ABSTRACT

Parallel processing divides data to be processed in each core on the CPU. One programming language that is widely used for machine learning applications and can accommodate the use of parallel processing is Python. Multiprocessing, mpi4py and CuPy are examples of Python libraries that can process data in parallel. In this research, we will compare the use of these libraries to process huge amounts of data. This research takes agricultural data for data image enhancement through color conversion from RGB to grayscale. The results showed that multiprocessing, mpi4py, and CuPy can increase the image enhancement speed three times faster than its singlecore execution. Then, using a combination of multiprocessing with CuPy, a 1.7 times performance improvement is achieved compared with multiprocessing-only. Also, using a combination of mpi4py with CuPy achieved 2.5 times performance improvement compared to mpi4py-only.

Keywords: - Core CPU, parallel processing, mpi4py, multiprocessing, CuPy