Releases: byuflowlab/FastMultipole.jl
Releases · byuflowlab/FastMultipole.jl
Multithreading adjustments
Changes
- improved multithreaded performance for small problems
v2.0.2
Relative Error Tolerance
This release contains the following updates:
- implemented relative error tolerance based on a user-defined estimate of potential or gradient magnitude (likely obtained from the previous step of a time-stepping or iterative algorithm)
- cleaner branch structure (which is why this is a breaking change)
Fully Automated Tuning
This release includes the following features:
- fully automated tuning with the
tune_fmmfunction - more conservative error constraint algorithm
- Fast Gauss Seidel solver as published in AIAA Aviation Conference 2024 (doi:10.2514/6.2024-4018)
- new data structures to avoid sorting the user defined system when generating the octree and to improve type stability
- lockless multithreading (on CPU)
- easy installation via the Julia Registry
Dynamic Expansion Order
Implement dynamic expansion order selection based on error tolerance.
Use O(p^3) multipole-to-local transformation
v0.3.0 v0.3.0
GPU for direct calculations
This release uses a syntax to make it possible to offload direct interactions to GPU.
Fast Gauss-Seidel Solver with FLOWPanel (tmp)
Temporary version for developing FLOWPanel with the Fast Gauss-Seidel solver (doi:10.2514/6.2024-4018).