[flang-commits] [flang] [llvm] [flang] Use backend fp128 support to determine REAL(16) availability (PR #221907)

[Shunsuke Watanabe via flang-commits]

s-watanabe314 wrote:

Related to #221904.

This patch makes `REAL(16)` availability follow backend fp128 support, allowing the frontend to recognize `REAL(16)` on targets such as x86_64 and AArch64 even when quadruple-precision math intrinsic library support is unavailable. As a consequence, programs like
```Fortran
integer, parameter :: k = merge(16, 4, selected_real_kind(p=33) .eq. 16)
real(kind=k) :: x
test = sin(x)
```
may select `REAL(16)` and later fail to link when quadruple-precision math library support is unavailable. #221904 introduces `-fdisable-real-16` as a way to disable REAL(16) support in such cases.

https://github.com/llvm/llvm-project/pull/221907