On Monday, 31 October 2016 at 09:52:55 UTC, Dicebot wrote:
On 10/30/2016 06:35 PM, Laeeth Isharc wrote:
But what I meant was LLVM will have a wasm backend.
Yes, but it is developed so slowly and conservatively, that
coming up with own proof-of-concept backend may be a chance to
win early interest. They may speed up greatly though when
WebAssembly design gets closer to MVP stage, but I am checking
that regularly.
I was disappointed that after early hype it all went quiet for
now.
I guess there is a window to grab attention, but language like
nim already has a JS backend (so they say - I haven't used), and
I think D is in a different space from languages where how it
plays on hacker news is most important. Because if you have a
wasm backend today for dmd, I guess it will be some time before
it's fast, stable, and has the basic stuff already in
emscripten (though maybe with Remedy work it gets easier to wrap
that). So given limited manpower maybe the easier job
(presumably most work needs to be done on phobos and runtime?)
makes sense because in this case it's also likely the long term
sensible way. It would be great if we had a dmd back end early
of course, and I would certainly use it as soon as it was stable
enough.
But right now it is mostly irrelevant because runtime
requirements have not been defined in WebAssembly at all, only
low level byte code stuff. It is all in very early stages
really.
You have looked into it more than me at a low level, but how is
it possible then to run an app today in nightly browser in wasm?
Is it like saying you don't need glibc, but it's probably a
better idea to use one written for your platform rather than have
some combination of assembly and hacked up library designed for
another architecture and platform (which is what I guess
emscripten do)?
Taking a step back, it's quite amusing how much ingenuity goes
into having to avoid writing Javascript...
