[klickverbot]

LDC dev here. All the credit for this goes to Ilya Yaroshenko for the expertly crafted implementation and the LLVM developers for efficient low-level code generation (inlining, register allocation and so on, and in some places auto-vectorization). LDC only needs to make sure not to "mess up" things too much.

D does play a significant role in this achievement, though – D's very powerful yet easy to use features for generics and introspection make it possible to finely tune the code for different parameters (sizes/dimensions/…) while still being easy to understand and modify.

[openasocket]

That's helpful, thanks!

I definitely appreciate the power of generics, especially when you can to generics over values instead of just types. But I'm having trouble seeing the value of the compile-time introspection, for the most part it doesn't seem to give you much more power than generics do. Could you give me an example of the "killer feature" introspection gives over generics? Bonus points if you can compare it to Rust-style generics with traits and specialization.

[nicwilson]

Many things, although introspection works glove in hand with generics. Below os a non-exhaustive list.

You can 'Write once - automate everywhere' all the boilerplate. See https://github.com/kaleidicassociates/excel-d/ , for an example of automating the interaction between D and Excel. I am in the process of doing something similar for OpenCL and CUDA and will be presenting it at Dconf.

You can make at compile time (additional) fast paths by checking to see if a type (or symbol or whatever else) provides a fast primitive to accelerate your algorithm. e.g. if Foo implements fastfoo use that otherwise fallback to a general algorithm.

See also Andrei Alexandrescu's 2016(15?) talk (IIRC the relevant section is about half way through https://www.youtube.com/watch?v=4oDK91E3VKs).