Hacker News Reader

Comment by moomin | Hacker News Reader

The reports from all of these micro benchmarks seems to be consistent on the subject of Rust: it doesn't yet have the performance that justifies its existence. I'm guessing you could get similar numbers out of Java.

I'm hoping Rust's numbers improve. It's incredibly promising, but without C++ level performance, it's too much semantic complexity.

View on HN

Achieving the performance that Rust does is amazing, given the focus so far been on language design and not performance. It's not even 1.0 yet!

I'm sure these numbers will improve.

With XorShift as the PRNG, Rust is now almost on par with C and beating the DMD and GDC D implementations. If the syntax were stable, and the standard library and documentation improved, then we'd seriously consider it as an alternative to C++ for parts of our current project.

You can write slow code in any language, that's exactly what the problem was (and still is) here. The implementations still aren't using the same random number generator.

Rust has C++ level performance if you compare the same code in both languages.

+1 The same is true probably also about D.

No because D has C performance, now. Have you seen the blog numbers?

As pointed out in the other comments on this thread, this benchmark benchmarks mostly standard RNG performance, not quality of the code generated by each compiler. Here you've got much more computationally intensive benchmark, showing that D is "as fast as C" like it could be said about Java: http://attractivechaos.wordpress.com/2013/04/06/performance-...

It is close, but Java is just as close, too. Not a tremendous performance advantage that someone would choose one over another for and not something I'd expect from a language marketed as "performance oriented".

Ok you live in a world where Java is "as fast as C". Let's terminate this discussion.

I didn't say it. You said D was as fast as C now. Then, according to your definition of "as fast as C", Java must be fast as C as well. Seems your as fast as C definition must be something between 30%-95% performance of C because those are the most common performance numbers reported for D in various benchmarks (I cited one of them, and the one in the blog also confirms this - only one D compiler managed to match the best C compiler, the others were significantly behind). The same numbers are consistently reported about Java, so I don't really understand what you're arguing about. So I conclude both are the same league (with two minor exceptions: 1. I saw Java beating GCC many times, once quite severely by almost 2x on simple array numerical code, while I never saw that happening for D and 2. D has much worse GC implementation than Java or C# have).

In the real world it's not nearly as fast as C, but in a micro-benchmark like this with a bunch of loops crunching fixed-size integers it would come close.

The difference in performance between the languages is almost entirely due to differences in the implementation of the benchmark.

Reply on news.ycombinator.com