J vs Perl Regex Performance Test

J Regex
+-+---------------+--------------+---------------+
| |A              |B             |C              |
+-+---------------+--------------+---------------+
|1|0.112824  19520|0.045339 14528|0.129472  28480|
|2|0.247294  28416|0.146528 16448|0.384141  53056|
|3|0.366651  52992|0.293741 22592|0.779365 102208|
|4|0.497595  52992|0.515006 24640| 1.32489 102208|
|5| 0.62397 102144|0.768388 34880| 1.96263 200512|
|6|0.748988 102144| 1.07931 36928| 2.71365 200512|
+-+---------------+--------------+---------------+

Perl Regex
+-+---------------+---------------+----------------+
| |A              |B              |C               |
+-+---------------+---------------+----------------+
|1| 0.11283  19520|0.0181568 14528|0.0585491  28480|
|2|0.222631  28416|0.0397327 16448| 0.122349  53056|
|3|0.357632  52992|0.0629762 22592| 0.191922 102208|
|4|0.503336  52992|0.0927869 24640| 0.264515 102208|
|5|0.655015 102144| 0.108931 34880| 0.335509 200512|
|6|0.774323 102144| 0.135322 36928|  0.40176 200512|
+-+---------------+---------------+----------------+

Discussion

Perl regex is faster, see 0.5 sec vs 2 sec.
Perl regex has linear, whereas J has at least polynomial time.
Perl regex fraction in total time is constantly lower than the J engine time, both being linear. J regex exceedingly surpasses J engine time. (J engine time is reflected by the first trivial case.)

Cases and series

Cases correspond to the following regex's:

pA=: '.'
pB=: '[0-9]+'
pC=: '(\+)|([0-9]+)|([a-z]+)'

Series are sequence TEST repeated N times:

TEST=: 1000$'1 + a '

Validity:

   load 'regex'
   P=:'(\+)|([0-9]+)|([a-z]+)'rxmatches '1 + a '
   load '~addons/pcre/regex.ijs'
   Q=:'(\+)|([0-9]+)|([a-z]+)'rxmatches '1 + a '
   P-:Q
1