REDUCTION METHOD IMPLEMENTATION ON MPM ALGORITHM FOR 3D FLUID SIMULATION
To achieve a good fluid simulation requires efficient algorithms and high computational power. An alternative method of doing fluid simulation is by using the MPM algorithm on the CUDA platform. Research done by Kusnadi (2018) has implemented a parallelized version of MPM on the CUDA platform. A...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/51254 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | To achieve a good fluid simulation requires efficient algorithms and high
computational power. An alternative method of doing fluid simulation is by using
the MPM algorithm on the CUDA platform. Research done by Kusnadi (2018)
has implemented a parallelized version of MPM on the CUDA platform. A few
things that can be improved from that implementation is to improve the
simulation from 2 dimensions to 3 dimensions to better represent real-life fluids
and to change the atomic method used on the algorithm to a reduction method.
To change from 2D simulation to 3D simulation just requires adding variables and
adding loops to calculations to represent the new dimension. Meanwhile, to
change the atomic method used it is required to firstly group the particles based
on the node they are going to be grouped into. This grouping is done by
implementing a sorting algorithm first.
Based on the result of testing, the new implementation using the reduction method
has better performance than the implementation using the atomic method.
|
|---|