tange/parallel
1
0
Fork 0
mirror of https://git.savannah.gnu.org/git/parallel.git synced 2024-11-21 21:47:54 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

parsort: --parallel=N does better what is expected.

Browse source
This commit is contained in:
Ole Tange 2023-02-18 22:23:46 +01:00
parent 5d5cdcf77f
commit c8e203dfeb
1 changed files with 36 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

43
src/parsort
View file

@ -24,6 +24,21 @@ multicore machine.
Hopefully these ideas will make it into GNU B<sort> in the future.
=head1 OPTIONS
Same as B<sort>. Except:
=over 4
=item B<--parallel=>I<N>
Change the number of sorts run concurrently to I<N>. I<N> will be
increased to number of files if B<parsort> is given more than I<N>
files.
=back
=head1 EXAMPLE
Sort files:
@ -37,7 +52,7 @@ Sort stdin (standard input) numerically:
=head1 PERFORMANCE
B<parsort> is faster on a file than on stdin (standard input), because
B<parsort> is faster on files than on stdin (standard input), because
different parts of a file can be read in parallel.
On a 48 core machine you should see a speedup of 3x over B<sort>.
@ -115,7 +130,7 @@ GetOptions(
"C" => \$opt::dummy,
"compress-program=s" => \$opt::dummy,
"T|temporary-directory=s" => \$opt::dummy,
"parallel=s" => \$opt::dummy,
"parallel=s" => \$opt::parallel,
"u|unique" => \$opt::dummy,
"S|buffer-size=s" => \$opt::dummy,
"s|stable" => \$opt::dummy,
@ -124,9 +139,20 @@ GetOptions(
$Global::progname = ($0 =~ m:(^|/)([^/]+)$:)[1];
$Global::version = 20230122;
if($opt::version) { version(); exit 0; }
@Global::sortoptions = grep { ! /^-D$/ }
shell_quote(@ARGV_before[0..($#ARGV_before-$#ARGV-1)]);
# Remove -D and --parallel=N
my @s = (grep { ! /^-D$|^--parallel=\S+$/ }
@ARGV_before[0..($#ARGV_before-$#ARGV-1)]);
my @sortoptions;
while(@s) {
my $o = shift @s;
# Remove '--parallel N'
if($o eq "--parallel") {
$o = shift @s;
} else {
push @sortoptions, $o;
}
}
@Global::sortoptions = shell_quote(@sortoptions);
$ENV{'TMPDIR'} ||= "/tmp";
sub merge {
@ -158,7 +184,9 @@ sub sort_files {
# Let GNU Parallel generate the commands to read parts of files
# The commands split at \n (or \0)
# and there will be at least one for each CPU thread
my @subopt = $opt::zero_terminated ? qw(--recend "\0") : ();
my @subopt;
if($opt::zero_terminated) { push @subopt, qw(--recend "\0"); }
if($opt::parallel) { push @subopt, qw(--jobs), $opt::parallel; }
# $uniq is needed because @files could contain \n
my $uniq = join "", map { (0..9,"a".."z","A".."Z")[rand(62)] } (1..20);
open(my $par,"-|",qw(parallel), @subopt,
@ -182,7 +210,8 @@ sub sort_stdin {
# Input is stdin
# Spread the input between n processes that each sort
# n = number of CPU threads
my $numthreads = `parallel --number-of-threads`;
my $numthreads;
$numthreads = $opt::parallel || `parallel --number-of-threads`;
my @fifos = map { tmpfifo() } 1..$numthreads;
map { mkfifo($_,0600) } @fifos;
# This trick removes the fifo as soon as it is connected in the other end