1#! /usr/bin/perl -w
2##--------------------------------------------------------------------##
3##--- The cache simulation framework: instrumentation, recording   ---##
4##--- and results printing.                                        ---##
5##---                                           callgrind_annotate ---##
6##--------------------------------------------------------------------##
7
8#  This file is part of Callgrind, a cache-simulator and call graph
9#  tracer built on Valgrind.
10#
11#  Copyright (C) 2003 Josef Weidendorfer
12#     Josef.Weidendorfer@gmx.de
13#
14#  This file is based heavily on cg_annotate, part of Valgrind.
15#  Copyright (C) 2002 Nicholas Nethercote
16#     njn@valgrind.org
17#
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.
22#
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.
27#
28#  You should have received a copy of the GNU General Public License
29#  along with this program; if not, write to the Free Software
30#  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
31#  02111-1307, USA.
32#
33#  The GNU General Public License is contained in the file COPYING.
34
35#----------------------------------------------------------------------------
36# Annotator for cachegrind/callgrind. 
37#
38# File format is described in /docs/techdocs.html.
39#
40# Performance improvements record, using cachegrind.out for cacheprof, doing no
41# source annotation (irrelevant ones removed):
42#                                                               user time
43# 1. turned off warnings in add_hash_a_to_b()                   3.81 --> 3.48s
44#    [now add_array_a_to_b()]
45# 6. make line_to_CC() return a ref instead of a hash           3.01 --> 2.77s
46#
47#10. changed file format to avoid file/fn name repetition       2.40s
48#    (not sure why higher;  maybe due to new '.' entries?)
49#11. changed file format to drop unnecessary end-line "."s      2.36s
50#    (shrunk file by about 37%)
51#12. switched from hash CCs to array CCs                        1.61s
52#13. only adding b[i] to a[i] if b[i] defined (was doing it if
53#    either a[i] or b[i] was defined, but if b[i] was undefined
54#    it just added 0)                                           1.48s
55#14. Stopped converting "." entries to undef and then back      1.16s
56#15. Using foreach $i (x..y) instead of for ($i = 0...) in
57#    add_array_a_to_b()                                         1.11s
58#
59# Auto-annotating primes:
60#16. Finding count lengths by int((length-1)/3), not by
61#    commifying (halves the number of commify calls)            1.68s --> 1.47s
62
63use strict;
64
65#----------------------------------------------------------------------------
66# Overview: the running example in the comments is for:
67#   - events = A,B,C,D
68#   - --show=C,A,D
69#   - --sort=D,C
70#----------------------------------------------------------------------------
71
72#----------------------------------------------------------------------------
73# Global variables, main data structures
74#----------------------------------------------------------------------------
75# CCs are arrays, the counts corresponding to @events, with 'undef'
76# representing '.'.  This makes things fast (faster than using hashes for CCs)
77# but we have to use @sort_order and @show_order below to handle the --sort and
78# --show options, which is a bit tricky.
79#----------------------------------------------------------------------------
80
81# Total counts for summary (an array reference).
82my $summary_CC;
83my $totals_CC;
84my $summary_calculated = 0;
85
86# Totals for each function, for overall summary.
87# hash(filename:fn_name => CC array)
88my %fn_totals;
89
90# Individual CCs, organised by filename and line_num for easy annotation.
91# hash(filename => hash(line_num => CC array))
92my %all_ind_CCs;
93
94# Files chosen for annotation on the command line.  
95# key = basename (trimmed of any directory), value = full filename
96my %user_ann_files;
97
98# Generic description string.
99my $desc = "";
100
101# Command line of profiled program.
102my $cmd = "";
103
104# Info on the profiled process.
105my $creator = "";
106my $pid = "";
107my $part = "";
108my $thread = "";
109
110# Positions used for cost lines; default: line numbers
111my $has_line = 1;
112my $has_addr = 0;
113
114# Events in input file, eg. (A,B,C,D)
115my @events;
116my $events;
117
118# Events to show, from command line, eg. (C,A,D)
119my @show_events;
120
121# Map from @show_events indices to @events indices, eg. (2,0,3).  Gives the
122# order in which we must traverse @events in order to show the @show_events, 
123# eg. (@events[$show_order[1]], @events[$show_order[2]]...) = @show_events.
124# (Might help to think of it like a hash (0 => 2, 1 => 0, 2 => 3).)
125my @show_order;
126
127# Print out the function totals sorted by these events, eg. (D,C).
128my @sort_events;
129
130# Map from @sort_events indices to @events indices, eg. (3,2).  Same idea as
131# for @show_order.
132my @sort_order;
133
134# Thresholds, one for each sort event (or default to 1 if no sort events
135# specified).  We print out functions and do auto-annotations until we've
136# handled this proportion of all the events thresholded.
137my @thresholds;
138
139my $default_threshold = 99;
140
141my $single_threshold  = $default_threshold;
142
143# If on, automatically annotates all files that are involved in getting over
144# all the threshold counts.
145my $auto_annotate = 0;
146
147# Number of lines to show around each annotated line.
148my $context = 8;
149
150# Directories in which to look for annotation files.
151my @include_dirs = ("");
152
153# Verbose mode
154my $verbose = "1";
155
156# Inclusive statistics (with subroutine events)
157my $inclusive = 0;
158
159# Inclusive totals for each function, for overall summary.
160# hash(filename:fn_name => CC array)
161my %cfn_totals;
162
163# hash( file:func => [ called file:func ])
164my $called_funcs;
165
166# hash( file:func => [ calling file:func ])
167my $calling_funcs;
168
169# hash( file:func,line => [called file:func ])
170my $called_from_line;
171
172# hash( file:func,line => file:func
173my %func_of_line;
174
175# hash (file:func => object name)
176my %obj_name;
177
178# Print out the callers of a function
179my $tree_caller = 0;
180
181# Print out the called functions
182my $tree_calling = 0;
183
184# hash( file:func,cfile:cfunc => call CC[])
185my %call_CCs;
186
187# hash( file:func,cfile:cfunc => call counter)
188my %call_counter;
189
190# hash(context, index) => realname for compressed traces
191my %compressed;
192
193# Input file name, will be set in process_cmd_line
194my $input_file = "";
195
196# Version number
197my $version = "@VERSION@";
198
199# Usage message.
200my $usage = <<END
201usage: callgrind_annotate [options] [callgrind-out-file [source-files...]]
202
203  options for the user, with defaults in [ ], are:
204    -h --help             show this message
205    --version             show version
206    --show=A,B,C          only show figures for events A,B,C [all]
207    --sort=A,B,C          sort columns by events A,B,C [event column order]
208    --threshold=<0--100>  percentage of counts (of primary sort event) we
209                          are interested in [$default_threshold%]
210    --auto=yes|no         annotate all source files containing functions
211                          that helped reach the event count threshold [no]
212    --context=N           print N lines of context before and after
213                          annotated lines [8]
214    --inclusive=yes|no    add subroutine costs to functions calls [no]
215    --tree=none|caller|   print for each function their callers,
216           calling|both   the called functions or both [none]
217    -I --include=<dir>    add <dir> to list of directories to search for 
218                          source files
219
220END
221;
222
223# Used in various places of output.
224my $fancy = '-' x 80 . "\n";
225
226#-----------------------------------------------------------------------------
227# Argument and option handling
228#-----------------------------------------------------------------------------
229sub process_cmd_line() 
230{
231    for my $arg (@ARGV) { 
232
233        # Option handling
234        if ($arg =~ /^-/) {
235
236            # --version
237            if ($arg =~ /^--version$/) {
238                die("callgrind_annotate-$version\n");
239
240            # --show=A,B,C
241            } elsif ($arg =~ /^--show=(.*)$/) {
242                @show_events = split(/,/, $1);
243
244            # --sort=A,B,C
245            } elsif ($arg =~ /^--sort=(.*)$/) {
246                @sort_events = split(/,/, $1);
247                my $th_specified = 0;
248                foreach my $i (0 .. scalar @sort_events - 1) {
249                    if ($sort_events[$i] =~ /.*:([\d\.]+)%?$/) {
250                        my $th = $1;
251                        ($th >= 0 && $th <= 100) or die($usage);
252                        $sort_events[$i] =~ s/:.*//;
253                        $thresholds[$i] = $th;
254                        $th_specified = 1;
255                    } else {
256                        $thresholds[$i] = 0;
257                    }
258                }
259                if (not $th_specified) {
260                    @thresholds = ();
261                }
262
263            # --threshold=X (tolerates a trailing '%')
264            } elsif ($arg =~ /^--threshold=([\d\.]+)%?$/) {
265                $single_threshold = $1;
266                ($1 >= 0 && $1 <= 100) or die($usage);
267
268            # --auto=yes|no
269            } elsif ($arg =~ /^--auto=(yes|no)$/) {
270                $auto_annotate = 1 if ($1 eq "yes");
271                $auto_annotate = 0 if ($1 eq "no");
272
273            # --context=N
274            } elsif ($arg =~ /^--context=([\d\.]+)$/) {
275                $context = $1;
276                if ($context < 0) {
277                    die($usage);
278                }
279
280            # --inclusive=yes|no
281            } elsif ($arg =~ /^--inclusive=(yes|no)$/) {
282                $inclusive = 1 if ($1 eq "yes");
283                $inclusive = 0 if ($1 eq "no");
284
285            # --tree=none|caller|calling|both
286            } elsif ($arg =~ /^--tree=(none|caller|calling|both)$/) {
287                $tree_caller  = 1 if ($1 eq "caller" || $1 eq "both");
288                $tree_calling = 1 if ($1 eq "calling" || $1 eq "both");
289
290            # --include=A,B,C
291            } elsif ($arg =~ /^(-I|--include)=(.*)$/) {
292                my $inc = $2;
293                $inc =~ s|/$||;         # trim trailing '/'
294                push(@include_dirs, "$inc/");
295
296            } else {            # -h and --help fall under this case
297                die($usage);
298            }
299
300        # Argument handling -- annotation file checking and selection.
301        # Stick filenames into a hash for quick 'n easy lookup throughout
302        } else {
303	  if ($input_file eq "") {
304	    $input_file = $arg;
305	  }
306	  else {
307            my $readable = 0;
308            foreach my $include_dir (@include_dirs) {
309                if (-r $include_dir . $arg) {
310                    $readable = 1;
311                }
312            }
313            $readable or die("File $arg not found in any of: @include_dirs\n");
314            $user_ann_files{$arg} = 1;
315        } 
316    }
317    }
318
319    if ($input_file eq "") {
320      $input_file = (<callgrind.out*>)[0];
321      if (!defined $input_file) {
322	  $input_file = (<cachegrind.out*>)[0];
323      }
324
325      (defined $input_file) or die($usage);
326      print "Reading data from '$input_file'...\n";
327    }
328}
329
330#-----------------------------------------------------------------------------
331# Reading of input file
332#-----------------------------------------------------------------------------
333sub max ($$) 
334{
335    my ($x, $y) = @_;
336    return ($x > $y ? $x : $y);
337}
338
339# Add the two arrays;  any '.' entries are ignored.  Two tricky things:
340# 1. If $a2->[$i] is undefined, it defaults to 0 which is what we want; we turn
341#    off warnings to allow this.  This makes things about 10% faster than
342#    checking for definedness ourselves.
343# 2. We don't add an undefined count or a ".", even though it's value is 0,
344#    because we don't want to make an $a2->[$i] that is undef become 0
345#    unnecessarily.
346sub add_array_a_to_b ($$) 
347{
348    my ($a1, $a2) = @_;
349
350    my $n = max(scalar @$a1, scalar @$a2);
351    $^W = 0;
352    foreach my $i (0 .. $n-1) {
353        $a2->[$i] += $a1->[$i] if (defined $a1->[$i] && "." ne $a1->[$i]);
354    }
355    $^W = 1;
356}
357
358# Is this a line with all events zero?
359sub is_zero ($)
360{
361    my ($CC) = @_;
362    my $isZero = 1;
363    foreach my $i (0 .. (scalar @$CC)-1) {
364	$isZero = 0 if ($CC->[$i] >0);
365    }
366    return $isZero;
367}
368
369# Add each event count to the CC array.  '.' counts become undef, as do
370# missing entries (implicitly).
371sub line_to_CC ($)
372{
373    my @CC = (split /\s+/, $_[0]);
374    (@CC <= @events) or die("Line $.: too many event counts\n");
375    return \@CC;
376}
377
378sub uncompressed_name($$)
379{
380   my ($context, $name) = @_;
381
382   if ($name =~ /^\((\d+)\)\s*(.*)$/) {
383     my $index = $1;
384     my $realname = $2;
385
386     if ($realname eq "") {
387       $realname = $compressed{$context,$index};
388     }
389     else {
390       $compressed{$context,$index} = $realname;
391     }
392     return $realname;
393   }
394   return $name;
395}
396
397sub read_input_file() 
398{
399    open(INPUTFILE, "< $input_file") || die "File $input_file not opened\n";
400
401    my $line;
402
403    # Read header
404    while(<INPUTFILE>) {
405
406      # remove comments
407      s/#.*$//;
408
409      if (/^$/) { ; }
410
411      elsif (/^version:\s*(\d+)/) {
412	# Can't read format with major version > 1
413	($1<2) or die("Can't read format with major version $1.\n");
414      }
415
416      elsif (/^pid:\s+(.*)$/) { $pid = $1;  }
417      elsif (/^thread:\s+(.*)$/) { $thread = $1;  }
418      elsif (/^part:\s+(.*)$/) { $part = $1;  }
419      elsif (/^desc:\s+(.*)$/) {
420	my $dline = $1;
421	# suppress profile options in description output
422	if ($dline =~ /^Option:/) {;}
423	else { $desc .= "$dline\n"; }
424      }
425      elsif (/^cmd:\s+(.*)$/)  { $cmd = $1; }
426      elsif (/^creator:\s+(.*)$/)  { $creator = $1; }
427      elsif (/^positions:\s+(.*)$/) {
428	my $positions = $1;
429	$has_line = ($positions =~ /line/);
430	$has_addr = ($positions =~ /(addr|instr)/);
431      }
432      elsif (/^events:\s+(.*)$/) {
433	$events = $1;
434	
435	# events line is last in header
436	last;
437      }
438      else {
439	warn("WARNING: header line $. malformed, ignoring\n");
440	if ($verbose) { chomp; warn("    line: '$_'\n"); }
441      }
442    }
443
444    # Read "events:" line.  We make a temporary hash in which the Nth event's
445    # value is N, which is useful for handling --show/--sort options below.
446    ($events ne "") or die("Line $.: missing events line\n");
447    @events = split(/\s+/, $events);
448    my %events;
449    my $n = 0;
450    foreach my $event (@events) {
451        $events{$event} = $n;
452        $n++
453    }
454
455    # If no --show arg give, default to showing all events in the file.
456    # If --show option is used, check all specified events appeared in the
457    # "events:" line.  Then initialise @show_order.
458    if (@show_events) {
459        foreach my $show_event (@show_events) {
460            (defined $events{$show_event}) or 
461                die("--show event `$show_event' did not appear in input\n");
462        }
463    } else {
464        @show_events = @events;
465    }
466    foreach my $show_event (@show_events) {
467        push(@show_order, $events{$show_event});
468    }
469
470    # Do as for --show, but if no --sort arg given, default to sorting by
471    # column order (ie. first column event is primary sort key, 2nd column is
472    # 2ndary key, etc).
473    if (@sort_events) {
474        foreach my $sort_event (@sort_events) {
475            (defined $events{$sort_event}) or 
476                die("--sort event `$sort_event' did not appear in input\n");
477        }
478    } else {
479        @sort_events = @events;
480    }
481    foreach my $sort_event (@sort_events) {
482        push(@sort_order, $events{$sort_event});
483    }
484
485    # If multiple threshold args weren't given via --sort, stick in the single
486    # threshold (either from --threshold if used, or the default otherwise) for
487    # the primary sort event, and 0% for the rest.
488    if (not @thresholds) {
489        foreach my $e (@sort_order) {
490            push(@thresholds, 0);
491        }
492        $thresholds[0] = $single_threshold;
493    }
494
495    # Current directory, used to strip from file names if absolute
496    my $pwd = `pwd`;
497    chomp $pwd;
498    $pwd .= '/';
499
500    my $curr_obj = "";
501    my $curr_file;
502    my $curr_fn;
503    my $curr_name;
504    my $curr_line_num = 0;
505    my $prev_line_num = 0;
506
507    my $curr_cobj = "";
508    my $curr_cfile = "";
509    my $curr_cfunc = "";
510    my $curr_cname;
511    my $curr_call_counter = 0;
512    my $curr_cfn_CC = [];
513
514    my $curr_fn_CC = [];
515    my $curr_file_ind_CCs = {};     # hash(line_num => CC)
516
517    # Read body of input file.
518    while (<INPUTFILE>) {
519	$prev_line_num = $curr_line_num;
520
521        s/#.*$//;   # remove comments
522        s/^\+(\d+)/$prev_line_num+$1/e;
523        s/^\-(\d+)/$prev_line_num-$1/e;
524        s/^\*/$prev_line_num/e;
525        if (s/^(-?\d+|0x\w+)\s+//) {
526            $curr_line_num = $1;
527	    if ($has_addr) {
528	      if ($has_line) {
529                s/^\+(\d+)/$prev_line_num+$1/e;
530	        s/^\-(\d+)/$prev_line_num-$1/e;
531                s/^\*/$prev_line_num/e;
532
533	        if (s/^(\d+)\s+//) { $curr_line_num = $1; }
534	      }
535	      else { $curr_line_num = 0; }
536	    }
537            my $CC = line_to_CC($_);
538
539	    if ($curr_call_counter>0) {
540#	      print "Read ($curr_name => $curr_cname) $curr_call_counter\n";
541
542	      if (!defined $call_CCs{$curr_name,$curr_cname}) {
543		$call_CCs{$curr_name,$curr_cname} = [];
544		$call_counter{$curr_name,$curr_cname} = 0;
545	      }
546	      add_array_a_to_b($CC, $call_CCs{$curr_name,$curr_cname});
547	      $call_counter{$curr_name,$curr_cname} += $curr_call_counter;
548
549	      my $tmp = $called_from_line->{$curr_file,$curr_line_num};
550	      if (!defined $tmp) {
551		$func_of_line{$curr_file,$curr_line_num} = $curr_name;
552	      }
553	      $tmp = {} unless defined $tmp;
554	      $$tmp{$curr_cname} = 1;
555	      $called_from_line->{$curr_file,$curr_line_num} = $tmp;
556	      if (!defined $call_CCs{$curr_name,$curr_cname,$curr_line_num}) {
557		$call_CCs{$curr_name,$curr_cname,$curr_line_num} = [];
558		$call_counter{$curr_name,$curr_cname,$curr_line_num} = 0;
559	      }
560	      add_array_a_to_b($CC, $call_CCs{$curr_name,$curr_cname,$curr_line_num});
561	      $call_counter{$curr_name,$curr_cname,$curr_line_num} += $curr_call_counter;
562
563	      $curr_call_counter = 0;
564
565	      # inclusive costs
566	      $curr_cfn_CC = $cfn_totals{$curr_cname};
567	      $curr_cfn_CC = [] unless (defined $curr_cfn_CC);
568	      add_array_a_to_b($CC, $curr_cfn_CC);
569	      $cfn_totals{$curr_cname} = $curr_cfn_CC;
570
571	      if ($inclusive) {
572		add_array_a_to_b($CC, $curr_fn_CC);
573	      }
574	      next;
575	    }
576
577            add_array_a_to_b($CC, $curr_fn_CC);
578
579            # If curr_file is selected, add CC to curr_file list.  We look for
580            # full filename matches;  or, if auto-annotating, we have to
581            # remember everything -- we won't know until the end what's needed.
582            if ($auto_annotate || defined $user_ann_files{$curr_file}) {
583                my $tmp = $curr_file_ind_CCs->{$curr_line_num};
584                $tmp = [] unless defined $tmp;
585                add_array_a_to_b($CC, $tmp);
586                $curr_file_ind_CCs->{$curr_line_num} = $tmp;
587            }
588
589        } elsif (s/^fn=(.*)$//) {
590            # Commit result from previous function
591            $fn_totals{$curr_name} = $curr_fn_CC if (defined $curr_name);
592
593            # Setup new one
594            $curr_fn = uncompressed_name("fn",$1);
595            $curr_name = "$curr_file:$curr_fn";
596	    $obj_name{$curr_name} = $curr_obj;
597            $curr_fn_CC = $fn_totals{$curr_name};
598            $curr_fn_CC = [] unless (defined $curr_fn_CC);
599
600        } elsif (s/^ob=(.*)$//) {
601            $curr_obj = uncompressed_name("ob",$1);
602
603        } elsif (s/^fl=(.*)$//) {
604            $all_ind_CCs{$curr_file} = $curr_file_ind_CCs 
605                if (defined $curr_file);
606
607            $curr_file = uncompressed_name("fl",$1);
608            $curr_file =~ s/^\Q$pwd\E//;
609            $curr_file_ind_CCs = $all_ind_CCs{$curr_file};
610            $curr_file_ind_CCs = {} unless (defined $curr_file_ind_CCs);
611
612        } elsif (s/^(fi|fe)=(.*)$//) {
613            (defined $curr_name) or die("Line $.: Unexpected fi/fe line\n");
614            $fn_totals{$curr_name} = $curr_fn_CC;
615            $all_ind_CCs{$curr_file} = $curr_file_ind_CCs;
616
617            $curr_file = uncompressed_name("fl",$2);
618            $curr_file =~ s/^\Q$pwd\E//;
619            $curr_name = "$curr_file:$curr_fn";
620            $curr_file_ind_CCs = $all_ind_CCs{$curr_file};
621            $curr_file_ind_CCs = {} unless (defined $curr_file_ind_CCs);
622            $curr_fn_CC = $fn_totals{$curr_name};
623            $curr_fn_CC = [] unless (defined $curr_fn_CC);
624
625        } elsif (s/^\s*$//) {
626            # blank, do nothing
627
628        } elsif (s/^cob=(.*)$//) {
629	  $curr_cobj = uncompressed_name("ob",$1);
630
631	} elsif (s/^cf[il]=(.*)$//) {
632	  $curr_cfile = uncompressed_name("fl",$1);
633
634	} elsif (s/^cfn=(.*)$//) {
635	  $curr_cfunc = uncompressed_name("fn",$1);
636	  if ($curr_cfile eq "") {
637	    $curr_cname = "$curr_file:$curr_cfunc";
638	  }
639	  else {
640	    $curr_cname = "$curr_cfile:$curr_cfunc";
641	    $curr_cfile = "";
642	  }
643
644	  my $tmp = $calling_funcs->{$curr_cname};
645	  $tmp = {} unless defined $tmp;
646	  $$tmp{$curr_name} = 1;
647	  $calling_funcs->{$curr_cname} = $tmp;
648		
649	  my $tmp2 = $called_funcs->{$curr_name};
650	  $tmp2 = {} unless defined $tmp2;
651	  $$tmp2{$curr_cname} = 1;
652	  $called_funcs->{$curr_name} = $tmp2;
653
654	} elsif (s/^calls=(\d+)//) {
655	  $curr_call_counter = $1;
656
657        } elsif (s/^(jump|jcnd)=//) {
658	  #ignore jump information
659
660        } elsif (s/^jfi=(.*)$//) {
661          # side effect needed: possibly add compression mapping 
662          uncompressed_name("fl",$1);
663          # ignore jump information	
664
665        } elsif (s/^jfn=(.*)$//) {
666          # side effect needed: possibly add compression mapping
667          uncompressed_name("fn",$1);
668          # ignore jump information
669
670        } elsif (s/^totals:\s+//) {
671	    $totals_CC = line_to_CC($_);
672
673        } elsif (s/^summary:\s+//) {
674            $summary_CC = line_to_CC($_);
675
676        } else {
677            warn("WARNING: line $. malformed, ignoring\n");
678	    if ($verbose) { chomp; warn("    line: '$_'\n"); }
679        }
680    }
681
682    # Finish up handling final filename/fn_name counts
683    $fn_totals{"$curr_file:$curr_fn"} = $curr_fn_CC
684	if (defined $curr_file && defined $curr_fn);
685    $all_ind_CCs{$curr_file} =
686	$curr_file_ind_CCs if (defined $curr_file);
687
688    # Correct inclusive totals
689    if ($inclusive) {
690      foreach my $name (keys %cfn_totals) {
691	$fn_totals{$name} = $cfn_totals{$name};
692      }
693    }
694
695    close(INPUTFILE);
696
697    if ((not defined $summary_CC) || is_zero($summary_CC)) {
698	$summary_CC = $totals_CC;
699
700	# if neither 'summary:' nor 'totals:' line is given,
701	# calculate summary from fn_totals hash
702	if ((not defined $summary_CC) || is_zero($summary_CC)) {
703	    $summary_calculated = 1;
704	    $summary_CC = [];
705	    foreach my $name (keys %fn_totals) {
706		add_array_a_to_b($fn_totals{$name}, $summary_CC);
707	    }
708	}
709    }
710}
711
712#-----------------------------------------------------------------------------
713# Print options used
714#-----------------------------------------------------------------------------
715sub print_options ()
716{
717    print($fancy);
718    print "Profile data file '$input_file'";
719    if ($creator ne "") { print " (creator: $creator)"; }
720    print "\n";
721
722    print($fancy);
723    print($desc);
724    my $target = $cmd;
725    if ($target eq "") { $target = "(unknown)"; }
726    if ($pid ne "") {
727      $target .= " (PID $pid";
728      if ($part ne "") { $target .= ", part $part"; }
729      if ($thread ne "") { $target .= ", thread $thread"; }
730      $target .= ")";
731    }
732    print("Profiled target:  $target\n");
733    print("Events recorded:  @events\n");
734    print("Events shown:     @show_events\n");
735    print("Event sort order: @sort_events\n");
736    print("Thresholds:       @thresholds\n");
737
738    my @include_dirs2 = @include_dirs;  # copy @include_dirs
739    shift(@include_dirs2);       # remove "" entry, which is always the first
740    unshift(@include_dirs2, "") if (0 == @include_dirs2); 
741    my $include_dir = shift(@include_dirs2);
742    print("Include dirs:     $include_dir\n");
743    foreach my $include_dir (@include_dirs2) {
744        print("                  $include_dir\n");
745    }
746
747    my @user_ann_files = keys %user_ann_files;
748    unshift(@user_ann_files, "") if (0 == @user_ann_files); 
749    my $user_ann_file = shift(@user_ann_files);
750    print("User annotated:   $user_ann_file\n");
751    foreach $user_ann_file (@user_ann_files) {
752        print("                  $user_ann_file\n");
753    }
754
755    my $is_on = ($auto_annotate ? "on" : "off");
756    print("Auto-annotation:  $is_on\n");
757    print("\n");
758}
759
760#-----------------------------------------------------------------------------
761# Print summary and sorted function totals
762#-----------------------------------------------------------------------------
763sub mycmp ($$) 
764{
765    my ($c, $d) = @_;
766
767    # Iterate through sort events (eg. 3,2); return result if two are different
768    foreach my $i (@sort_order) {
769        my ($x, $y);
770        $x = $c->[$i];
771        $y = $d->[$i];
772        $x = -1 unless defined $x;
773        $y = -1 unless defined $y;
774
775        my $cmp = $y <=> $x;        # reverse sort
776        if (0 != $cmp) {
777            return $cmp;
778        }
779    }
780    # Exhausted events, equal
781    return 0;
782}
783
784sub commify ($) {
785    my ($val) = @_;
786    1 while ($val =~ s/^(\d+)(\d{3})/$1,$2/);
787    return $val;
788}
789
790# Because the counts can get very big, and we don't want to waste screen space
791# and make lines too long, we compute exactly how wide each column needs to be
792# by finding the widest entry for each one.
793sub compute_CC_col_widths (@) 
794{
795    my @CCs = @_;
796    my $CC_col_widths = [];
797
798    # Initialise with minimum widths (from event names)
799    foreach my $event (@events) {
800        push(@$CC_col_widths, length($event));
801    }
802    
803    # Find maximum width count for each column.  @CC_col_width positions
804    # correspond to @CC positions.
805    foreach my $CC (@CCs) {
806        foreach my $i (0 .. scalar(@$CC)-1) {
807            if (defined $CC->[$i]) {
808                # Find length, accounting for commas that will be added
809                my $length = length $CC->[$i];
810                my $clength = $length + int(($length - 1) / 3);
811                $CC_col_widths->[$i] = max($CC_col_widths->[$i], $clength); 
812            }
813        }
814    }
815    return $CC_col_widths;
816}
817
818# Print the CC with each column's size dictated by $CC_col_widths.
819sub print_CC ($$) 
820{
821    my ($CC, $CC_col_widths) = @_;
822
823    foreach my $i (@show_order) {
824        my $count = (defined $CC->[$i] ? commify($CC->[$i]) : ".");
825        my $space = ' ' x ($CC_col_widths->[$i] - length($count));
826        print("$space$count ");
827    }
828}
829
830sub print_events ($)
831{
832    my ($CC_col_widths) = @_;
833
834    foreach my $i (@show_order) { 
835        my $event       = $events[$i];
836        my $event_width = length($event);
837        my $col_width   = $CC_col_widths->[$i];
838        my $space       = ' ' x ($col_width - $event_width);
839        print("$space$event ");
840    }
841}
842
843# Prints summary and function totals (with separate column widths, so that
844# function names aren't pushed over unnecessarily by huge summary figures).
845# Also returns a hash containing all the files that are involved in getting the
846# events count above the thresholds (ie. all the interesting ones).
847sub print_summary_and_fn_totals ()
848{
849    my @fn_fullnames = keys   %fn_totals;
850
851    # Work out the size of each column for printing (summary and functions
852    # separately).
853    my $summary_CC_col_widths = compute_CC_col_widths($summary_CC);
854    my      $fn_CC_col_widths = compute_CC_col_widths(values %fn_totals);
855
856    # Header and counts for summary
857    print($fancy);
858    print_events($summary_CC_col_widths);
859    print("\n");
860    print($fancy);
861    print_CC($summary_CC, $summary_CC_col_widths);
862    print(" PROGRAM TOTALS");
863    if ($summary_calculated) {
864	print(" (calculated)");
865    }
866    print("\n\n");
867
868    # Header for functions
869    print($fancy);
870    print_events($fn_CC_col_widths);
871    print(" file:function\n");
872    print($fancy);
873
874    # Sort function names into order dictated by --sort option.
875    @fn_fullnames = sort {
876        mycmp($fn_totals{$a}, $fn_totals{$b})
877    } @fn_fullnames;
878
879
880    # Assertion
881    (scalar @sort_order == scalar @thresholds) or 
882        die("sort_order length != thresholds length:\n",
883            "  @sort_order\n  @thresholds\n");
884
885    my $threshold_files       = {};
886    # @curr_totals has the same shape as @sort_order and @thresholds
887    my @curr_totals = ();
888    foreach my $e (@thresholds) {
889        push(@curr_totals, 0);
890    }
891
892    # Print functions, stopping when the threshold has been reached.
893    foreach my $fn_name (@fn_fullnames) {
894
895        # Stop when we've reached all the thresholds
896        my $reached_all_thresholds = 1;
897        foreach my $i (0 .. scalar @thresholds - 1) {
898            my $prop = $curr_totals[$i] * 100;
899	    if ($summary_CC->[$sort_order[$i]] >0) {
900	      $prop = $prop / $summary_CC->[$sort_order[$i]];
901	    }
902            $reached_all_thresholds &&= ($prop >= $thresholds[$i]);
903        }
904        last if $reached_all_thresholds;
905
906	if ($tree_caller || $tree_calling) { print "\n"; }
907
908	if ($tree_caller && ($fn_name ne "???:???")) {
909	  # Print function callers
910	  my $tmp1 = $calling_funcs->{$fn_name};
911	  if (defined $tmp1) {
912	    foreach my $calling (keys %$tmp1) {
913	      if (defined $call_counter{$calling,$fn_name}) {
914		print_CC($call_CCs{$calling,$fn_name}, $fn_CC_col_widths);
915		print" < $calling (";
916		print $call_counter{$calling,$fn_name} . "x)";
917		if (defined $obj_name{$calling}) {
918		  print " [$obj_name{$calling}]";
919		}
920		print "\n";
921	      }
922	    }
923	  }
924	}
925
926        # Print function results
927        my $fn_CC = $fn_totals{$fn_name};
928        print_CC($fn_CC, $fn_CC_col_widths);
929	if ($tree_caller || $tree_calling) { print " * "; }
930        print(" $fn_name");
931	if ((defined $obj_name{$fn_name}) &&
932	    ($obj_name{$fn_name} ne "")) {
933	  print " [$obj_name{$fn_name}]";
934	}
935	print "\n";
936
937	if ($tree_calling && ($fn_name ne "???:???")) {
938	  # Print called functions
939	  my $tmp2 = $called_funcs->{$fn_name};
940	  if (defined $tmp2) {
941	    foreach my $called (keys %$tmp2) {
942	      if (defined $call_counter{$fn_name,$called}) {
943		print_CC($call_CCs{$fn_name,$called}, $fn_CC_col_widths);
944		print" >   $called (";
945		print $call_counter{$fn_name,$called} . "x)";
946		if (defined $obj_name{$called}) {
947		  print " [$obj_name{$called}]";
948		}
949		print "\n";
950	      }
951	    }
952	  }
953	}
954
955        # Update the threshold counts
956        my $filename = $fn_name;
957        $filename =~ s/:.+$//;    # remove function name
958        $threshold_files->{$filename} = 1;
959        foreach my $i (0 .. scalar @sort_order - 1) {
960	  if ($inclusive) {
961	    $curr_totals[$i] = $summary_CC->[$sort_order[$i]] -
962                               $fn_CC->[$sort_order[$i]]
963	      if (defined $fn_CC->[$sort_order[$i]]);
964	  } else {
965            $curr_totals[$i] += $fn_CC->[$sort_order[$i]] 
966                if (defined $fn_CC->[$sort_order[$i]]);
967        }
968    }
969    }
970    print("\n");
971
972    return $threshold_files;
973}
974
975#-----------------------------------------------------------------------------
976# Annotate selected files
977#-----------------------------------------------------------------------------
978
979# Issue a warning that the source file is more recent than the input file. 
980sub warning_on_src_more_recent_than_inputfile ($)
981{
982    my $src_file = $_[0];
983
984    my $warning = <<END
985@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
986@@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@
987@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
988@ Source file '$src_file' is more recent than input file '$input_file'.
989@ Annotations may not be correct.
990@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
991
992END
993;
994    print($warning);
995}
996
997# If there is information about lines not in the file, issue a warning
998# explaining possible causes.
999sub warning_on_nonexistent_lines ($$$)
1000{
1001    my ($src_more_recent_than_inputfile, $src_file, $excess_line_nums) = @_;
1002    my $cause_and_solution;
1003
1004    if ($src_more_recent_than_inputfile) {
1005        $cause_and_solution = <<END
1006@@ cause:    '$src_file' has changed since information was gathered.
1007@@           If so, a warning will have already been issued about this.
1008@@ solution: Recompile program and rerun under "valgrind --cachesim=yes" to 
1009@@           gather new information.
1010END
1011    # We suppress warnings about .h files
1012    } elsif ($src_file =~ /\.h$/) {
1013        $cause_and_solution = <<END
1014@@ cause:    bug in the Valgrind's debug info reader that screws up with .h
1015@@           files sometimes
1016@@ solution: none, sorry
1017END
1018    } else {
1019        $cause_and_solution = <<END
1020@@ cause:    not sure, sorry
1021END
1022    }
1023
1024    my $warning = <<END
1025@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1026@@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@ WARNING @@
1027@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1028@@
1029@@ Information recorded about lines past the end of '$src_file'.
1030@@
1031@@ Probable cause and solution:
1032$cause_and_solution@@
1033@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
1034END
1035;
1036    print($warning);
1037}
1038
1039sub annotate_ann_files($)
1040{
1041    my ($threshold_files) = @_; 
1042
1043    my %all_ann_files;
1044    my @unfound_auto_annotate_files;
1045    my $printed_totals_CC = [];
1046
1047    # If auto-annotating, add interesting files (but not "???")
1048    if ($auto_annotate) {
1049        delete $threshold_files->{"???"};
1050        %all_ann_files = (%user_ann_files, %$threshold_files) 
1051    } else {
1052        %all_ann_files = %user_ann_files;
1053    }
1054
1055    # Track if we did any annotations.
1056    my $did_annotations = 0;
1057
1058    LOOP:
1059    foreach my $src_file (keys %all_ann_files) {
1060
1061        my $opened_file = "";
1062        my $full_file_name = "";
1063        foreach my $include_dir (@include_dirs) {
1064            my $try_name = $include_dir . $src_file;
1065            if (open(INPUTFILE, "< $try_name")) {
1066                $opened_file    = $try_name;
1067                $full_file_name = ($include_dir eq "" 
1068                                  ? $src_file 
1069                                  : "$include_dir + $src_file"); 
1070                last;
1071            }
1072        }
1073        
1074        if (not $opened_file) {
1075            # Failed to open the file.  If chosen on the command line, die.
1076            # If arose from auto-annotation, print a little message.
1077            if (defined $user_ann_files{$src_file}) {
1078                die("File $src_file not opened in any of: @include_dirs\n");
1079
1080            } else {
1081                push(@unfound_auto_annotate_files, $src_file);
1082            }
1083
1084        } else {
1085            # File header (distinguish between user- and auto-selected files).
1086            print("$fancy");
1087            my $ann_type = 
1088                (defined $user_ann_files{$src_file} ? "User" : "Auto");
1089            print("-- $ann_type-annotated source: $full_file_name\n");
1090            print("$fancy");
1091
1092            # Get file's CCs
1093            my $src_file_CCs = $all_ind_CCs{$src_file};
1094            if (!defined $src_file_CCs) {
1095                print("  No information has been collected for $src_file\n\n");
1096                next LOOP;
1097            }
1098        
1099            $did_annotations = 1;
1100            
1101            # Numeric, not lexicographic sort!
1102            my @line_nums = sort {$a <=> $b} keys %$src_file_CCs;  
1103
1104            # If $src_file more recent than cachegrind.out, issue warning
1105            my $src_more_recent_than_inputfile = 0;
1106            if ((stat $opened_file)[9] > (stat $input_file)[9]) {
1107                $src_more_recent_than_inputfile = 1;
1108                warning_on_src_more_recent_than_inputfile($src_file);
1109            }
1110
1111            # Work out the size of each column for printing
1112            my $CC_col_widths = compute_CC_col_widths(values %$src_file_CCs);
1113
1114            # Events header
1115            print_events($CC_col_widths);
1116            print("\n\n");
1117
1118            # Shift out 0 if it's in the line numbers (from unknown entries,
1119            # likely due to bugs in Valgrind's stabs debug info reader)
1120            shift(@line_nums) if (0 == $line_nums[0]);
1121
1122            # Finds interesting line ranges -- all lines with a CC, and all
1123            # lines within $context lines of a line with a CC.
1124            my $n = @line_nums;
1125            my @pairs;
1126            for (my $i = 0; $i < $n; $i++) {
1127                push(@pairs, $line_nums[$i] - $context);   # lower marker
1128                while ($i < $n-1 && 
1129                       $line_nums[$i] + 2*$context >= $line_nums[$i+1]) {
1130                    $i++;
1131                }
1132                push(@pairs, $line_nums[$i] + $context);   # upper marker
1133            }
1134
1135            # Annotate chosen lines, tracking total counts of lines printed
1136            $pairs[0] = 1 if ($pairs[0] < 1);
1137            while (@pairs) {
1138                my $low  = shift @pairs;
1139                my $high = shift @pairs;
1140                while ($. < $low-1) {
1141                    my $tmp = <INPUTFILE>;
1142                    last unless (defined $tmp);     # hack to detect EOF
1143                }
1144                my $src_line;
1145                # Print line number, unless start of file
1146                print("-- line $low " . '-' x 40 . "\n") if ($low != 1);
1147                while (($. < $high) && ($src_line = <INPUTFILE>)) {
1148                    if (defined $line_nums[0] && $. == $line_nums[0]) {
1149                        print_CC($src_file_CCs->{$.}, $CC_col_widths);
1150                        add_array_a_to_b($src_file_CCs->{$.}, 
1151                                         $printed_totals_CC);
1152                        shift(@line_nums);
1153
1154                    } else {
1155                        print_CC( [], $CC_col_widths);
1156                    }
1157
1158                    print(" $src_line");
1159
1160		    my $tmp  = $called_from_line->{$src_file,$.};
1161		    my $func = $func_of_line{$src_file,$.};
1162		    if (defined $tmp) {
1163		      foreach my $called (keys %$tmp) {
1164			if (defined $call_CCs{$func,$called,$.}) {
1165			  print_CC($call_CCs{$func,$called,$.}, $CC_col_widths);
1166			  print " => $called (";
1167			  print $call_counter{$func,$called,$.} . "x)\n";
1168			}
1169		      }
1170		    }
1171                }
1172                # Print line number, unless EOF
1173                if ($src_line) {
1174                    print("-- line $high " . '-' x 40 . "\n");
1175                } else {
1176                    last;
1177                }
1178            }
1179
1180            # If there was info on lines past the end of the file...
1181            if (@line_nums) {
1182                foreach my $line_num (@line_nums) {
1183                    print_CC($src_file_CCs->{$line_num}, $CC_col_widths);
1184                    print(" <bogus line $line_num>\n");
1185                }
1186                print("\n");
1187                warning_on_nonexistent_lines($src_more_recent_than_inputfile,
1188                                             $src_file, \@line_nums);
1189            }
1190            print("\n");
1191
1192            # Print summary of counts attributed to file but not to any
1193            # particular line (due to incomplete debug info).
1194            if ($src_file_CCs->{0}) {
1195                print_CC($src_file_CCs->{0}, $CC_col_widths);
1196                print(" <counts for unidentified lines in $src_file>\n\n");
1197            }
1198            
1199            close(INPUTFILE);
1200        }
1201    }
1202
1203    # Print list of unfound auto-annotate selected files.
1204    if (@unfound_auto_annotate_files) {
1205        print("$fancy");
1206        print("The following files chosen for auto-annotation could not be found:\n");
1207        print($fancy);
1208        foreach my $f (@unfound_auto_annotate_files) {
1209            print("  $f\n");
1210        }
1211        print("\n");
1212    }
1213
1214    # If we did any annotating, print what proportion of events were covered by
1215    # annotated lines above.
1216    if ($did_annotations) {
1217        my $percent_printed_CC;
1218        foreach (my $i = 0; $i < @$summary_CC; $i++) {
1219            $percent_printed_CC->[$i] = 
1220                sprintf("%.0f", 
1221                        $printed_totals_CC->[$i] / $summary_CC->[$i] * 100);
1222        }
1223        my $pp_CC_col_widths = compute_CC_col_widths($percent_printed_CC);
1224        print($fancy);
1225        print_events($pp_CC_col_widths);
1226        print("\n");
1227        print($fancy);
1228        print_CC($percent_printed_CC, $pp_CC_col_widths);
1229        print(" percentage of events annotated\n\n");
1230    }
1231}
1232
1233#----------------------------------------------------------------------------
1234# "main()"
1235#----------------------------------------------------------------------------
1236process_cmd_line();
1237read_input_file();
1238print_options();
1239my $threshold_files = print_summary_and_fn_totals();
1240annotate_ann_files($threshold_files);
1241
1242##--------------------------------------------------------------------##
1243##--- end                                           vg_annotate.in ---##
1244##--------------------------------------------------------------------##
1245
1246
1247