DavidTruby wrote: > > Do we really need to have all 4 variants of the 3 fortran runtime > > libraries? That's a lot of complexity. Can we pare it down to just > > static/dynamic? It's also sometimes possible to generate code that works in > > both the static and dynamic context, depending on what is in those > > libraries. We don't create 4 variants of clang_rt.builtins, for examle. > > From glancing at the fortran runtime code, I think the answer is probably > "no". There is too much C++ standard library usage. If you wish to avoid this > build complexity, you may consider writing code in the STL-less style that is > used for C++ code in the sanitizers in compiler-rt.
I don't think we can avoid it if we want to allow anyone to link flang-generated object files into a C/C++ application. I don't think we could even get it down to static/dynamic reliably without committing to not only not using the STL but not using any C/C++ functions that might call into the runtime (as compiler-rt builtins does). I don't think that's a route we want to go down with the flang runtime; I think we'd generally put build complexity secondary to code complexity in here ( https://github.com/llvm/llvm-project/pull/70833 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
