Update bug status.
[valgrind.git] / callgrind / callgrind_annotate.in
blob16c0edf97c547451e74405f86ee37efbc8bb5b89
1 #! /usr/bin/perl -w
2 ##--------------------------------------------------------------------##
3 ##--- The cache simulation framework: instrumentation, recording ---##
4 ##--- and results printing. ---##
5 ##--- callgrind_annotate ---##
6 ##--------------------------------------------------------------------##
8 # This file is part of Callgrind, a cache-simulator and call graph
9 # tracer built on Valgrind.
11 # Copyright (C) 2003-2017 Josef Weidendorfer
12 # Josef.Weidendorfer@gmx.de
14 # This file is based heavily on cg_annotate, part of Valgrind.
15 # Copyright (C) 2002-2017 Nicholas Nethercote
16 # njn@valgrind.org
18 # This program is free software; you can redistribute it and/or
19 # modify it under the terms of the GNU General Public License as
20 # published by the Free Software Foundation; either version 2 of the
21 # License, or (at your option) any later version.
23 # This program is distributed in the hope that it will be useful, but
24 # WITHOUT ANY WARRANTY; without even the implied warranty of
25 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
26 # General Public License for more details.
28 # You should have received a copy of the GNU General Public License
29 # along with this program; if not, see <http://www.gnu.org/licenses/>.
31 # The GNU General Public License is contained in the file COPYING.
33 #----------------------------------------------------------------------------
34 # Annotator for cachegrind/callgrind.
36 # File format is described in /docs/techdocs.html.
38 # Performance improvements record, using cachegrind.out for cacheprof, doing no
39 # source annotation (irrelevant ones removed):
40 # user time
41 # 1. turned off warnings in add_hash_a_to_b() 3.81 --> 3.48s
42 # [now add_array_a_to_b()]
43 # 6. make line_to_CC() return a ref instead of a hash 3.01 --> 2.77s
45 #10. changed file format to avoid file/fn name repetition 2.40s
46 # (not sure why higher; maybe due to new '.' entries?)
47 #11. changed file format to drop unnecessary end-line "."s 2.36s
48 # (shrunk file by about 37%)
49 #12. switched from hash CCs to array CCs 1.61s
50 #13. only adding b[i] to a[i] if b[i] defined (was doing it if
51 # either a[i] or b[i] was defined, but if b[i] was undefined
52 # it just added 0) 1.48s
53 #14. Stopped converting "." entries to undef and then back 1.16s
54 #15. Using foreach $i (x..y) instead of for ($i = 0...) in
55 # add_array_a_to_b() 1.11s
57 # Auto-annotating primes:
58 #16. Finding count lengths by int((length-1)/3), not by
59 # commifying (halves the number of commify calls) 1.68s --> 1.47s
61 use strict;
63 #----------------------------------------------------------------------------
64 # Overview: the running example in the comments is for:
65 # - events = A,B,C,D
66 # - --show=C,A,D
67 # - --sort=D,C
68 #----------------------------------------------------------------------------
70 #----------------------------------------------------------------------------
71 # Global variables, main data structures
72 #----------------------------------------------------------------------------
73 # CCs are arrays, the counts corresponding to @events, with 'undef'
74 # representing '.'. This makes things fast (faster than using hashes for CCs)
75 # but we have to use @sort_order and @show_order below to handle the --sort and
76 # --show options, which is a bit tricky.
77 #----------------------------------------------------------------------------
79 # Total counts for summary (an array reference).
80 my $summary_CC;
81 my $totals_CC;
82 my $summary_calculated = 0;
84 # Totals for each function, for overall summary.
85 # hash(filename:fn_name => CC array)
86 my %fn_totals;
88 # Individual CCs, organised by filename and line_num for easy annotation.
89 # hash(filename => hash(line_num => CC array))
90 my %all_ind_CCs;
92 # Files chosen for annotation on the command line.
93 # key = basename (trimmed of any directory), value = full filename
94 my %user_ann_files;
96 # Generic description string.
97 my $desc = "";
99 # Command line of profiled program.
100 my $cmd = "";
102 # Info on the profiled process.
103 my $creator = "";
104 my $pid = "";
105 my $part = "";
106 my $thread = "";
108 # Positions used for cost lines; default: line numbers
109 my $has_line = 1;
110 my $has_addr = 0;
112 # Events in input file, eg. (A,B,C,D)
113 my @events;
114 my $events;
116 # Events to show, from command line, eg. (C,A,D)
117 my @show_events;
119 # Map from @show_events indices to @events indices, eg. (2,0,3). Gives the
120 # order in which we must traverse @events in order to show the @show_events,
121 # eg. (@events[$show_order[1]], @events[$show_order[2]]...) = @show_events.
122 # (Might help to think of it like a hash (0 => 2, 1 => 0, 2 => 3).)
123 my @show_order;
125 # Print out the function totals sorted by these events, eg. (D,C).
126 my @sort_events;
128 # Map from @sort_events indices to @events indices, eg. (3,2). Same idea as
129 # for @show_order.
130 my @sort_order;
132 # Thresholds, one for each sort event (or default to 1 if no sort events
133 # specified). We print out functions and do auto-annotations until we've
134 # handled this proportion of all the events thresholded.
135 my @thresholds;
137 my $default_threshold = 99;
139 my $single_threshold = $default_threshold;
141 # If on, show a percentage for each non-zero count.
142 my $show_percs = 1;
144 # If on, automatically annotates all files that are involved in getting over
145 # all the threshold counts.
146 my $auto_annotate = 1;
148 # Number of lines to show around each annotated line.
149 my $context = 8;
151 # Directories in which to look for annotation files.
152 my @include_dirs = ("");
154 # Verbose mode
155 my $verbose = "1";
157 # Inclusive statistics (with subroutine events)
158 my $inclusive = 0;
160 # Inclusive totals for each function, for overall summary.
161 # hash(filename:fn_name => CC array)
162 my %cfn_totals;
164 # hash( file:func => [ called file:func ])
165 my $called_funcs;
167 # hash( file:func => [ calling file:func ])
168 my $calling_funcs;
170 # hash( file:func,line => [called file:func ])
171 my $called_from_line;
173 # hash( file:func,line => file:func
174 my %func_of_line;
176 # hash (file:func => object name)
177 my %obj_name;
179 # Print out the callers of a function
180 my $tree_caller = 0;
182 # Print out the called functions
183 my $tree_calling = 0;
185 # hash( file:func,cfile:cfunc => call CC[])
186 my %call_CCs;
188 # hash( file:func,cfile:cfunc => call counter)
189 my %call_counter;
191 # hash(context, index) => realname for compressed traces
192 my %compressed;
194 # Input file name, will be set in process_cmd_line
195 my $input_file = "";
197 # Version number
198 my $version = "@VERSION@";
200 # Usage message.
201 my $usage = <<END
202 usage: callgrind_annotate [options] [callgrind-out-file [source-files...]]
204 options for the user, with defaults in [ ], are:
205 -h --help show this message
206 --version show version
207 --show=A,B,C only show figures for events A,B,C [all]
208 --threshold=<0--100> percentage of counts (of primary sort event) we
209 are interested in [$default_threshold%]
210 --sort=A,B,C sort columns by events A,B,C [event column order]
211 Each event can optionally be followed by a :
212 and a threshold percentage. If some event specific
213 threshold are given, --threshold value is ignored.
214 --show-percs=yes|no show a percentage for each non-zero count [yes]
215 --auto=yes|no annotate all source files containing functions
216 that helped reach the event count threshold [yes]
217 --context=N print N lines of context before and after
218 annotated lines [8]
219 --inclusive=yes|no add subroutine costs to functions calls [no]
220 --tree=none|caller| print for each function their callers,
221 calling|both the called functions or both [none]
222 -I --include=<dir> add <dir> to list of directories to search for
223 source files
228 # Used in various places of output.
229 my $fancy = '-' x 80 . "\n";
231 sub safe_div($$)
233 my ($x, $y) = @_;
234 return ($y == 0 ? 0 : $x / $y);
237 #-----------------------------------------------------------------------------
238 # Argument and option handling
239 #-----------------------------------------------------------------------------
240 sub process_cmd_line()
242 for my $arg (@ARGV) {
244 # Option handling
245 if ($arg =~ /^-/) {
247 # --version
248 if ($arg =~ /^--version$/) {
249 die("callgrind_annotate-$version\n");
251 # --show=A,B,C
252 } elsif ($arg =~ /^--show=(.*)$/) {
253 @show_events = split(/,/, $1);
255 # --sort=A,B,C
256 } elsif ($arg =~ /^--sort=(.*)$/) {
257 @sort_events = split(/,/, $1);
258 my $th_specified = 0;
259 foreach my $i (0 .. scalar @sort_events - 1) {
260 if ($sort_events[$i] =~ /.*:([\d\.]+)%?$/) {
261 my $th = $1;
262 ($th >= 0 && $th <= 100) or die($usage);
263 $sort_events[$i] =~ s/:.*//;
264 $thresholds[$i] = $th;
265 $th_specified = 1;
266 } else {
267 $thresholds[$i] = 0;
270 if (not $th_specified) {
271 @thresholds = ();
274 # --threshold=X (tolerates a trailing '%')
275 } elsif ($arg =~ /^--threshold=([\d\.]+)%?$/) {
276 $single_threshold = $1;
277 ($1 >= 0 && $1 <= 100) or die($usage);
279 # --show-percs=yes|no
280 } elsif ($arg =~ /^--show-percs=yes$/) {
281 $show_percs = 1;
282 } elsif ($arg =~ /^--show-percs=no$/) {
283 $show_percs = 0;
285 # --auto=yes|no
286 } elsif ($arg =~ /^--auto=(yes|no)$/) {
287 $auto_annotate = 1 if ($1 eq "yes");
288 $auto_annotate = 0 if ($1 eq "no");
290 # --context=N
291 } elsif ($arg =~ /^--context=([\d\.]+)$/) {
292 $context = $1;
293 if ($context < 0) {
294 die($usage);
297 # --inclusive=yes|no
298 } elsif ($arg =~ /^--inclusive=(yes|no)$/) {
299 $inclusive = 1 if ($1 eq "yes");
300 $inclusive = 0 if ($1 eq "no");
302 # --tree=none|caller|calling|both
303 } elsif ($arg =~ /^--tree=(none|caller|calling|both)$/) {
304 $tree_caller = 1 if ($1 eq "caller" || $1 eq "both");
305 $tree_calling = 1 if ($1 eq "calling" || $1 eq "both");
307 # --include=A,B,C
308 } elsif ($arg =~ /^(-I|--include)=(.*)$/) {
309 my $inc = $2;
310 $inc =~ s|/$||; # trim trailing '/'
311 push(@include_dirs, "$inc/");
313 } else { # -h and --help fall under this case
314 die($usage);
317 # Argument handling -- annotation file checking and selection.
318 # Stick filenames into a hash for quick 'n easy lookup throughout
319 } else {
320 if ($input_file eq "") {
321 $input_file = $arg;
323 else {
324 my $readable = 0;
325 foreach my $include_dir (@include_dirs) {
326 if (-r $include_dir . $arg) {
327 $readable = 1;
330 $readable or die("File $arg not found in any of: @include_dirs\n");
331 $user_ann_files{$arg} = 1;
336 if ($input_file eq "") {
337 $input_file = (<callgrind.out*>)[0];
338 if (!defined $input_file) {
339 $input_file = (<cachegrind.out*>)[0];
342 (defined $input_file) or die($usage);
343 print "Reading data from '$input_file'...\n";
347 #-----------------------------------------------------------------------------
348 # Reading of input file
349 #-----------------------------------------------------------------------------
350 sub max ($$)
352 my ($x, $y) = @_;
353 return ($x > $y ? $x : $y);
356 # Add the two arrays; any '.' entries are ignored. Two tricky things:
357 # 1. If $a2->[$i] is undefined, it defaults to 0 which is what we want; we turn
358 # off warnings to allow this. This makes things about 10% faster than
359 # checking for definedness ourselves.
360 # 2. We don't add an undefined count or a ".", even though it's value is 0,
361 # because we don't want to make an $a2->[$i] that is undef become 0
362 # unnecessarily.
363 sub add_array_a_to_b ($$)
365 my ($a1, $a2) = @_;
367 my $n = max(scalar @$a1, scalar @$a2);
368 $^W = 0;
369 foreach my $i (0 .. $n-1) {
370 $a2->[$i] += $a1->[$i] if (defined $a1->[$i] && "." ne $a1->[$i]);
372 $^W = 1;
375 # Is this a line with all events zero?
376 sub is_zero ($)
378 my ($CC) = @_;
379 my $isZero = 1;
380 foreach my $i (0 .. (scalar @$CC)-1) {
381 $isZero = 0 if ($CC->[$i] >0);
383 return $isZero;
386 # Add each event count to the CC array. '.' counts become undef, as do
387 # missing entries (implicitly).
388 sub line_to_CC ($)
390 my @CC = (split /\s+/, $_[0]);
391 (@CC <= @events) or die("Line $.: too many event counts\n");
392 return \@CC;
395 sub uncompressed_name($$)
397 my ($context, $name) = @_;
399 if ($name =~ /^\((\d+)\)\s*(.*)$/) {
400 my $index = $1;
401 my $realname = $2;
403 if ($realname eq "") {
404 $realname = $compressed{$context,$index};
406 else {
407 $compressed{$context,$index} = $realname;
409 return $realname;
411 return $name;
414 sub read_input_file()
416 open(INPUTFILE, "< $input_file") || die "File $input_file not opened\n";
418 my $line;
420 # Read header
421 while(<INPUTFILE>) {
423 # Skip comments and empty lines.
424 if (/^\s*$/ || /^\#/) { ; }
426 elsif (/^version:\s*(\d+)/) {
427 # Can't read format with major version > 1
428 ($1<2) or die("Can't read format with major version $1.\n");
431 elsif (/^pid:\s+(.*)$/) { $pid = $1; }
432 elsif (/^thread:\s+(.*)$/) { $thread = $1; }
433 elsif (/^part:\s+(.*)$/) { $part = $1; }
434 elsif (/^desc:\s+(.*)$/) {
435 my $dline = $1;
436 # suppress profile options in description output
437 if ($dline =~ /^Option:/) {;}
438 else { $desc .= "$dline\n"; }
440 elsif (/^cmd:\s+(.*)$/) { $cmd = $1; }
441 elsif (/^creator:\s+(.*)$/) { $creator = $1; }
442 elsif (/^positions:\s+(.*)$/) {
443 my $positions = $1;
444 $has_line = ($positions =~ /line/);
445 $has_addr = ($positions =~ /(addr|instr)/);
447 elsif (/^event:\s+.*$/) {
448 # ignore lines giving a long name to an event
450 elsif (/^events:\s+(.*)$/) {
451 $events = $1;
453 # events line is last in header
454 last;
456 else {
457 warn("WARNING: header line $. malformed, ignoring\n");
458 if ($verbose) { chomp; warn(" line: '$_'\n"); }
462 # Read "events:" line. We make a temporary hash in which the Nth event's
463 # value is N, which is useful for handling --show/--sort options below.
464 ($events ne "") or die("Line $.: missing events line\n");
465 @events = split(/\s+/, $events);
466 my %events;
467 my $n = 0;
468 foreach my $event (@events) {
469 $events{$event} = $n;
470 $n++
473 # If no --show arg give, default to showing all events in the file.
474 # If --show option is used, check all specified events appeared in the
475 # "events:" line. Then initialise @show_order.
476 if (@show_events) {
477 foreach my $show_event (@show_events) {
478 (defined $events{$show_event}) or
479 die("--show event `$show_event' did not appear in input\n");
481 } else {
482 @show_events = @events;
484 foreach my $show_event (@show_events) {
485 push(@show_order, $events{$show_event});
488 # Do as for --show, but if no --sort arg given, default to sorting by
489 # column order (ie. first column event is primary sort key, 2nd column is
490 # 2ndary key, etc).
491 if (@sort_events) {
492 foreach my $sort_event (@sort_events) {
493 (defined $events{$sort_event}) or
494 die("--sort event `$sort_event' did not appear in input\n");
496 } else {
497 @sort_events = @events;
499 foreach my $sort_event (@sort_events) {
500 push(@sort_order, $events{$sort_event});
503 # If multiple threshold args weren't given via --sort, stick in the single
504 # threshold (either from --threshold if used, or the default otherwise) for
505 # the primary sort event, and 0% for the rest.
506 if (not @thresholds) {
507 foreach my $e (@sort_order) {
508 push(@thresholds, 0);
510 $thresholds[0] = $single_threshold;
511 } else {
512 # setting $single_threshold to 0 to ensure the 'per event'
513 # threshold logic is used.
514 $single_threshold = 0;
517 # Current directory, used to strip from file names if absolute
518 my $pwd = `pwd`;
519 chomp $pwd;
520 $pwd .= '/';
522 my $curr_obj = "";
523 my $curr_file;
524 my $curr_fn;
525 my $curr_name;
526 my $curr_line_num = 0;
527 my $prev_line_num = 0;
529 my $curr_cobj = "";
530 my $curr_cfile = "";
531 my $curr_cfunc = "";
532 my $curr_cname;
533 my $curr_call_counter = 0;
534 my $curr_cfn_CC = [];
536 my $curr_fn_CC = [];
537 my $curr_file_ind_CCs = {}; # hash(line_num => CC)
539 # Read body of input file.
540 while (<INPUTFILE>) {
541 # Skip comments and empty lines.
542 next if /^\s*$/ || /^\#/;
544 $prev_line_num = $curr_line_num;
546 s/^\+(\d+)/$prev_line_num+$1/e;
547 s/^\-(\d+)/$prev_line_num-$1/e;
548 s/^\*/$prev_line_num/e;
549 if (s/^(-?\d+|0x\w+)\s+//) {
550 $curr_line_num = $1;
551 if ($has_addr) {
552 if ($has_line) {
553 s/^\+(\d+)/$prev_line_num+$1/e;
554 s/^\-(\d+)/$prev_line_num-$1/e;
555 s/^\*/$prev_line_num/e;
557 if (s/^(\d+)\s+//) { $curr_line_num = $1; }
559 else { $curr_line_num = 0; }
561 my $CC = line_to_CC($_);
563 if ($curr_call_counter>0) {
564 # print "Read ($curr_name => $curr_cname) $curr_call_counter\n";
566 if (!defined $call_CCs{$curr_name,$curr_cname}) {
567 $call_CCs{$curr_name,$curr_cname} = [];
568 $call_counter{$curr_name,$curr_cname} = 0;
570 add_array_a_to_b($CC, $call_CCs{$curr_name,$curr_cname});
571 $call_counter{$curr_name,$curr_cname} += $curr_call_counter;
573 my $tmp = $called_from_line->{$curr_file,$curr_line_num};
574 if (!defined $tmp) {
575 $func_of_line{$curr_file,$curr_line_num} = $curr_name;
577 $tmp = {} unless defined $tmp;
578 $$tmp{$curr_cname} = 1;
579 $called_from_line->{$curr_file,$curr_line_num} = $tmp;
580 if (!defined $call_CCs{$curr_name,$curr_cname,$curr_line_num}) {
581 $call_CCs{$curr_name,$curr_cname,$curr_line_num} = [];
582 $call_counter{$curr_name,$curr_cname,$curr_line_num} = 0;
584 add_array_a_to_b($CC, $call_CCs{$curr_name,$curr_cname,$curr_line_num});
585 $call_counter{$curr_name,$curr_cname,$curr_line_num} += $curr_call_counter;
587 $curr_call_counter = 0;
589 # inclusive costs
590 $curr_cfn_CC = $cfn_totals{$curr_cname};
591 $curr_cfn_CC = [] unless (defined $curr_cfn_CC);
592 add_array_a_to_b($CC, $curr_cfn_CC);
593 $cfn_totals{$curr_cname} = $curr_cfn_CC;
595 if ($inclusive) {
596 add_array_a_to_b($CC, $curr_fn_CC);
598 next;
601 add_array_a_to_b($CC, $curr_fn_CC);
603 # If curr_file is selected, add CC to curr_file list. We look for
604 # full filename matches; or, if auto-annotating, we have to
605 # remember everything -- we won't know until the end what's needed.
606 if ($auto_annotate || defined $user_ann_files{$curr_file}) {
607 my $tmp = $curr_file_ind_CCs->{$curr_line_num};
608 $tmp = [] unless defined $tmp;
609 add_array_a_to_b($CC, $tmp);
610 $curr_file_ind_CCs->{$curr_line_num} = $tmp;
613 } elsif (s/^fn=(.*)$//) {
614 # Commit result from previous function
615 $fn_totals{$curr_name} = $curr_fn_CC if (defined $curr_name);
617 # Setup new one
618 $curr_fn = uncompressed_name("fn",$1);
619 $curr_name = "$curr_file:$curr_fn";
620 $obj_name{$curr_name} = $curr_obj;
621 $curr_fn_CC = $fn_totals{$curr_name};
622 $curr_fn_CC = [] unless (defined $curr_fn_CC);
624 } elsif (s/^ob=(.*)$//) {
625 $curr_obj = uncompressed_name("ob",$1);
627 } elsif (s/^fl=(.*)$//) {
628 $all_ind_CCs{$curr_file} = $curr_file_ind_CCs
629 if (defined $curr_file);
631 $curr_file = uncompressed_name("fl",$1);
632 $curr_file =~ s/^\Q$pwd\E//;
633 $curr_file_ind_CCs = $all_ind_CCs{$curr_file};
634 $curr_file_ind_CCs = {} unless (defined $curr_file_ind_CCs);
636 } elsif (s/^(fi|fe)=(.*)$//) {
637 (defined $curr_name) or die("Line $.: Unexpected fi/fe line\n");
638 $fn_totals{$curr_name} = $curr_fn_CC;
639 $all_ind_CCs{$curr_file} = $curr_file_ind_CCs;
641 $curr_file = uncompressed_name("fl",$2);
642 $curr_file =~ s/^\Q$pwd\E//;
643 $curr_name = "$curr_file:$curr_fn";
644 $curr_file_ind_CCs = $all_ind_CCs{$curr_file};
645 $curr_file_ind_CCs = {} unless (defined $curr_file_ind_CCs);
646 $curr_fn_CC = $fn_totals{$curr_name};
647 $curr_fn_CC = [] unless (defined $curr_fn_CC);
649 } elsif (s/^cob=(.*)$//) {
650 $curr_cobj = uncompressed_name("ob",$1);
652 } elsif (s/^cf[il]=(.*)$//) {
653 $curr_cfile = uncompressed_name("fl",$1);
655 } elsif (s/^cfn=(.*)$//) {
656 $curr_cfunc = uncompressed_name("fn",$1);
657 if ($curr_cfile eq "") {
658 $curr_cname = "$curr_file:$curr_cfunc";
660 else {
661 $curr_cname = "$curr_cfile:$curr_cfunc";
662 $curr_cfile = "";
665 my $tmp = $calling_funcs->{$curr_cname};
666 $tmp = {} unless defined $tmp;
667 $$tmp{$curr_name} = 1;
668 $calling_funcs->{$curr_cname} = $tmp;
670 my $tmp2 = $called_funcs->{$curr_name};
671 $tmp2 = {} unless defined $tmp2;
672 $$tmp2{$curr_cname} = 1;
673 $called_funcs->{$curr_name} = $tmp2;
675 } elsif (s/^calls=(\d+)//) {
676 $curr_call_counter = $1;
678 } elsif (s/^(jump|jcnd)=//) {
679 #ignore jump information
681 } elsif (s/^jfi=(.*)$//) {
682 # side effect needed: possibly add compression mapping
683 uncompressed_name("fl",$1);
684 # ignore jump information
686 } elsif (s/^jfn=(.*)$//) {
687 # side effect needed: possibly add compression mapping
688 uncompressed_name("fn",$1);
689 # ignore jump information
691 } elsif (s/^totals:\s+//) {
692 $totals_CC = line_to_CC($_);
694 } elsif (s/^summary:\s+//) {
695 $summary_CC = line_to_CC($_);
697 } else {
698 warn("WARNING: line $. malformed, ignoring\n");
699 if ($verbose) { chomp; warn(" line: '$_'\n"); }
703 # Finish up handling final filename/fn_name counts
704 $fn_totals{"$curr_file:$curr_fn"} = $curr_fn_CC
705 if (defined $curr_file && defined $curr_fn);
706 $all_ind_CCs{$curr_file} =
707 $curr_file_ind_CCs if (defined $curr_file);
709 # Correct inclusive totals
710 if ($inclusive) {
711 foreach my $name (keys %cfn_totals) {
712 $fn_totals{$name} = $cfn_totals{$name};
716 close(INPUTFILE);
718 if ((not defined $summary_CC) || is_zero($summary_CC)) {
719 $summary_CC = $totals_CC;
721 # if neither 'summary:' nor 'totals:' line is given,
722 # calculate summary from fn_totals hash
723 if ((not defined $summary_CC) || is_zero($summary_CC)) {
724 $summary_calculated = 1;
725 $summary_CC = [];
726 foreach my $name (keys %fn_totals) {
727 add_array_a_to_b($fn_totals{$name}, $summary_CC);
733 #-----------------------------------------------------------------------------
734 # Print options used
735 #-----------------------------------------------------------------------------
736 sub print_options ()
738 print($fancy);
739 print "Profile data file '$input_file'";
740 if ($creator ne "") { print " (creator: $creator)"; }
741 print "\n";
743 print($fancy);
744 print($desc);
745 my $target = $cmd;
746 if ($target eq "") { $target = "(unknown)"; }
747 if ($pid ne "") {
748 $target .= " (PID $pid";
749 if ($part ne "") { $target .= ", part $part"; }
750 if ($thread ne "") { $target .= ", thread $thread"; }
751 $target .= ")";
753 print("Profiled target: $target\n");
754 print("Events recorded: @events\n");
755 print("Events shown: @show_events\n");
756 print("Event sort order: @sort_events\n");
757 print("Thresholds: @thresholds\n");
759 my @include_dirs2 = @include_dirs; # copy @include_dirs
760 shift(@include_dirs2); # remove "" entry, which is always the first
761 unshift(@include_dirs2, "") if (0 == @include_dirs2);
762 my $include_dir = shift(@include_dirs2);
763 print("Include dirs: $include_dir\n");
764 foreach my $include_dir (@include_dirs2) {
765 print(" $include_dir\n");
768 my @user_ann_files = keys %user_ann_files;
769 unshift(@user_ann_files, "") if (0 == @user_ann_files);
770 my $user_ann_file = shift(@user_ann_files);
771 print("User annotated: $user_ann_file\n");
772 foreach $user_ann_file (@user_ann_files) {
773 print(" $user_ann_file\n");
776 my $is_on = ($auto_annotate ? "on" : "off");
777 print("Auto-annotation: $is_on\n");
778 print("\n");
781 #-----------------------------------------------------------------------------
782 # Print summary and sorted function totals
783 #-----------------------------------------------------------------------------
784 sub mycmp ($$)
786 my ($c, $d) = @_;
788 # Iterate through sort events (eg. 3,2); return result if two are different
789 foreach my $i (@sort_order) {
790 my ($x, $y);
791 $x = $c->[$i];
792 $y = $d->[$i];
793 $x = -1 unless defined $x;
794 $y = -1 unless defined $y;
796 my $cmp = $y <=> $x; # reverse sort
797 if (0 != $cmp) {
798 return $cmp;
801 # Exhausted events, equal
802 return 0;
805 sub commify ($) {
806 my ($val) = @_;
807 1 while ($val =~ s/^(\d+)(\d{3})/$1,$2/);
808 return $val;
811 # Because the counts can get very big, and we don't want to waste screen space
812 # and make lines too long, we compute exactly how wide each column needs to be
813 # by finding the widest entry for each one.
814 sub compute_CC_col_widths (@)
816 my @CCs = @_;
817 my $CC_col_widths = [];
819 # Initialise with minimum widths (from event names)
820 foreach my $event (@events) {
821 push(@$CC_col_widths, length($event));
824 # Find maximum width count for each column. @CC_col_width positions
825 # correspond to @CC positions.
826 foreach my $CC (@CCs) {
827 foreach my $i (0 .. scalar(@$CC)-1) {
828 if (defined $CC->[$i]) {
829 # Find length, accounting for commas that will be added, and
830 # possibly a percentage.
831 my $length = length $CC->[$i];
832 my $width = $length + int(($length - 1) / 3);
833 if ($show_percs) {
834 $width += 9; # e.g. " (12.34%)" is 9 chars
836 $CC_col_widths->[$i] = max($CC_col_widths->[$i], $width);
840 return $CC_col_widths;
843 # Print the CC with each column's size dictated by $CC_col_widths.
844 sub print_CC ($$)
846 my ($CC, $CC_col_widths) = @_;
848 foreach my $i (@show_order) {
849 my $count = (defined $CC->[$i] ? commify($CC->[$i]) : ".");
851 my $perc = "";
852 if ($show_percs) {
853 if (defined $CC->[$i] && $CC->[$i] != 0) {
854 # Try our best to keep the number fitting into 5 chars. This
855 # requires dropping a digit after the decimal place if it's
856 # sufficiently negative (e.g. "-10.0") or positive (e.g.
857 # "100.0"). Thanks to diffs it's possible to have even more
858 # extreme values, like "-100.0" or "1000.0"; those rare case
859 # will end up with slightly wrong indenting, oh well.
860 $perc = safe_div($CC->[$i] * 100, $summary_CC->[$i]);
861 $perc = (-9.995 < $perc && $perc < 99.995)
862 ? sprintf(" (%5.2f%%)", $perc)
863 : sprintf(" (%5.1f%%)", $perc);
864 } else {
865 # Don't show percentages for "." and "0" entries.
866 $perc = " ";
870 # $reps will be negative for the extreme values mentioned above. The
871 # use of max() avoids a possible warning about a negative repeat count.
872 my $text = $count . $perc;
873 my $len = length($text);
874 my $reps = $CC_col_widths->[$i] - length($text);
875 my $space = ' ' x max($reps, 0);
876 print("$space$text ");
880 sub print_events ($)
882 my ($CC_col_widths) = @_;
884 foreach my $i (@show_order) {
885 my $event = $events[$i];
886 my $event_width = length($event);
887 my $col_width = $CC_col_widths->[$i];
888 my $space = ' ' x ($col_width - $event_width);
889 print("$event$space ");
893 # Prints summary and function totals (with separate column widths, so that
894 # function names aren't pushed over unnecessarily by huge summary figures).
895 # Also returns a hash containing all the files that are involved in getting the
896 # events count above the thresholds (ie. all the interesting ones).
897 sub print_summary_and_fn_totals ()
899 my @fn_fullnames = keys %fn_totals;
901 # Work out the size of each column for printing (summary and functions
902 # separately).
903 my $summary_CC_col_widths = compute_CC_col_widths($summary_CC);
904 my $fn_CC_col_widths = compute_CC_col_widths(values %fn_totals);
906 # Header and counts for summary
907 print($fancy);
908 print_events($summary_CC_col_widths);
909 print("\n");
910 print($fancy);
911 print_CC($summary_CC, $summary_CC_col_widths);
912 print(" PROGRAM TOTALS");
913 if ($summary_calculated) {
914 print(" (calculated)");
916 print("\n\n");
918 # Header for functions
919 print($fancy);
920 print_events($fn_CC_col_widths);
921 print(" file:function\n");
922 print($fancy);
924 # Sort function names into order dictated by --sort option.
925 @fn_fullnames = sort {
926 mycmp($fn_totals{$a}, $fn_totals{$b}) || $a cmp $b
927 } @fn_fullnames;
930 # Assertion
931 (scalar @sort_order == scalar @thresholds) or
932 die("sort_order length != thresholds length:\n",
933 " @sort_order\n @thresholds\n");
935 my $threshold_files = {};
936 # @curr_totals has the same shape as @sort_order and @thresholds
937 my @curr_totals = ();
938 foreach my $e (@thresholds) {
939 push(@curr_totals, 0);
942 # Print functions, stopping when the threshold has been reached.
943 foreach my $fn_name (@fn_fullnames) {
944 # if $single_threshold is 100 the user want to see everything,
945 # so do not enter the filtering logic, as truncation can cause
946 # some functions to not be shown.
947 if ($single_threshold < 100) {
948 # Stop when we've reached all the thresholds
949 my $reached_all_thresholds = 1;
950 foreach my $i (0 .. scalar @thresholds - 1) {
951 my $prop = $curr_totals[$i] * 100;
952 if (defined $summary_CC->[$sort_order[$i]] &&
953 $summary_CC->[$sort_order[$i]] >0) {
954 $prop = $prop / $summary_CC->[$sort_order[$i]];
956 $reached_all_thresholds &&= ($prop >= $thresholds[$i]);
958 last if $reached_all_thresholds;
961 if ($tree_caller || $tree_calling) { print "\n"; }
963 if ($tree_caller && ($fn_name ne "???:???")) {
964 # Print function callers
965 my $tmp1 = $calling_funcs->{$fn_name};
966 if (defined $tmp1) {
967 # Sort calling functions into order dictated by --sort option.
968 my @callings = sort {
969 mycmp($call_CCs{$a,$fn_name}, $call_CCs{$b,$fn_name})
970 } keys %$tmp1;
971 foreach my $calling (@callings) {
972 if (defined $call_counter{$calling,$fn_name}) {
973 print_CC($call_CCs{$calling,$fn_name}, $fn_CC_col_widths);
974 print" < $calling (";
975 print commify($call_counter{$calling,$fn_name}) . "x)";
976 if (defined $obj_name{$calling}) {
977 print " [$obj_name{$calling}]";
979 print "\n";
985 # Print function results
986 my $fn_CC = $fn_totals{$fn_name};
987 print_CC($fn_CC, $fn_CC_col_widths);
988 if ($tree_caller || $tree_calling) { print " * "; }
989 print(" $fn_name");
990 if ((defined $obj_name{$fn_name}) &&
991 ($obj_name{$fn_name} ne "")) {
992 print " [$obj_name{$fn_name}]";
994 print "\n";
996 if ($tree_calling && ($fn_name ne "???:???")) {
997 # Print called functions
998 my $tmp2 = $called_funcs->{$fn_name};
999 if (defined $tmp2) {
1000 # Sort called functions into order dictated by --sort option.
1001 my @calleds = sort {
1002 mycmp($call_CCs{$fn_name,$a}, $call_CCs{$fn_name,$b})
1003 } keys %$tmp2;
1004 foreach my $called (@calleds) {
1005 if (defined $call_counter{$fn_name,$called}) {
1006 print_CC($call_CCs{$fn_name,$called}, $fn_CC_col_widths);
1007 print" > $called (";
1008 print commify($call_counter{$fn_name,$called}) . "x)";
1009 if (defined $obj_name{$called}) {
1010 print " [$obj_name{$called}]";
1012 print "\n";
1018 # Update the threshold counts
1019 my $filename = $fn_name;
1020 $filename =~ s/:.+$//; # remove function name
1021 $threshold_files->{$filename} = 1;
1022 foreach my $i (0 .. scalar @sort_order - 1) {
1023 if ($inclusive) {
1024 $curr_totals[$i] = $summary_CC->[$sort_order[$i]] -
1025 $fn_CC->[$sort_order[$i]]
1026 if (defined $fn_CC->[$sort_order[$i]]);
1027 } else {
1028 $curr_totals[$i] += $fn_CC->[$sort_order[$i]]
1029 if (defined $fn_CC->[$sort_order[$i]]);
1033 print("\n");
1035 return $threshold_files;
1038 #-----------------------------------------------------------------------------
1039 # Annotate selected files
1040 #-----------------------------------------------------------------------------
1042 # Issue a warning that the source file is more recent than the input file.
1043 sub warning_on_src_more_recent_than_inputfile ($)
1045 my $src_file = $_[0];
1047 my $warning = <<END
1048 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1049 @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@
1050 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1051 @ Source file '$src_file' is more recent than input file '$input_file'.
1052 @ Annotations may not be correct.
1053 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1057 print($warning);
1060 # If there is information about lines not in the file, issue a warning
1061 # explaining possible causes.
1062 sub warning_on_nonexistent_lines ($$$)
1064 my ($src_more_recent_than_inputfile, $src_file, $excess_line_nums) = @_;
1065 my $cause_and_solution;
1067 if ($src_more_recent_than_inputfile) {
1068 $cause_and_solution = <<END
1069 @@ cause: '$src_file' has changed since information was gathered.
1070 @@ If so, a warning will have already been issued about this.
1071 @@ solution: Recompile program and rerun under "valgrind --cachesim=yes" to
1072 @@ gather new information.
1074 # We suppress warnings about .h files
1075 } elsif ($src_file =~ /\.h$/) {
1076 $cause_and_solution = <<END
1077 @@ cause: bug in the Valgrind's debug info reader that screws up with .h
1078 @@ files sometimes
1079 @@ solution: none, sorry
1081 } else {
1082 $cause_and_solution = <<END
1083 @@ cause: not sure, sorry
1087 my $warning = <<END
1088 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1089 @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@
1090 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1092 @@ Information recorded about lines past the end of '$src_file'.
1094 @@ Probable cause and solution:
1095 $cause_and_solution@@
1096 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1099 print($warning);
1102 sub annotate_ann_files($)
1104 my ($threshold_files) = @_;
1106 my %all_ann_files;
1107 my @unfound_auto_annotate_files;
1108 my $printed_totals_CC = [];
1110 # If auto-annotating, add interesting files (but not "???")
1111 if ($auto_annotate) {
1112 delete $threshold_files->{"???"};
1113 %all_ann_files = (%user_ann_files, %$threshold_files)
1114 } else {
1115 %all_ann_files = %user_ann_files;
1118 # Track if we did any annotations.
1119 my $did_annotations = 0;
1121 LOOP:
1122 foreach my $src_file (keys %all_ann_files) {
1124 my $opened_file = "";
1125 my $full_file_name = "";
1126 foreach my $include_dir (@include_dirs) {
1127 my $try_name = $include_dir . $src_file;
1128 if (open(INPUTFILE, "< $try_name")) {
1129 $opened_file = $try_name;
1130 $full_file_name = ($include_dir eq ""
1131 ? $src_file
1132 : "$include_dir + $src_file");
1133 last;
1137 if (not $opened_file) {
1138 # Failed to open the file. If chosen on the command line, die.
1139 # If arose from auto-annotation, print a little message.
1140 if (defined $user_ann_files{$src_file}) {
1141 die("File $src_file not opened in any of: @include_dirs\n");
1143 } else {
1144 push(@unfound_auto_annotate_files, $src_file);
1147 } else {
1148 # File header (distinguish between user- and auto-selected files).
1149 print("$fancy");
1150 my $ann_type =
1151 (defined $user_ann_files{$src_file} ? "User" : "Auto");
1152 print("-- $ann_type-annotated source: $full_file_name\n");
1153 print("$fancy");
1155 # Get file's CCs
1156 my $src_file_CCs = $all_ind_CCs{$src_file};
1157 if (!defined $src_file_CCs) {
1158 print(" No information has been collected for $src_file\n\n");
1159 next LOOP;
1162 $did_annotations = 1;
1164 # Numeric, not lexicographic sort!
1165 my @line_nums = sort {$a <=> $b} keys %$src_file_CCs;
1167 # If $src_file more recent than cachegrind.out, issue warning
1168 my $src_more_recent_than_inputfile = 0;
1169 if ((stat $opened_file)[9] > (stat $input_file)[9]) {
1170 $src_more_recent_than_inputfile = 1;
1171 warning_on_src_more_recent_than_inputfile($src_file);
1174 # Work out the size of each column for printing
1175 my $CC_col_widths = compute_CC_col_widths(values %$src_file_CCs);
1177 # Events header
1178 print_events($CC_col_widths);
1179 print("\n\n");
1181 # Shift out 0 if it's in the line numbers (from unknown entries,
1182 # likely due to bugs in Valgrind's stabs debug info reader)
1183 shift(@line_nums) if (0 == $line_nums[0]);
1185 # Finds interesting line ranges -- all lines with a CC, and all
1186 # lines within $context lines of a line with a CC.
1187 my $n = @line_nums;
1188 my @pairs;
1189 for (my $i = 0; $i < $n; $i++) {
1190 push(@pairs, $line_nums[$i] - $context); # lower marker
1191 while ($i < $n-1 &&
1192 $line_nums[$i] + 2*$context >= $line_nums[$i+1]) {
1193 $i++;
1195 push(@pairs, $line_nums[$i] + $context); # upper marker
1198 # Annotate chosen lines, tracking total counts of lines printed
1199 $pairs[0] = 1 if ($pairs[0] < 1);
1200 while (@pairs) {
1201 my $low = shift @pairs;
1202 my $high = shift @pairs;
1203 while ($. < $low-1) {
1204 my $tmp = <INPUTFILE>;
1205 last unless (defined $tmp); # hack to detect EOF
1207 my $src_line;
1208 # Print line number, unless start of file
1209 print("-- line $low " . '-' x 40 . "\n") if ($low != 1);
1210 while (($. < $high) && ($src_line = <INPUTFILE>)) {
1211 if (defined $line_nums[0] && $. == $line_nums[0]) {
1212 print_CC($src_file_CCs->{$.}, $CC_col_widths);
1213 add_array_a_to_b($src_file_CCs->{$.},
1214 $printed_totals_CC);
1215 shift(@line_nums);
1217 } else {
1218 print_CC([], $CC_col_widths);
1221 print(" $src_line");
1223 my $tmp = $called_from_line->{$src_file,$.};
1224 my $func = $func_of_line{$src_file,$.};
1225 if (defined $tmp) {
1226 # Sort called functions into order dictated by --sort option.
1227 my @calleds = sort {
1228 mycmp($call_CCs{$func,$a}, $call_CCs{$func,$b})
1229 } keys %$tmp;
1230 foreach my $called (@calleds) {
1231 if (defined $call_CCs{$func,$called,$.}) {
1232 print_CC($call_CCs{$func,$called,$.}, $CC_col_widths);
1233 print " => $called (";
1234 print commify($call_counter{$func,$called,$.}) . "x)\n";
1239 # Print line number, unless EOF
1240 if ($src_line) {
1241 print("-- line $high " . '-' x 40 . "\n");
1242 } else {
1243 last;
1247 # If there was info on lines past the end of the file...
1248 if (@line_nums) {
1249 foreach my $line_num (@line_nums) {
1250 print_CC($src_file_CCs->{$line_num}, $CC_col_widths);
1251 print(" <bogus line $line_num>\n");
1253 print("\n");
1254 warning_on_nonexistent_lines($src_more_recent_than_inputfile,
1255 $src_file, \@line_nums);
1257 print("\n");
1259 # Print summary of counts attributed to file but not to any
1260 # particular line (due to incomplete debug info).
1261 if ($src_file_CCs->{0}) {
1262 print_CC($src_file_CCs->{0}, $CC_col_widths);
1263 print(" <counts for unidentified lines in $src_file>\n\n");
1266 close(INPUTFILE);
1270 # Print list of unfound auto-annotate selected files.
1271 if (@unfound_auto_annotate_files) {
1272 print("$fancy");
1273 print("The following files chosen for auto-annotation could not be found:\n");
1274 print($fancy);
1275 foreach my $f (sort @unfound_auto_annotate_files) {
1276 print(" $f\n");
1278 print("\n");
1281 # If we did any annotating, show how many events were covered by annotated
1282 # lines above.
1283 if ($did_annotations) {
1284 foreach (my $i = 0; $i < @$summary_CC; $i++) {
1285 # Some files (in particular the files produced by --xtree-memory)
1286 # have non additive self costs, so have a special case for these
1287 # to print all functions and also to avoid a division by 0.
1288 if ($summary_CC->[$i] == 0
1289 || $printed_totals_CC->[$i] > $summary_CC->[$i]) {
1290 # Set the summary_CC value equal to the printed_totals_CC value
1291 # so that the percentage printed by the print_CC call below is
1292 # 100%. This is ok because the summary_CC value is not used
1293 # again afterward.
1294 $summary_CC->[$i] = $printed_totals_CC->[$i];
1297 my $CC_col_widths = compute_CC_col_widths($printed_totals_CC);
1298 print($fancy);
1299 print_events($CC_col_widths);
1300 print("\n");
1301 print($fancy);
1302 print_CC($printed_totals_CC, $CC_col_widths);
1303 print(" events annotated\n\n");
1307 #----------------------------------------------------------------------------
1308 # "main()"
1309 #----------------------------------------------------------------------------
1310 process_cmd_line();
1311 read_input_file();
1312 print_options();
1313 my $threshold_files = print_summary_and_fn_totals();
1314 annotate_ann_files($threshold_files);
1316 ##--------------------------------------------------------------------##
1317 ##--- end vg_annotate.in ---##
1318 ##--------------------------------------------------------------------##