Perform a cross-benchmark on between 8 Tesla A800 GPU server and GeForce RTX 3090. Single GPU performances on the basic matrix operations (Python C++) and real life large-scale neural net

Build a Linear Autoencoder (LAE) comparable to Proper orthogonal decomposition (POD) in two ways. The modes extracted by the two algorithms are compared on the wide (MINST [28*28, 60000])

Try to clarify the connections between reduce order models and finite element methods, especially in terms of the weighted residual and the Galerkin projection methods. Concepts of finite

A method of launching Paraview as a web server, and visualizing in localhost.

Attend to validate a turbulence modeling, need a quick recap on the classical theorem.

under construction... After inspecting the CUDA Particles simulation template. A simplest SPH model is implemented based on it.

This is a code review of particle simulation CUDA template . Several changes are also documented afterwards.

The limits of performance and way of optimisation are introduced with examples, mostly memory optimisation and warp divergence.

More details of the basics, the matrix multiply example is carefully analysed and a optimised version is coded.

Basic CUDA programming methods, with a simple matrix multiply example constructed from the scratch.