mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Mahler <bmah...@apache.org>
Subject Re: Review Request 50044: Updated Sorter::sort to return a vector rather than list.
Date Fri, 15 Jul 2016 17:55:13 GMT


> On July 14, 2016, 11:49 p.m., Klaus Ma wrote:
> > src/master/allocator/sorter/drf/sorter.hpp, line 22
> > <https://reviews.apache.org/r/50044/diff/1/?file=1444158#file1444158line22>
> >
> >     This's not necessary, `sorter/sorter.hpp` had included it.
> 
> Guangya Liu wrote:
>     I think the reason here is we generally don't rely on transitive includes as it's
harder to maintain.
> 
> Klaus Ma wrote:
>     We should keep small list of header; if anyone missed removing, it'll impact compile
performance.
> 
> Jiang Yan Xu wrote:
>     We follow the rules here: https://google.github.io/styleguide/cppguide.html#Names_and_Order_of_Includes
so in this case I think it's fine for drf/sorter.hpp to include vector again even if sorter/sorter.hpp
already includes it. However if we follow the guide strictly, then drf/sorter.cpp doesn't
need to include again because it's own header includes it. :)

We generally don't rely on transitive includes, even in the case that the header clearly demonstrates
intent (e.g. the interface function returns vector), because it's easier to maintain (quickly
scan a file and add includes for all symbols present).

Klaus, note that this is a redundant include, so either way the compilation units with sorter.hpp
are going to have one or many vector includes, but not zero. Zero -> one definitely adds
a significant compile-time performance impact, but here we're talking about N -> M where
N >= 1, M > N. I don't think we found that redundant includes had a significant impact
on compile time performance, but if you have found that to be the case, please show us!


- Benjamin


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/50044/#review142308
-----------------------------------------------------------


On July 14, 2016, 7:35 p.m., Benjamin Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/50044/
> -----------------------------------------------------------
> 
> (Updated July 14, 2016, 7:35 p.m.)
> 
> 
> Review request for mesos, Guangya Liu, Klaus Ma, and Jiang Yan Xu.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> We now prefer to use vector rather than list in general for
> efficiency reasons, unless we need to take advantage of the
> operations that are efficient on a linked-list.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/sorter/drf/sorter.hpp e29feebd70277c79f7c3f6fb233e7a36501cf220

>   src/master/allocator/sorter/drf/sorter.cpp 7df4dd641b21ea0705368861bf4679fed1ef078d

>   src/master/allocator/sorter/sorter.hpp f5f0b086cb95eb2ab70b3f67e5b20814925bf702 
>   src/tests/sorter_tests.cpp bdd4355bfcd7b1fa1c22983f8e0ee6f20906917a 
> 
> Diff: https://reviews.apache.org/r/50044/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> benchmarks: with 1000 clients, sort time is reduced by 72 us
> 
> 
> Thanks,
> 
> Benjamin Mahler
> 
>


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