This repostory contains the implementation, data, scripts, and report for the final assignment for the course of High Performance Computing.
The project consists of:
- Exercise 1: Implementing a scalable, hybrid MPI + OMP version of Conway's game of life and analyzing the scalability.
- Exercise 2: Benchmarking the three high performance math libraries: MKL, OpenBLAS, and BLIS.
The full problem description can be found in the repository of the course.
All the problems were analyzed using the Orfeo super computer using:
- "THIN" nodes: 24 cores of Intel Xeon Gold 6126 @ 2.6 GHz.
- "EPYC" nodes: 128 cores of AMD EPYC
README.md
report.pdf
report/all the files used for the report
exercise1/README.md
exercise1/all the files used for exercise 1
exercise2/README.md
exercise2/all the files used for exercise 2
Where:
- README.md briefly describes what was done and how to navigate the directory.
- report.pdf contains a detailed report of all the exercises.
- exercise1/README.md contains all the necessary details for exercise 1
- exercise2/README.md contains all the necessary details for exercise 2
Exercise 1 consists in the implementation of a scalable, hybrid MPI + OMP version of Conway's game of life. Furthermore, an analysis of the strong OMP, strong MPI, and weak MPI scalability will be conducted in the report.
Exercise 2 consists of benchmarking the level 3 math function gemm for matrix-matrix
multiplication for three HPC math libraries, namely, MKL, OpenBLAS, and BLIS.
We compare the obtained GFLOPS with the expected peak performance for the two types
of architectures in Orfeo in two scenarios.