WIP: Enable 64bit `long double`

Windows-on-ARM-Experiments / gcc-woarm64

Fork of gcc containing fixes for Windows on ARM64.

GNU General Public License v2.0

8 stars 1 forks source link

WIP: Enable 64bit `long double` #3

Open Blackhex opened 10 months ago

Blackhex commented 10 months ago

I'd split enabling 64bit long double and enabling _Float128 support into separate commits/PRs.
If long double size is not 128bit, libgfortran requires having _Float32. For this reason, 10 and 16 byte REAL has been disabled in this PR until _Float32 support will be added in the next PR.

Closes https://github.com/ZacWalk/mingw-woarm64-build/issues/9

Blackhex commented 10 months ago

@ZacWalk In the context of the above, is _Float128 actually needed for something?

ZacWalk commented 10 months ago

If _Float128 works now I believe we should preserve that functionality.

The main issue I see with these changes is that they need to be specific to aarch64-coff. All other build targets should be unchanged.

For example, aarch64.cc is a file also used by many aarch64 targets. Code changes in this file need to be coff specific - maybe behind an ifdef.

Blackhex commented 10 months ago

If _Float128 works now I believe we should preserve that functionality.

_Float128 is supported only for some targets. Whether it should be supported on aarch64-w64-mingw32 is a question to find out. I'd say it should be suported only if it is supported both on aarch64-pc-linux and x86_64-w64-mingw32.

eukarpov commented 9 months ago

here is a link to CI check https://github.com/Windows-on-ARM-Experiments/mingw-woarm64-build/actions/runs/7007818760

eukarpov commented 9 months ago

Let's postpone review of this PR for a while as it was discussed