Abstract:We propose a novel approach to extracting crack-free iso-surfaces from Structured AMR data that is more general than previous techniques, is trivially simple to implement, requires no information other than the list of AMR cells, and works, in particular, for different AMR formats including octree AMR, block-structured AMR with arbitrary level differences at level boundaries, and AMR data that consist of individual cells without any existing grid structure. We describe both the technique itself and a CUDA-based GPU implementation of this technique, and evaluate it on several non-trivial AMR data sets.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to efficiently and accurately extract crack - free iso - surfaces from Adaptive Mesh Refinement (AMR) data. Specifically, due to its irregularity and multi - level structure, AMR data makes the calculation of crack - free iso - surfaces very challenging. Existing methods, such as the "stitching" method proposed by Weber et al., rely on specific mesh hierarchies and are complex to implement. Moreover, these methods may encounter difficulties when dealing with complex AMR data sets, such as those containing multiple - level boundaries or "holes".
### Main contributions of the paper
1. **Proposed a new method**: This method can generate a dual mesh for any AMR data set and can extract crack - free iso - surfaces from it. This method does not require any additional information, only a list of AMR cells is needed.
2. **Simplified the construction of the dual mesh**: By using a simple integer - labeling scheme and a fast cell location kernel, this method avoids complex epsilon - offset operations and simplifies the method proposed by Moran et al.
3. **Easy to parallelize**: This method is completely independent at each cell level, so it is very suitable for parallel implementation on GPUs and multi - core CPUs.
4. **Applicable to multiple AMR formats**: Including octree AMR, block - structured AMR, and AMR data sets composed of single cells without an existing mesh structure.
### Specific problem description
- **Characteristics of AMR data**: The characteristic of AMR data is that the initial coarse grid is adaptively refined according to simulation requirements, which significantly reduces the number of discretization elements required to reach a certain precision. However, this irregular grid structure makes visualization and iso - surface extraction complex.
- **Limitations of existing methods**: Although the "stitching" method of Weber et al. is widely used, it relies on a specific mesh hierarchy and requires more complex case tables when dealing with complex situations. The method proposed by Moran et al. is more general, but less well - known.
### Solution
The method proposed in the paper solves the above problems through the following steps:
1. **Define logical cells and dual cells**: A set of simple terms is introduced to distinguish between logical cells and actual cells, and the concept of dual cells is defined.
2. **Efficient cell location**: Use a fast cell location kernel (SNAP(C)) to find the cells that actually contain these coordinates given the coordinates.
3. **Construction of the dual mesh**: Generate a dual mesh by "snapping" the vertices of the logical dual cells, that is, "adsorbing" them to the centers of the actual cells.
4. **Avoid repeated generation**: Ensure that each dual shape is generated only once through three simple rules, avoiding repeated calculations.
5. **Iso - surface extraction**: Input the generated dual mesh into the Marching Cubes algorithm, and crack - free iso - surfaces can be correctly generated even if some dual cells degenerate into wedges, tetrahedra or line segments.
### Experimental results
The paper shows the application of this method on several complex AMR data sets, including the NASA Exajet model, Astro model, Cloud model and Landing Gear model, etc. The experimental results show that this method can generate iso - surfaces of millions or even hundreds of millions of triangles within a few seconds, with high efficiency and accuracy.
### Summary
This paper proposes a simple, general and efficient method for extracting crack - free iso - surfaces from AMR data. This method not only simplifies the construction of the dual mesh, but is also very suitable for parallel computing and is applicable to data sets of multiple AMR formats.