Re:The Resurgence of C Programming
by shepardrtc on Tuesday December 27, 2016 @02:23PM
(#13265448)
Regardless of what glue language you use to write ML stuff, if you follow it down to its core, most of it is going to use a BLAS library to do matrix multiplication. That's where the heavy lifting is. And BLAS libraries are in C/C++ or Fortran. For example, OpenBLAS or Intel MKL.
BLAS Level 1 was released in 1979, thirty seven years ago.
There was no other comparable language to implement it in besides Fortran. C was a newcomer at the time, like Rust currently is.
The amount of optimization and edge-case handling work that went into BLAS since then is enormous.
BTW, Fortran, being mostly aliasing-free, is a safer language than C, and more optimizaton-friendly.
I thought Fortran's aliasing-free rule was just an assumption of the compiler, not something that it actually enforced.. so you could very easily write code that contained aliasing bugs.
The very high-performance things are a relatively narrow area, and an area where C can easily lose to e.g. Fortran.
Also, for ML stuff you likely don't have a huge entrenched C codebase yet. You are freer to choose a more expressive language, and only hand-optimize / rewrite in C a small amount of hot paths.