Beefy Boxes and Bandwidth Generously Provided by pair Networks
There's more than one way to do things
 
PerlMonks  

Question about threads

by Gwalchmai (Novice)
on Aug 26, 2003 at 12:46 UTC ( [id://286681]=perlquestion: print w/replies, xml ) Need Help??

Gwalchmai has asked for the wisdom of the Perl Monks concerning the following question:

I am trying to fire off 13 simultaneous threads. I have noticed that when I add a new thread to the code below, the wait time increases. I have checked each of the threads individually and they only take 10 - 23 seconds. But after I had put in the eight one, it was up to 100 seconds total. It seems like the code is going sequential instead of parallel.

Please help. 

#!/opt/perl5.8t/bin/perl

######################################################################
+##########
# #
# Program: distributedSearch.pl #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This script allows for the simulataneous searching of W
+eb- #
# based databases through the use of threads. The results are #
# collected and displayed together. #
# #
# Updates:                              #
#     Date Name Description of update #
# 08/04/03    JDB    Added entry for European Fusion Development Agree
+ment #
# (EFDA). #
# 08/06/03    JDB    Added the following 5 databases located at the #
# University of Manchester Institute of Science and #
# Technology: #
# Astro Physics #
# Atmospheric Physics #
# Condensed Matter #
# Plasma Physics #
# Theoretical Physics #
# 08/15/03    JDB    Added entries for Advanced Knowledge Technologies
+ and #
# Australian National University #
#                                      #
######################################################################
+##########
use threads;
use threads::shared;
use CGI qw/:standard/;

# Arrays containing search key words.
@title    = param('keywords');
@abstract = param('keywords');
@author   = qw( );

# Hash of search terms
my %searchTerms = (
    'title'    => [@title],
    'abstract' => [@abstract],
    'author'   => [@author],
);

# Maximum number of hits to return from any one site
my $maxHits = param('hits');

# Array used by all threads to store hits
my @searchResults : shared = undef;

# Create threads
$thr1 = threads->new( \&citeSeer )
  if param('CiteSeer');
$thr2 = threads->new( \&spiresHEP )
  if param('SLAC SPIRES-HEP');
$thr3 = threads->new( \&sfb288Preprints )
  if param('SFB 288 Preprints');
$thr4 = threads->new( \&chalmers )
  if param('Chalmers University of Technology');
$thr5 = threads->new( \&cleo )
  if param('CLEO');
$thr6 = threads->new( \&efda )
  if param('European Fusion Development Agreement');
$thr7 = threads->new( \&umist_AstroPhysics )
  if param('UMIST - Astro Physics');
$thr8 = threads->new( \&umist_AtmosphericPhysics )
  if param('UMIST - Atmospheric Physics');
$thr9 = threads->new( \&umist_CondensedMatter )
  if param('UMIST - Condensed Matter');
$thr10 = threads->new( \&umist_PlasmaPhysics )
  if param('UMIST - Plasma Physics');
$thr11 = threads->new( \&umist_TheoreticalPhysics )
  if param('UMIST - Theoretical Physics');
$thr12 = threads->new( \&advancedKnowledgeTechnologies )
  if param('Advanced Knowledge Technologies');
$thr13 = threads->new( \&australianNationalUniversity )
  if param('Australian National University');

# Do not end program until all threads have finished
$thr1->join()  if param('CiteSeer');
$thr2->join()  if param('SLAC SPIRES-HEP');
$thr3->join()  if param('SFB 288 Preprints');
$thr4->join()  if param('Chalmers University of Technology');
$thr5->join()  if param('CLEO');
$thr6->join()  if param('European Fusion Development Agreement');
$thr7->join()  if param('UMIST - Astro Physics');
$thr8->join()  if param('UMIST - Atmospheric Physics');
$thr9->join()  if param('UMIST - Condensed Matter');
$thr10->join() if param('UMIST - Plasma Physics');
$thr11->join() if param('UMIST - Theoretical Physics');
$thr12->join() if param('Advanced Knowledge Technologies');
$thr13->join() if param('Australian National University');

# Display results                                                     
+                                                                     
+    
&displayResults;

######################################################################
+##########
# #
# Subroutine: displayResults #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine displays the results returned by the th
+reads. #
# #
# Updates:                              #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub displayResults {

    my $reportNo;
    my $title;
    my $author;
    my $publisher;
    my $source;
    my $keywords = param('keywords');

    @searchResults = rank(@searchResults);

    print header,
      start_html(
        -title      => 'Distributed Search Results',
        -background => 'images/happyface.gif'
      ),
      h1("Distributed Search Results"), h3("Search term(s): $keywords"
+);

    foreach (@searchResults) {
        next unless ($_);
        ( $reportNo, $title, $author, $publisher, $source ) = split (/
+\t/);

        # Display
        print "Report No: ", $reportNo,  "<br>" if ($reportNo);
        print "Title: ",     $title,     "<br>" if ($title);
        print "Author: ",    $author,    "<br>" if ($author);
        print "Publisher: ", $publisher, "<br>" if ($publisher);
        print "Source: ", $source, "<br><br>";
    }
    print hr, a( { -href => './search.pl' }, "Return to search page." 
+),
      end_html;
}

######################################################################
+##########
# #
# Subroutine: citeSeer #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub citeSeer {
    use CiteSeer;
    CiteSeer::search( $maxHits, %searchTerms );
    my @results = CiteSeer::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: spiresHEP #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub spiresHEP {
    use SpiresHEP;
    SpiresHEP::search( $maxHits, %searchTerms );
    my @results = SpiresHEP::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: sfb288Preprints #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub sfb288Preprints {
    use SFB288Preprints;
    SFB288Preprints::search( $maxHits, %searchTerms );
    my @results = SFB288Preprints::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: chalmers #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub chalmers {
    use Chalmers;
    Chalmers::search( $maxHits, %searchTerms );
    my @results = Chalmers::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: cleo #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: June 10, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub cleo {
    use CLEO;
    CLEO::search( $maxHits, %searchTerms );
    my @results = CLEO::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: efda #                   
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 # 
# # 
# Version: 1.0.0 #
# # 
# Release Date: August 4, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub efda {
    use EFDA;
    EFDA::search( $maxHits, %searchTerms );
    my @results = EFDA::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
} 
 
######################################################################
+##########
# #   
# Subroutine: umist_AstroPhysics #     
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 #
# # 
# Version: 1.0.0 #
# # 
# Release Date: August 6, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub umist_AstroPhysics {
    use UMIST_AstroPhysics;
    UMIST_AstroPhysics::search( $maxHits, %searchTerms );
    my @results = UMIST_AstroPhysics::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
} 
 
######################################################################
+##########
# #   
# Subroutine: umist_AtmosphericPhysics #
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 #
# # 
# Version: 1.0.0 #
# # 
# Release Date: August 6, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub umist_AtmosphericPhysics {
    use UMIST_AtmosphericPhysics;
    UMIST_AtmosphericPhysics::search( $maxHits, %searchTerms );
    my @results = UMIST_AtmosphericPhysics::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
} 
 
######################################################################
+##########
# #   
# Subroutine: umist_CondensedMatter #  
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 #
# # 
# Version: 1.0.0 #
# # 
# Release Date: August 6, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub umist_CondensedMatter {
    use UMIST_CondensedMatter;
    UMIST_CondensedMatter::search( $maxHits, %searchTerms );
    my @results = UMIST_CondensedMatter::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: umist_PlasmaPhysics #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: August 6, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub umist_PlasmaPhysics {
    use UMIST_PlasmaPhysics;
    UMIST_PlasmaPhysics::search( $maxHits, %searchTerms );
    my @results = UMIST_PlasmaPhysics::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: umist_TheoreticalPhysics #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: August 6, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub umist_TheoreticalPhysics {
    use UMIST_TheoreticalPhysics;
    UMIST_TheoreticalPhysics::search( $maxHits, %searchTerms );
    my @results = UMIST_TheoreticalPhysics::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

######################################################################
+##########
# #
# Subroutine: advancedKnowledgeTechnologies #
# #
# Programmer: Doug Bales #
# #
# Language: Perl 5.8 #
# #
# Version: 1.0.0 #
# #
# Release Date: August 15, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub advancedKnowledgeTechnologies {
    use AdvancedKnowledgeTechnologies;
    AdvancedKnowledgeTechnologies::search( $maxHits, %searchTerms );
    my @results = AdvancedKnowledgeTechnologies::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
} 
 
######################################################################
+##########
# #   
# Subroutine: australianNationalUniversity #
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 #
# # 
# Version: 1.0.0 #
# # 
# Release Date: August 15, 2003 #
# #
# Description: This subroutine uses the individual site module to form
+ulate a #
# site specific search request, connect to and search the site, #
# collect and return data. The subroutine placed the returned #
# data in the shared array @searchResults. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub australianNationalUniversity {
    use AustralianNationalUniversity;
    AustralianNationalUniversity::search( $maxHits, %searchTerms );
    my @results = AustralianNationalUniversity::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
} 
 
######################################################################
+##########
# #   
# Subroutine: rank #                   
# #
# Programmer: Doug Bales #
# # 
# Language: Perl 5.8 #
# # 
# Version: 1.0.0 #
# # 
# Release Date: June 12, 2003 #
# #
# Description: This subroutine takes an array containing tab delimited
+ data #
# and performs a reverse sort on that data. An array containing #
# the sorted data is returned. #
# #
# Updates:                          #
#     Date Name Description of update #
#                                      #
######################################################################
+##########
sub rank { 
    my (@origArray) = @_;
    my @keywords  = param('keywords');
    my $keyword   = "";
    my @tempArray = undef;
    my @newArray  = undef;
    my %hash      = undef; 
    my $temp      = "";
    my $score     = 0;
    my $count     = 0;              

    foreach my $data (@origArray) {    

        # Increment count
        $count++;
    
        # Reset score value to zero
        $score = 0;
    
        # Loop for each keyword
        foreach $keyword (@keywords) {

            # Score value for exact word matches
            $score += wordCount( $keyword, $data ) * .1;

            # Score value for plural word matches
            $temp = $keyword . "s";
            $score += wordCount( $temp, $data ) * .01;
            $temp = $keyword . "es";
            $score += wordCount( $temp, $data ) * .01;
            $temp = $keyword . "ies";
            $score += wordCount( $temp, $data ) * .01;

            # Scored value for past or action matches
            $temp = $keyword . "ed";
            $score += wordCount( $temp, $data ) * .001;
            $temp = $keyword . "ing";
            $score += wordCount( $temp, $data ) * .001;
        }

        # Score value for placement in returned list
        $score *= ( 1 - ( $count * .0001 ) );
    
        # Populate temp array with score and original information
        $data = $score . "\t" . $data;
        push ( @tempArray, $data );
    }
    
    # Sort temp array
    @tempArray = reverse( sort (@tempArray) );

    # Remove scores from sorted data
    foreach (@tempArray) {
        m/\A.*?\t(.*)/;
        push ( @newArray, $1 ) unless ( $1 =~ /\A\s*?\Z/ );
    }
    return (@newArray);
}

sub wordCount {
    my ( $word, $data ) = @_;
    my $num = 0;
    while ( $data =~ m/\b$word\b/ig ) { $num++; }
    return ($num);
}

[download]

update (broquaint): added <readmore> + <code> tags and ran through perltidy

Replies are listed 'Best First'.
Re: Question about threads
by liz (Monsignor) on Aug 26, 2003 at 13:46 UTC
    I haven't looked at your code yet. I might later. It's pretty massive ;-)

    Perl threads have that problem, yes. The problem is mainly caused by the fact that all data structures of Perl are copied to each new thread, shared and unshared. Every time you start a new thread. Only opcode trees appear are shared (although I have to take Arthur Bergman's word for that as I never dared go into the actual Perl core to see what's going on there).

    This means that you should try to reduce your data-structures before firing off threads. There are several ways to do that: one of them is Thread::Use which allows you to use modules only in threads. Or you could use Thread::Pool to start up a number of worker threads way at the beginning of your program. And you might want to look at monitored queues, either Thread::Queue::Monitored or Thread::Conveyor::Monitored.

    Arthur is now working on using a different approach to shared variables which involves some COW (Copy On Write) techniques. But this is for Ponie so it will take some time before that becomes mainstream.

    Finally, there is one other alternative that I can recommend if you're not on Win32: forks. Instead of threading, it forks so the OS is taking care of all the COWing. It does this at the expense of using a TCP connection for communication between threads. YMMV

    Hope this helps.

    Liz

[reply]
Re: Question about threads
by jeffa (Bishop) on Aug 26, 2003 at 14:17 UTC
    If you are on a *NIX box, type cat /proc/cpuinfo to see how many processors you have on the machine this code runs on. Remember, multi-threading works better on a multi-processor box.

    Having said that, why aren't you being more general? You have a subroutine for each thread that does the same thing: 

    sub australianNationalUniversity {
    use AustralianNationalUniversity;
    AustralianNationalUniversity::search( $maxHits, %searchTerms );
    my @results = AustralianNationalUniversity::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

    [download]
    You should generalize that. Something like this untested piece of code: 
    # at the beginning of your code:
use GenericSearch;

# and then just ONE sub for all threads
sub get_search {
    my $group = shift;
    GenericSearch::search($group, $maxHits, %searchTerms );
    my @results = GenericSearch::getResults();
    lock(@searchResults);
    push ( @searchResults, @results );
}

    [download]
    Also, why hold each thread in a seperate variable? Why not store them in an array instead? Because you have 13 seperate search items? Nonesense! Just "tag" them with a leading underscore (or whatever) so you can grep them out of CGI::param() like so: 
    my @search = grep /^_/, param();

# now remove the underscores
$_ = substr($_,1) for @search;

# now create the threads;
my @thread = map { threads->new(sub {get_search($_)})}, @search;

# and finally, join them when done
$_->join for @thread;

    [download]
    Code smart, not hard. ;)

    jeffa

    L-LL-L--L-LL-L--L-LL-L--
-R--R-RR-R--R-RR-R--R-RR
B--B--B--B--B--B--B--B--
H---H---H---H---H---H---
(the triplet paradiddle with high-hat)
[reply]
[d/l]
[select]
Re: Question about threads
by benn (Vicar) on Aug 26, 2003 at 14:33 UTC
    Another factor may be simply your network connection - if a single search is taking 10-20 seconds and this *isn't* to do with the server that you're querying, then it looks like there's either a lot of data being transmitted back and forth, or you've got a slow connection. No matter how many threads you fire off, if they're all trying to receive 100K though a single 56K line then you're going to have to wait a while for all the results to come back.

    Cheers, Ben.

[reply]

Back to Seekers of Perl Wisdom

Log In?
Username:
Password:

What's my password?
Create A New User
Domain Nodelet?

www.com | www.net | www.org
Node Status?
node history
Node Type: perlquestion [id://286681]
Approved by Corion
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? • Last hour • Other CB clients
Other Users?
Others exploiting the Monastery: (6)
As of 2024-04-24 09:41 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found