> On July 20, 2018, 3:58 a.m., Benno Evers wrote: > > 3rdparty/stout/include/stout/jsonify.hpp > > Line 370 (original), 278 (patched) > > <https://reviews.apache.org/r/67988/diff/1/?file=2061761#file2061761line403> > > > > If we care enough about performance to make a templatized overload for > > `const char(&)[N]`, maybe we should also add one for `std::string&&`?
The `const char (&)[N]` exists to avoid constructing a temporary `std::string` from a string literal argument at the moment. What cases would adding `std::string&&` improve? We can't exactly move the `string` anyway right? - Michael ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/67988/#review206267 ----------------------------------------------------------- On July 19, 2018, 8:38 p.m., Benjamin Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/67988/ > ----------------------------------------------------------- > > (Updated July 19, 2018, 8:38 p.m.) > > > Review request for mesos, Alexander Rukletsov, Benno Evers, and Michael Park. > > > Bugs: MESOS-9092 > https://issues.apache.org/jira/browse/MESOS-9092 > > > Repository: mesos > > > Description > ------- > > This reduces the time needed for the client to finish receiving a > master's /state response by 50% in the `StateQuery` benchmark: > > min q1 q3 max > baseline 6.52 6.76 7.33 8.26 > rapidjson w/ SIMD 3.48 3.54 4.12 4.4 > rapidjson 3.29 3.32 3.65 3.85 > > SIMD is left disabled for now since it showed slightly slower > results. > > > Diffs > ----- > > 3rdparty/stout/include/stout/jsonify.hpp > 2314980e185ee61cc2ea54f1b4d2a8b35e58121c > > > Diff: https://reviews.apache.org/r/67988/diff/1/ > > > Testing > ------- > > Tested at the end of this chain, since this is split across > stout/libprocess/mesos. > > > Thanks, > > Benjamin Mahler > >
