Dan Gohman

Yes. The spec testsuite has tests for negative zero, infinity, and subnormals, cases which would round differently in 80-bit precision, cases which would round differently on x87's double-precision mode, and...

Interesting question. Here's what we know: - IEEE 754 implementations of `sqrt` are required to operate as if they computed the infinitely precise result and then rounded it to the...

Promoting 32-bit or 64-bit inputs to 80-bit, computing `sqrt` results rounded to 80 bits, and then rounding those results down to 32 or 64 bits in a separate step suffers...

Yeah, something liike `__memory_base` and `__table_base` is possible. They might not be imports from `env`; I've been contemplating a scheme where modules load themselves, using `malloc`, `memory.init`, `table.grow`, and `table.init`,...

I'm open to the idea of ISA-specific APIs. That's an interesting conversation to have. However, it still makes sense to have an "intersection"-ish API to serve as a common shared...

On the topic of _mm_max_ps in particular: We won't be mapping _mm_max_ps onto the JS-SIMD max function. _mm_max_ps has defined behavior on NaN, and no matter what we do in...

As a follow-up, I just added implementations of _mm_max_ps and _mm_min_ps to Emscripten's xmmintrin.h using compare+select as described above. This makes the NaN and -0.0 handling exactly match that of...

I expect we'll beat 2x in many cases with the portable API. Even though they have dominated the discussion here, min and max are a sideshow compared to add and...

@juj, @andhow, and @kripken and I discussed this earlier today. The conclusion was that if we're going to embark on a bold new strategy here, we'll need some compelling arguments...

Hi @chadaustin. I see that you're passionate about this issue, which is great, because we would benefit from some help :-). One thing that would help would be testcases, preferably...