Desbordante is a high-performance data profiler that is capable of discovering many different patterns in data using various algorithms. It also allows to run data cleaning scenarios using these algorithms. Desbordante has a console version and an easy-to-use web application.
GNU Affero General Public License v3.0
382
stars
70
forks
source link
Multi-threaded implementation of HyFD algorithm #409
The HyFD algorithm was originally single-threaded,
which goes against its intended implementation.
If ThreadNumber option is specified in the algorithm configuration step,
the algorithm will utilize the specified number of threads.
This implementation was benchmarked on several large datasets such as:
iowa1kk
EpicMeds
flight
The general trend is displayed in the following graph:
(data for iowa1kk dataset on i9-13905h)
Alternative implementations showing worse performance:
Desbordante's utils::ParallelForEach
std::future + std::async
std::execution::par
OpenMP parallel for macro
TBB parallel for function
The final implementation showed high time consumption by memory allocation/deallocation.
In attempt of optimizing them the following allocators were benchmarked:
tbb::scalable_allocator
tcmalloc
jemalloc
mimalloc
All of which showed similar/worse performance,
therefore implementing custom allocator is probably useless.
The HyFD algorithm was originally single-threaded, which goes against its intended implementation. If ThreadNumber option is specified in the algorithm configuration step, the algorithm will utilize the specified number of threads.
This implementation was benchmarked on several large datasets such as:
The general trend is displayed in the following graph: (data for iowa1kk dataset on i9-13905h)
Alternative implementations showing worse performance:
The final implementation showed high time consumption by memory allocation/deallocation. In attempt of optimizing them the following allocators were benchmarked: