-
-
Notifications
You must be signed in to change notification settings - Fork 16
Description
My idea was to create an app which runs native and on the internet
using KernelAbstractions to do work. And use a tool like juliaC to to create shared libraries or webassembly modules which then include the shaders and can be loaded or updated. Now I have arm and intel and AMD computers. Which means I need cross compilation
I dreamd about a lisp like language which can go down to native speed more like C/Rust/.. and that's what Julia is very close to.
I also wondered how to bridge the gab eg shell vs high level language
or why do we have to bear that JS is 10x slower than eg ocaml.
And tools like QwikKS the need backend and frontend (js/webassembly) derived from the same code.
If I understand julia is meant to run on the architecture it runs on. And for that it compiles and optimizes. So what I as for is little bit alien, right ?
The question is a different one ? Would it be doable with static analysis.
Eg analyse assuming you run on Metal/Vulkan/.. and on arm .. and now create binaries following reachable code with entry point X ?
Then also Android could be targeted. Now jitting integration is huge.
https://github.com/vnmakarov/mir tries to find a fast minimal alternative.
And shaders could be sent to a server and cached.