- General resiliency under heavy task load, when tasks > all(cores, threads)
- Correct results
- No crashes
- No "cliff-like" performance as overload point is passed
- Note: better to OOM than GC thrash
- Similarly better to serve some tasks well than all badly via e.g. context switch storms
- Idle/block on inputs without burning CPU
- (and/or use idle time for something productive like GC)
- Good scalability for high core count (including multi-socket/NUMA)
- Better straight-line performance for e.g. Supply/tap chains
- Proc::Async:
- Rock solid
- see http://irclog.perlgeek.de/perl6/2015-03-31#i_10370872 and following
- ... including the impressive http://irclog.perlgeek.de/perl6/2015-03-31#i_10371057
- Working on r-j
- Rock solid
- Promise cancellation (that actually stops associated tasks)
- Union support, at least for num32/64 <-> [u]int32/64
- Unsigned integer return values
- C++ bindings
- Inheritance handling in OO::Monitors
- I/O:
- Better performance
- TTY/console user (non-piped I/O) detection
- More complete support
- Precompiling -> caching