chore: add benchmark analysis scripts

analyze_bench.py

@@ -0,0 +1,70 @@
+import csv
+import sys
+
+def analyze(filename, label):
+    durations = []
+    lengths = []
+    nodes = []
+    failed = 0
+    total = 0
+    stutters = 0
+    try:
+        with open(filename, 'r') as f:
+            reader = csv.reader(f)
+            next(reader)
+            for row in reader:
+                if not row or row[0].startswith('#'):
+                    if 'total_paths' in ','.join(row) if row else '':
+                        continue
+                    continue
+                total += 1
+                try:
+                    d = int(row[1])
+                    durations.append(d)
+                    lengths.append(int(row[2]))
+                    nodes.append(int(row[3]))
+                    if d > 1000: stutters += 1
+                except:
+                    pass
+        
+        if not durations:
+            return
+        durations.sort()
+        lengths.sort()
+        nodes.sort()
+        
+        p50 = durations[len(durations)//2]
+        p90 = durations[int(len(durations)*0.90)]
+        p99 = durations[int(len(durations)*0.99)]
+        max_d = durations[-1]
+        
+        avg_nodes = sum(nodes) / len(nodes) if nodes else 0
+        avg_len = sum(lengths) / len(lengths) if lengths else 0
+        
+        print(f"\n=== {label} ===")
+        print(f"Total paths: {total}")
+        print(f"P50: {p50}µs | P90: {p90}µs | P99: {p99}µs | Max: {max_d}µs")
+        print(f"Stutters (>1ms): {stutters} ({100*stutters/total:.2f}%)")
+        print(f"Avg nodes: {avg_nodes:.1f} | Avg length: {avg_len:.1f}")
+        print(f"Median length: {lengths[len(lengths)//2]}")
+        return durations, lengths, nodes, stutters, total
+    except Exception as e:
+        print(f"Error reading {filename}: {e}")
+        return None
+
+base = analyze("/home/popertots/bench/pathfinding_benchmark_baseline_release.csv", "ORIGINAL BASELINE (>6 months ago, 5x5 map)")
+cur = analyze("pathfinding_benchmark_current.csv", "CURRENT (15x15 map, after all fixes)")
+
+if base and cur:
+    print("\n=== COMPARISON ===")
+    b_dur, b_len, b_nodes, b_stut, b_tot = base
+    c_dur, c_len, c_nodes, c_stut, c_tot = cur
+    
+    p50_imp = (b_dur[len(b_dur)//2] - c_dur[len(c_dur)//2]) / b_dur[len(b_dur)//2] * 100
+    p99_imp = (b_dur[int(len(b_dur)*0.99)] - c_dur[int(len(c_dur)*0.99)]) / b_dur[int(len(b_dur)*0.99)] * 100
+    stutter_imp = (b_stut - c_stut) / b_stut * 100 if b_stut > 0 else 0
+    
+    print(f"P50 improvement: {p50_imp:+.1f}%")
+    print(f"P99 improvement: {p99_imp:+.1f}%")
+    print(f"Stutter reduction: {stutter_imp:+.1f}%")
+    print(f"Map size: 5x5 -> 15x15 (9x larger area)")

check_lengths.py

@@ -0,0 +1,47 @@
+import csv
+
+def check_lengths(filename, label):
+    lengths = []
+    failed = 0
+    total = 0
+    with open(filename, 'r') as f:
+        reader = csv.reader(f)
+        next(reader)
+        for row in reader:
+            if not row or row[0].startswith('#'):
+                continue
+            total += 1
+            try:
+                length = int(row[2])
+                lengths.append(length)
+                if length == 0:
+                    failed += 1
+            except:
+                pass
+    
+    print(f"\n=== {label} ===")
+    print(f"Total paths: {total}")
+    print(f"Failed paths (length=0): {failed} ({100*failed/total if total else 0:.1f}%)")
+    print(f"Paths with length > 0: {total - failed}")
+    
+    if lengths:
+        avg = sum(lengths) / len(lengths)
+        print(f"Avg length (all): {avg:.1f}")
+        nonzero = [l for l in lengths if l > 0]
+        if nonzero:
+            print(f"Avg length (non-zero): {sum(nonzero)/len(nonzero):.1f}")
+            print(f"Max length: {max(lengths)}")
+    
+    # Length buckets
+    print("\nLength distribution:")
+    buckets = [(0, 0), (1, 5), (6, 10), (11, 50), (51, 100), (100, 1000)]
+    for lo, hi in buckets:
+        if lo == hi == 0:
+            count = sum(1 for l in lengths if l == 0)
+        else:
+            count = sum(1 for l in lengths if lo <= l <= hi)
+        pct = 100*count/len(lengths) if lengths else 0
+        print(f"  {lo}-{hi if hi < 1000 else 'inf'} nodes: {count} ({pct:.1f}%)")
+
+check_lengths("/home/popertots/bench/pathfinding_benchmark_baseline_release.csv", "BASELINE")
+check_lengths("pathfinding_benchmark_current.csv", "CURRENT")

check_nodes.py

@@ -0,0 +1,36 @@
+import csv
+
+def check_nodes(filename, label):
+    nodes_list = []
+    durations = []
+    with open(filename, 'r') as f:
+        reader = csv.reader(f)
+        next(reader)
+        for row in reader:
+            if not row or row[0].startswith('#'):
+                continue
+            try:
+                nodes_list.append(int(row[3]))
+                durations.append(int(row[1]))
+            except:
+                pass
+    
+    print(f"\n=== {label} NODES EXPANDED ===")
+    print(f"Total paths: {len(nodes_list)}")
+    
+    # Group by node count
+    node_buckets = [(0, 0), (1, 64), (65, 200), (201, 500), (501, 1000), (1001, 10000)]
+    for lo, hi in node_buckets:
+        count = sum(1 for n in nodes_list if lo <= n <= hi)
+        pct = 100*count/len(nodes_list) if nodes_list else 0
+        print(f"  {lo}-{hi if hi < 10000 else 'inf'} nodes: {count} ({pct:.1f}%)")
+    
+    # Show duration correlation with nodes
+    print("\nDuration by node count:")
+    for lo, hi in [(1, 64), (65, 200), (201, 500), (501, 1000), (1001, 10000)]:
+        subset_durations = [d for n, d in zip(nodes_list, durations) if lo <= n <= hi]
+        if subset_durations:
+            avg = sum(subset_durations) / len(subset_durations)
+            print(f"  {lo}-{hi if hi < 10000 else 'inf'} nodes: avg {avg:.0f}µs, count {len(subset_durations)}")
+
+check_nodes("pathfinding_benchmark_current.csv", "CURRENT")

deep_analysis.py

@@ -0,0 +1,42 @@
+import csv
+
+def analyze_tail(filename):
+    durations = []
+    with open(filename, 'r') as f:
+        reader = csv.reader(f)
+        next(reader)
+        for row in reader:
+            if not row or row[0].startswith('#'):
+                continue
+            try:
+                durations.append(int(row[1]))
+            except:
+                pass
+    
+    durations.sort()
+    n = len(durations)
+    
+    print(f"\n=== TAIL ANALYSIS ({filename}) ===")
+    print(f"Total: {n}")
+    print(f"P50: {durations[n//2]}µs")
+    print(f"P90: {durations[int(n*0.90)]}µs")
+    print(f"P95: {durations[int(n*0.95)]}µs")
+    print(f"P99: {durations[int(n*0.99)]}µs")
+    print(f"P99.5: {durations[int(n*0.995)]}µs")
+    print(f"P99.9: {durations[int(n*0.999)]}µs")
+    print(f"Max: {durations[-1]}µs")
+    
+    # Count stutters
+    for threshold in [500, 1000, 2000, 3000, 5000]:
+        count = sum(1 for d in durations if d > threshold)
+        print(f"Paths > {threshold}µs: {count} ({100*count/n:.2f}%)")
+    
+    # Distribution buckets
+    print("\n=== DISTRIBUTION ===")
+    buckets = [(0, 50), (50, 100), (100, 200), (200, 500), (500, 1000), (1000, 2000), (2000, 5000), (5000, float('inf'))]
+    for lo, hi in buckets:
+        count = sum(1 for d in durations if lo <= d < hi)
+        print(f"{lo}-{hi if hi != float('inf') else '∞'}µs: {count} ({100*count/n:.1f}%)")
+
+analyze_tail("/home/popertots/bench/pathfinding_benchmark_baseline_release.csv")
+analyze_tail("pathfinding_benchmark_current.csv")