cuezel VS Dagger.jl

I played with a similar idea a while ago: https://github.com/ecordell/cuezel/ (cuezel as in: "Bazel but with CUE"), but I was never sure that what I was doing was in the spirit of CUE.

CUE pushes nondeterminism into "_tool.cue"[0] files that are allowed to do things like IO and run external processes. Tool files scratch a similar itch to Makefiles, but they lack an integrated plugin system like Bazel (hence why I played with the idea of CUE + Bazel).

With Dagger you seem to be restricted to the set of things that the dagger tool can interpret just with like my Cuezel tool you are limited to what I happened to implement.

In CUE `_tool` files you are also limited to the set of things that the tool builtins provide, but the difference is that you know that the rest of the CUE program is deterministic/pure (everything not in a _tool file).

There's clearly value in tooling that reads CUE definitions, and dagger is the first commercial interest in CUE that I've seen, which is exciting.

But I'm most interested in some CUE-interpreter meta-tool that would allow you to import cue definitions + their interpreters and version them together, but for use in `_tool` files to keep the delineation clear. Maybe this is where dagger is heading? (if so it wasn't clear from the docs)

[0]: https://pkg.go.dev/cuelang.org/[email protected]/pkg/tool

Firstly, I'll say that we already have work started to implement out-of-core directly in Dagger: https://github.com/JuliaParallel/Dagger.jl/pull/289.

With that PR in place, it should be possible to define a "storage device" which is backed by a database. I haven't had a chance to actually try this, since the PR still needs quite some work and testing, but it's definitely something on my radar!

I'm not sure that's necessarily the domain of a low-level package like CUDA.jl though (which I assume you're referring to). That kind of interface is more the domain of higher-level packages like https://github.com/JuliaParallel/Dagger.jl/ and to a lesser extent https://juliagpu.github.io/KernelAbstractions.jl/stable/. Moreover, the jury is still out on whether the built-in Distributed module is an ideal abstraction for every use-case (clusters, heterogeneous compute, etc.)

WRT Nx, my biggest question is how they'll crack the problem of still needing big balls of C++ and the shims everywhere to get acceleration. Creating a compiler that generates efficient GPU or other accelerator code is a massive research project with no clear winners, never mind the challenge of reconciling the very mutation-heavy needs of GPU compute with a mostly immutable language model.