MSBG - GitLite

KO | EN

gitlite — search

#security #api #devops #vue #linux #graphql #machine-learning #javascript #html #claude-code

MSBG

(★ 178)

Multiresolution Sparse Block Grids (MSBG) library

File Explorer

ZIP Download

- bun_zipper.ply
- bun_zipper_res2.ply
- - dispatch_example.cpp
  - instrset.h
  - instrset_detect.cpp
  - LICENSE
  - README.md
  - vector_convert.h
  - vectorclass.h
  - vectorf128.h
  - vectorf256.h
  - vectorf256e.h
  - vectorf512.h
  - vectorf512e.h
  - vectori128.h
  - vectori256.h
  - vectori256e.h
  - vectori512.h
  - vectori512e.h
  - vectori512s.h
  - vectori512se.h
  - vectormath_common.h
  - vectormath_exp.h
  - vectormath_hyp.h
  - vectormath_lib.h
  - vectormath_trig.h
- comparison_MSBG_octree.png
- msbg_comparison_vdb.png
- msbg_teaser_lores.png
- bitmap.c
- bitmap.h
- bitmap2.cpp
- blockpool.cpp
- blockpool.h
- common_headers.h
- fastmath.cpp
- fastmath.h
- globdef.h
- grid.c
- grid.h
- gwx.c
- gwx.h
- halo.cpp
- halo.h
- internal.h
- kernels_ispc.h
- log.h
- main.cpp
- MGWin.h
- mm.c
- mm.h
- mm2.cpp
- msbg.cpp
- msbg.h
- msbg2.cpp
- msbg3.cpp
- msbg4.cpp
- msbg_demo.cpp
- msbg_demo.h
- msbgaux.cpp
- msbgcell.h
- mtool.c
- mtool.h
- mtool2.h
- noise.cpp
- noise.h
- panel.c
- panel.h
- plot.c
- plot.h
- pnoise.c
- pnoise.h
- pnoise2.cpp
- pnoise3.cpp
- rand.c
- rand.h
- readpds.h
- readpng.c
- render.cpp
- render.h
- sbg.cpp
- sbg.h
- stdmath.h
- thread.cpp
- thread.h
- util.c
- util.h
- util2.cpp
- VEC3.h
- vectorclass_util.h
- vectorclass_util2.h
- visualizeSlices.cpp
- windows_linux.h
- wturbglob.h
.gitignore
LICENSE
makefile
mk
README.md

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<p align="left">
  <img src="images/msbg_teaser_lores.png" alt="MSBG teaser image" />
</p>

Implementation of the **Multiresolution Sparse Block Grid (MSBG)** data-structure as introduced in the SIGGRAPH 2025 paper [Adaptive Phase-Field-FLIP for Very Large Scale Two-Phase Fluid
Simulation](https://ge.in.tum.de/download/Adaptive_Phase_Field_FLIP_preprint.pdf) by Bernhard Braun, Jan Bender and Nils Thuerey.

---

## Key Features

* Extremely efficient treeless spatially adaptive data structure supporting both sparseness and multiresolution adaptivity.  
* No GPU needed. Designed for maximum efficiency on the CPU in order to leverage maximum RAM ressources for very large scale scenarios.   
* On a single CPU (32 cores, 256 GB RAM), MSBG is able to process sparse 3D Volumes at an effective resolution of 32768^3 containing 100 billion active (out of 30 trillion virtual) voxels at 16 bit precisison.
* Highly efficient PDE solver for sparse volumes that can advance the solution at a speed of ~10 billion unknowns per second per iteration. 
* Command-line demo program demonstrating sparse surface reconstruction from particles and sparse PDE smoothing (mean curvature flow).

## Bunny-of-bunnies example

The included demo executable implements the bunny-of-bunnies example (left-hand side of the teaser image above), as described in Section 8.1 of the paper  [Adaptive Phase-Field-FLIP for Very Large Scale Two-Phase Fluid
Simulation](https://ge.in.tum.de/download/Adaptive_Phase_Field_FLIP_preprint.pdf): A surface, sparsely embedded in an MSBG volume is reconstructed via lockless 8-color particle splatting and MSBGs multicolor-in-place mean curvature flow PDE solver from 32,000 Stanford bunnies—each of which, in turn, consists of 32,000 point samples. The example was ran with an effective resolution of 32,768 × 32,768 × 32,768 on a single CPU with 256 GB of RAM. Here, the MSBG base block resolution was set to 32x32x32.

## Comparison with OpenVDB

<p align="left">
  <img src="images/msbg_comparison_vdb.png" alt="comparison_MSBG_VDB" />
</p>

## Comparison with Block Octrees

Unlike OpenVDB, MSBG supports not only sparsity but also multiresolution adaptivity—traditionally implemented using block based tree structures. MSBG offers an attractive, tree-less alternative, as demonstrated in the following comparison.

<p align="left">
  <img src="images/comparison_MSBG_octree.png" alt="comparison_MSBG_octree" />
</p>

## Prerequisites

MSBG has only a few dependencies: Just C++11 and TBB <https://github.com/uxlfoundation/oneTBB>. The header-only VectorClass library <https://github.com/vectorclass/version1> is
already included for convenience. The MSBG library compiles on Linux as well as on Windows via MSYS2/MinGW64.

## Limitations

At this stage, the MSBG library is still in an experimental phase (research code). Some features may be insufficiently tested and lack comprehensive documentation. It is recommended to refer to msbg_demo.cpp for usage examples of the library.

---

## Building from Source

```bash
git clone https://github.com/your-org/MSBG.git
cd MSBG

mkdir build && cd build

# simple Make-based build provided for convenience
../mk               # builds the library and 'msbg_demo' executable
msbg_demo -h        # Show options