Be careful with microbenchmarks. From my experience, there tends to be very little correlation between performance improvements on microbenchmarks and real applications. If you improve performance on the microbench suite by a factor of two, you definitely will not get a factor of two on real applications -- more likely, it will be something like 2%.
Basically, if the benchmark is not testing end-to-end performance on Wordpress (or similar), then it's a useless benchmark.
Has anybody used it? The PHP code is full of questionable code which somehow works in PHP but I am not sure if it's possible to map PHP's behavior into .NET..
Error handling is a big issue - in PHP every notice and warning can be handled in user code, in .NET we use Debug Assertions or exceptions. But .NET has threads and async which shares one thread by more requests, so in case of error we just don't know what request has to handle the error. We would have to pass some request context object to every function in runtime in order to be able to let the correct request's error_handler to handle the error - or we just assume warnings and errors should not happen and crash when it does.
Garbage Collection - for performance reasons (and sanity) we let .NET GC to do its job. But unlike PHP's GC it is undeterministic and does its job on a different thread. For now, __destruct is not handled in Peachpie yet cause there are more ways on how to simulate PHPs behavior.
Changing existing closure's $this to anything - well.. besides it would make type analysis useless, it totaly violates strong .NET method access checks.
https://github.com/peachpiecompiler/peachpie/wiki/Peachpie-R...
Looks like PHP 5.4 and 7.1 roadmap goals have been achieved.
Though http://www.peachpie.io/about references Phalanger as the inspiration which apparently got as far as PHP 5.6.
There's also compatibility page that looks useful:
https://github.com/peachpiecompiler/peachpie/wiki/Compatibil...