Alexander Root

> > I don't add tests to simd_op_check_x86 > > Understood, but is there any other reasonable way to add checks for this? This sort of stuff is often pretty...

We can do the same thing for HVX saturating narrow instructions, which I was going to do in a separate PR, but it requires the same changes to Bounds.cpp and...

@steven-johnson Luckily, the improvement to HVX can be tested (unlike x86). Unluckily, this revealed that constant bounds inference does not work on u32 like I expected (which is part of...

This has gotten stale, but I am trying to find time to update it. Andrew and I discussed some issues with using bounds inference in this way, but I think...

Yeah, realized it wasn't a ton of work, and was bored during an ASPLOS workshop. This is good for review now! Also, the u32 HVX tests now work, because the...

@abadams Sorry to add a change after your approval, but I just saw a wasm talk and was reminded that we can do this for wasm too. It's basically a...

> The big TODO for wasm now is exploiting constant_integer_bounds to hit the relaxed simd instructions in case where we know they don't have UB Yep, I think #7312 is...

I don't think relaxed SIMD has actually shipped in wasm yet, so that's still waiting on the wasm folks