How fast is pp/PAR::Packer at packing DateTime::Format::ISO8601 for you?

[Anonymous Monk]

If you have pp/PAR::Packer, how long does this take?

perl -le " print ~~gmtime "
pp -e " use strict; use warnings; use Getopt::Long; use DateTime::Form
+at::ISO8601; print qq{$_\n} for %INC; "
perl -le " print ~~gmtime "

On my old laptop it took 32min at 98% cpu usage

It doesn't seem to get faster (on second run) even after adding  -cd pp-scandeps-cachefile which has about 2084 files

I thought, hey, the cache does md5summing, all that md5summing is taking extra time, so I produced this patch

diff -ruN Module-ScanDeps-1.09/lib/Module/ScanDeps/Cache.pm Module-Sca
+nDeps-1.09-new/lib/Module/ScanDeps/Cache.pm
--- Module-ScanDeps-1.09/lib/Module/ScanDeps/Cache.pm    2012-02-21 07
+:25:42.000000000 -0800
+++ Module-ScanDeps-1.09-new/lib/Module/ScanDeps/Cache.pm    2012-10-1
+8 04:31:16.312500000 -0700
@@ -57,7 +57,7 @@
 }
 
 ### check for existence of the entry
-### check for identity of the file
+### check for identity of the file, ONLY if mtime is different
 ### pass cached value in $mod_aref
 ### return true in case of a hit
 
@@ -66,6 +66,14 @@
     my ($key, $file, $mod_aref) = @args{qw/key file modules/};
     return 0 unless (exists $cache->{$key});
     my $entry = $cache->{$key};
+    
+    my $mtime = ( stat $file )[9];
+    if (exists $entry->{mtime} and $entry->{mtime}  == $mtime ){
+        ## if mtime hasn't changed, don't bother with md5, its faster
+        @$mod_aref = @{$entry->{modules}};
+        return 1;
+    }
+    
     my $checksum = _file_2_md5($file);
     if ($entry->{checksum} eq $checksum){
         @$mod_aref = @{$entry->{modules}};
@@ -81,6 +89,7 @@
     my $checksum = _file_2_md5($file);
     $entry->{checksum} = $checksum;
     $entry->{modules} = [@$mod_aref];
+    $entry->{mtime} = ( stat $file )[9]; ## cause md5 is slow
     $cache_dirty = 1;
     return 1;
 }
diff -ruN Module-ScanDeps-1.09/lib/Module/ScanDeps.pm Module-ScanDeps-
+1.09-new/lib/Module/ScanDeps.pm
--- Module-ScanDeps-1.09/lib/Module/ScanDeps.pm    2012-09-04 13:38:32
+.000000000 -0700
+++ Module-ScanDeps-1.09-new/lib/Module/ScanDeps.pm    2012-10-18 04:3
+3:48.578125000 -0700
@@ -4,7 +4,7 @@
 use warnings;
 use vars qw( $VERSION @EXPORT @EXPORT_OK @ISA $CurrentPackage @Includ
+eLibs $ScanFileRE );
 
-$VERSION   = '1.09';
+$VERSION   = '1.0901';
 @EXPORT    = qw( scan_deps scan_deps_runtime );
 @EXPORT_OK = qw( scan_line scan_chunk add_deps scan_deps_runtime path
+_to_inc_name );

But that doesn't seem to speed anything up (I killed it after 9min)

Ideas?

[marto]

C:\>perl -le " print ~~gmtime "
Thu Oct 18 12:06:43 2012

C:\>pp -e " use strict; use warnings; use Getopt::Long; use DateTime::
+Format::IS
O8601; print qq{$_\n} for %INC; "

C:\>perl -le " print ~~gmtime "
Thu Oct 18 12:07:42 2012

My laptop setup is Intel Core2Duo @2.1GHz, 1GB RAM, Windows XP SP3, Strawberry Perl 5.12.3.

Update: I didn't notice that Radia was running, which usually slows down the laptop a fair bit. I'll retest once it's done.

Updated: results from running the above after Radia stops.

C:\>perl -le " print ~~gmtime "
Thu Oct 18 12:43:16 2012

C:\>pp -e " use strict; use warnings; use Getopt::Long; use DateTime::
+Format::IS
O8601; print qq{$_\n} for %INC; "

C:\>perl -le " print ~~gmtime "
Thu Oct 18 12:43:39 2012

[daxim]

$ time pp -e " use strict; use warnings; use Getopt::Long; use DateTime::Format::ISO8601; print qq{$_\n} for %INC; "

real    0m11.527s
user    0m6.153s
sys     0m0.900s

I suspect something else is slowing down file access on your system. Use Sysinternals Filemon or similar to detect this.

[Anonymous Monk]

I suspect something else is slowing down file access on your system. Use Sysinternals Filemon or similar to detect this.

I have, it doesn't appear to :) Though I will schedule some downtime later to run chkdsk etc

[Anonymous Monk]

The program runs in ~1 second (perl -e ... ) like you'd expect, its the packing that takes forever -- if something else was slowing things down, the running would take forever also