user0/rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Enable target_has_reliable_f16_math on x86

Browse source 
This has been disabled due to an LLVM misoptimization with `powi.f16`
[1]. This was fixed upstream and the fix is included in LLVM20, so tests
no longer need to be disabled.

`f16` still remains disabled on MinGW due to the ABI issue.

[1]: https://github.com/llvm/llvm-project/issues/98665
This commit is contained in:
Trevor Gross 2025-03-11 08:08:53 +00:00
parent 1b8ab72680
commit 19e82b43eb
1 changed files with 3 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

10
compiler/rustc_codegen_llvm/src/llvm_util.rs
View file

@ -466,13 +466,9 @@ fn update_target_reliable_float_cfg(sess: &Session, cfg: &mut TargetConfig) {
_ => true,
};
cfg.has_reliable_f16_math = match (target_arch, target_os) {
// x86 has a crash for `powi`: <https://github.com/llvm/llvm-project/issues/105747>
("x86" | "x86_64", _) => false,
// Assume that working `f16` means working `f16` math for most platforms, since
// operations just go through `f32`.
_ => true,
} && cfg.has_reliable_f16;
// Assume that working `f16` means working `f16` math for most platforms, since
// operations just go through `f32`.
cfg.has_reliable_f16_math = cfg.has_reliable_f16;
cfg.has_reliable_f128_math = match (target_arch, target_os) {
// LLVM lowers `fp128` math to `long double` symbols even on platforms where