Im curious what you call signifigant, and also how you benchmarked them. Please post your benchmark. In my tests, with a predeclared swap variable, I see a %50 difference when using dynamic variables, and a %25 difference when using lexical variables. Personally I call that signifigant, especially if its inside a loop. Oh the difference would probably be even more drammatic for lexicals but for the fact of having to repeatedly declare them, which wouldnt be necessary in a more realisitic situation.

use Benchmark 'cmpthese';

our $a=1;
our $b=2;
our $tmp;

print "Dynamic variables, swap benchmark.\n";
cmpthese -10, {standard => '$::tmp=$::a; $::a=$::b; $::b=$tmp;',
              list     => '($::a,$::b)=($::b,$::a)'
             };
print "Lexical variables, swap benchmark.\n";
cmpthese -10, {
              standard_my     => sub{my $x=1;my $y=2; my $z =$x; $x=$y
+; $y=$z},
              standard_local  => sub{my $x=1;my $y=2; local $_=$x; $x=
+$y; $y=$_},
              standard_global => sub{my $x=1;my $y=2; $_=$x; $x=$y; $y
+=$_},
              list            => sub{my $x=1;my $y=2; ($x,$y)=($y,$x)}
+,
             };
__END__
Dynamic variables, swap benchmark.
Benchmark: running list, standard, each for at least 10 CPU seconds...
      list: 11 wallclock secs (10.39 usr +  0.00 sys = 10.39 CPU) @ 66
+4534.94/s (n=6904518)
  standard: 11 wallclock secs (10.84 usr +  0.00 sys = 10.84 CPU) @ 98
+3324.54/s (n=10662188)
             Rate     list standard
list     664535/s       --     -32%
standard 983325/s      48%       --
Lexical variables, swap benchmark.
Benchmark: running list, standard_global, standard_local, standard_my,
+ 
  each for at least 10 CPU seconds...
      list: 10 wallclock secs (10.13 usr +  0.00 sys = 10.13 CPU) @ 40
+4027.75/s (n=4090781)
standard_global:
  9 wallclock secs (10.05 usr +  0.00 sys = 10.05 CPU) @ 484686.67/s (
+n=4869647)
standard_local: 
 10 wallclock secs (10.03 usr +  0.00 sys = 10.03 CPU) @ 402621.17/s (
+n=4038693)
standard_my: 
 10 wallclock secs (10.80 usr +  0.00 sys = 10.80 CPU) @ 503630.27/s (
+n=5437696)

                    Rate standard_local         list standard_global s
+tandard_my
standard_local  402621/s             --          -0%            -17%  
+      -20%
list            404028/s             0%           --            -17%  
+      -20%
standard_global 484687/s            20%          20%              --  
+       -4%
standard_my     503630/s            25%          25%              4%  
+        --
[download]

Yves / DeMerphq
---
Software Engineering is Programming when you can't. -- E. W. Dijkstra (RIP)