Dan Gohman

It seems out of scope for v1 to me. Some people close to TC39 have encouraged me that future iterations of SIMD.js should relatively simple and quick to standardize, because...

I appologize; I've been busy and haven't reviewed this in detail yet. Also, FWIW, an interesting analogous conversation is happening in LLVM right now: http://reviews.llvm.org/D9029

The opaque_type solution here looks nice, though I haven't yet determined whether this will be usable for our use cases. I do still plan on digging into this further. It...

My new proposal is as follows: we drop sumOfAbsoluteDifferences and add absoluteDifference. This is `vabd` on ARM, and on x86, it might expand to this sequence (AT&T operand ordering): ```...

A more complete version of my proposal is now available, in #209.

Compatible with @simhos01's observation, my full proposal in #209 includes a widening form of absoluteDifference as well as a horizontal sum function. I realize that this is adding yet more...

Right now, yes, it's shuffle and repeat. I'm open to refining this, but I think this is a good start.

The next step in the TC-39 process is to request Stage 3 approval. At Stage 3, the task is to "Indicate that further refinement will require feedback from implementations", and...

> Although obviously no committee permission is needed to get implementation experience, I guess it'll be easier to remove things from Stage 3 to Stage 4 than to add them....

Work on a spec for SIMD.js in asm.js is presently being done on [this Specification page](http://discourse.specifiction.org/t/request-for-comments-simd-js-in-asm-js/676).