This repository contains code for our paper: Parallel Batch-Dynamic kd-Trees. It contains an implementation of a logarithmically-structured parallel batch-dynamic kd-tree as well as several baseline implementations of parallel batch-dynamic kd-trees.
We use ParlayLib, a wonderful library developed at CMU for multi-core parallel programming. We also use xxHash for a 64-bit hash function. These modules are included in the project as submodules, initialize them before building the project:
git submodule init
git submodule update
We use CMake as the build system. To build the project:
mkdir build
cd build
cmake ..
make -j
See the top level CMakeLists for various build options that can be included with the cmake command above.
After the build completes, you will find executables in the nested directories of build/. Comprehensive tests can be found under build/test/ for all the implemented data structures. Benchmarks are written using Google Benchmark and can be found under build/benchmark/. They can be run using GoogleBenchmark syntax; for example, to run the construction benchmark over the BDL-tree with a 10M, 2D dataset, use
./bench_construction --benchmark_filter="bench_construction<2, LogTree_t<2>>/10000000/100/0/real_time"
You can view all the different benchmarks under a given executable by passing the flag --benchmark_list_tests. Example datasets can be found here.
This code is still in a research-phase. Please contact the developers or paper authors if you encounter any problems.