1#!/usr/bin/env python
2# Copyright 2019 the V8 project authors. All rights reserved.
3# Use of this source code is governed by a BSD-style license that can be
4# found in the LICENSE file.
5
6"""
7python %prog trace-file
8
9Parses output generated by v8 with flag --trace-regexp-bytecodes and generates
10a list of the most common sequences.
11"""
12
13from __future__ import print_function
14
15import sys
16import re
17import collections
18
19def parse(file, seqlen):
20  # example:
21  # pc = 00, sp = 0, curpos = 0, curchar = 0000000a ..., bc = PUSH_BT, 02, 00, 00, 00, e8, 00, 00, 00 .......
22  rx = re.compile(r'pc = (?P<pc>[0-9a-f]+), sp = (?P<sp>\d+), '
23                  r'curpos = (?P<curpos>\d+), curchar = (?P<char_hex>[0-9a-f]+) '
24                  r'(:?\.|\()(?P<char>\.|\w)(:?\.|\)), bc = (?P<bc>\w+), .*')
25  total = 0
26  bc_cnt = [None] * seqlen
27  for i in range(seqlen):
28    bc_cnt[i] = {}
29  last = [None] * seqlen
30  with open(file) as f:
31    l = f.readline()
32    while l:
33      l = l.strip()
34      if l.startswith("Start bytecode interpreter"):
35        for i in range(seqlen):
36          last[i] = collections.deque(maxlen=i+1)
37
38      match = rx.search(l)
39      if match:
40        total += 1
41        bc = match.group('bc')
42        for i in range(seqlen):
43          last[i].append(bc)
44          key = ' --> '.join(last[i])
45          bc_cnt[i][key] = bc_cnt[i].get(key,0) + 1
46
47      l = f.readline()
48  return bc_cnt, total
49
50def print_most_common(d, seqlen, total):
51  sorted_d = sorted(d.items(), key=lambda kv: kv[1], reverse=True)
52  for (k,v) in sorted_d:
53    if v*100/total < 1.0:
54      return
55    print("{}: {} ({} %)".format(k,v,(v*100/total)))
56
57def main(argv):
58  max_seq = 7
59  bc_cnt, total = parse(argv[1],max_seq)
60  for i in range(max_seq):
61    print()
62    print("Most common of length {}".format(i+1))
63    print()
64    print_most_common(bc_cnt[i], i, total)
65
66if __name__ == '__main__':
67  main(sys.argv)
68