I see from the paper that you compute the fields from wire configurations which require 1-D integrations. In this case I suggest to use the QuadQK package (https://github.com/JuliaMath/QuadGK.jl) which impements state-of-the art adaptive integration routines, with the benefit that it is very well suited to integrands with singularities (which can be passed to the routine). To increase performance I would not resort to GPUs but rather parallelize the computation within standard Julia by either parallelizing on the conductor pieces (for all points) or on the evaluation points (for all conductors). Hope this helps.
Pier
6
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Is Julia integral calculation not as fast as MATLAB? | 20 | 1560 | April 22, 2022 | |
| Unreasonable slow speed in numerical integration | 16 | 2958 | November 12, 2022 | |
| Numerical integration over 3D domain of vector-valued function | 33 | 1667 | April 6, 2024 | |
| Double integration with hcubature slower than nested quadgk | 8 | 1734 | January 19, 2023 | |
| Problem with small integrals | 19 | 1108 | August 12, 2023 |