mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dmitry Zhuk <dz...@twopensource.com>
Subject Re: Review Request 60003: Reduced copying in defer, dispatch and Future.
Date Mon, 17 Jul 2017 11:44:14 GMT


> On July 4, 2017, 10:55 a.m., Ilya Pronin wrote:
> > 3rdparty/libprocess/include/process/deferred.hpp
> > Line 77 (original), 79 (patched)
> > <https://reviews.apache.org/r/60003/diff/1/?file=1748112#file1748112line79>
> >
> >     This is more a question than an issue. Do we plan to forward functor references
through `_Deferred`, hence the use of `std::forward<F>()` here instead of `std::move()`?
> >     
> >     Carrying a functor reference is unsafe. `template <typename F> _Deferred<F>
defer(F&&)` overload silently captures references to lvalues in `_Deferred<F&>`.
But as long as `_Deferred` is immediately used for conversion to `Deferred` or `std::function`
it should be fine.

This is intentional, as in this case move/copy happens only once on conversion. Saving `f`
by value would add an extra move/copy on `_Deferred` construction.
This implementation is relatively safe, as `_Deferred` is documented as an intermediate type,
which is not supposed to be used directly, and it's somewhat protected from misuse by having
public methods callable on rvalues only.


> On July 4, 2017, 10:55 a.m., Ilya Pronin wrote:
> > 3rdparty/libprocess/include/process/deferred.hpp
> > Lines 83-88 (original), 85-90 (patched)
> > <https://reviews.apache.org/r/60003/diff/1/?file=1748112#file1748112line85>
> >
> >     Can we eliminate more copyings here? `f_` is captured by value by lambda and
then passed as a `const` to `dispatch`. We can make the lambda `mutable` and `move()` captured
`f_`. Or, for example, we could store `f` value and `pid` in a structure, that will be captured
by a `std::shared_ptr`.

We can't assume that lambda is called only once here, so we must make a copy. Avoiding this
last copy is doable, but I left it out of scope of this review.


- Dmitry


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/60003/#review179373
-----------------------------------------------------------


On July 17, 2017, 11:42 a.m., Dmitry Zhuk wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/60003/
> -----------------------------------------------------------
> 
> (Updated July 17, 2017, 11:42 a.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, haosdent huang, James Peach, and Michael
Park.
> 
> 
> Bugs: MESOS-7713
>     https://issues.apache.org/jira/browse/MESOS-7713
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This reduces number of copies made for each parameter in a code like this:
> future.then(defer(pid, &SomeProcess::someMethod, param1, param2));
> 
> For the objects that do not support move semantics (e.g. protobuf messages), number of
copies is reduced from 8-10 to 6. If move semantics is supported, then number of copies is
reduced from 6-7 to 1 if parameter is passed with std::move, or 2 otherwise.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/defer.hpp 7f3369e723cb244e97930621cbba89cf7873567d

>   3rdparty/libprocess/include/process/deferred.hpp e446621be11ac51f5f91c417cc8975e363c2f715

>   3rdparty/libprocess/include/process/dispatch.hpp 3a0793888dc0df5e3ec31b06f47cd920c71e0db9

>   3rdparty/libprocess/include/process/future.hpp cce950509f58022e79bb51a6e72ea1a005b9cb50

>   3rdparty/libprocess/include/process/http.hpp f637999174d92a98208b5fc49a65f9929efb11a0

> 
> 
> Diff: https://reviews.apache.org/r/60003/diff/2/
> 
> 
> Testing
> -------
> 
> make check
> 
> Number of copies was checked by using defer to subscribe process for Future callbacks,
and passing parameters that count number of copies made.
> 
> 
> Thanks,
> 
> Dmitry Zhuk
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message