6
sub parse_objdump_file {
9
open (INPUT, $filename) or die "$filename: $!\n";
10
print "opened objdump output file $filename\n" if $DEBUG;
12
if (/\s*([0-9a-f]*):\t(([0-9a-f]{2} )+) *\t(.*)$/) {
13
my ($addr, $bytes, $instr) = ($1, $2, $4);
15
$bytes =~ s/\s*(.*\S)\s*/$1/; # trim any remaining whitespace
16
$instr =~ s/\s*(.*\S)\s*/$1/;
17
push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
18
print "addr=$addr bytes='$bytes' instr='$instr'\n" if $DEBUG;
29
open (INPUT, $filename) or die "$filename: $!\n";
30
print "opened gdb output file $filename\n" if $DEBUG;
32
if (/^(0x[0-9a-f]*):\t([^\t]*)\t[^:]*:\t((0x[0-9a-f]{2}\s*)+)\s*$/) {
33
my ($addr, $bytes, $instr) = ($1, $3, $2);
35
$bytes =~ s/\s+/ /g; # regularize whitespace
36
$bytes =~ s/\s*(.*\S)\s*/$1/; # trim any remaining whitespace
37
$instr =~ s/\s*(.*\S)\s*/$1/;
38
push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
39
print "addr=$addr bytes='$bytes' instr='$instr'\n" if $DEBUG;
40
} elsif (/^(0x[0-9a-f]*):\t$/) { # deal with gdb's line breaker
42
} elsif ($got_addr && /^ ([^\t]*)\t[^:]*:\t((0x[0-9a-f]{2}\s*)+)\s*$/) {
43
my ($addr, $bytes, $instr) = ($got_addr, $2, $1);
45
$bytes =~ s/\s+/ /g; # regularize whitespace
46
$bytes =~ s/\s*(.*\S)\s*/$1/; # trim any remaining whitespace
47
$instr =~ s/\s*(.*\S)\s*/$1/;
48
push (@result, {'addr' => $addr, 'bytes' => $bytes, 'instr' => $instr});
49
print "addr=$addr bytes='$bytes' instr='$instr'\n" if $DEBUG;
58
my ($objdump_file, $gdb_file) = @_;
59
my @file1 = parse_objdump_file ($objdump_file);
60
my @file2 = parse_gdb_file ($gdb_file);
61
my $lastrecord = ($#file1 >= $#file2) ? ($#file1) : ($#file2);
62
for (my $i = 0; $i <= $lastrecord; ++$i) {
65
if ($d1->{'bytes'} ne $d2->{'bytes'}) {
66
next if (($d1->{'instr'} eq $d2->{'instr'}) && $opt_d);
67
printf "0x%08x:\t%30s \t%s\n", 0+$d1->{'addr'}, $d1->{'bytes'}, $d1->{'instr'};
68
printf "0x%08x:\t%30s \t%s\n\n", 0+$d2->{'addr'}, $d2->{'bytes'}, $d2->{'instr'};
74
$objdump_file = $ARGV[0];
76
binary_diffs ($objdump_file, $gdb_file);
87
codegen-diff [-d] I<OBJDUMP-OUTPUT-FILE> I<GDB-DISASSEMBLY-FILE>
91
B<codegen-diff> is a program that tries to show you the differences
92
between the code that B<llc> generated and the code that B<lli> generated.
94
The way you use it is as follows: first, you create I<OBJDUMP-OUTPUT-FILE>
95
by running B<objdump> on the B<llc> compiled and linked binary. You need to
96
trim down the result so it contains only the function of interest.
98
Second, you create I<GDB-DISASSEMBLY-FILE> by running B<gdb>, with my patch
99
to print out hex bytes in the B<disassemble> command output, on
100
B<lli>. Set a breakpoint in C<Emitter::finishFunction()> and wait until
101
the function you want is compiled. Then use the B<disassemble> command
102
to print out the assembly dump of the function B<lli> just compiled.
103
(Use C<lli -debug> to find out where the function starts and ends in memory.)
104
It's easiest to save this output by using B<script>.
106
Finally, you run B<codegen-diff>, as indicated in the Synopsis section of
107
this manpage. It will print out a two-line stanza for each mismatched
108
instruction, with the B<llc> version first, and the B<lli> version second.
116
Don't show instructions where the bytes are different but they
117
disassemble to the same thing. This puts a lot of trust in the
118
disassembler, but it might help you highlight the more egregious cases
125
B<codegen-diff> was written by Brian Gaeke.
129
L<gdb(1)>, L<objdump(1)>, L<script(1)>.
131
You will need my B<gdb> patch:
133
http://llvm.cs.uiuc.edu/~gaeke/gdb-disassembly-print-bytes.patch