Case sharing on the use of c++ map and finding performance tests

Recently I am doing performance tuning for the recommendation service. The main logic of this service is to use offline calculated model data to score each advertisement in the request, and then return the sorting results of these advertisements. The scoring process here actually uses The data in the request is combined into various keys to check a large map. This kind of calculation is very much and has become a major performance bottleneck. The code is relatively old and uses boost::unordered_map. In order to solve this problem, I found some Comparing the third-party library and the standard library

The following are the test results on an aws r4.xlarge machine (note that -O2 must be added when compiling):

std::map<int, int>                                 => 51866903
std::unordered_map<int, int>                       => 3838175
std::unordered_map<int, int, nohashint>            => 3508570
std::unordered_map<int, int>(N)                    => 3804471
boost::unordered_map<int, int>                     => 3291384
boost::unordered_map<int, int, nohashint>          => 3293934
boost::unordered_map<int, int>(N)                  => 3265856
google::dense_hash_map<int, int>                   => 785969
google::dense_hash_map<int, int, nohashint>        => 784455
google::dense_hash_map<int, int>(N)                => 899262
tsl::hopscotch_map<int, int>                       => 654668
tsl::hopscotch_map<int, int, nohashint>            => 680964
tsl::hopscotch_map<int, int>(N)                    => 663607
tsl::robin_map<int, int>                           => 406176
tsl::robin_map<int, int, nohashint>                => 411358
tsl::robin_map<int, int>(N)                        => 409993

You can see that the performance of tsl::robin_map can basically reach 10 times that of std::unordered_map. This performance is also related to the operating system and library version. In the actual production environment, it is recommended to pull down the code and run it in your own environment. Let’s test it next

We replaced the original boost::unordered_map with tsl::robin_map online, and the overall performance increased by 5 times. Of course, it also includes some other optimizations. This optimization is relatively large. The optimization point

Related articles:

MySQL slow query search and tuning test

php function search performance test

Related videos:

C Manual Tutorial

The above is the detailed content of Case sharing on the use of c++ map and finding performance tests. For more information, please follow other related articles on the PHP Chinese website!