Re^4: Can this code be optimized further?

[dragonchild]

sub hash_style2 {
    my %values;
    my @temp = @master;
    push @{$values{substr($_,0,1)}}, substr($_,2)
        for @temp;
    print "@{$values{a}} <==> @{$values{b}}\n" if $testing;
}

sub ikegami {
    my @temp = @master;
    local $_ = "@temp";
    my @a = /\ba_(\d+)\b/g;
    my @b = /\bb_(\d+)\b/g;
    print "@a <==> @b\n" if $testing;
}

When I rerun adding this method (removing the three worst performers), here's what I get:

               Rate Trinary New_Grep Original ikegami Hash2 Tri_Substr
+ Switch Tri_Substr2
Trinary     10273/s      --      -5%     -22%    -34%  -40%       -41%
+   -41%        -50%
New_Grep    10766/s      5%       --     -18%    -31%  -37%       -38%
+   -38%        -48%
Original    13096/s     27%      22%       --    -16%  -23%       -25%
+   -25%        -37%
ikegami     15683/s     53%      46%      20%      --   -8%       -10%
+   -10%        -24%
Hash2       17016/s     66%      58%      30%      8%    --        -2%
+    -3%        -18%
Tri_Substr  17437/s     70%      62%      33%     11%    2%         --
+    -0%        -16%
Switch      17453/s     70%      62%      33%     11%    3%         0%
+     --        -15%
Tri_Substr2 20651/s    101%      92%      58%     32%   21%        18%
+    18%          --

Changing to substr() brings the hash into respectable range.

[Roy Johnson]

Yes, it looks like using substr instead of matching with capture is the most significant improvement to be made. In fact, changing the original to do that causes it to run faster than anything except Tri_Substr2.

sub orig_substr {
    my (@a, @b) ;
    my @temp = @master;
    foreach my $value (@temp) {
    push @a, substr($value, 2) if (substr($value,0,2) eq 'a_') ;
    push @b, substr($value, 2) if (substr($value,0,2) eq 'b_') ;
    }
    print "@a <==> @b\n" if $testing;
}