[dzaima]

That just removes a tiny tiny bit of dynamic dispatch overhead. Which is still needed anyways, as array languages can often dynamically switch between 1-bit, 8-bit, 16-bit, 32-bit integer (and 64-bit float) arrays, depending on the elements, completely transparently to the user.

[slver]

Most compilers can inline statically defined closures in these contexts. And tracing JITs do this even when the closure is not define statically (but is stable).

It's more about allowing the SIMD goodness without the ambiguity and restrictions of "scalar operators work on arrays" implemented naively.

[klibertp]

> without the ambiguity

There's no ambiguity! In J, all (with some exceptions) operations work on arrays, period. `1 + 1` is not an addition of two scalars, but instead of two arrays of length 1. As I mentioned, you can have scalars, but a) they still live in arrays and b) you can't do anything with them without unboxing. So there's no ambiguity, as far as I can tell.

Also, APL and J are implemented as intepreters, but that doesn't preclude using SIMD instructions when executing expressions. I'm 100% sure the operations are not "implemented naively" :)

[rscho]

> `1 + 1` is not an addition of two scalars, but instead of two arrays of length 1

I think that's true of (Dyalog) APL, but not of J [1]. APL follows the nested array model (that you describe), while J follows the flat array model that does have true scalars.

[1] https://aplwiki.com/wiki/Array_model#Flat_array_theory

[dzaima]

AFAIK, J considers `1` to be an array - `L. 1` and `L. 1 2` both give 0 (`L. (1;2)` gives 1).

APL's simple scalar numbers are much more like regular numbers in non-array languages.

[rscho]

Traditional array languages (APL,J) are not amenable to static compilation due to other dynamic issues, though. I think Dyalog APL is experimenting with a bytecode interpreter, but no JITs in sight either (that I know of). The very dynamic aspect of those languages makes that difficult. I'd love a compileable similar language, though. To replace R/Python for statistics that are sooooo annoying when exploring data due to verbosity.

[rklmno]

That's not really the issue.

It's certainly true that, depending on how the code is written, you may not be able to optimize out the language implementation from the compiled program. But that just turns into a link against a library with support for the language for your hypothetical compiled program.

That said, the compiler being hypothetical is a very real obstacle. But even there the problem is not that the language can't be compiled -- it's that no one has bothered to implement a compiler for it.

Anyways, if you throw in some ML type inference, and some tools for characterizing the resulting code and its interfaces, you can generate code from an array language which is quite similar to the code you would get from a variety of other languages.

[dzaima]

A compiler could inline SIMD with or without an operation having an explicit map, the dynamic type checks needed are gonna be the same. (and, since this is an array language, the tiny extra overhead of completely dynamic dispatch is gonna be small compared to the executed SIMD afterwards anyways)

You lose a lot of simplicity & brevity by requiring explicit mapping, and gain close to nothing.