Mozilla's Servo Is Whooping The Other Browsers In Performance

[Ferdinand]

Yay. How about also have a responsive UI?

Still waiting on that feature in Firefox.

[rabcor]

WebGL Performance*

Not the same as browser performance. The initial reason why I switched to chrome over firefox (AND STILL USE IT) was faster browser performance, I don't care about the memory hogging, it loads pages faster than firefox ever could.

[codensity]

people who can't learn c++ have to invent their own toy languages
there are thousands of languages with same claims. they have no tools, no developers, no libraries and die in obscurity

servo is fast on this benchmark not because of rust, but because of gpu rendering. every other browser with gpu rendering will be as fast.
because of rust ( i am quoting youtube video description ) "there are a lot of features it doesn't support, and it crashes a lot, and the output it a bit aliased"
so much for fast, safe and featureful

I agree that there are a great number of languages which offer little incentive to move away from c++. For me at least, it therefore doesn't make sense to spend time and effort mastering a new language that's largely just a skin while perhaps solving some small problem such as issues with C integer types (granted I accept that overflow semantics is perhaps more than a small issue). That being said I find rust to be one of the few languages that brings something legitimately interesting to the table while being a sufficiently conservative, low-level language. There is one specific feature from c++ I miss, however I don't expect many others would.
I'm curious about some of the specifics you prefer about c++ over rust?

[Kushan]

it is only relevant c-replacement language. of course you don't need c++ to replace bash scripts.because this forum is full of kids who think that rust will magically write programs for them

Shit, I better not mention Golang, then...

[codensity]

You've really drunk the kool-aid didn't you? There is no safe and fast language. You are either fast and unsafe or safe but no fast. Don't let anybody lie to you. If you knew anything about programming you'd know why my statement is true.

I'm curious why you believe safety (perhaps we should agree on what safety means) precludes performance?

[mmstick]

Wow! Maybe this time Mozilla literally hit the spot!
I wonder, if Rust is as complete as C and so blazing faster'n'safer, will be possible to write a really, really, REALLY huge and critical project in Rust?
Something is usually written in C and which security and speed are really crucial features?
Something like.....
...LINUX KERNEL?
Or, well, just A kernel?

How about an entire operating system and surrounding suite? https://github.com/redox-os
We have our own kernel, display server, GUI toolkit, shell and more

[mmstick]

Well, it is already possible to write simple bootable programs with Rust: http://blog.theincredibleholk.org/bl...oting-to-rust/

Also, I found this: https://github.com/charliesome/rustboot

I think it would indeed be a better idea to design a new kernel from scratch rather than trying to re-write an old C kernel in Rust.

Edit: Also this seems to be interesting and more recent: https://github.com/ryanra/RustOS

What you may have missed is that we have our own operating system in development that is a lot farther ahead than the ones you mentioned above. Check out Redox OS

https://github.com/redox-os

[rob11311]

WebGL Performance*

Not the same as browser performance. The initial reason why I switched to chrome over firefox (AND STILL USE IT) was faster browser performance, I don't care about the memory hogging, it loads pages faster than firefox ever could.

Strange, for me, every time I've tried Chrome out I've wound up uninstalling it .. it's never loaded pages faster for me ~shrugs~

[mmstick]

You've really drunk the kool-aid didn't you? There is no safe and fast language. You are either fast and unsafe or safe but no fast. Don't let anybody lie to you. If you knew anything about programming you'd know why my statement is true.

There is no kool-aid because Rust is precisely what it states: safe and fast. The language completely disallows writing code that utilizes unsafe functionality unless the user drops into an unsafe scope, which mainly only happens when the user is calling C libraries which cannot be guaranteed to be safe. Because of the concept of ownership and borrowing, we can call our code safe because no variable on any thread could possibly write or read a variable that is already being borrowed, which variables can only be borrowed once. One of the larger issues with C and C++ is the lack of a concept for keeping data safe between threads so as to not allow threads to step on each other's toes. All of the higher level features in Rust are also designed in a way that they are translated into much lower level code so there is zero overhead. In example, the Iterators in Rust will be compiled lazily into a loop that is often times much more efficient than a loop manually created by a C++ programmer.

[Michael_S]

So why does Rust perform more like Java than C++? Safety comes with a performance cost. It is a rule of nature. Bounds checking an array is slower than just iterating through it checking fuck-all.

Where did that assertion come from? In the recent benchmarks I've seen, Rust is as fast as C or C++.

Rust does most of its checking in the compiler, not the runtime, so there is no heavy runtime overhead. And you can compile with flags to disable bounds-checking.

(Edit: Later in the thread, TomBomb corrects me. You can't compile Rust to disable bounds-checking. Sorry for the misinformation.)