[perl #114918] infix: greatly slower than nested for loops

[via RT]

# New Ticket Created by  jn...@jnthn.net 
# Please include the string:  [perl #114918]
# in the subject line of all future correspondence about this issue. 
# <URL: https://rt.perl.org:443/rt3/Ticket/Display.html?id=114918 >


Given the following code:

~~
sub ltp_1 {
     for (9...1) X (9...1) X (9...1) X (9...1) X (9...1) X (9,7,3) -> 
$a,$b,$c,$d,$e,$f {
     }
}

sub ltp_2 {
     for 9...1 -> $a {
         for 9...1 -> $b {
             for 9...1 -> $c {
                 for 9...1 -> $d {
                     for 9...1 -> $e {
                         for 9,7,3 -> $f {
                             1
                         }
                     }
                 }
             }
         }
     }
}

say now;
ltp_1;
say now;

say now;
ltp_2;
say now;
~~

The output is:

Instant:1347656155.562
Instant:1347657301.502
Instant:1347657301.508
Instant:1347657351.156

That is, 146s vs. 50s (that's after an initial optimization to 
infix:<X>). Looking at the profile at C level, we spend a large amount 
of time in perl6_rpa_find_type (a third of total execution time) in the 
infix:<X> case, and a further 27% of the time is spent doing 
boolification (presumably the if statement in the implementation of 
infix:<X>).

Curiously, in the problem that showed this up:

     http://rosettacode.org/wiki/Truncatable_primes#Perl_6

The difference was even more pronounced after switching to the for loop 
approach.

     http://irclog.perlgeek.de/perl6/2012-09-14#i_5988681

For the record, it used to read:

     for (9...1) X (9...1) X (9...1) X (9...1) X (9...1) X (9,7,3) -> 
$a,$b,$c,$d,$e,$f {
         my @x := [\+] $f, $e, $d, $c, $b, $a Z* (1,10,100 ... *);
         return @x[*-1] if prime @x[0] and prime @x[1] and prime @x[2] and
               prime @x[3] and prime @x[4] and prime @x[5];
     }

Would be good to get to the bottom of why it's so much slower and see if 
something can be done about it.

/jnthn

Functions