 - alpha / 2.0, dof)
             # Standard error of the parameters
             perr = np.sqrt(np.diag(pcov))
             k_stderr = perr[1]
             k_ci_lower = k - t_val * k_stderr
             k_ci_upper = k + t_val * k_stderr
 
-            report.append(f"- The performance of the analysis for this case scales approximately as **O(n<sup>{k:.2f}</sup>)**.")
-            report.append(f"- **Goodness of Fit (R²):** `{r_squared:.4f}` (closer to 1.0 is better).")
-            report.append(f"- **95% Confidence Interval for exponent 'k':** `[{k_ci_lower:.2f}, {k_ci_upper:.2f}]`.")
+            report.append(
+                f"- The performance of the analysis for this case scales approximately as **O(n<sup>{k:.2f}</sup>)**."
+            )
+            report.append(
+                f"- **Goodness of Fit (R<sup>2</sup>):** `{r_squared:.4f}` (closer to 1.0 is better)."
+            )
+            report.append(
+                f"- **95% Confidence Interval for exponent 'k':** `[{k_ci_lower:.2f}, {k_ci_upper:.2f}]`."
+            )
 
         except RuntimeError:
             report.append("- Could not determine a best-fit curve for the data.")
@@ -149,67 +197,82 @@ def generate_markdown_report(results: dict) -> str:
 
     return "\n".join(report)
 
+
 def run_single_test(clang_binary: str, test_type: str, n: int) -> tuple[float, float]:
     """Generates, compiles, and benchmarks a single test case."""
     print(f"--- Running Test: {test_type.capitalize()} with N={n} ---")
-    
+
     generated_code = ""
-    if test_type == 'cycle':
+    if test_type == "cycle":
         generated_code = generate_cpp_cycle_test(n)
-    else: # merge
+    else:  # merge
         generated_code = generate_cpp_merge_test(n)
 
-    with tempfile.NamedTemporaryFile(mode='w+', suffix='.cpp', delete=False) as tmp_cpp:
-        tmp_cpp.write(generated_code)
-        source_file = tmp_cpp.name
-    
-    trace_file = os.path.splitext(source_file)[0] + '.json'
-
-    clang_command = [
-        clang_binary, '-c', '-o', '/dev/null', '-ftime-trace=' + trace_file,
-        '-Wexperimental-lifetime-safety', '-std=c++17', source_file
-    ]
-
-    result = subprocess.run(clang_command, capture_output=True, text=True)
-    
-    if result.returncode != 0:
-        print(f"Compilation failed for N={n}!", file=sys.stderr)
-        print(result.stderr, file=sys.stderr)
-        os.remove(source_file)
-        return 0.0, 0.0
-    
-    lifetime_us, total_us = analyze_trace_file(trace_file)
-    os.remove(source_file)
-    os.remove(trace_file)
-    
-    return lifetime_us / 1000.0, total_us / 1000.0
+    # Use a temporary directory to manage the source and trace files.
+    # The directory and its contents will be cleaned up automatically on exit.
+    with tempfile.TemporaryDirectory() as tmpdir:
+        base_name = f"test_{test_type}_{n}"
+        source_file = os.path.join(tmpdir, f"{base_name}.cpp")
+        trace_file = os.path.join(tmpdir, f"{base_name}.json")
+
+        with open(source_file, "w") as f:
+            f.write(generated_code)
+
+        clang_command = [
+            clang_binary,
+            "-c",
+            "-o",
+            "/dev/null",
+            "-ftime-trace=" + trace_file,
+            "-Wexperimental-lifetime-safety",
+            "-std=c++17",
+            source_file,
+        ]
+
+        result = subprocess.run(clang_command, capture_output=True, text=True)
+
+        if result.returncode != 0:
+            print(f"Compilation failed for N={n}!", file=sys.stderr)
+            print(result.stderr, file=sys.stderr)
+            # No need for manual cleanup, the 'with' statement handles it.
+            return 0.0, 0.0
+
+        lifetime_us, total_us = analyze_trace_file(trace_file)
+
+        return lifetime_us / 1000.0, total_us / 1000.0
+
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="Generate, compile, and benchmark C++ test cases for Clang's lifetime analysis.")
-    parser.add_argument("--clang-binary", type=str, required=True, help="Path to the Clang executable.")
-    
+    parser = argparse.ArgumentParser(
+        description="Generate, compile, and benchmark C++ test cases for Clang's lifetime analysis."
+    )
+    parser.add_argument(
+        "--clang-binary", type=str, required=True, help="Path to the Clang executable."
+    )
+
     args = parser.parse_args()
 
     n_values = [10, 25, 50, 75, 100, 150, 200]
     results = {
-        'cycle': {'n': [], 'lifetime_ms': [], 'total_ms': []},
-        'merge': {'n': [], 'lifetime_ms': [], 'total_ms': []}
+        "cycle": {"n": [], "lifetime_ms": [], "total_ms": []},
+        "merge": {"n": [], "lifetime_ms": [], "total_ms": []},
     }
 
     print("Running performance benchmarks...")
-    for test_type in ['cycle', 'merge']:
+    for test_type in ["cycle", "merge"]:
         for n in n_values:
             lifetime_ms, total_ms = run_single_test(args.clang_binary, test_type, n)
             if total_ms > 0:
-                results[test_type]['n'].append(n)
-                results[test_type]['lifetime_ms'].append(lifetime_ms)
-                results[test_type]['total_ms'].append(total_ms)
-                print(f"    Total: {human_readable_time(total_ms)} | Analysis: {human_readable_time(lifetime_ms)}")
-    
-    print("\n\n" + "="*80)
+                results[test_type]["n"].append(n)
+                results[test_type]["lifetime_ms"].append(lifetime_ms)
+                results[test_type]["total_ms"].append(total_ms)
+                print(
+                    f"    Total: {human_readable_time(total_ms)} | Analysis: {human_readable_time(lifetime_ms)}"
+                )
+
+    print("\n\n" + "=" * 80)
     print("Generating Markdown Report...")
-    print("="*80 + "\n")
-    
+    print("=" * 80 + "\n")
+
     markdown_report = generate_markdown_report(results)
     print(markdown_report)
-