Open ponyatov opened 6 years ago
As for aarch64 crosompilation in docker with qemu-aarch64-static works. For 32bit arm results in very large package.
As a follow up, would it be possible to create a yocto script for llvmlite and numba? I would be interested in using numba on an ARM platform, but would like to create the ARM binaries on my X86_64 desktop PC using bitbake.
Would this be feasible at all?
Has there been any activity on this? I'd also like to get llvmlite and numba building via yocto for an aarch64 system built on my X86_64 system.
I'd also welcome any help on how to just build a wheel for each that's compatible with my system. I have a Petalinux 2020.1 system (Yocto Zeus) and I'm struggling to get through the llvmlite build issues. At the moment I'm trying to install llvmlite==0.36.0 since my LLVM==9.0.1, but I'm getting a "SVML not detected" message along with g++ errors of "unrecognized command line option '-Qunused-arguments'".
Any guidance would be appreciated!
@proximous thank you for asking about this. The wheels on PyPI for llvmlite include aarch64
wheels by now:
https://pypi.org/project/llvmlite/#files
Will this suffice for you use-case?
Need some direct code composing (managed compiling) for embedded targets. Is any samples exist on using llvmlite for cross targets?