Scripts and documents from Computational Methods for Physics course at University of Studies of Milan (2019)
Inside this folder you can find all data, documents and scripts used to evaluate the performances of the Gauss-Jordan algorithm to reduce a large NxN matrix to triangular form.
For the full written report (in Italian) see: /Computational-Methods-for-Physics-2019/relazione/PDF/Relazione_Metodi_Computazionali_della_Fisica.pdf
Here a brief summary: The experience descripted consists of the characterization of the Gauss-Jordan algorithm to reduce a large NxN matrix to triangular form, where N gets up to 4000-5000. This has been done comparing time and error performances on a simple CPU as well as on an Nvidia GPU (Geforce GTX 480), mainly focucing on the last one as requested by the course. The report is divided in three section: in the first one the related theory is presented, while in the second and third sections the CPU and GPU performances are described respectively.
This experience has been completed by myself under the supervision of Professor A. Vicini.