add Intel OneApi compiler to CI
Changes done to CMake:
- Disable MemoryUsage test for intel LLVM. Apparently, the intel compiler allocates more memory than needed. But as far as I know, there are no "rules" or standards on how much memory will be allocated with a
std::vector
- Add a check if "the OpenMP sign conversion bug" (https://bugs.llvm.org/show_bug.cgi?id=48387) exists and disable the corresponding warning
- add
WALBERLA_BUILD_WITH_INTELLLVM
variable - disable Fast Math if it is not explicitly turned on. Otherwise, it would be enabled with
O1
and above
Depends on https://i10git.cs.fau.de/walberla/buildenvs/-/merge_requests/32
Edited by Dominik Thoennes