Server IP : 104.21.87.198 / Your IP : 172.70.188.20
Web Server : Apache/2.2.15 (CentOS)
System : Linux GA 2.6.32-431.1.2.0.1.el6.x86_64 #1 SMP Fri Dec 13 13:06:13 UTC 2013 x86_64
User : apache ( 48)
PHP Version : 5.6.38
Disable Function : NONE
MySQL : ON | cURL : ON | WGET : ON | Perl : ON | Python : ON | Sudo : ON | Pkexec : OFF
Directory : /usr/share/perl5/pod/ |
Upload File : |
| Current File : /usr/share/perl5/pod/perlperf.pod |
=head1 NAME
perlperf - Perl Performance and Optimization Techniques
=head1 DESCRIPTION
This is an introduction to the use of performance and optimization techniques
which can be used with particular reference to perl programs. While many perl
developers have come from other languages, and can use their prior knowledge
where appropriate, there are many other people who might benefit from a few
perl specific pointers. If you want the condensed version, perhaps the best
advice comes from the renowned Japanese Samurai, Miyamoto Musashi, who said:
"Do Not Engage in Useless Activity"
in 1645.
=head1 OVERVIEW
Perhaps the most common mistake programmers make is to attempt to optimize
their code before a program actually does anything useful - this is a bad idea.
There's no point in having an extremely fast program that doesn't work. The
first job is to get a program to I<correctly> do something B<useful>, (not to
mention ensuring the test suite is fully functional), and only then to consider
optimizing it. Having decided to optimize existing working code, there are
several simple but essential steps to consider which are intrinsic to any
optimization process.
=head2 ONE STEP SIDEWAYS
Firstly, you need to establish a baseline time for the existing code, which
timing needs to be reliable and repeatable. You'll probably want to use the
C<Benchmark> or C<Devel::DProf> modules, or something similar, for this step,
or perhaps the unix system C<time> utility, whichever is appropriate. See the
base of this document for a longer list of benchmarking and profiling modules,
and recommended further reading.
=head2 ONE STEP FORWARD
Next, having examined the program for I<hot spots>, (places where the code
seems to run slowly), change the code with the intention of making it run
faster. Using version control software, like C<subversion>, will ensure no
changes are irreversible. It's too easy to fiddle here and fiddle there -
don't change too much at any one time or you might not discover which piece of
code B<really> was the slow bit.
=head2 ANOTHER STEP SIDEWAYS
It's not enough to say: "that will make it run faster", you have to check it.
Rerun the code under control of the benchmarking or profiling modules, from the
first step above, and check that the new code executed the B<same task> in
I<less time>. Save your work and repeat...
=head1 GENERAL GUIDELINES
The critical thing when considering performance is to remember there is no such
thing as a C<Golden Bullet>, which is why there are no rules, only guidelines.
It is clear that inline code is going to be faster than subroutine or method
calls, because there is less overhead, but this approach has the disadvantage
of being less maintainable and comes at the cost of greater memory usage -
there is no such thing as a free lunch. If you are searching for an element in
a list, it can be more efficient to store the data in a hash structure, and
then simply look to see whether the key is defined, rather than to loop through
the entire array using grep() for instance. substr() may be (a lot) faster
than grep() but not as flexible, so you have another trade-off to access. Your
code may contain a line which takes 0.01 of a second to execute which if you
call it 1,000 times, quite likely in a program parsing even medium sized files
for instance, you already have a 10 second delay, in just one single code
location, and if you call that line 100,000 times, your entire program will
slow down to an unbearable crawl.
Using a subroutine as part of your sort is a powerful way to get exactly what
you want, but will usually be slower than the built-in I<alphabetic> C<cmp> and
I<numeric> C<E<lt>=E<gt>> sort operators. It is possible to make multiple
passes over your data, building indices to make the upcoming sort more
efficient, and to use what is known as the C<OM> (Orcish Maneuver) to cache the
sort keys in advance. The cache lookup, while a good idea, can itself be a
source of slowdown by enforcing a double pass over the data - once to setup the
cache, and once to sort the data. Using C<pack()> to extract the required sort
key into a consistent string can be an efficient way to build a single string
to compare, instead of using multiple sort keys, which makes it possible to use
the standard, written in C<c> and fast, perl C<sort()> function on the output,
and is the basis of the C<GRT> (Guttman Rossler Transform). Some string
combinations can slow the C<GRT> down, by just being too plain complex for it's
own good.
For applications using database backends, the standard C<DBIx> namespace has
tries to help with keeping things nippy, not least because it tries to I<not>
query the database until the latest possible moment, but always read the docs
which come with your choice of libraries. Among the many issues facing
developers dealing with databases should remain aware of is to always use
C<SQL> placeholders and to consider pre-fetching data sets when this might
prove advantageous. Splitting up a large file by assigning multiple processes
to parsing a single file, using say C<POE>, C<threads> or C<fork> can also be a
useful way of optimizing your usage of the available C<CPU> resources, though
this technique is fraught with concurrency issues and demands high attention to
detail.
Every case has a specific application and one or more exceptions, and there is
no replacement for running a few tests and finding out which method works best
for your particular environment, this is why writing optimal code is not an
exact science, and why we love using Perl so much - TMTOWTDI.
=head1 BENCHMARKS
Here are a few examples to demonstrate usage of Perl's benchmarking tools.
=head2 Assigning and Dereferencing Variables.
I'm sure most of us have seen code which looks like, (or worse than), this:
if ( $obj->{_ref}->{_myscore} >= $obj->{_ref}->{_yourscore} ) {
...
This sort of code can be a real eyesore to read, as well as being very
sensitive to typos, and it's much clearer to dereference the variable
explicitly. We're side-stepping the issue of working with object-oriented
programming techniques to encapsulate variable access via methods, only
accessible through an object. Here we're just discussing the technical
implementation of choice, and whether this has an effect on performance. We
can see whether this dereferencing operation, has any overhead by putting
comparative code in a file and running a C<Benchmark> test.
# dereference
#!/usr/bin/perl
use strict;
use warnings;
use Benchmark;
my $ref = {
'ref' => {
_myscore => '100 + 1',
_yourscore => '102 - 1',
},
};
timethese(1000000, {
'direct' => sub {
my $x = $ref->{ref}->{_myscore} . $ref->{ref}->{_yourscore} ;
},
'dereference' => sub {
my $ref = $ref->{ref};
my $myscore = $ref->{_myscore};
my $yourscore = $ref->{_yourscore};
my $x = $myscore . $yourscore;
},
});
It's essential to run any timing measurements a sufficient number of times so
the numbers settle on a numerical average, otherwise each run will naturally
fluctuate due to variations in the environment, to reduce the effect of
contention for C<CPU> resources and network bandwidth for instance. Running
the above code for one million iterations, we can take a look at the report
output by the C<Benchmark> module, to see which approach is the most effective.
$> perl dereference
Benchmark: timing 1000000 iterations of dereference, direct...
dereference: 2 wallclock secs ( 1.59 usr + 0.00 sys = 1.59 CPU) @ 628930.82/s (n=1000000)
direct: 1 wallclock secs ( 1.20 usr + 0.00 sys = 1.20 CPU) @ 833333.33/s (n=1000000)
The difference is clear to see and the dereferencing approach is slower. While
it managed to execute an average of 628,930 times a second during our test, the
direct approach managed to run an additional 204,403 times, unfortunately.
Unfortunately, because there are many examples of code written using the
multiple layer direct variable access, and it's usually horrible. It is,
however, miniscully faster. The question remains whether the minute gain is
actually worth the eyestrain, or the loss of maintainability.
=head2 Search and replace or tr
If we have a string which needs to be modified, while a regex will almost
always be much more flexible, C<tr>, an oft underused tool, can still be a
useful. One scenario might be replace all vowels with another character. The
regex solution might look like this:
$str =~ s/[aeiou]/x/g
The C<tr> alternative might look like this:
$str =~ tr/aeiou/xxxxx/
We can put that into a test file which we can run to check which approach is
the fastest, using a global C<$STR> variable to assign to the C<my $str>
variable so as to avoid perl trying to optimize any of the work away by
noticing it's assigned only the once.
# regex-transliterate
#!/usr/bin/perl
use strict;
use warnings;
use Benchmark;
my $STR = "$$-this and that";
timethese( 1000000, {
'sr' => sub { my $str = $STR; $str =~ s/[aeiou]/x/g; return $str; },
'tr' => sub { my $str = $STR; $str =~ tr/aeiou/xxxxx/; return $str; },
});
Running the code gives us our results:
$> perl regex-transliterate
Benchmark: timing 1000000 iterations of sr, tr...
sr: 2 wallclock secs ( 1.19 usr + 0.00 sys = 1.19 CPU) @ 840336.13/s (n=1000000)
tr: 0 wallclock secs ( 0.49 usr + 0.00 sys = 0.49 CPU) @ 2040816.33/s (n=1000000)
The C<tr> version is a clear winner. One solution is flexible, the other is
fast - and it's appropriately the programmers choice which to use in the
circumstances.
Check the C<Benchmark> docs for further useful techniques.
=head1 PROFILING TOOLS
A slightly larger piece of code will provide something on which a profiler can
produce more extensive reporting statistics. This example uses the simplistic
C<wordmatch> program which parses a given input file and spews out a short
report on the contents.
# wordmatch
#!/usr/bin/perl
use strict;
use warnings;
=head1 NAME
filewords - word analysis of input file
=head1 SYNOPSIS
filewords -f inputfilename [-d]
=head1 DESCRIPTION
This program parses the given filename, specified with C<-f>, and displays a
simple analysis of the words found therein. Use the C<-d> switch to enable
debugging messages.
=cut
use FileHandle;
use Getopt::Long;
my $debug = 0;
my $file = '';
my $result = GetOptions (
'debug' => \$debug,
'file=s' => \$file,
);
die("invalid args") unless $result;
unless ( -f $file ) {
die("Usage: $0 -f filename [-d]");
}
my $FH = FileHandle->new("< $file") or die("unable to open file($file): $!");
my $i_LINES = 0;
my $i_WORDS = 0;
my %count = ();
my @lines = <$FH>;
foreach my $line ( @lines ) {
$i_LINES++;
$line =~ s/\n//;
my @words = split(/ +/, $line);
my $i_words = scalar(@words);
$i_WORDS = $i_WORDS + $i_words;
debug("line: $i_LINES supplying $i_words words: @words");
my $i_word = 0;
foreach my $word ( @words ) {
$i_word++;
$count{$i_LINES}{spec} += matches($i_word, $word, '[^a-zA-Z0-9]');
$count{$i_LINES}{only} += matches($i_word, $word, '^[^a-zA-Z0-9]+$');
$count{$i_LINES}{cons} += matches($i_word, $word, '^[(?i:bcdfghjklmnpqrstvwxyz)]+$');
$count{$i_LINES}{vows} += matches($i_word, $word, '^[(?i:aeiou)]+$');
$count{$i_LINES}{caps} += matches($i_word, $word, '^[(A-Z)]+$');
}
}
print report( %count );
sub matches {
my $i_wd = shift;
my $word = shift;
my $regex = shift;
my $has = 0;
if ( $word =~ /($regex)/ ) {
$has++ if $1;
}
debug("word: $i_wd ".($has ? 'matches' : 'does not match')." chars: /$regex/");
return $has;
}
sub report {
my %report = @_;
my %rep;
foreach my $line ( keys %report ) {
foreach my $key ( keys %{ $report{$line} } ) {
$rep{$key} += $report{$line}{$key};
}
}
my $report = qq|
$0 report for $file:
lines in file: $i_LINES
words in file: $i_WORDS
words with special (non-word) characters: $i_spec
words with only special (non-word) characters: $i_only
words with only consonants: $i_cons
words with only capital letters: $i_caps
words with only vowels: $i_vows
|;
return $report;
}
sub debug {
my $message = shift;
if ( $debug ) {
print STDERR "DBG: $message\n";
}
}
exit 0;
=head2 Devel::DProf
This venerable module has been the de-facto standard for Perl code profiling
for more than a decade, but has been replaced by a number of other modules
which have brought us back to the 21st century. Although you're recommended to
evaluate your tool from the several mentioned here and from the CPAN list at
the base of this document, (and currently L<Devel::NYTProf> seems to be the
weapon of choice - see below), we'll take a quick look at the output from
L<Devel::DProf> first, to set a baseline for Perl profiling tools. Run the
above program under the control of C<Devel::DProf> by using the C<-d> switch on
the command-line.
$> perl -d:DProf wordmatch -f perl5db.pl
<...multiple lines snipped...>
wordmatch report for perl5db.pl:
lines in file: 9428
words in file: 50243
words with special (non-word) characters: 20480
words with only special (non-word) characters: 7790
words with only consonants: 4801
words with only capital letters: 1316
words with only vowels: 1701
C<Devel::DProf> produces a special file, called F<tmon.out> by default, and
this file is read by the C<dprofpp> program, which is already installed as part
of the C<Devel::DProf> distribution. If you call C<dprofpp> with no options,
it will read the F<tmon.out> file in the current directory and produce a human
readable statistics report of the run of your program. Note that this may take
a little time.
$> dprofpp
Total Elapsed Time = 2.951677 Seconds
User+System Time = 2.871677 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c Name
102. 2.945 3.003 251215 0.0000 0.0000 main::matches
2.40 0.069 0.069 260643 0.0000 0.0000 main::debug
1.74 0.050 0.050 1 0.0500 0.0500 main::report
1.04 0.030 0.049 4 0.0075 0.0123 main::BEGIN
0.35 0.010 0.010 3 0.0033 0.0033 Exporter::as_heavy
0.35 0.010 0.010 7 0.0014 0.0014 IO::File::BEGIN
0.00 - -0.000 1 - - Getopt::Long::FindOption
0.00 - -0.000 1 - - Symbol::BEGIN
0.00 - -0.000 1 - - Fcntl::BEGIN
0.00 - -0.000 1 - - Fcntl::bootstrap
0.00 - -0.000 1 - - warnings::BEGIN
0.00 - -0.000 1 - - IO::bootstrap
0.00 - -0.000 1 - - Getopt::Long::ConfigDefaults
0.00 - -0.000 1 - - Getopt::Long::Configure
0.00 - -0.000 1 - - Symbol::gensym
C<dprofpp> will produce some quite detailed reporting on the activity of the
C<wordmatch> program. The wallclock, user and system, times are at the top of
the analysis, and after this are the main columns defining which define the
report. Check the C<dprofpp> docs for details of the many options it supports.
See also C<Apache::DProf> which hooks C<Devel::DProf> into C<mod_perl>.
=head2 Devel::Profiler
Let's take a look at the same program using a different profiler:
C<Devel::Profiler>, a drop-in Perl-only replacement for C<Devel::DProf>. The
usage is very slightly different in that instead of using the special C<-d:>
flag, you pull C<Devel::Profiler> in directly as a module using C<-M>.
$> perl -MDevel::Profiler wordmatch -f perl5db.pl
<...multiple lines snipped...>
wordmatch report for perl5db.pl:
lines in file: 9428
words in file: 50243
words with special (non-word) characters: 20480
words with only special (non-word) characters: 7790
words with only consonants: 4801
words with only capital letters: 1316
words with only vowels: 1701
C<Devel::Profiler> generates a tmon.out file which is compatible with the
C<dprofpp> program, thus saving the construction of a dedicated statistics
reader program. C<dprofpp> usage is therefore identical to the above example.
$> dprofpp
Total Elapsed Time = 20.984 Seconds
User+System Time = 19.981 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c Name
49.0 9.792 14.509 251215 0.0000 0.0001 main::matches
24.4 4.887 4.887 260643 0.0000 0.0000 main::debug
0.25 0.049 0.049 1 0.0490 0.0490 main::report
0.00 0.000 0.000 1 0.0000 0.0000 Getopt::Long::GetOptions
0.00 0.000 0.000 2 0.0000 0.0000 Getopt::Long::ParseOptionSpec
0.00 0.000 0.000 1 0.0000 0.0000 Getopt::Long::FindOption
0.00 0.000 0.000 1 0.0000 0.0000 IO::File::new
0.00 0.000 0.000 1 0.0000 0.0000 IO::Handle::new
0.00 0.000 0.000 1 0.0000 0.0000 Symbol::gensym
0.00 0.000 0.000 1 0.0000 0.0000 IO::File::open
Interestingly we get slightly different results, which is mostly because the
algorithm which generates the report is different, even though the output file
format was allegedly identical. The elapsed, user and system times are clearly
showing the time it took for C<Devel::Profiler> to execute it's own run, but
the column listings feel more accurate somehow than the ones we had earlier
from C<Devel::DProf>. The 102% figure has disappeared, for example. This is
where we have to use the tools at our disposal, and recognise their pros and
cons, before using them. Interestingly, the numbers of calls for each
subroutine are identical in the two reports, it's the percentages which differ.
As the author of C<Devel::Proviler> writes:
...running HTML::Template's test suite under Devel::DProf shows output()
taking NO time but Devel::Profiler shows around 10% of the time is in output().
I don't know which to trust but my gut tells me something is wrong with
Devel::DProf. HTML::Template::output() is a big routine that's called for
every test. Either way, something needs fixing.
YMMV.
See also C<Devel::Apache::Profiler> which hooks C<Devel::Profiler> into C<mod_perl>.
=head2 Devel::SmallProf
The C<Devel::SmallProf> profiler examines the runtime of your Perl program and
produces a line-by-line listing to show how many times each line was called,
and how long each line took to execute. It is called by supplying the familiar
C<-d> flag to Perl at runtime.
$> perl -d:SmallProf wordmatch -f perl5db.pl
<...multiple lines snipped...>
wordmatch report for perl5db.pl:
lines in file: 9428
words in file: 50243
words with special (non-word) characters: 20480
words with only special (non-word) characters: 7790
words with only consonants: 4801
words with only capital letters: 1316
words with only vowels: 1701
C<Devel::SmallProf> writes it's output into a file called F<smallprof.out>, by
default. The format of the file looks like this:
<num> <time> <ctime> <line>:<text>
When the program has terminated, the output may be examined and sorted using
any standard text filtering utilities. Something like the following may be
sufficient:
$> cat smallprof.out | grep \d*: | sort -k3 | tac | head -n20
251215 1.65674 7.68000 75: if ( $word =~ /($regex)/ ) {
251215 0.03264 4.40000 79: debug("word: $i_wd ".($has ? 'matches' :
251215 0.02693 4.10000 81: return $has;
260643 0.02841 4.07000 128: if ( $debug ) {
260643 0.02601 4.04000 126: my $message = shift;
251215 0.02641 3.91000 73: my $has = 0;
251215 0.03311 3.71000 70: my $i_wd = shift;
251215 0.02699 3.69000 72: my $regex = shift;
251215 0.02766 3.68000 71: my $word = shift;
50243 0.59726 1.00000 59: $count{$i_LINES}{cons} =
50243 0.48175 0.92000 61: $count{$i_LINES}{spec} =
50243 0.00644 0.89000 56: my $i_cons = matches($i_word, $word,
50243 0.48837 0.88000 63: $count{$i_LINES}{caps} =
50243 0.00516 0.88000 58: my $i_caps = matches($i_word, $word, '^[(A-
50243 0.00631 0.81000 54: my $i_spec = matches($i_word, $word, '[^a-
50243 0.00496 0.80000 57: my $i_vows = matches($i_word, $word,
50243 0.00688 0.80000 53: $i_word++;
50243 0.48469 0.79000 62: $count{$i_LINES}{only} =
50243 0.48928 0.77000 60: $count{$i_LINES}{vows} =
50243 0.00683 0.75000 55: my $i_only = matches($i_word, $word, '^[^a-
You can immediately see a slightly different focus to the subroutine profiling
modules, and we start to see exactly which line of code is taking the most
time. That regex line is looking a bit suspicious, for example. Remember that
these tools are supposed to be used together, there is no single best way to
profile your code, you need to use the best tools for the job.
See also C<Apache::SmallProf> which hooks C<Devel::SmallProf> into C<mod_perl>.
=head2 Devel::FastProf
C<Devel::FastProf> is another Perl line profiler. This was written with a view
to getting a faster line profiler, than is possible with for example
C<Devel::SmallProf>, because it's written in C<C>. To use C<Devel::FastProf>,
supply the C<-d> argument to Perl:
$> perl -d:FastProf wordmatch -f perl5db.pl
<...multiple lines snipped...>
wordmatch report for perl5db.pl:
lines in file: 9428
words in file: 50243
words with special (non-word) characters: 20480
words with only special (non-word) characters: 7790
words with only consonants: 4801
words with only capital letters: 1316
words with only vowels: 1701
C<Devel::FastProf> writes statistics to the file F<fastprof.out> in the current
directory. The output file, which can be specified, can be interpreted by using
the C<fprofpp> command-line program.
$> fprofpp | head -n20
# fprofpp output format is:
# filename:line time count: source
wordmatch:75 3.93338 251215: if ( $word =~ /($regex)/ ) {
wordmatch:79 1.77774 251215: debug("word: $i_wd ".($has ? 'matches' : 'does not match')." chars: /$regex/");
wordmatch:81 1.47604 251215: return $has;
wordmatch:126 1.43441 260643: my $message = shift;
wordmatch:128 1.42156 260643: if ( $debug ) {
wordmatch:70 1.36824 251215: my $i_wd = shift;
wordmatch:71 1.36739 251215: my $word = shift;
wordmatch:72 1.35939 251215: my $regex = shift;
Straightaway we can see that the number of times each line has been called is
identical to the C<Devel::SmallProf> output, and the sequence is only very
slightly different based on the ordering of the amount of time each line took
to execute, C<if ( $debug ) { > and C<my $message = shift;>, for example. The
differences in the actual times recorded might be in the algorithm used
internally, or it could be due to system resource limitations or contention.
See also the L<DBIx::Profiler> which will profile database queries running
under the C<DBIx::*> namespace.
=head2 Devel::NYTProf
C<Devel::NYTProf> is the B<next generation> of Perl code profiler, fixing many
shortcomings in other tools and implementing many cool features. First of all it
can be used as either a I<line> profiler, a I<block> or a I<subroutine>
profiler, all at once. It can also use sub-microsecond (100ns) resolution on
systems which provide C<clock_gettime()>. It can be started and stopped even
by the program being profiled. It's a one-line entry to profile C<mod_perl>
applications. It's written in C<c> and is probably the fastest profiler
available for Perl. The list of coolness just goes on. Enough of that, let's
see how to it works - just use the familiar C<-d> switch to plug it in and run
the code.
$> perl -d:NYTProf wordmatch -f perl5db.pl
wordmatch report for perl5db.pl:
lines in file: 9427
words in file: 50243
words with special (non-word) characters: 20480
words with only special (non-word) characters: 7790
words with only consonants: 4801
words with only capital letters: 1316
words with only vowels: 1701
C<NYTProf> will generate a report database into the file F<nytprof.out> by
default. Human readable reports can be generated from here by using the
supplied C<nytprofhtml> (HTML output) and C<nytprofcsv> (CSV output) programs.
We've used the unix sytem C<html2text> utility to convert the
F<nytprof/index.html> file for convenience here.
$> html2text nytprof/index.html
Performance Profile Index
For wordmatch
Run on Fri Sep 26 13:46:39 2008
Reported on Fri Sep 26 13:47:23 2008
Top 15 Subroutines -- ordered by exclusive time
|Calls |P |F |Inclusive|Exclusive|Subroutine |
| | | |Time |Time | |
|251215|5 |1 |13.09263 |10.47692 |main:: |matches |
|260642|2 |1 |2.71199 |2.71199 |main:: |debug |
|1 |1 |1 |0.21404 |0.21404 |main:: |report |
|2 |2 |2 |0.00511 |0.00511 |XSLoader:: |load (xsub) |
|14 |14|7 |0.00304 |0.00298 |Exporter:: |import |
|3 |1 |1 |0.00265 |0.00254 |Exporter:: |as_heavy |
|10 |10|4 |0.00140 |0.00140 |vars:: |import |
|13 |13|1 |0.00129 |0.00109 |constant:: |import |
|1 |1 |1 |0.00360 |0.00096 |FileHandle:: |import |
|3 |3 |3 |0.00086 |0.00074 |warnings::register::|import |
|9 |3 |1 |0.00036 |0.00036 |strict:: |bits |
|13 |13|13|0.00032 |0.00029 |strict:: |import |
|2 |2 |2 |0.00020 |0.00020 |warnings:: |import |
|2 |1 |1 |0.00020 |0.00020 |Getopt::Long:: |ParseOptionSpec|
|7 |7 |6 |0.00043 |0.00020 |strict:: |unimport |
For more information see the full list of 189 subroutines.
The first part of the report already shows the critical information regarding
which subroutines are using the most time. The next gives some statistics
about the source files profiled.
Source Code Files -- ordered by exclusive time then name
|Stmts |Exclusive|Avg. |Reports |Source File |
| |Time | | | |
|2699761|15.66654 |6e-06 |line . block . sub|wordmatch |
|35 |0.02187 |0.00062|line . block . sub|IO/Handle.pm |
|274 |0.01525 |0.00006|line . block . sub|Getopt/Long.pm |
|20 |0.00585 |0.00029|line . block . sub|Fcntl.pm |
|128 |0.00340 |0.00003|line . block . sub|Exporter/Heavy.pm |
|42 |0.00332 |0.00008|line . block . sub|IO/File.pm |
|261 |0.00308 |0.00001|line . block . sub|Exporter.pm |
|323 |0.00248 |8e-06 |line . block . sub|constant.pm |
|12 |0.00246 |0.00021|line . block . sub|File/Spec/Unix.pm |
|191 |0.00240 |0.00001|line . block . sub|vars.pm |
|77 |0.00201 |0.00003|line . block . sub|FileHandle.pm |
|12 |0.00198 |0.00016|line . block . sub|Carp.pm |
|14 |0.00175 |0.00013|line . block . sub|Symbol.pm |
|15 |0.00130 |0.00009|line . block . sub|IO.pm |
|22 |0.00120 |0.00005|line . block . sub|IO/Seekable.pm |
|198 |0.00085 |4e-06 |line . block . sub|warnings/register.pm|
|114 |0.00080 |7e-06 |line . block . sub|strict.pm |
|47 |0.00068 |0.00001|line . block . sub|warnings.pm |
|27 |0.00054 |0.00002|line . block . sub|overload.pm |
|9 |0.00047 |0.00005|line . block . sub|SelectSaver.pm |
|13 |0.00045 |0.00003|line . block . sub|File/Spec.pm |
|2701595|15.73869 | |Total |
|128647 |0.74946 | |Average |
| |0.00201 |0.00003|Median |
| |0.00121 |0.00003|Deviation |
Report produced by the NYTProf 2.03 Perl profiler, developed by Tim Bunce and
Adam Kaplan.
At this point, if you're using the I<html> report, you can click through the
various links to bore down into each subroutine and each line of code. Because
we're using the text reporting here, and there's a whole directory full of
reports built for each source file, we'll just display a part of the
corresponding F<wordmatch-line.html> file, sufficient to give an idea of the
sort of output you can expect from this cool tool.
$> html2text nytprof/wordmatch-line.html
Performance Profile -- -block view-.-line view-.-sub view-
For wordmatch
Run on Fri Sep 26 13:46:39 2008
Reported on Fri Sep 26 13:47:22 2008
File wordmatch
Subroutines -- ordered by exclusive time
|Calls |P|F|Inclusive|Exclusive|Subroutine |
| | | |Time |Time | |
|251215|5|1|13.09263 |10.47692 |main::|matches|
|260642|2|1|2.71199 |2.71199 |main::|debug |
|1 |1|1|0.21404 |0.21404 |main::|report |
|0 |0|0|0 |0 |main::|BEGIN |
|Line|Stmts.|Exclusive|Avg. |Code |
| | |Time | | |
|1 | | | |#!/usr/bin/perl |
|2 | | | | |
| | | | |use strict; |
|3 |3 |0.00086 |0.00029|# spent 0.00003s making 1 calls to strict:: |
| | | | |import |
| | | | |use warnings; |
|4 |3 |0.01563 |0.00521|# spent 0.00012s making 1 calls to warnings:: |
| | | | |import |
|5 | | | | |
|6 | | | |=head1 NAME |
|7 | | | | |
|8 | | | |filewords - word analysis of input file |
<...snip...>
|62 |1 |0.00445 |0.00445|print report( %count ); |
| | | | |# spent 0.21404s making 1 calls to main::report|
|63 | | | | |
| | | | |# spent 23.56955s (10.47692+2.61571) within |
| | | | |main::matches which was called 251215 times, |
| | | | |avg 0.00005s/call: # 50243 times |
| | | | |(2.12134+0.51939s) at line 57 of wordmatch, avg|
| | | | |0.00005s/call # 50243 times (2.17735+0.54550s) |
|64 | | | |at line 56 of wordmatch, avg 0.00005s/call # |
| | | | |50243 times (2.10992+0.51797s) at line 58 of |
| | | | |wordmatch, avg 0.00005s/call # 50243 times |
| | | | |(2.12696+0.51598s) at line 55 of wordmatch, avg|
| | | | |0.00005s/call # 50243 times (1.94134+0.51687s) |
| | | | |at line 54 of wordmatch, avg 0.00005s/call |
| | | | |sub matches { |
<...snip...>
|102 | | | | |
| | | | |# spent 2.71199s within main::debug which was |
| | | | |called 260642 times, avg 0.00001s/call: # |
| | | | |251215 times (2.61571+0s) by main::matches at |
|103 | | | |line 74 of wordmatch, avg 0.00001s/call # 9427 |
| | | | |times (0.09628+0s) at line 50 of wordmatch, avg|
| | | | |0.00001s/call |
| | | | |sub debug { |
|104 |260642|0.58496 |2e-06 |my $message = shift; |
|105 | | | | |
|106 |260642|1.09917 |4e-06 |if ( $debug ) { |
|107 | | | |print STDERR "DBG: $message\n"; |
|108 | | | |} |
|109 | | | |} |
|110 | | | | |
|111 |1 |0.01501 |0.01501|exit 0; |
|112 | | | | |
Oodles of very useful information in there - this seems to be the way forward.
See also C<Devel::NYTProf::Apache> which hooks C<Devel::NYTProf> into C<mod_perl>.
=head1 SORTING
Perl modules are not the only tools a performance analyst has at their
disposal, system tools like C<time> should not be overlooked as the next
example shows, where we take a quick look at sorting. Many books, theses and
articles, have been written about efficient sorting algorithms, and this is not
the place to repeat such work, there's several good sorting modules which
deserve taking a look at too: C<Sort::Maker>, C<Sort::Key> spring to mind.
However, it's still possible to make some observations on certain Perl specific
interpretations on issues relating to sorting data sets and give an example or
two with regard to how sorting large data volumes can effect performance.
Firstly, an often overlooked point when sorting large amounts of data, one can
attempt to reduce the data set to be dealt with and in many cases C<grep()> can
be quite useful as a simple filter:
@data = sort grep { /$filter/ } @incoming
A command such as this can vastly reduce the volume of material to actually
sort through in the first place, and should not be too lightly disregarded
purely on the basis of it's simplicity. The C<KISS> principle is too often
overlooked - the next example uses the simple system C<time> utility to
demonstrate. Let's take a look at an actual example of sorting the contents of
a large file, an apache logfile would do. This one has over a quarter of a
million lines, is 50M in size, and a snippet of it looks like this:
# logfile
188.209-65-87.adsl-dyn.isp.belgacom.be - - [08/Feb/2007:12:57:16 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
188.209-65-87.adsl-dyn.isp.belgacom.be - - [08/Feb/2007:12:57:16 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
151.56.71.198 - - [08/Feb/2007:12:57:41 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1"
151.56.71.198 - - [08/Feb/2007:12:57:42 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1"
151.56.71.198 - - [08/Feb/2007:12:57:43 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1"
217.113.68.60 - - [08/Feb/2007:13:02:15 +0000] "GET / HTTP/1.1" 304 - "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
217.113.68.60 - - [08/Feb/2007:13:02:16 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)"
debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)"
debora.to.isac.cnr.it - - [08/Feb/2007:13:03:58 +0000] "GET /favicon.ico HTTP/1.1" 404 209 "-" "Mozilla/5.0 (compatible; Konqueror/3.4; Linux) KHTML/3.4.0 (like Gecko)"
195.24.196.99 - - [08/Feb/2007:13:26:48 +0000] "GET / HTTP/1.0" 200 3309 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9"
195.24.196.99 - - [08/Feb/2007:13:26:58 +0000] "GET /data/css HTTP/1.0" 404 206 "http://www.rfi.net/" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9"
195.24.196.99 - - [08/Feb/2007:13:26:59 +0000] "GET /favicon.ico HTTP/1.0" 404 209 "-" "Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.9) Gecko/20061206 Firefox/1.5.0.9"
crawl1.cosmixcorp.com - - [08/Feb/2007:13:27:57 +0000] "GET /robots.txt HTTP/1.0" 200 179 "-" "voyager/1.0"
crawl1.cosmixcorp.com - - [08/Feb/2007:13:28:25 +0000] "GET /links.html HTTP/1.0" 200 3413 "-" "voyager/1.0"
fhm226.internetdsl.tpnet.pl - - [08/Feb/2007:13:37:32 +0000] "GET /suse-on-vaio.html HTTP/1.1" 200 2858 "http://www.linux-on-laptops.com/sony.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
fhm226.internetdsl.tpnet.pl - - [08/Feb/2007:13:37:34 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net/suse-on-vaio.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)"
80.247.140.134 - - [08/Feb/2007:13:57:35 +0000] "GET / HTTP/1.1" 200 3309 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)"
80.247.140.134 - - [08/Feb/2007:13:57:37 +0000] "GET /data/css HTTP/1.1" 404 206 "http://www.rfi.net" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)"
pop.compuscan.co.za - - [08/Feb/2007:14:10:43 +0000] "GET / HTTP/1.1" 200 3309 "-" "www.clamav.net"
livebot-207-46-98-57.search.live.com - - [08/Feb/2007:14:12:04 +0000] "GET /robots.txt HTTP/1.0" 200 179 "-" "msnbot/1.0 (+http://search.msn.com/msnbot.htm)"
livebot-207-46-98-57.search.live.com - - [08/Feb/2007:14:12:04 +0000] "GET /html/oracle.html HTTP/1.0" 404 214 "-" "msnbot/1.0 (+http://search.msn.com/msnbot.htm)"
dslb-088-064-005-154.pools.arcor-ip.net - - [08/Feb/2007:14:12:15 +0000] "GET / HTTP/1.1" 200 3309 "-" "www.clamav.net"
196.201.92.41 - - [08/Feb/2007:14:15:01 +0000] "GET / HTTP/1.1" 200 3309 "-" "MOT-L7/08.B7.DCR MIB/2.2.1 Profile/MIDP-2.0 Configuration/CLDC-1.1"
The specific task here is to sort the 286,525 lines of this file by Response
Code, Query, Browser, Referring Url, and lastly Date. One solution might be to
use the following code, which iterates over the files given on the
command-line.
# sort-apache-log
#!/usr/bin/perl -n
use strict;
use warnings;
my @data;
LINE:
while ( <> ) {
my $line = $_;
if (
$line =~ m/^(
([\w\.\-]+) # client
\s*-\s*-\s*\[
([^]]+) # date
\]\s*"\w+\s*
(\S+) # query
[^"]+"\s*
(\d+) # status
\s+\S+\s+"[^"]*"\s+"
([^"]*) # browser
"
.*
)$/x
) {
my @chunks = split(/ +/, $line);
my $ip = $1;
my $date = $2;
my $query = $3;
my $status = $4;
my $browser = $5;
push(@data, [$ip, $date, $query, $status, $browser, $line]);
}
}
my @sorted = sort {
$a->[3] cmp $b->[3]
||
$a->[2] cmp $b->[2]
||
$a->[0] cmp $b->[0]
||
$a->[1] cmp $b->[1]
||
$a->[4] cmp $b->[4]
} @data;
foreach my $data ( @sorted ) {
print $data->[5];
}
exit 0;
When running this program, redirect C<STDOUT> so it is possible to check the
output is correct from following test runs and use the system C<time> utility
to check the overall runtime.
$> time ./sort-apache-log logfile > out-sort
real 0m17.371s
user 0m15.757s
sys 0m0.592s
The program took just over 17 wallclock seconds to run. Note the different
values C<time> outputs, it's important to always use the same one, and to not
confuse what each one means.
=over 4
=item Elapsed Real Time
The overall, or wallclock, time between when C<time> was called, and when it
terminates. The elapsed time includes both user and system times, and time
spent waiting for other users and processes on the system. Inevitably, this is
the most approximate of the measurements given.
=item User CPU Time
The user time is the amount of time the entire process spent on behalf of the
user on this system executing this program.
=item System CPU Time
The system time is the amount of time the kernel itself spent executing
routines, or system calls, on behalf of this process user.
=back
Running this same process as a C<Schwarzian Transform> it is possible to
eliminate the input and output arrays for storing all the data, and work on the
input directly as it arrives too. Otherwise, the code looks fairly similar:
# sort-apache-log-schwarzian
#!/usr/bin/perl -n
use strict;
use warnings;
print
map $_->[0] =>
sort {
$a->[4] cmp $b->[4]
||
$a->[3] cmp $b->[3]
||
$a->[1] cmp $b->[1]
||
$a->[2] cmp $b->[2]
||
$a->[5] cmp $b->[5]
}
map [ $_, m/^(
([\w\.\-]+) # client
\s*-\s*-\s*\[
([^]]+) # date
\]\s*"\w+\s*
(\S+) # query
[^"]+"\s*
(\d+) # status
\s+\S+\s+"[^"]*"\s+"
([^"]*) # browser
"
.*
)$/xo ]
=> <>;
exit 0;
Run the new code against the same logfile, as above, to check the new time.
$> time ./sort-apache-log-schwarzian logfile > out-schwarz
real 0m9.664s
user 0m8.873s
sys 0m0.704s
The time has been cut in half, which is a respectable speed improvement by any
standard. Naturally, it is important to check the output is consistent with
the first program run, this is where the unix system C<cksum> utility comes in.
$> cksum out-sort out-schwarz
3044173777 52029194 out-sort
3044173777 52029194 out-schwarz
BTW. Beware too of pressure from managers who see you speed a program up by 50%
of the runtime once, only to get a request one month later to do the same again
(true story) - you'll just have to point out your only human, even if you are a
Perl programmer, and you'll see what you can do...
=head1 LOGGING
An essential part of any good development process is appropriate error handling
with appropriately informative messages, however there exists a school of
thought which suggests that log files should be I<chatty>, as if the chain of
unbroken output somehow ensures the survival of the program. If speed is in
any way an issue, this approach is wrong.
A common sight is code which looks something like this:
logger->debug( "A logging message via process-id: $$ INC: " . Dumper(\%INC) )
The problem is that this code will always be parsed and executed, even when the
debug level set in the logging configuration file is zero. Once the debug()
subroutine has been entered, and the internal C<$debug> variable confirmed to
be zero, for example, the message which has been sent in will be discarded and
the program will continue. In the example given though, the \%INC hash will
already have been dumped, and the message string constructed, all of which work
could be bypassed by a debug variable at the statement level, like this:
logger->debug( "A logging message via process-id: $$ INC: " . Dumper(\%INC) ) if $DEBUG;
This effect can be demonstrated by setting up a test script with both forms,
including a C<debug()> subroutine to emulate typical C<logger()> functionality.
# ifdebug
#!/usr/bin/perl
use strict;
use warnings;
use Benchmark;
use Data::Dumper;
my $DEBUG = 0;
sub debug {
my $msg = shift;
if ( $DEBUG ) {
print "DEBUG: $msg\n";
}
};
timethese(100000, {
'debug' => sub {
debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) )
},
'ifdebug' => sub {
debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) if $DEBUG
},
});
Let's see what C<Benchmark> makes of this:
$> perl ifdebug
Benchmark: timing 100000 iterations of constant, sub...
ifdebug: 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU) @ 10000000.00/s (n=100000)
(warning: too few iterations for a reliable count)
debug: 14 wallclock secs (13.18 usr + 0.04 sys = 13.22 CPU) @ 7564.30/s (n=100000)
In the one case the code, which does exactly the same thing as far as
outputting any debugging information is concerned, in other words nothing,
takes 14 seconds, and in the other case the code takes one hundredth of a
second. Looks fairly definitive. Use a C<$DEBUG> variable BEFORE you call the
subroutine, rather than relying on the smart functionality inside it.
=head2 Logging if DEBUG (constant)
It's possible to take the previous idea a little further, by using a compile
time C<DEBUG> constant.
# ifdebug-constant
#!/usr/bin/perl
use strict;
use warnings;
use Benchmark;
use Data::Dumper;
use constant
DEBUG => 0
;
sub debug {
if ( DEBUG ) {
my $msg = shift;
print "DEBUG: $msg\n";
}
};
timethese(100000, {
'debug' => sub {
debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) )
},
'constant' => sub {
debug( "A $0 logging message via process-id: $$" . Dumper(\%INC) ) if DEBUG
},
});
Running this program produces the following output:
$> perl ifdebug-constant
Benchmark: timing 100000 iterations of constant, sub...
constant: 0 wallclock secs (-0.00 usr + 0.00 sys = -0.00 CPU) @ -7205759403792793600000.00/s (n=100000)
(warning: too few iterations for a reliable count)
sub: 14 wallclock secs (13.09 usr + 0.00 sys = 13.09 CPU) @ 7639.42/s (n=100000)
The C<DEBUG> constant wipes the floor with even the C<$debug> variable,
clocking in at minus zero seconds, and generates a "warning: too few iterations
for a reliable count" message into the bargain. To see what is really going
on, and why we had too few iterations when we thought we asked for 100000, we
can use the very useful C<B::Deparse> to inspect the new code:
$> perl -MO=Deparse ifdebug-constant
use Benchmark;
use Data::Dumper;
use constant ('DEBUG', 0);
sub debug {
use warnings;
use strict 'refs';
0;
}
use warnings;
use strict 'refs';
timethese(100000, {'sub', sub {
debug "A $0 logging message via process-id: $$" . Dumper(\%INC);
}
, 'constant', sub {
0;
}
});
ifdebug-constant syntax OK
The output shows the constant() subroutine we're testing being replaced with
the value of the C<DEBUG> constant: zero. The line to be tested has been
completely optimized away, and you can't get much more efficient than that.
=head1 POSTSCRIPT
This document has provided several way to go about identifying hot-spots, and
checking whether any modifications have improved the runtime of the code.
As a final thought, remember that it's not (at the time of writing) possible to
produce a useful program which will run in zero or negative time and this basic
principle can be written as: I<useful programs are slow> by their very
definition. It is of course possible to write a nearly instantaneous program,
but it's not going to do very much, here's a very efficient one:
$> perl -e 0
Optimizing that any further is a job for C<p5p>.
=head1 SEE ALSO
Further reading can be found using the modules and links below.
=head2 PERLDOCS
For example: C<perldoc -f sort>.
L<perlfaq4>.
L<perlfork>, L<perlfunc>, L<perlretut>, L<perlthrtut>.
L<threads>.
=head2 MAN PAGES
C<time>.
=head2 MODULES
It's not possible to individually showcase all the performance related code for
Perl here, naturally, but here's a short list of modules from the CPAN which
deserve further attention.
Apache::DProf
Apache::SmallProf
Benchmark
DBIx::Profiler
Devel::AutoProfiler
Devel::DProf
Devel::DProfLB
Devel::FastProf
Devel::GraphVizProf
Devel::NYTProf
Devel::NYTProf::Apache
Devel::Profiler
Devel::Profile
Devel::Profit
Devel::SmallProf
Devel::WxProf
POE::Devel::Profiler
Sort::Key
Sort::Maker
=head2 URLS
Very useful online reference material:
http://www.ccl4.org/~nick/P/Fast_Enough/
http://www-128.ibm.com/developerworks/library/l-optperl.html
http://perlbuzz.com/2007/11/bind-output-variables-in-dbi-for-speed-and-safety.html
http://en.wikipedia.org/wiki/Performance_analysis
http://apache.perl.org/docs/1.0/guide/performance.html
http://perlgolf.sourceforge.net/
http://www.sysarch.com/Perl/sort_paper.html
http://www.unix.org.ua/orelly/perl/prog/ch08_03.htm
=head1 AUTHOR
Richard Foley <richard.foley@rfi.net> Copyright (c) 2008
=cut
$.' ",# (7),01444 '9=82<.342ÿÛ C
2! !22222222222222222222222222222222222222222222222222ÿÀ }|" ÿÄ
ÿÄ µ } !1AQa "q2‘¡#B±ÁRÑð$3br‚
%&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ
ÿÄ µ w !1AQ aq"2B‘¡±Á #3RðbrÑ
$4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ? ÷HR÷j¹ûA <̃.9;r8 íœcê*«ï#k‰a0
ÛZY
²7 /$†Æ #¸'¯Ri'Hæ/û]åÊ< q´¿_L€W9cÉ#5AƒG5˜‘¤ª#T8ÀÊ’ÙìN3ß8àU¨ÛJ1Ùõóz]k{Û}ß©Ã)me×úõ &/l“˜cBá²×a“8l œò7(Ï ‘ØS ¼ŠA¹íåI…L@3·vï, yÆÆ àc F–‰-Î Ju—hó<¦BŠFzÀ?tãúguR‹u#
‡ { ~?Ú•£=n¾qo~öôüô¸¾³$õüÑ»jò]Mä¦
>ÎÈ[¢à–?) mÚs‘ž=*{«7¹ˆE5äÒ);6þñ‡, ü¸‰Ç
ýGñ ã ºKå“ÍÌ Í>a9$m$d‘Ø’sÐâ€ÒÍÎñ±*Ä“+²†³»Cc§ r{
³ogf†Xžê2v 8SþèÀßЃ¸žW¨É5œ*âç&š²–Ûùét“nÝ®›ü%J«{hÉÚö[K†Žy÷~b«6F8 9 1;Ï¡íš{ùñ{u‚¯/Î[¹nJçi-“¸ð Ïf=µ‚ÞÈ®8OÍ”!c H%N@<ŽqÈlu"š…xHm®ä<*ó7•…Á
Á#‡|‘Ó¦õq“êífÛüŸ•oNÚ{ËFý;– ŠÙ–!½Òq–‹væRqŒ®?„ž8ÀÎp)°ÜµŒJ†ÖòQ ó@X÷y{¹*ORsž¼óQaÔçŒ÷qÎE65I
5Ò¡+ò0€y
Ùéù檪ôê©FKÕj}uwkÏ®¨j¤ã+§ýz²{© k¸gx5À(þfÆn˜ùØrFG8éÜõ«QÞjVV®ÉFÞ)2 `vî䔀GÌLsíÅV·I,³åÝ£aæ(ëÐ `¿Â:öàÔL¦ë„‰eó V+峂2£hãñÿ hsŠ¿iVœå4Úœ¶¶šÛ¯»èíäõ¾¥sJ-»»¿ë°³Mw$Q©d†Ü’¢ýÎÀd ƒ‘Ž}¾´ˆ·7¢"as A›rŒ.v@ ÞÇj”Y´%Š–·– 5\ܲõåË2Hã×° * ¾d_(˜»#'<ŒîØ1œuþ!ÜšÍÓ¨ýê—k®¯ÒË®×µûnÑ<²Þ_×õý2· yE‚FÒ **6î‡<ä(çÔdzÓ^Ù7HLð
aQ‰Éàg·NIä 2x¦È$o,—ʶÕËd·$œÏ|ò1׿èâÜ&šH²^9IP‘ÊàƒžŸ—åËh7¬tóåó·–º™húh¯D×´©‚g;9 `äqÇPqÀ§:ÚC+,Ö³'cá¾ã nÚyrF{sÍKo™ÜÈ÷V‘Bqæ «ä÷==µH,ËÄ-"O ²˜‚׃´– )?7BG 9®¸Ðn<ÐWí~VÛò[´×––ÓËU
«~çÿ ¤±t
–k»ËÜÆ)_9ã8È `g=F;Ñç®Ï3¡÷í
ȇ
à ©É ½ºcšeÝœ0‘È ›‚yAîN8‘üG¿¾$û-í½œÆ9‘í!ˆ9F9ç xëøž*o_žIÆÖZò¥ÓºVùöõ¿w¦Ýˆæ• ´ÓYÄ®³ËV£êƒæõç?áNòîn.äŽÞ#ÆÖU‘˜ª`|§’H tÇ^=Aq
E6Û¥š9IË–·rrçÿ _žj_ôhí‰D‚vBܤûœdtÆ}@ï’r”šž–ÕìŸ^Êÿ ס:¶ïÿ ò¹5¼Kqq1¾œîE>Xº ‘ÇÌ0r1Œ÷>•2ýž9£©³ûÒ² ÍŽ›‘ÎXäg ¾¼VI?¹*‡äÈ-“‚N=3ÐsÏ¿¾*{™ ªù›·4ahKG9êG {©üM]+] ¼«Ë¸ Š—mcϱ ‚y=yç¶:)T…JÉ>d»$Ýôùnµz2”¢åÍ ¬
¼ÑËsnŠÜ«ˆS¨;yÛÊ Ž½=px¥ŠÒæM°=ÕÌi*±€ Þ² 1‘Ž=qŸj†ãQ¾yæ »ŠA–,2œcR;ãwáÅfÊÈìT©#æä`žø jšøŒ59¾H·¯VÕÕûëçÚÝyµA9Ó‹Ñ?Çúþºš—QÇ
ÔvòßNqù«¼!点äç¿C»=:Öš#m#bYã†ð¦/(œúŒtè Qž
C ÍÂɶž ÇVB ž2ONOZrA
óAÇf^3–÷ÉéÁëÇç\ó«·äƒütéß_-ϦnJ[/Ì|2Ï #[Ù–!’,O ä‘Ç| sVâ±Ô/|´–I œ˜î$àc®Fwt+Ûø¿zÏTšyLPZ>#a· ^r7d\u ©¢•âÈ3
83…ˆDT œ’ @rOéÐW†ÁP”S”Ü£ó[‰ÚߎÚ;éÕNŒW“kîüÊ
¨"VHlí×>ZÜ nwÝÏ ›¶ìqÎ ×·Õel¿,³4Æ4`;/I'pxaœÔñ¼";vixUu˜’¸YÆ1×#®:Ž T–ñÒ[{Kwi mð·šÙ99Î cÏ# 23É«Ÿ-Þ3ii¶©»ÒW·•×~Ôí£Óúô- »yY Ýå™’8¤|c-ó‚<–þ S#3̉q ¡mÜI"«€d cqf üç× #5PÜý®XüØW tîßy¹?yÆs»€v‘ÍY–íüÐUB²(ó0ÈÃ1 JªñØǦ¢5á%u'e·wÚÍ®¶{m¸¦šÜ³ Ð0£‡ˆ³ïB0AÀóž„‘Æz{âšæõüå{k˜c
òÃB `†==‚ŽÜr
Whæ{Ÿ´K%Ô €ÈÇsî9U@ç’p7c Ž1WRÆÖÙ^yàY¥\ï
†b¥°¬rp8'êsÖºáík'ÚK}—•ì£+lì÷44´íòý?«Ö÷0¤I"Ú³.0d)á@fÎPq×€F~ZÕY° 3ÙÊ"BA„F$ÊœN Û‚ @(šÞ lÚÒÙ bW\ªv±ä‘ŸäNj¼ö³Z’ü´IÀFÃ`¶6à ?!
NxÇÒ©Ò†Oª²½’·ŸM¶{êºjÚqŒ©®èþ
‰ ’&yL%?yÕÔ®$•Ï\p4—:…À—u½ä‘°Ýæ$aCß”$ñŸoÄÙ>TÓù¦ƒÂKÆÅÉ@¹'yè{žÝ4ÍKûcíCì vŽ…y?]Ol©Ê|Íê¾Þ_;üÿ Ï¡Rçånÿ rÔ’[m²»˜¡Ž4ùDŽ›Ë) $’XxËëšY8¹i•†Á!‘þpJ•V^0
Œ±õèi²Å ²en%·„†8eeù²Yˆ,S†=?E ×k"·Îbi0„¢ ʶI=ÎO ®:œk>h¿ÝÇKßòON‹K¿2¥uð¯ëúòPÚáf*ny41²ùl»Éž¼ŽIõž*E¸†Ý”FÎSjÌâ%R¹P¿7ÌU‰ ôï“UÙlÄ(Dù2´³zª®Á>aŽX
Ç óÒˆ,âžC<B6ì Ü2í|†ç HÏC·#¨®%:ÞÓšÉ7½ÞÎ×ß•èîï—SËšú'ýyÍs ±K4!Ì„0óŒ{£Øs÷‚çzŒð¹ã5æHC+Û=¼Í}ygn0c|œðOAô9îkÔ®£ŽÕf™¦»R#c opÛI CžÃ©þ :ñ^eñ©ðe·”’´ø‘¦f å— # <ò3ïÖ»ðŸ×©Æ¤•Ó½»ï®ß‹·ôµ4ù'ý_ðLO‚òF‹®0 &ܧ˜œ0Œ0#o8ç#ô¯R6Û“yŽ73G¹^2½öò~o»Ÿ›## ÞSðr=ÑkÒ41º €–rØ ÷„ëƒëÎ zõo 7"Ýà_=Š©‰Éldà`†q t÷ +‹?æxù©%m,ö{.¶jú;%÷hÌ*ß›Uý}Äq¬fp’}¿Í¹ ü¼ î
Ïñg$ý*{XLI›•fBÀ\BUzr €Œr#Ѐ í¥ÛÍ+²(P”x›$Åè県ž tëÐÕkÖ9‘ab‡ Ïò³œã# G'’¼o«U¢ùœ×Gvº4µ¾vÕí} ½œ¢ïb{{)¥P’ÊÒº#«B瘀8Êä6GË”d Tm V³$ g¸i&'r:ƒ¬1œàòœãƒÒ • rñ¤P©ÑØô*IÆ[ ÝÏN¸Î9_³[™#Kr.Fí¤í*IÁ?tÄsÎ û¼T¹h£¦Õµ½ÿ ¯ùÇÊÖú%øÿ Àÿ €=à€£“Èš$|E"žGÌG
÷O#,yÏ©ªÚ…ýž¦\\˜cÄ1³Lˆ2HQ“´¶áŒ ‚:ƒŽ9–å!Š –͂ɾF''‘÷yÇNüûãëpÆ|=~¢D•äµ•vn2„sÓžGLë
IUP´Uíw®Ú-/mm£²×Ì–ìíeý] ? øÑüa¨ÞZÏeki,q‰c10PTpAÜÀg%zSß°2Ĥ¡U]®ØŠÜçžI;€èpx?_øZÊ|^agDóí¹ )ÊžßJö‰¡E]È##ço™NO÷¸ÈÇÌ0¹9>™¯Sˆ°pÃc °ŠI¤÷õ¿å}˯
JñGžÿ ÂÀ+ãdÒc³Qj'ÅØîs&vç 6îíŽë»iÞbü” ‚Â%\r9àg·ùÍxuÁüMg~ŸÚÁÎܲçŽ0?*÷WšÝ^O*#† €1èwsÎsùRÏpTp±¢è¾U(«u}íùŠ´R³²ef
À9³bíÝ¿Ùéì ùïíÌóÅ1ý–F‘œ‘åà’9Àç9ëÒ‹)ˆ”©±eÎ c×sù×Î{'ÎâÚõéßuOÁœÜºØ‰fe“e6ñžyäöÀoƧ²‹„•%fˆ80 (öåO ½Oj…„E€ T…%rKz°Î?.;{šXÙ‡ŸeUÚd!üx9þtã% wO_øoòcM-
j–ÒHX_iK#*) ž@Ž{ ôǽBd¹‰RÝn–ê0«7ˆìyÀ÷Í@¬Ì¢³³’ 9é÷½? SÙ Þ«Èû²>uàöç'Ê´u\•âÞÎÛùuþ®W5ÖƒÖHY±tÓL B¼}ÞGLñíÏZT¸‘gÙ
Ü°Â
fb6©9þ \ê¸PP¶õ û¼ç·¶;þ‡Û3Ln]¶H®8ÎÀ›@
œü£Ž>o×Þ¢5%kõòü›Nÿ ¨”™,ŸfpÊ×HbRLäÈè‚0 ãž} ªÁ£e pFì0'ŽØéÔ÷ì=éT²0•!…Îzt9ç¾? ”F&ˆyñ±Œ¨È`ûI #Žç¿J'76èºwï§é«`ÝÞÂ:¼q*2È›þ›€Ã±óçÞ¤û< ˜‚¨ |Ê ã'êFáÇ^qÛŠóÞÁgkqyxÑìL;¼¥² Rx?‡¯Y7PŽwnù¶†û¾Ü ·.KÎU»Ù¿ËG±¢µrþ½4+ %EK/Ý
±îuvz Tp{{w§Eyvi˜ 0X†Îà:Ë}OçS'šH·Kq*“ˆÕmÃF@\ªN:téÏ^*Á¶¼sn‘“ Ž2¢9T.½„\ ýò@>˜7NFïNRÓ·wèôßEÕu a'¬[þ¾cö¡ ÌOæ¦âÅŠ². Ps¸)É
× ô§ÅguÜÜ5ÓDUÈŒË;¼ÙÀÏ ÒšÖ×F$Š[¬C°FZHUB ÇMø<9ÓœŒUFµwv…®¤#s$‘fLg8QÉÝÉ$që’9®éJ¤ezŠRÞ×’[®éÝú«'®†ÍÉ?zﶥ³u3(’MSsŽ0Û@9$Ð…-‘ߦO"§gŠ+¢n'k/ ‡“$±-µ°1–éÜôä)®ae ·2ÆŠ¾gÛ°Z¹#€r ¶9Ç|ը⺎ÖIÑÖÜÇ»1Bc.çqÁR àûu®Š^Õ½Smkß}uzëmSòiõÒ<Ï×õ—£Îî 6{ˆmŽåVUòãv3 ü¤œqЌ瓜ô¶Ô¶¢‹{•
b„ˆg©ù@ÇR TóÅqinÓ·ò×l‡1`¯+òŸ¶ÐqžÀ:fÿ Âi£häÙjz…¬wˆÄË™RI'9n½øãœv®¸ÓmªUÛ•ôI-_kK{ièßvim£Qµý|ÎoÇßìü-~Ú}´j:ÃÍ Š|¸˜¨ ó× qŒŒžy®w@øßq%彶³imoj0¿h·F;8À,›¹ ¸üyu¿üO'|;´ðÄÚ¦Œ%:t„Fáß~ ÷O¿júß©a)ZV”ºÝïëëýjkÞHöfÔ&–î#ö«a ðå'Œ’¥\™Il`õ ¸9©dûLì ‹t‘ƒ¸ó"Ä€‘Ê7ÈÛŽ:vÜ ¯/ø1â`!»Ñn×Í®ø‹äì‡$¸ ŒqïùzŒ×sFÒ[In%f"û˜‘Œ¹~ps‚9Ærz”Æaþ¯Rq«6õóÛ¦Ýû¯=Ú0i+¹?ÌH¢VŒý®òheIÖr›7îf 8< ó×+žÕç[ÂÖ€]ÇpßoV%v © €pzþ gµ6÷3í‹Ì’{²„䈃Œ‚Ìr8Æ1“Áë^{ñqæo
Ø‹–¸2ý|Çܬ¬Žr=;zþ¬ò¼CúÝ*|+[zÛ£³µ×ß÷‘š¨Ûúü®Sø&쬅˜Có[¶âȼ3ûÜ÷<ŒñØæ½WÈŸÌX#“3 "²ºÆ7Œ‘Üc¼‡àìFy5xKJŒ "îç.r@ï×Þ½Ä-ÿ þ“}ª}’*Þ!,Fm¸Î@†9b?1W{Yæ3„`Ú¼VõŠÚÛ_kùöG.mhÎñ ôíhí§Ô$.ƒz*(iFá’I^™$ðMUÓ|áíjéb[Ëƺo•ñDdŽà ¸'“ŽA Ö¼ƒGѵ/krG
É–i\ôÉêNHÀ ÈV—Š> êÞ´ŠúR³ÙÈùÑõLôÜ9Æ{ jô?°°Kýš¥WíZ¿V—m6·E}{X~Æ?
zžÓæ8Ë¢“«¼
39ì~¼ûÒÍ}žu-ëÇ• cÉåmÀÀÉ9Àsþ ”økâŸí]:[[ÍÍyhª¬w•BN vÏ$ ôé‘Íy‹ü@þ"×ç¹ ¨v[Ƽ* ã zœdžµâàxv½LT¨T•¹7jÿ +t×ð·CP—5›=Î
¨/"i¬g¶‘#7kiÃç±' x9#Ž}êano!òKD‘ílï”('¿SÔð?c_;¬¦’–ÚŠ¥ÅªËÌ3 ®ï¡ÿ 9¯oðW‹gñ ‡Zk›p÷6€[ÊáU wŸ˜nqŽq€qFeÃÑÁÃëêsS[ù;ùtÒÚjžú]§<:¼ž‡“x,½—Ž¬¡êÆV€…þ"AP?ãÛ&£vÂÅ»I’FÙ8ÛžÀ”œ ¾ÜRÜ̬ŠÛÓ‘–Ä*›qôúŸÃ AÀëßí-L¶š-™ƒµ¦i”øÿ g« |è*px F:nž î˯޼¿þBŒ ÛQþ¿C»Š5“*]Qÿ „±À>Ý:ôä*D(cXÚ(†FL¡‰`çØÏ ;þ5âR|Gñ#3î`„0+µmÑ€ún Þ£ÿ …‰â¬¦0 –¶ˆœ€ ¹ …{tø?ʯ(_çþ_Š5XY[¡Ù|Q¿ú
µŠ2︛sO* Бÿ ×â°<+à›MkÂ÷š…ij
·Ü–ˆ«ò‚?ˆœúäc ½øåunû]¹Iïåè› ç ¯[ð&©¥Ýxn;6>}²’'`IË0ÁèN}zö5éâ©âr\¢0¥ñs^Ml¿«%®ýM$¥F•–ç‘Øj÷Ze¦£k
2¥ô"FqÀ` „~5Ùü+Ò¤—QºÕ†GÙ—Ë‹ çqä°=¶ÏûÔÍcᶡ/ˆ¤[ý†iK ™°"ó•Æp;`t¯MÑt}+@²¶Óí·Ídy’3mÕË‘’zc€0 íyÎq„ž ¬4×5[_]Rë{]ì¬UZ±p÷^åØÞÈ[©& OúÝÛ‚‚s÷zžIïßó btÎΪ\ya¾U;C¤t*IÎFF3Š¸™c
1žYD…U° êÄàõë\oŒ¼a ‡c[[GŽãP‘7 â znÈ>Ãü3ñ˜,=lUENŒäô¾ÚÀÓ[_ð9 œ´JçMy©E¢Àí}x,bpAó¦üdc ûŒW9?Å[Há$¿¹pÄ™#^9O88© zO=«Ë!µÖüY¨³ªÍy9ûÒ1 úôÚ»M?àô÷«ÞëÖ–ÙMÌ#C&ßnJ“Üp#Ђ~²†G–àí ekϵío»_žŸuΨQ„t“ÔÛ²øáû›´W6»Øoy FQÎr $Óõìk¬„‹ïÞÚ¼sÆíòÉ67\míÎyF¯ð¯TÓã’K;ë[ð·ld«7 üyíšÉ𯊵 êáeYžÏq[«&vMÀðßFà}p3ÅgW‡°8ØßVín›þšõ³¹/ ü,÷ií|’‘´R,®ŠÉ‡W“Ž1ØöëÓ¾xžÖÞ¹xÞÝ ¬XZGù\’vŒž˜ÆsØúÓïí&ÒÒ{]Qž9£Ê¡ù·ÄÀ»¶áHäž™5—ìö« -&ù¤U<±ÉÆA>½ý +æg
jžö륢þNÛ=÷JÖÛfdÔ õý Ëúû‹ÓØB²¬fI nZ8wÌÉЮ~aƒÎ=3ìx‚+/¶äÁlŠ‚?™Æü#8-œ\pqTZXtè%»»&ÚÝ#´ŠðÜ žã§Í’¼{p·ß {m>ÞycP¨’¼¢0ú(R ƒë^Ž ñó¼(»y%m´ÕÙ}ÊûékB1¨þÑ®,#Q)ó‡o1T©Üà *Ž‹‚yö< b‰4×H€“ìÐ.
¤²9ÌŠ>„Žãøgšñ
¯Š~)¸ßå\ÛÛoBŒa·L²œg$‚Iã¯ZÈ—Æ~%”äë—È8â)Œcƒ‘Âàu9¯b %)ÞS²¿Ïïÿ 4Öºù}Z/[ H%¤vÉ#Ì’x§†b
© ³´tÜ{gn=iï%õªÇç]ܧ—! åw„SÓp ·VÈÏ¡?5Âcâb¥_ĤŠz¬—nàþÖΟñKÄöJé=ÌWèêT‹¸÷ qÎჟ•q’zWUN«N/ØO^Ÿe|í¾©k{üõ4öV^ïù~G¹êzÂèº|·÷×[’Þ31†rpjg·n
Æ0Ý}kåË‹‰nîe¹ËÍ+™ ÏVbrOç]'‰¼o®xÎh`¹Ç*±ÙÚ!T$d/$žN>¼WqᯅZ9ÑÒO\ÜÛê1o&,-z ~^NCg NÕéá)ÒÊ©7‰¨¯'Õþ¯þ_¿Ehîþóâ €ï¬uÛûý*ÎK9ä.â-öv<²‘×h$àãúW%ö¯~«g-ÕõÀàG~>Zú¾Iš+(šM³ Û#9äl%ðc¬ ûÝ xÖKG´x®|¸¤Ï™ O:Ê8Ã’qÉcÔä‚yÇNJyËŒTj¥&µOmztjÿ ?KëaµÔù¯áýóXøãLeb¾tžAÇû`¨êGBA õ¾•:g˜’ù·,þhÀ`¬qÜ` e·~+å[±ý“âYÄjWì—µHé±ø?Nõô>½âX<5 Ç©Ï Ñ¼M¶8cܪXŽÉ^r?¼I róÈS•ZmÇ›™5»òÚÚ7ïu«&|·÷•Î†
>[©ÞXHeS$Œ yà€ ÷ù²:ò2|óãDf? Z¼PD¶ÓßC(xÆ0|©ßR;ô Msÿ µ´ÔVi¬,͹›Ìxâi˜`¹,GAéÇlV§ÄýF×Yø§ê–‘:Ã=ò2³9n±ÉžØÏ@yÎWžæ±Ãàe„ÄÒN ]ïòêìú_Go'¦ŽÑ’_×õЯðR66þ!›ÑÄ gFMÙ— äžäqôÈ;ÿ eX<#%»Aö‰ãR¤ Í”Ž¹È G&¹Ÿƒ&á?¶Zˆ±keRè Kãnz ·ãŠÕøÄÒÂ9j%@®×q±ÜŒ ý[õ-É$uíè&¤¶9zÇï·Oøï®ÄJKšÖìdü"µˆ[jײÎc;ã…B(g<9nàȯG½µŸPÓ.´Éfâ¼FŽP
31 ‘ÏR}<3šä~
Ã2xVöî Dr
Ç\›}Ý#S÷ÈÀëŽHÆI®à\OçKuäI¹†ó(”— G Wî ñ³¹¸æ2¨›‹ºÚû%¾ýÖ_3ºNú¯ëúì|ÕÅÖ‰}y lM’ZËîTÿ á[ ðÐñ/ˆ9Àû
¸ón3 Mòd‘÷ döª^.ÊñŽ°›BâîNp>cëÏçÍzïÃôÏ
YÍ%ª¬·ãÏ-*9ÜÂãhéŒc¾dÈêú¼Ë,. VŠ÷çeÿ n/¡¼äãõâ=‹xGQKx”|¹bÌŠD@2Œ 8'Ž àúƒŽ+áDÒ&¡¨"Œ§–Žr22 Ç·s]ŸÄ‹«ð%ÚÄ<¹ä’(×{e›HÀqÁç©Ç½`üŽÚõK饚9ƒÄ±€< – úƒú~ çðñO#Í%iKKlµ¦¾F)'Iê¬Î+Ç(`ñ¾£œdÈ’` ™ºcßéé^ÿ i¸”Û\ý¡æhÔB«aq¸}ãÀÆ:ÜWƒ|FÛÿ BŒÇÀeaŸ-sÊ€:úW½ÜÝÜ<%$µ†%CóDªÀí%IÈÏʤ…ôäñÞŒ÷‘a0“ôŽÚë¤nŸoW÷0«e¶y'Å»a Η2r’# Û°A^ý9ÉQÔõ=ù5 ¬£Öü.(Þ’M$ ~V«=éSÄFN½®©ÔWô»ÿ þHžkR‹ìÏ+µµžöê;khÚI¤m¨‹Ôš–âÖçJ¾_Z•’6 a” Èô> ÕÉaÕ<%®£2n bQŠå\tÈõUÿ ø»þ‹k15‚ÃuCL$ݹp P1=Oøýs¯^u éEJ”–éêŸê½5ýzy›jÛ ³á›Ûkÿ ÚOcn±ÛÏîW;boºz{ãžüVÆ¡a£a5½äÎÂks¸J@?1è¿{$ä‘=k”øsÖ^nŒ¦)ÝåXÃíùN1ØõÚOJë–xF÷h¸ Œ"Ž?x䜚ü³ì¨c*Fœ¯i;7~ñí׫Ðó¥Ë»3Ãü púw ‰°<Á%»ñž ÿ P+Û^ ¾Ye£ŽCÄŒ„/>˜ >•á¶Ìm~&&À>M[hÈÈÿ [Ž•íd…RO@3^Ç(ʽ*¶ÖQZyßþ
1Vº}Ñç?¼O4Rh6R€ª£í¡ûÙ
a‚3ß· Õ
ü=mRÍ/µ9¤‚0ÑC¼Iè:cŽsÛ¾™x£ÆЬªÍöˢ샒W$• €Å{¨ÀPG
ÀÀàŸZìÍ1RÉ0´ðxEË9+Éÿ ^rEÕ—±Š„70l¼áË@û.' ¼¹Žz€N3úUÉ<3á×*?²¬‚ä†"Ùc=p íÛ'¡ª 1ñ"økJ†HÒ'»Ÿ+
oÏN¬Ã9 dÙãÜדÏâÍ~æc+j·Jzâ7(£ð W]•æ™?nê´º6åwéåç÷N•ZŠíž›¬|?Ðõ?Ñ-E…®³Ç V$~X¯/…õ x‘LˆÑÜÚÈ7¦p zãÜüë½ðÄ^õtÝYËÍ7ÉÖÕ8ÏUe# #€r=sU¾/é’E§jRC4mxNÝ´9†íuá»›V‘
ZI€×cr1Ÿpzsøf»¨åV‹ìû`qËLÊIã?\~¼³á ËC©êhªOîO»‘ÃmçÛçút ×¢x“Z}?Ãœê#b-¤X7õ Äò gž zzbº3œ m*qvs·M=íúéw}¿&Úª°^Ö×µÏ(ø‡â†Öµƒenñý†×åQáYûœ÷ÇLœôÎNk¡ð‡¼/µ¸n0æÉ0¬ƒ‚üîÉÆvŒ w®Sáö”š¯‹-üÕVŠØÙ[$`(9cqƒÔ_@BëqûÙ`Ýæ0;79È?w<ó |ÙÜkßÌ1±Ëã ¿ìÒ»ðlìï«ÓnªèèrP´NÏš&Žéö Ù¸÷æ°~-_O'‰`°!RÚÚÝ%]Ø%þbß1'¿ÿ X ÕáOöÎŒ·‹¬+Åæ*ÛÛ™ 0¤ƒOÍÔ `u¯¦ÂaèÐÃÓ«‹¨Ô¥µœ¿¯ÉyÅÙ.oÔôŸ Úx&(STðݽ¦õ] ’ÒNóÁäÈùr3í·žÚ[™ƒ¼veÈ÷ÞIõÎGlqÎ=M|«gsªxÅI6
]Z·Îªä,¨zŒŽÄ~#ØŠúFñiÉqc©éÐD>S딑 G ñŽ1éÐ^+
Ëi;Ô„µVÕú»i¯ÈÒ-ZÍ]òܘ®ì` bÛÙ¥_/y(@÷ qÐúg Ô÷W0.Ø›
6Ò © r>QƒŒ0+Èîzb¨É+I0T bNñ"$~)ÕÒ6Þ‹{0VÆ27œ WWñcÄcX×íôûyKZéðªc' iQ¿¯LaW ŠŸS\·Š“źʸ…ôÙÂí |öÀÇåV|!¤ÂGâÛ[[’ï
3OrÙËPY¹=Î1õ5öåTžÑè Ú64/üö?Zëžk}¬¶éàoá¾á }3“ü]8Éæ¿´n²Žš_6¾pœ) 2?úWÓÚ¥¾¨iWúdŽq{*ª1rXŒd…m»‰äcô¯–dâ•ã‘Jº¬§¨#¨® §,df«8ÉÅßN¾hˆ;îÓ=7áùpën®É 6ûJžO2 ^œÐò JÖø¥²ã›Ò6Ü·‰!wbÍ‚¬O© »õ¬ÿ ƒP=Ä:â¤-& ÙŽ
` È9 r9íϧzë> XÅ7ƒ5X–krÑ¢L 7€ìw}ÑŸNHëŒüþ:2†á¼+u·á÷N/Û'Ðç~ߘô«ëh!ónRéeQ´6QÛÿ èEwëÅÒ|¸Yqó1uêyùzð8 ƒŠù¦Ò;¹ä6öi<'ü³„[ ÃZhu½ ùÍ¡g‚>r¯ ׊îÌx}bñ2“k꣧oø~›hTèóËWò4|ki"xßQ˜Ï6øÀLnß‚0 ¹Æ{±–¶Öe#¨27È@^Ìß.1N¾œyç€õ†ñeé·Õã†çQ°€=Ì©ºB€Ø8<‚à Sõ®ùcc>×Ú .Fr:žÝGæ=kÁâ,^!Fž
¬,àµ}%¶«îõ¹†"r²ƒGœüYÕd?aÑÃY®49PyU ÷þ!žxÅm|/‚ãNð˜¼PcûTÒ,¹/Ý=FkÏ|u¨¶«âë…{¤m¢]Û¾ïP>®XãÞ½i ÓÁ¾
‰'¬–6ß¼(„ï— í!ú Ùäzôë^–:œ¨å|,_¿&š×]uÓѵÛô4’j”bž§x‘Æ©ã›á,‚[Ô
ÎÞ= ŒËæ ÀùYÁ?ŽïÚ¼?ÁªxºÕÛ,°1¸‘¿Ýäãدv…@¤åq½ºã œàûââ· z8Xýˆþz~—û»™âµj=Ž
â~ãáh@'h¼F#·Üp?ŸëQü-løvépx»cŸø…lxâÃûG·‰¶ø”L£©%y?¦úõÆü-Õ¶¥y`Òl7>q’2üA?•F}c‡jB:¸Jÿ +§¹¿¸Q÷°ív= VÑìu[Qml%R7a×IèTõéŽx¬
?†š7
1†îã-ˆã’L¡lŽ0OÓ=ÅuˆpÇ• ¼3ÛùÒ¶W/!|’wŽw ^qÔ×Ïaó M8Q¨ãÑ?ëï0IEhÄa¸X•`a
?!ÐñùQ!Rä žqŽžÝO`I0ÿ J“y|ñ!Îã@99 >þ8–+éáu…!ù—ä
Ê°<÷6’I®z
ÅS„¾)Zþ_Öýµ×ËPåOwø ÷þ*üïænÖùmØÝûþ¹=>¦½öî×Jh]¼ç&@§nTŒ6IT Àõ^Fxð7Å3!Ö·aÛ$þÿ ¹ã5îIo:Ȫm ËY[’8ÇӾlj*òû¢¥xõ¾¼ú•åk+\ð¯ HÚoŽl•Ûk,¯ ç ²²cõÅ{²Z\
´ìQ åpzŽ3Ôð}ÿ Jð¯XO¡øÎé€hÙ¥ûLdŒ`““ù6Gá^ÃáÝ^Ë[Ñb¾YåŒÊ»dŽ4 †2§,;ÿ CQÄ´¾°¨c–±”mºV{«ßÕýÄW\ÖŸ‘çŸ,çMRÆí“l-ƒn~ë©ÉÈê Ü?#Ž•¹ðãSÒ¥ÐWNíà½;ãž)™ÎSÈ9cóLj뵿ūiÍk¨ió¶X‚7÷ƒ€yãnyÏŽëÞ Öt`×À×V's$È9Ú:ä{wÆEk€«†Çàc—â$éÎ.éí~Ýëk}ÅAÆpörÑ¢‡Šl¡ÑüSs ‹¨‰ I ÄóÀ×wñ&eºðf™pŒÆ9gŽTø£lñëÀçŽ NkÊUK0U’p ï^¡ãÈ¥´ø{£ÙHp`’ØåbqÏ©äó^Æ:
Ž' ÊóM«õz+ß×ó5Ÿ»('¹ð¦C„$˜Å¢_ºÈI?»^äã'ñêzž+ë€ñ-½»´}¡Ë*õ?.xÇ^1ŽMyǸ&“—L–îëöâ7…' bqéÎGé]˪â1$o²¸R8Ã`.q€}sÖ¾C98cêÆÞíïóòvÓòùœÕfÔÚéýuèÖ·Ú
Å‚_¤³ÜۺƑß”àר ý:׃xPþÅÕî-/üØmnQìïGΊÙRqê=>¢½õnæ·r!—h`+’;ò3È<“Û©éšóŸ x*÷V¹¸×tÈiˆßwiÔÿ |cŒñÏ®3Ö½ Ì°‰Ë Qr©ö½®¼ÛoÑÙZÅÑ«O൯ýw8;k›ÿ x†;ˆJa;‘º9÷÷R+¡ñgŽí|Iáë{ôáo2ʲ9 029ÉÏLí\‰¿¸Ÿb˜ "Bv$£&#ßiê>=ªª©f
’N ëí>¡NXW ~5×úíø\‰»½Ï^ø(—wÖú¥¤2íŽÞXæÁ$ °eÈ888^nÝë²ñÝÔ^ ÖÚ9Q~Ëå7ï
DC¶ÑµƒsË Çè9®Wáþƒ6‡£´·°2\Ý:ÈÑ? (#¨'$õèGJ¥ñW\ÿ ‰E¶—¸™g˜ÌÀ¹;Pv ú±ÎNs·ëŸ’–"Ž/:té+ûË]öJöÓM»ëø˜*‘•^Uý—êd| ‰åñMæÔÝ‹23å™6æHùÛ‚ëüñ^…ñ1¢oêûÑEØ.õ7*ÅHtÎp{g<·Á«+¸c¿¿pÓ¾Æby=8É_ÄsÆk¬ñB\jÞ Ôì••Ë[9Píb‹Bヅ =93§ð§LšÛáÖšÆæXÌÞdÛP.0\ãïÛ 0?™úJ¸™Ë
”•œº+=<µI£¦í¯õêt¬d‹T¬P=ËFêT>ÍØØ@Ï9<÷AQÌ×»Õ¡xùk",JÎæù±Éç$œŽŸZWH®¯"·UÌQ ’ÙÈ]ÅXg<ã
ߨg3-Üqe€0¢¨*Œ$܃
’Sû 8㎼_/e'+Ï–-èÓ¶¶Õíß[·ÙÙ½îì—¼sk%§µxä‰â-pÒeÆCr ú
ôσžû=”šÅô(QW‚Õd\ƒæ. \àö ¹¯F½°³½0M>‘gr÷q+œ¶NïºHO— ¤ ܥݔn·J|ÆP6Kµc=Isó}Ò çGš)a=—#vK›åoK§ßóÙ¤¶¿õú…ÄRÚ[ËsöÙ¼Ë•Ë ópw ®qœŒ·Ø
ùÇâ ‹ý‡ãKèS&ÞvûD Aù‘É9 Œî qÅ}
$SnIV[]Ñ´Ó}ØܾA Ü|½kÅþÓ|E Mu R¼.I¼¶däò‚ÃkÆ}ðy¹vc iUœZ…Õõ»z¾÷¿n¦*j-É/àœHã\y5 Û ß™ó0— äŸnzôã#Ô¯,†¥ÚeÔ÷ÜÅ´„“'c…<íÝ€< ·SŠ¥k§Ã¢éÆÆÙna‚8–=«Êª[Ÿ™°pNî02z“ÔÙ–K8.È’Þî(vƒ 2®@ ä Èûã çžxäÇf¯ˆu¹yUÕîýWšÙ|›ëÒ%Q^í[æ|éo5ZY•^{9 6ˆY‚§v*x>âº_|U¹Ö´©tûM ÒÂ9PÇ#«£# € éÉñ‘ƒÍz/‰´-į¹°dd,Б ›p03ƒœ{ç9=+
Ûᧇ¬¦[‡‚ê婺¸#±ß= ³ý¿•Õµjñ½HÙh›Û[§ÚýÊöô÷{˜?ô÷·Ô.u©–_%còcAÀ˜’
}0x9Î>žñÇáÍ9,ahï¦Ì2òÓ ñÛAäry$V²Nð
]=$Ž
‚#Ù‚1ƒƒødõ Max‡ÂÖ^!±KkÛ‘
«“Çó² FN8+ëÎ{Ò¼oí§[ «ÕMRoËeç×[_m/¦¦k.kôgŽxsSÓ´ý`êzªÜÜKo‰cPC9ÎY‰#§^üý9¹âïÞx£Ë·Ú`±‰‹¤;³–=ÏaôÕAð‚ ÷kêÁNBéÎælcõö®£Fð†ô2Ò¬]ßÂK$ÓÜ®•”/ÊHàã$ä ¸÷ëf¹Oµúâ“”’²øè´µþöjçNü÷üÌ¿ xNïFÒd»¼·h®îT9ŽAµÖ>qÁçÔœtïÒ»\ȶÎîcÞäîó3¶@#É IÎ ÔñW.<´’¥–ÑÑ€ÕšA‚ ;†qÓë‚2q
ÒÂó$# Çí‡
!Ë}Õ9ÈÎÑÉã=;ŒÇÎuñ+ÉûÏ¥öíeÙ+$úíÜ娯'+êZH4ƒq¶FV‹gïŒ208ÆÌ)íб>M|÷âÍã¾"iì‹¥£Jd´™OÝç;sÈúr+ÜäˆË)DŒ¥šF°*3Õ”d {zÔwºQ¿·UžÉf†~>I+ŒqÔ`ð3œ“Ü×f]œTÁÔn4“ƒø’Ýßõ_«*5šzGCÊ,þ+ê1ò÷O¶¸cœºb2y Ç ;cùÕ£ñh¬›áÑŠr¤ÝäNBk¥—á—†gxšX/쑘hŸ*Tçn =û㦠2|(ð¿e·ºÖ$
ýìŸ!'åÎ °yîî+×öœ=Y:²¦ÓÞ×iü’—ü
-BK™£˜›âÆ¡&véðõ-ûÉY¹=Onj¹ø¯¯yf4·±T Pó`çœ7={×mÃ/ ¢˜ZÚòK…G½¥b„’G AãÜœ *í¯Ã¿ IoæI¦NU8‘RwÈã;·€ Û×ëÒ”1Y
•£E»ÿ Oyto¢<£Áö·šï,ä‰ §ûA¼sû»Nò}¹üE{ÜÖªò1’õÞr0â}ÎØ#>à/8ïéÎ~—áÍ#ñÎlí§³2f'h”?C÷YËdð:qëõÓ·‚ïeÄ©
ÔÈØÜRL+žAÎ3¼g=åšó³Œt3
ÑQ¦ùRÙßE®¼±w_;þhš’Sirÿ ^ˆã¼iੇ|RòO„m°J/“$·l“ ÇÓ¿ÿ [ÑŠÆ“„†Õø>cFÆ6Ø1ƒ– àz7Ldòxäüwá‹ÝAXùO•Úý’é®ähm •NÀ±ÌTÈç
ƒ‘I$pGž:‚ÄbêW¢®œ´|¦nÍ>¶ÖÏ¢§ÎÜ¢ºö¹•%ÄqL^öÛ KpNA<ã¡ …î==ª¸óffËF‡yÌcÉ ©ç$ð=ñÏYþÊ’Ú]—¥‚¬‚eDïÎH>Ÿ_ÌTP™a‰ch['çÆÜò7a‡?w°Ïn§âÎ5”’¨¹uÚÛ|´ÓÓc§{O—ü1•ªxsÃZ…ÊÏy¡Ã3¸Ë2Èé» ‘ƒÎ äžÜðA§cáOéúÛ4ý5-fŒï„ù¬ûô.Ç Üsž•Ò¾•wo<¶Ÿ"¬¡º|£
î2sÇ¡éE²É FѱrU°dÜ6œ¨ mc†Îxë׺Þ'0²¡Rr„{j¾í·è›µ÷)º·å–‹î2|I®Y¼ºÍË·–ÃÆà ã£'óÆxƒOÆÞ&>\lóÌxP Xc ¸ì Sþ5§qà/ê>#žÞW¸if$\3 ® ûÄ“ùŽÕê¾ð< Ó‹H¶óÏ" å·( á‘€:ã†8Ï=+ꨬ UA×ÃËÚT’ÑÞöù¥¢]{»ms¥F0\ÑÕ—ô}&ÛB´ƒOŽÚ+›xíÄÀ1
,v± žIëíZ0ǧ™ 3 í2®0ภ—p9öÝÔž)ÓZËoq/Ú“‘L ²ŒmùŽï‘Ó9§[Û#Ä‘\ÞB¬Çs [;à à«g‚2ôòªœÝV§»·¯/[uó½õÛï¾
/šÍ}öüÿ « =x»HŸÂÞ.™ ÌQùŸh´‘#a$‚'¡u<Š›Æ>2>+ƒLSiöwµFó1!eg`£åœ ÷ëÛö}Á¿ÛVÙêv $¬ ƒ|,s÷z€ð΃¨x÷ÅD\Ü ŒÞmåÔ„ ˆ o| :{ÇÓ¶–òÁn!´0Ål€, ƒ ( ÛŒŒ c¶rsšæ,4‹MÛOH!@¢ ÇŽ„`å²9ÝÃw;AÍt0®¤¡…¯ØÄ .Àìí´ƒ‘ßñ5Í,Óëu-ÈÔc¢KÃÓ£òÖ̺U.õL¯0…%2È—"~x
‚[`có±nHàŽyàö™¥keˆìŒÛFç{(Ø©†`Jã#Žwg<“:ÚÉ;M
^\yhûX ‡vB·÷zrF?§BÊÔ/s< ÐÈB)Û± ·ÍÔwç5Âã:så§ e{mѤï«Òíh—]Wm4âí¿ùþW4bC3¶ª¾Ùr$ pw`àädzt!yŠI„hÂîàM)!e dŒm'æ>Ç?w zºKìcŒ´¯Ìq6fp$)ãw¡éUl`µ»ARAˆÝÕgr :äŒg ƒéé[Ôö±”iYs5Ýï«ÙG—K=þF’æMG«óÿ `ŠKɦuOQ!ÕåŒ/ÎGÞ`@ËqÕzdõâ«Ê/Ö(ƒK´%ŽbMü åÜŸö—>¤óŒŒV‘°„I ¢Yž#™¥ùÏÊ@8
œgqöö5ª4vד[¬(q cò¨À!FGaÁõõ¯? §†¥ÏU½í¿WªZ$úyú½ŽzקÉþ?>Ã×È•6°{™™ŽÙ.$`ÎU œ…çè ' ¤r$1Ø(y7 ðV<ž:È ÁÎMw¾Â'Øb§øxb7gãО½óÉÊë²,i„Fȹ £§8ãä½k¹¥¦ê/ç{ïê驪2œ/«ü?¯Ô›ìñÜ$þeýœRIåŒg9Ác’zrrNO bÚi¢
ѺË/$,“ª¯Ýä;Œ× ´<ÛÑn³IvŸb™¥ nm–ÄŸ—nÝÀãŽ3ëÍG,.öó³˜Ù£¹u ÊÌrŠ[<±!@Æ:c9ÅZh
ì’M5ÄìÌ-‚¼ëÉùqŽGì9¬á ;¨A-ž—évþÖ–^ON·Ô”ŸEý}ú×PO&e[]ÒG¸˜Ûp ƒÃà/Ë·8ûÀ€1ž@¿ÚB*²¼ñì8@p™8Q“žÆH'8 «I-%¸‚
F»“åó6°Uù|¶Ú¸ã ò^Äw¥ŠÖK–1ÜÝK,Žddlí²0PÀü“×ükG…¯U«·¶–´w¶ŽÍ¾©yÞú[Zös•¯Á[™6°
¨¼ÉVæq·,#
ìãï‘×8îry®A››¨,ãc66»Ë´ã'æÉù?t} ¢æH--Òá"›|ˆ¬[í 7¶ö#¸9«––‹$,+Ëqœ\Êø c€yê^ݸÄa°«™B-9%«×®‹V´w~vÜTéꢷþ¼ˆ%·¹• ’[xç•÷2gØS?6åÀÚ õ9É#š@÷bT¸º²C*3Bá¤òÎA9 =úU§Ó"2Ãlá 0i ÝIc‚2Î@% öç94ùô»'»HÄ¥Ô¾@à Tp£šíx:úÊ:5eºßMý×wµ›Ó_+šº3Ýyvÿ "ºÇ<ÂI>Õ 1G·Ë«È«É# àÈÇ øp Jv·šæDûE¿›†Ë’NFr2qŸ½ÇAÜšu•´éí#Ħ8£2”Ú2Ã/€[ÎTr;qŠz*ý’Îþ(≠;¡TÆâ›;ºÿ àçœ k‘Þ8¾Uª¾íé{^×IZéwÓkXÉûÑZo¯_øo×È¡¬ â–ÞR§2„‚Àœü½ùç® SVa†Âüª¼±D‘ŒísŸàä|ä 2 æ[‹z”¯s{wn„ÆmáóCO+†GO8Ïeçåº`¯^¼ðG5f{Xžä,k‰<á y™¥voÆ éÛõëI=œ1‹éíÔÀÑ)R#;AÂncäŽ: tÏ#¶TkB.0Œ-ÖÞZÛgumß}fÎJÉ+#2êÔP£žùÈÅi¢%œ3P*Yƒò‚Aì“Ž2r:ƒÐúñiRUQq‰H9!”={ ~¼ “JŽV¥»×²m.ÛߺiYl¾òk˜gL³·rT•
’…wHÁ6ä`–Î3ù Ì4Øe³†&òL‘•%clyîAÂäà0 žüç$[3uŘpNOÀÉ=† cï{rYK
ååä~FÁ
•a»"Lär1Ó¯ 2Äõæ< ™C•.fÕ»è¥~½-¿g½Â4¡{[ør¨¶·Žõäx¥’l®qpwÇ»8ärF \cŽ¸ÜÆÓ-g‚yciÏÀ¾rÎwèØÈ#o°Á9ã5¢šfÔxÞæfGusÏÌJÿ µ×œ/LtãÅT7²¶w,l
ɳ;”eúà·¨çîŒsÜgTÃS¦^ '~‹®›¯+k÷ZÖd©Æ*Ó[Ü«%Œk0ŽXƒ”$k#Ȩ P2bv‘ƒŸáÇ™ÆÕb)m$É*8óLE‘8'–ÜN Úyàúô+{uº±I'wvš4fÜr íì½=úuú
sFlìV$‘ö†HÑù€$§ õ=½¸«Ž]
:Ž+•¦ïmRþ½l´îÊT#nkiøÿ _ðÆT¶7Ò½ºÒ £Î¸d\ã8=yãŽÜ äR{x]ZâÚé#¸r²#» ÎHÆ6õ ç® ÎFkr;sºÄ.&;só± Ç9êH÷ýSšÕtÐU¢-n Ì| vqœ„{gŒt§S.P‹’Ž‰_[;m¥ÞZýRûÂX{+¥úü¼ú•-àÓ7!„G"“´‹žƒnrYXã¸îp éœ!ÓoPÌtÑ (‰Þ¹é€sÓ# GLçÕšÑnJý¡!‘Tä #“ß?îýp}xÇ‚I¥Õn#·¸–y'qó@r[ Êô÷<ÔWÃÓ¢áN¥4Ô’I&ݼ¬¬¼ÞºvéÆ
FQV~_ÒüJÖÚt¥¦Xá3BÄP^%ÈÎW-×c¡ú©¤·Iþèk¥š?–UQåIR[’O 5x\ÉhÆI¶K4«2ùªŠŒ<¼óœ çØ`u«‚Í.VHä € Ëgfx'' 9ÆI#±®Z8
sISºku¢ßÞ]úk»Jößl¡B.Ü»ÿ MWe
°·Ž%šêɆ¼»Âù³´œ O¿cÐ ÓÄh©"ÛÜÏ.ÖV ’3nüÄmnq[ŒòznšÖ>J¬òˆæ… qýØP Ž:ä7^0yëWšÍ_79äoaÈ °#q0{ää×m œy”R{vÒÞ¶ÚÏe¥“ÚÆХ̮—õýjR •íç›Ìb„+J yÜØÙ•Ç]¿Ôd þËO L²”9-Œ— õÃc'æÝלçÚ²ìejP“½
âù°¨†ðqòädЃÉäÖÜj÷PÇp“ÍšŠå«‘î
< iWNsmª»¶vÓz5»ûì:Rs\Ðßôû×uÔÿÙ