[PATCH] D94557: [ARM] Fixed incorrect lowering when using GNUEABI (libgcc) and 16bit floats

[Fernando De La Garza via Phabricator via llvm-commits]

fdelagarza added a comment.

That said, I don't know if making `FPROUND_F64_F16` consistent with `FPEXT_F16_F32` and `FPROUND_F32_F16` is necessarily the right thing to do

Being explicit with a `Subtarget->isTargetGNUAEABI()` code branch also seems like a good option and would minimize change to other subtargets

For example, I'm not sure what the android triplet is, but the swift stdlib is already adding an implementation for `__aeabi_d2h` for android amv7 platforms: https://github.com/apple/swift/blob/cd2d4cc287814c39f33ee1f57ec05aac3f0537e4/stdlib/public/runtime/Float16Support.cpp#L168


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D94557/new/

https://reviews.llvm.org/D94557