On Thu, Mar 21, 2019 at 12:58 PM Dmitry Stogov <dmi...@zend.com> wrote:

> Hey,
>
> I'm starting the vote on JIT RFC.
>
>
>  https://wiki.php.net/rfc/jit<https://wiki.php.net/rfc/typed_properties_v2
> >
>
>
> The voting period is one week, until Thursday 28-03-2019 GMT.
>
>
> Since the initial announcement and following discussions, RFC was imprved
> and implementation extended with support for Clang, Windows and ZTS builds.
> Please reread RFC carefully.
>
>
> Thanks. Dmitry.
>

Thanks a lot for your work on this Dmitry! I didn't expect to see full
Windows & ZTS support so soon :)

A few people have already commented on the question of 7.4 vs 8.0, here are
my thoughts on that.

Pros:
 * If Linux distros build with the JIT compiled in, we may get much more
testing, as not many people compile PHP themselves.
 * It allows testing the JIT on code that is not compatible with PHP 8.0,
but is compatible with PHP 7.4.

Cons:
 * The JIT for PHP 8 will (presumably) see a lot more work to improve
performance, in particular based on runtime profiling and speculative
optimization. This means that tests run with the PHP 7.4 JIT may not be
representative (and may show worse than actual performance).
 * It makes for better marketing to include it in PHP 8. PHP 7.4 is already
a very packed release (FFI, preloading, typed properties, type covariance,
??=, maybe short closures), probably the most important release since 7.0.
 * More maintenance burden: I think you are right that we won't be seeing
many major changes land in 7.4 anymore. But with the number of other big
features already in 7.4, I think we have enough new bug surface area to
deal with already. (Unfortunately we are still not on top of all stability
issues from PHP 7.3...)

Overall, I think overall we're better served with introducing the JIT in
PHP 8.0.

Nikita

Reply via email to