krwc / volume-modeler

OpenCL implementation of dual contouring

Introduction

This is an implementation of the grid-based Dual Contouring algorithm entirely on the GPU. It lacks many important features, like materials, some actual shading, mesh simplification - to name a few.

It's no longer under development, and likely would never be again.

What can you do with it

At the moment it allows to add / subtract two predefined surfaces (sphere and a cube). Result of the operations are saved on the fly under scene/ subdirectory automatically. So, when restarted, the scene created previously will get loaded.

It implements a QEF solver, allowing to reproduce sharp features relatively well.

Pictures

Keyboard & mouse control

• W, S, A, D moves the camera around the scene,

• 1, 2 switches between first (sphere) / second (cube) brush,

• Left ALT causes the brush to rotate with camera,

• F1, F2 switches between wireframe and solid rendering,

• ESC causes the mouse cursor to not be grabbed by the application anymore.

• Mouse Left adds the brush at the position indicated by the rendered bounding box,

• Mouse Right subtracts the brush at the position indicated by the rendered bounding box.

Compilation and configuration

There are a number of CMake configuration options that can be tweaked:

• VM_CHUNK_SIZE - size of the single voxel chunk (64x64x64 by default),
• VM_VOXEL_SIZE - distance in world-space unit between two voxels (0.02 by default),
• WITH_FEATURES - allows to enable reproduction of sharp features (off by default),
• WITH_TEST - enables compilation of unit tests (on by default).

Compilation

• git submodule update --init
• mkdir build && cd build
• cmake .. -DWITH_FEATURES=ON
• make -j$(nproc)

Dependencies

• OpenCL 1.2
• GLFW3
• glm
• glew
• zlib
• Boost filesystem, regex, iostreams
• pthreads