Skip to content

Bempp-cl v0.2.0
Compare
Choose a tag to compare
@tbetcke tbetcke released this 09 Sep 14:19
v0.2.0
22f567b
This commit was created on github.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

This release contains a number of major improvements:

  • In addition to OpenCL all operators are now also implemented in Numba, allowing Bempp to be run on systems without OpenCL stack.
    However, OpenCL is preferrable due to its better performance.
  • Internally, the exact implementation of operators has been fully abstracted out so that it is straight forward to write other compute backends
    (e.g. Cuda, Sycl, etc.)
  • All integral operators and potentials can now be assembled via FMM via interfaces to the Exafmm-t library
    (https://github.com/exafmm/exafmm-t).
  • Decorators for grid function callables have been worked over so that it is easy to disable Just-In-Time compilation. Alternatively, a vectorised
    of grid function callables is now possible.
  • The FEniCS interface now allows to obtain trace matrices for Nedelec function spaces for Maxwell FEM/BEM problems.
  • Bempp-cl now has automatic deployment with generation of Docker images and a fully automated testing infrastructure.
  • Many smaller bug fixes and performance improvements.
Assets 2
Loading