Tsunami Extension for Rem's Engine

Original Project Idea

My project for the Tsunami Lab is split into two parts, which are developed together. On the one hand, the solver and simulation parts will be integrated into my game engine "Rem's Engine" as an extension. Rem's Engine is a Kotlin¹, Entity-Component based engine, which is built on top of OpenGL (using LWJGL). The goal is to visualize and simulate smaller tsunami simulations in real time using OpenGL. This will allow the user to change and investigate setup and simulation parameters directly.

On the other hand, the solver currently only works on the CPU and the project will investigate potential performance improvements by using the GPU. There are two straightforward ways, which can be relatively easily used with OpenGL: graphics shaders and compute shaders. The project will investigate which one is easier to implement, and how performance differs.

Graphics shaders offer the advantage of broader hardware support, because they have been around for longer. Compute shaders offer more flexible memory accesses², shared memory and atomic operations.

The testing hardware for this project will be my desktop computer (Ryzen 5 2600 + Radeon RX 580 8 GB), and optionally my laptop (Ryzen 7 4700U + integrated graphics).

Building

To build this project, Rem's Engine needs to be built (until I create an official release of it). The IDE Intellij Idea is recommended, because all library configuration is already done with it.

Running

This project can be run is different ways:

• run benchmarks using me.anno.tsunamis.perf.Performance
• create a video of the results of such a performance run with me.anno.tsunamis.VideoRenderer
• run this mod with Rem's Engine: me.anno.tsunamis.FluidSimMod
• run Rem's Engine with more mods than just this one: build it, and place the resulting jar in the mods folder (<user>/.config/RemsEngine/mods)

1 Kotlin is a high level language with different compiler targets. The target for Rem's Engine is Java Virtual Machine. ↩
2 Using the graphics pipeline, you can only write in the shape of triangles, lines or points. Writing to arbitrary memory locations directly is not possible from the fragment shader. ↩