Sliced Data Structure for Particle Simulation Memory Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Particle-based fluid simulations using a fixed calculation grid face inefficiencies in memory usage and calculation area limitations, particularly in proximate particle searches, due to the need for extensive memory allocation for voxels with no particles, limiting the scale of simulations.
Innovation Solution
A sliced data structure divides the three-dimensional calculation space into two-dimensional areas perpendicular to a single axis, determining rectangular ranges (bounding boxes) for particles, allowing memory allocation only for voxels containing particles, and utilizing this structure for efficient loading into a GPU for improved memory efficiency and proximate particle search performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed calculation grid is used for proximate particle search in particle-based simulation, then the search efficiency is improved, but the memory usage increases significantly due to allocation for voxels containing no particles
Solution Approach 1:
The three-dimensional calculation space is divided into multiple two-dimensional sliced areas (slices) perpendicular to a single arbitrary axis. This segmentation allows the system to process and manage voxels in a more granular and efficient manner, reducing the need to allocate memory for entire 3D grids that may contain many empty regions.
Solution Approach 2:
The patent transitions from a traditional three-dimensional grid structure to a two-dimensional sliced area structure by introducing slicing along an arbitrary axis. This dimensional reduction allows for more efficient memory utilization while maintaining the ability to perform proximate particle searches, as slices can be processed independently and only contain voxels that are relevant to the simulation.
2Ease of operation
If a fixed calculation grid is used to enclose the calculation area, then the proximate particle search can be performed, but the calculation area is limited and memory efficiency deteriorates
Solution Approach 1:
The calculation area is no longer fixed by a static 3D grid but becomes dynamic through the use of multiple two-dimensional slices that can be configured independently. Each slice can be adjusted to match the actual distribution of particles, allowing the calculation area to adapt to different simulation scenarios and particle configurations.
Solution Approach 2:
Instead of applying a uniform grid structure throughout the entire calculation space, the patent allows each two-dimensional slice to have its own specific characteristics and boundaries. This enables local optimization where memory is allocated only for voxels that actually contain particles in each slice, improving both memory efficiency and adaptability to different calculation areas.
Data Source
AI summary
The sliced data structure used for a particle-based simulation using a CPU or GPU is a data structure for a calculation space. The space is a three-dimensional calculation space constructed from numerous voxels; a plurality of slices perpendicular to the Y axis is formed; numerous voxels are divided by a plurality of two-dimensional slices; the respective starting coordinates of the maximum and minimum voxels are calculated for a range of voxels in which particles are present in each of a plurality of two-dimensional slices; the voxel range is determined as a bounding box surrounded by a rectangular shape; and memory is provided for the voxels contained in the bounding boxes of each of the plurality of two-dimensional slices.