Implemented in C++, used boost::python for python bindings.
supports execution in serial mode, eager GPU execution mode and delayed GPU execution mode.
GPU execution mode works best with larger datasets and larger number of continuous operations.
clients files in src have a few examples.
src/example.csv has the expected csv file format.
run src/make_so.sh, creating a file DataFrame.so.
one can import DataFrame from the folder containing the .so file.
additionally, one can use run.sh to convert any code written for eager GPU execution to delayed execution code.