But I agree with the overal message, and I'm looking forward to Rust maturing. I think it is the better language although I am no expert at either.
[1]: but really, when you start using AVX instructions and the like beyond what the compiler will automate for you, you are just using wrappers around assembly instructions in languages like C today anyway (__mm256_shuffpd, __mm256_perm128f2d)
That's not to say we don't have some platform-specific code in the "preferred" language (in some cases there is no choice in the matter), but most of the system is C++.
Over the years several languages have been the 'C++ killer'. For quite a long time it was D, and I know several people who converted. Then the standard library split in two, and the language got fairly unusable for a while. Now things have calmed down with D 2.0, but it's not clear it will still be around in 10-15 years time.
Similarly with Rust -- at the moment it seems to be a one company product (Mozilla). What if they get bored of it? Will my Rust code still work in 10 years time?
previous discussion: https://news.ycombinator.com/item?id=10890310
Rust's ecosystem is still a bit anemic. I've played around with Rust, and I like the language, but C++ has a much wider array of battle-tested libraries at one's beck and call. Rust's development tools are also somewhat of a monoculture, at the moment.
Care to elaborate? I'm curious
Writing in other languages means you end up using ineffecient abstractions. C++ affords very efficient abstractions and really doesn't force a different code structure than other languages, so it is hard to see why there'd be an advantage with something else.
If somebody really needs to have the performace or be that close to the hardware, they can still probably find a better language today (Rust, Go?). But all others will certainly be better served by more modern, higher level languages. Just reading the examples in the PDF makes my head hurt.