EXPECTED

NOTE: THIS FILE IS EXTREMELY OUT OF DATE.  VOLUNTEERS TO UPDATE IT WELCOME.

* General
  + Startup times
    - For simple tests, many implementations suffer from slow startup times
      . Niecza on mono and NQP on JVM suffer the most, with a very slow startup
      . Rakudo on Parrot and Perlito* on JS follow with medium-high startup times
      . Perlito* on perl5 is a bit faster with medium-low startup times
      . NQP on Parrot has relatively quick startup times
      . perl5 (native) has extremely fast startup times
    - Use the '--ignore-startup' option to `analyze` or `bench` to subtract
      startup times so that pure compile and run time can be compared
    - Use the '--ignore-compile' option to `analyze` or `bench` to subtract
      script compile time as well
  + Perl 6 v. NQP idioms
    - Many idioms have to be changed to translate Perl 6 code to NQP code
    - Because Rakudo can (mostly) compile both languages and Niecza can compile
      some NQP code, this can be used to separate differences due to the
      compiler versus differences due to the language idioms
* Individual tests
  + empty_loop_*
    - Rakudo prefers the Perl 6 range iteration idiom to the NQP while loop
    - Niecza *greatly* prefers the NQP while loop to the Perl 6 range iteration
      idiom; it is unusually slow handling the Perl 6 idiom because:
      - "niecza's Range.iterator has seen much less love"
        http://irclog.perlgeek.de/perl6/2012-10-10#i_6050334
  + rc-forest-fire
    - Perlito6/* cannot yet compile this test
    - Niecza can compile but not run the NQP version of this test since it does
      not (for obvious reasons) recognize 'nqp::' syntax for raw NQP opcodes