This allow clang/MSVC to use __int128 and get runtime functions from clang_rt.builtins-<arch>.lib.
The following tests are temporarily disabled for MSVC as both pure MSVC and clang/MSVC do not have __float128 and 80-bit precision long double. Fix them when clang/MSVC exposes __float128 and 80-bit precision long double.
- fixunsxfti_test.c
- fixxfti_test.c
- floattixf_test.c
- floatuntixf_test.c
I find this and related changes to be not acceptable. It hard-codes the assumption that the division is not lowered natively by the compiler, which is IMO not valid.