thvnx / clang-fco Goto Github PK

Hi!

While looking for tools that could perform source-to-source transformation of C/C++ arithmetic expressions for speed optimization purposes, I came into this little tool that I find intriguing. After looking its source, may I ask if there's some LLVM (or thirdparty) documentation (hopefully with examples) that I could follow for learning doing this myself?

Do you know of any C and/or C++ compiler, or static analyzer, or similar tool, that can optimize the source code expressions for speed issuing a source-to-source result, rather than compiled object code?

I suppose that a proper optimization would require to tell the optimizer about the computing cost of each operation, because depending on the device you use, the cost of additions/subtractions/multiplications/divisions/etc... might be different. However, by looking at this source code, I found no traces of the assumed cost for each operation...

Thanks!!