fp128 implementation on x86


I am interested in language and compiler design so I am currently using LLVM to implement a C++ like language.

I have a question about fp128 support on an x86 target.
Am I correct to assume that the implementation is software only? Has any analysis been done to compare the computational speed of fp64 vs. fp128 on x86?

Any information including references to relevant source files would be very useful.

Thank you very much.