steffenlarsen wrote: > +329 -19 doesn't feel right to me....
It was previously suggested that I split the PR up, but I wanted to wait for general concensus on the approach. If we think this is the best option (even if it is not a great option) I would be happy to go split it. > > To avoid the size restriction that may be imposed by the address spaces of > > the target > > What's the size restriction? The AST would previously embed into a global variable to get it into the corresponding sections, then rely on the memory not being instantiated. With https://github.com/llvm/llvm-project/pull/179625, this doesn't fly with the verifier as it requires memory to fit in its corresponding address space, i.e. be addressible, so if an address space uses 32-bit pointers any memory object inside it can be at most ~4GB, which might not be enough for embedded AST on some projects. Note that https://github.com/llvm/llvm-project/pull/212733 was the initially proposed alternative. https://github.com/llvm/llvm-project/pull/214135 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
