[rust-dev] net::tcp::TcpSocket slow?

Erick Tryzelaar erick.tryzelaar at gmail.com
Fri Dec 21 15:22:52 PST 2012


Would we gain any significant performance if we made evented IO syscalls
like Go instead of going through libuv?


On Fri, Dec 21, 2012 at 11:30 AM, Brian Anderson <banderson at mozilla.com>wrote:

> On 12/21/2012 09:19 AM, Patrick Walton wrote:
>
>> On 12/21/12 12:17 AM, Brian Anderson wrote
>>
>>> In order to have competitive performance in high-throughput applications
>>> we need to use uv as intended and run our code inside the uv callbacks.
>>> Right now there is no high-level API for this, so I think that when we
>>> revisit this stuff one of the first things we need to do is add those
>>> modules and get them solid and fast. Then rebuild everything on top of
>>> that.
>>>
>>
>> How do we deal with failure? Do you mean that the code inside the uv
>> callbacks should be trusted not to fail and that user-level code should
>> still be synchronous?
>>
>
> To start with we would just have to trust the code not to fail, but I
> expect we will solve this problem somehow. User-level code that wants to be
> super fast would by asynchronous and need to understand the failure issue.
>
> ______________________________**_________________
> Rust-dev mailing list
> Rust-dev at mozilla.org
> https://mail.mozilla.org/**listinfo/rust-dev<https://mail.mozilla.org/listinfo/rust-dev>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/rust-dev/attachments/20121221/c4dee5fc/attachment.html>


More information about the Rust-dev mailing list