This is a header library containing a parallel implementation of an approximate priority queue. It is lock-free and uses only a single global atomic to keep track of whether the queue is empty or not as well as several local atomics for channels between workers.
The queue works on a work-sharing principle whose dynamics is determined by the underlying queue network, i.e., the directed graph of work-sharing channels between workers including frequencies of sharing their work with their neighbours, and sending batchsizes. This allows flexibility in the mixing rate of high-priority tasks across domains in non-uniform memory architectures (NUMA) to balance memory movement.
The following bash script will automatically download dependencies on GoogleTest and GoogleBenchmark and then build the tests and benchmarks in the folders build/tests and build/benchmarks, respectively.
mkdir -p build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j$(nproc)To run all tests, run the following command.
make run_testsTo run all benchmarks, run the following command.
make run_BMTo generate documentation, run the following command.
make docThe library is licensed under the Apache License, Version 2.0.
TBA