1#!/usr/bin/perl
2
3sub parse_objdump_file {
4  my ($filename) = @_;
5  my @result;
6  open (INPUT, $filename) or die "$filename: $!\n";
7  while (<INPUT>) {
8    if (/\s*([0-9a-f]*):\t(([0-9a-f]{2} )+) *\t(.*)$/) {
9      my ($addr, $bytes, $instr) = ($1, $2, $4);
10      $addr = "0x" . $addr;
11      $bytes =~ s/\s*(.*\S)\s*/$1/; # trim any remaining whitespace
12      $instr =~ s/\s*(.*\S)\s*/$1/;
13      push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
14    }
15  }
16  close INPUT;
17  return @result;
18}
19
20sub parse_gdb_file {
21  my ($filename) = @_;
22  my @result;
23  my $got_addr;
24  open (INPUT, $filename) or die "$filename: $!\n";
25  while (<INPUT>) {
26    if (/^(0x[0-9a-f]*):\t([^\t]*)\t[^:]*:\t((0x[0-9a-f]{2}\s*)+)\s*$/) {
27      my ($addr, $bytes, $instr) = ($1, $3, $2);
28      $bytes =~ s/0x//g;
29      $bytes =~ s/\s+/ /g;           # regularize whitespace
30      $bytes =~ s/\s*(.*\S)\s*/$1/;  # trim any remaining whitespace
31      $instr =~ s/\s*(.*\S)\s*/$1/;
32      push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
33    } elsif (/^(0x[0-9a-f]*):\t$/) { # deal with gdb's line breaker
34      $got_addr = $1;
35    } elsif ($got_addr && /^    ([^\t]*)\t[^:]*:\t((0x[0-9a-f]{2}\s*)+)\s*$/) {
36      my ($addr, $bytes, $instr) = ($got_addr, $2, $1);
37      $bytes =~ s/0x//g;
38      $bytes =~ s/\s+/ /g;           # regularize whitespace
39      $bytes =~ s/\s*(.*\S)\s*/$1/;  # trim any remaining whitespace
40      $instr =~ s/\s*(.*\S)\s*/$1/;
41      push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
42      undef $got_addr;
43    }
44  }
45  close INPUT;
46  return @result;
47}
48
49sub binary_diffs {
50  my ($objdump_file, $gdb_file) = @_;
51  my @file1 = parse_objdump_file ($objdump_file);
52  my @file2 = parse_gdb_file ($gdb_file);
53  my $lastrecord = ($#file1 >= $#file2) ? ($#file1) : ($#file2);
54  for (my $i = 0; $i <= $lastrecord; ++$i) {
55    my $d1 = $file1[$i];
56    my $d2 = $file2[$i];
57    if ($d1->{'bytes'} ne $d2->{'bytes'}) {
58      printf "0x%08x:\t%30s \t%s\n", 0+$d1->{'addr'}, $d1->{'bytes'}, $d1->{'instr'};
59      printf "0x%08x:\t%30s \t%s\n\n", 0+$d2->{'addr'}, $d2->{'bytes'}, $d2->{'instr'};
60    }
61  }
62}
63
64$objdump_file = $ARGV[0];
65$gdb_file = $ARGV[1];
66binary_diffs ($objdump_file, $gdb_file);
67exit (0);
68__END__
69=pod
70
71=head1 NAME
72
73codegen-diff
74
75=head1 SYNOPSIS
76
77codegen-diff I<OBJDUMP-OUTPUT-FILE> I<GDB-DISASSEMBLY-FILE>
78
79=head1 DESCRIPTION
80
81B<codegen-diff> is a program that tries to show you the differences
82between the code that B<llc> generated and the code that B<lli> generated.
83
84The way you use it is as follows: first, you create I<OBJDUMP-OUTPUT-FILE>
85by running B<objdump> on the B<llc> compiled and linked binary. You need to
86trim down the result so it contains only the function of interest.
87
88Second, you create I<GDB-DISASSEMBLY-FILE> by running B<gdb>, with my patch
89to print out hex bytes in the B<disassemble> command output, on
90B<lli>.  Set a breakpoint in C<Emitter::finishFunction()> and wait until
91the function you want is compiled.  Then use the B<disassemble> command
92to print out the assembly dump of the function B<lli> just compiled.
93(Use C<lli -debug> to find out where the function starts and ends in memory.)
94It's easiest to save this output by using B<script>.
95
96Finally, you run B<codegen-diff>, as indicated in the Synopsis section of
97this manpage. It will print out a two-line stanza for each mismatched
98instruction, with the  B<llc> version first, and the  B<lli> version second.
99
100=head1 AUTHOR
101
102B<codegen-diff> was written by Brian Gaeke.
103
104=head1 SEE ALSO
105
106L<gdb(1)>, L<objdump(1)>, L<script(1)>.
107
108You will need my B<gdb> patch:
109
110  http://llvm.cs.uiuc.edu/~gaeke/gdb-disassembly-print-bytes.patch
111
112=cut
113