Batch Parallel Data Structures with Obatcher

This is the repository for various case studies of batch parallel data structures with OBatcher (https://github.com/verse-lab/obatcher).

As this requires a custom version of the library domainslib for Multicore OCaml, it is advisable that you create a new opam switch to run this.

Build instructions

First, set up OBatcher along with a new opam switch:

opam switch create obatcher_test ocaml.5.0.0
eval $(opam env)
opam update
opam install dune batteries progress ptime cmdliner
opam pin add domainslib https://github.com/phongulus/obatcher.git\#wait-for-batch

Note: for VSCode users, you will want to run the following as well to reinstall the prerequisites for the OCaml Platform extension in the new switch:

opam install ocamlformat ocaml-lsp-server

You can now clone and build the code in this repository.

git clone https://github.com/phongulus/obatcher_ds.git obatcher_ds
cd obatcher_ds
dune build

Running tests and benchmarks

Provided tests can be simply run after building with:

dune runtest

For benchmarking, navigate to the benchmark/ directory and use make commands:

cd benchmark
make avltree-batched

See the Makefile for the full list of benchmarks that can be run this way. A more advisable way to run the benchmarks would be to use the provided Jupyter Notebook. It is recommended to set up a Python virtual environment for this purpose:

python3 -m venv venv
source ./venv/bin/activate
pip install -r requirements.txt

When running the Jupyter Notebook, use the virtual environment set up above.

Adding more data structures

TODO.