Crate wasmtime_internal_math

A minimal helper crate for implementing float-related operations for WebAssembly in terms of the native platform primitives.

⚠️ Warning ⚠️: this crate is an internal-only crate for the Wasmtime project and is not intended for general use. APIs are not strictly reviewed for safety and usage outside of Wasmtime may have bugs. If you’re interested in using this feel free to file an issue on the Wasmtime repository to start a discussion about doing so, but otherwise be aware that your usage of this crate is not supported.

This crate is intended to assist with solving the portability issues such as:

• Functions like f32::trunc are not available in #![no_std] targets.
• The f32::trunc function is likely faster than the libm fallback.
• Behavior of f32::trunc differs across platforms, for example it’s different on Windows and glibc on Linux. Additionally riscv64’s implementation of libm seems to have different NaN behavior than other platforms.
• Some wasm functions are in the Rust standard library, but not stable yet.

There are a few locations throughout the codebase that these functions are needed so they’re implemented only in a single location here rather than multiple.

Traits

WasmFloat

Functions

f32_cvt_to_int_bounds

Returns the bounds for guarding a trapping f32-to-int conversion.

f64_cvt_to_int_bounds

Same as f32_cvt_to_int_bounds but used for f64-to-int conversions.