[flang-commits] [flang] [flang] optimize `tand` precision (PR #155629)

[Connector Switch via flang-commits]

c8ef wrote:

BTW, I have a question about the f128 provider in Flang.

In the current main branch, `FLANG_RUNTIME_F128_MATH_LIB` can only be set to `libquadmath`, which is shipped with gcc for glibc versions below 2.26. With newer glibc versions, we can use the `f128` variant of math functions directly. However, I came across some code that indicates it is not supported. In issue #83874, it is mentioned that `FLANG_RUNTIME_F128_MATH_LIB` can be set to `libm`. I am unsure why it is currently not supported.

https://github.com/llvm/llvm-project/blob/f3a5c16b9810fc12e7be35ff719be10427338256/flang-rt/lib/quadmath/complex-math.h#L59

https://github.com/llvm/llvm-project/pull/155629