How do scalar objects behave in benchmark?

[HallofFamer]

Did anyone try that with userland implementation of scalar objects? I am trying to create a scalar classes extension in C with Boolean, Int, Float, Null and String, I suppose it will be even faster than userland implementation. I wonder if the speed of using scalar objects such as 2->abs(), and "my string"->toUpper() will be on par with just calling native function abs(2) and strtoupper("my string"). Or at least, it will not be magnitudes slower like if I wrap primitives in classes?

[nikic]

The performance (at least with the current implementation) will be worse than directly calling a native function, primarily because we have to go through an extra indirection function to handle the extra argument. But it should not be magnitudes slower either.

[HallofFamer]

I see, thanks for the explanation Nikic. And how does it compare to using a wrapper object? Lets say I have an Array object or String object, which stores an array or string literal as its property internally. Will scalar objects be faster than wrapper objects?

[nikic]

Yes, scalar objects will be more efficient than wrapper objects because you don't need to actually instantiate objects.

[HallofFamer]

I see, thanks for the explanation. So its possible to use this in production already, correct? I am also working on an extension called Scalar Classes, which provide implementation for Scalar Objects but its written in C as PHP extension, so I think it should be even faster this way.

And if I may ask, have you looked at the other issue I posted(Can you bring this to PHP RFC)? Any plan for starting an RFC for this? Or maybe bring it to PECL first?