Calling LLVM IRbuilder from the LLVM JIT compiler

Hello,

A friend and I am making a parser generator and in order to streamline the language development process, we were going to make the LLVM IR builder directly accessible using the LLVM Assembly language as a scripting language to generate the code. I had read somewhere on this list that the JIT compiler can execute native code using function pointers.

My questions are these: What kind of name mangling problems would we run into by calling the LLVM libraries directly from an instance of the LLVM JIT compiler? Is this approach feasible? Is anyone else interested in the techniques we're trying to use?

Thanks in advance,

--Sam Crow

A friend and I am making a parser generator and in order to streamline the language development process, we were going to make the LLVM IR builder directly accessible using the LLVM Assembly language as a scripting language to generate the code.

You propose to write scripts in LLVM IR? To save time?

What kind of name mangling problems would we run into by calling the LLVM libraries directly from an instance of the LLVM JIT compiler?

Don't try to call C++ functions from not-C++. Use the C bindings instead.

— Gordon