是否可以在内存中使用Clang C ++ API使用Clang编译源代码？

Question

I'd like to compile the sources using Clang with no executables running, but in-memory via API. I've found post about how to do it but Driver seems to find and run executables like clang / clang++ and ld behind the scene:

The output shows that the Driver is invoking clang to do the compiling and ld to do the linking. As you can see, the Driver adds arguments of its own to each invocation, in addition to the ones we passed in. The -v flag shows you exactly how the compiler and linker are being invoked.

As i'm going to do it on iOS running new process is prohibited ( fork is prohibited) and paths are completely different.

So it there any way to compile the sources using Clang API with no executables running?

Answer 1

Yes, this is possible. It's how most OpenCL implementations work - they use the clang and llvm libraries to build a single executable (typically) that produces code for the hardware that CL runs on. I believe CUDA does something very similar, but I haven't studied it.

It's not terribly easy, but I guess I could have something up and running in a few days, maybe a couple of weeks worst case. I think the biggest problem will be the linking. In particular if you don't actually want to produce a executable file, but make an "in memory execution of the code" - which involves allocating memory that has execute rights, and then load the code, which is normally tasks that the linker and loader are responsible for.