I am looking to develop an LLVM backend for an accelerator architecture. The end goal is the have this backend integrated into LLVM as a accelerator target for OpenMP 4.x for compute intensive code sections.
I hoping that someone on the list could point me to some documentation and/or code examples of how this integration is done?
I found the information from
https://github.com/clang-omp/OffloadingDesign is useful. For LLVM
Backend in general, you may want to have a look at
Please note that OpenMP 4.x offloading implementation in trunk is still under development. It can’t compile “hello, world” that executes on a non-host device at the moment.