I wish that I could more easily compare the assembly outputs of the different compilers! Three or more of the rightmost panel would be superb! I want to see how the assembly compares between them without having to switch back and forth using the compiler dropdown! Make it so my good man! Make it so!
Thanks! It's a frequently requested feature but I've yet to find a decent way to present it. There's an issue tracker on GitHub that you're welcome to chime in on if you have ideas :)
This is very interesting! There is a huge difference between the assembly code generated by the C compilers and the assembly code generated by the Rust compiler for the equivalent code samples! Sometimes there is 10 to 20 times more lines of assembly generated for the Rust code than there is for the C code! The C assembly output is typically very tight while the Rust assembly output is much more verbose! Doesn't the much greater amount of Rust assembly code have a very negative impact on its performance? I just don't see how it can be claimed that Rust can be nearly as fast as C when executing the Rust assembly code requires so many times more instructions to be executed! I don't think it is a case of many fast instructions being chosen instead of a small number of less fast instructions. The Rust assembly to me just looks very unoptimized! I know Rust is a young language still but there is still a very big gap between the C assembly and the Rust assembly!
> The Rust assembly to me just looks very unoptimized!
If you're not adding a flag like `-C opt-level=3` to the compiler invocation this is literally true. (For one, Rust's unoptimised code does more work than C's, e.g. arithmetic operations are checked for overflow in debug mode, which is the default when optimisations are off, but is disabled when optimisations are on.)
With optimisations on, Rust seems to do a better job, e.g. the loop in the max_array example is vectorised and unrolled by Rust, but not by gcc or clang.
> I just don't see how it can be claimed that Rust can be nearly as
> fast as C when executing the Rust assembly code requires so many
> times more instructions to be executed!
Well, beyond what others have said here, with performance, it depends on the exact details. For example, our regex library has gotten some upgrades recently, and so on the benchmark game:
8 comments
[ 2.8 ms ] story [ 33.3 ms ] threadRust, Go, and D language explorers are https://<lang>.godbolt.org, e.g. https://rust.godbolt.org/ and so on
If you're not adding a flag like `-C opt-level=3` to the compiler invocation this is literally true. (For one, Rust's unoptimised code does more work than C's, e.g. arithmetic operations are checked for overflow in debug mode, which is the default when optimisations are off, but is disabled when optimisations are on.)
With optimisations on, Rust seems to do a better job, e.g. the loop in the max_array example is vectorised and unrolled by Rust, but not by gcc or clang.
http://xania.org/201505/on-rust-performance
At best, a claim like "Rust is as nearly as fast as C" can only be said for broad categories of things. The devil is in the details.