Add extract.py for rocHPL

projects/rocHPL/extract.py

+import argparse
+import re
+
+parser = argparse.ArgumentParser(
+    description="Aggregate extrema for selected columns in HPL tables"
+)
+parser.add_argument(
+    "-i", "--input", required=True, help="Input file path, e.g. log/result.txt"
+)
+parser.add_argument(
+    "-t",
+    "--target",
+    action="append",
+    required=True,
+    help="Column name to aggregate; repeatable, supports regex, e.g. -t 'GFLOPS' -t 'pdfact'",
+)
+parser.add_argument(
+    "--op",
+    choices=("max", "min"),
+    default="max",
+    help="Aggregation: max or min (default: max)",
+)
+args = parser.parse_args()
+
+# Merge patterns into a single regex, using non-capturing group (?:) to ensure safe priority
+pattern = re.compile("|".join(f"(?:{t})" for t in args.target))
+agg_op = max if args.op == "max" else min
+
+tables = []
+cur_cols = None
+cur_best = {}
+
+
+def flush():
+    if cur_best:
+        tables.append(cur_best.copy())
+
+
+with open(args.input) as f:
+    for line in f:
+        if "|" in line:
+            if pattern.search(line):
+                flush()
+                # Split by | and strip whitespace, then filter columns by regex pattern
+                parts = [c.strip() for c in line.split("|")]
+                cur_cols = {
+                    name: i
+                    for i, name in enumerate(parts)
+                    if name and pattern.search(name)
+                }
+                cur_best = {}
+                continue
+
+        if not cur_cols:
+            continue
+
+        # Table end: strictly match {"-", "|"} combination, ignore pure dashes ---
+        stripped = line.strip()
+        if not stripped or set(stripped) == {"-", "|"}:
+            flush()
+            cur_cols = None
+            cur_best = {}
+            continue
+
+        parts = line.split("|")
+        for name, idx in cur_cols.items():
+            try:
+                val = float(parts[idx])
+                # Dictionary update logic
+                cur_best[name] = agg_op(val, cur_best.get(name, val))
+            except (IndexError, ValueError):
+                pass
+
+flush()
+
+if not tables:
+    print("No tables detected.")
+    raise SystemExit(0)
+
+actual_cols = list(tables[0].keys())
+print(f"op: {args.op}")
+
+# Build data rows in one step. str(t.get(name, "")) handles None conversion to empty string.
+headers = ["Table"] + actual_cols
+rows = [
+    [str(i)] + [str(t.get(name, "")) for name in actual_cols]
+    for i, t in enumerate(tables, 1)
+]
+
+# Use zip(*) matrix transpose to calculate column widths in one line
+all_data = [headers] + rows
+widths = [max(len(cell) for cell in col) for col in zip(*all_data)]
+
+
+def fmt(row):
+    return " | ".join(cell.ljust(w) for cell, w in zip(row, widths))
+
+
+print(fmt(headers))
+print("-+-".join("-" * w for w in widths))
+for row in rows:
+    print(fmt(row))