Is anyone else seeing a big performance boost on PHP 8?
A couple of tests in my small library dealing with reflection and a bit of regexps are running 30-50% faster ๐ช!
It is great, but the isolated unit tests are by no means representative.
Has anyone successfully benchmarked a more complex solution already?
As the third party library support for the new PHP version is far from sufficient, I was not yet able to run more complex tests.
This is the hobby lib that's seeing the boost.
dakujem / wire-genie
Autowiring Tool & Dependency Provider.
Wire Genie ๐ง
Autowiring Tool & Dependency Provider for PSR-11 service containers Wire with genie powers.
๐ฟ
composer require dakujem/wire-genie
๐ Changelog
What?
A superpowered call_user_func
? Yup! And more.
Wire Genie uses your PSR-11 service container to "magically" provide arguments (dependencies).
Allows you to:
- invoke any callables
- construct any objects
... with high level of control over the arguments. ๐ช
Usage
$container = new Any\Psr11\Container([
Thing::class => new Thing()
MyService::class => new MyService(),
]);
$callable = function (MyService $service, Thing $thing){ ... };
class Something {
public function __construct(MyService $service, Thing $thing) { ... }
}
$g = new Dakujem\Wire\Genie($container);
// Magic! The dependencies are resolved from the container.
$value = $g
โฆAnd since the attributes can now be used instead of doc-comment annotation parsing (which requires regexp parsing most of the time), the upcoming version will be much faster by design.
Top comments (10)
I don't expect to see much improvement on the frontend side, jit is meant for long running scripts. I think I may get some nice improvement in the cli, perhaps this will make php based server a viable option as well?
PHP used to be Personal Homepage Preprocessor and I'll continue to treat it as such. My main website uses PHP but it's only for rendering pages.
I would not make long running scripts out of PHP. It's what Node.js is for.
PHP actually has better type checking, better support for OOP, and similar if not faster performance compared to Node. I use both of them at the same time for their purposes.
I don't know about PHP8, but PHP7 indeed has good OOP.
As for type checking: you are not supposed to write JavaScript directly. Write TypeScript, compile and run in Node.
Performance depends on use case. Node would perform better on I/O intensive workloads; PHP would perform better if you don't need much I/O.
Neither would beat C++, but you would spend 3x coding time for C++.
If you like the performance of C++, try the Swoole extension for PHP, it brings the Go Coroutine syntax and power to PHP :). You can start writing async code with PHP as well and get the same I/O performance you have with Node if not more.
In any case, the point is not to say which language is better, PHP can be used for serious work, it's not just meant for personal toy project anymore. Even if you do not want to give that a try, I hope that you should not spread out-dated information that may discourage new beginners to try it out. Both PHP and Node are fantastic languages and they can live happily together :).
How widespread is the hosting/platform support for Swoole? Do you run it in docker, kubernetes or something like that? I have yet to experiment with it. Though i lived the coroutines in Go. Still, Go is a pain in the ass to write. ๐
If you can use pecl or run phpize to compile php then you can install swoole. It's no different than install other php extensions IMHO. I think you should have absolutely no issue on Amazon,, Google, etc...
Just checked my CI job for a small project (and I mean tiny) but the time to run the unit tests on PHP 8 was cut in half compared to 7.4.
That's what I'm seeing. Maybe Travis deals better with php 8 ๐
Interesting ... I'd like to benchmark PHP 8 vs PHP 7 with a Laravel app and then measure how the TPS (transactions/requests per second) is affected, because processing an HTTP request in Laravel involves executing a LOT of PHP code - I can imagine that a faster 'engine' executes all of that code a lot more quickly & efficiently, meaning that more requests can be handled using the same percentage of CPU ...