a2b28f9472
Due to SIMD's v128 operations, vectors that may have had an explicit type (e.g. f32x4) before a v128 operation will subsequently have CLIF's v128 stand-in type: i8x16. In order for follow-on operations (that may be more stricly typed, e.g. f32x4.add) to avoid CLIF errors, we must bitcast them back to the operation type. The raw_bitcast operation used to do this emits no machine code but does incur some small compile-time cost; it would be nice to avoid this in the future.
This crate performs the translation from a wasm module in binary format to the in-memory form of the Cranelift IR.
If you're looking for a complete WebAssembly implementation that uses this library, see Wasmtime.