Question: Is wasm-opt production ready?

[Robbepop]

This might sound like a super stupid question but I am here to ask if the usage of Binaryen's wasm-opt specifically is production ready?
We have a great deal of motivation to use it to further optimize our Wasm binaries (produced by the Rust compiler mainly) since wasm-opt seems to do a really good job at this. For instance it was capable of reducing the size of an already Rust-optimized Wasm binary from 27kB to roughly 9kB which is nice because it means that we do not have to introduce certain work arounds on the Rust code in order to improve the resulting code quality. However, we are unsure if the optimizations done by wasm-opt are "guaranteed" to not break or slightly change semantics of the optimized code. Of course, this is a declared anti-goal of every optimizer, however, if this is a real constraint for production usage: Would you still recommend Binaryen's wasm-opt for this use-case?

[MaxGraey]

In my experience it's LLVM do some optimizations stuff which disallow in binaryen due wasm by spec. See this for example: #1911

[kripken]

@Robbepop

Yes, definitely! Being robust and production-ready is a top priority in wasm-opt. We test it very heavily with our own internal test suite, the emscripten test suite, and using fuzzing. It is used by many people in production (anything using emscripten uses it, for example).

(wasm-opt does have optional optimizations that may change semantics; those would not be turned on by default, of course.)

[kripken]

This has been answered.