[NativeAOT-LLVM] Implement saturating semantics for float to int conversions

dotnet / runtimelab

This repo is for experimentation and exploring new ideas that may or may not make it into the main dotnet/runtime repo.

MIT License

1.36k stars 188 forks source link

[NativeAOT-LLVM] Implement saturating semantics for float to int conversions #2552

Closed SingleAccretion closed 2 months ago

SingleAccretion commented 2 months ago

Using the same intrinsics Rust uses, and which conveniently correspond exactly both to .NET semantics and WASM's non-trapping conversion instructions, which we now enable by default via -mnontrapping-fptoint.

Ref: https://github.com/dotnet/runtimelab/issues/2455#issuecomment-2067534258.

SingleAccretion commented 2 months ago

@dotnet/nativeaot-llvm