They can also do matrix multiplication, which is the relevant thing here.

Mojo is not an OpenCL-style library; it is a programming language, and it doesn't really abstract much. However, it is designed to support compilation to exotic hardware. Cuda CPP can also be thought of as a language; however, the compilation model and core primitives are fundamentally tied to Nvidia hardware.

> Like what? None of this is a rhetorical question, explain to me what information would help bring GPUs and TPUs up to parity with CUDA.

One primitive would be a data/memory abstraction libary like CuTe but for more than Nvidia: https://mojolang.org/nightly/docs/layout/

Another is a programming model that allows you to specialise based on hardware and hardware capabilities: - https://github.com/modular/modular/blob/c71933db1b22b6a91749... - https://github.com/modular/modular/blob/c71933db1b22b6a91749...

And finally, a novel compilation model that knows how to compile and link code for esoceteric hardwares: - https://www.youtube.com/watch?v=77Z_kngHs00

Ah, it seems impossible to you. These are very different hardwares... It is hard enough to make compatibility among different hardwares of the same vendor. Very difficult to imagine building primitives for hardwares with completely different memory layouts.

> How much performance do you perceive to be left on the table with native CUDA-based implimentations?

Zero is the idea. And I wasn't saying there should be a native cuda-based implementation, I'm asking you to imagine how much easier everything would have been if Cuda was cross-platform without any performance or ergonomic penalties.

Mojo is a foundational step here. The big HOW is powerful parametric programming. So much information could be passed during compile time which the compiler uses to specialize.

The messaging was changed because people got sold too hard on that, and kept trying Mojo with the expectation that it could compile existing Python code when it couldn't. What Modular did was change the messaging to reflect what Mojo is today, and provide a roadmap[1] of what they hope it'll turn into in the future. As it evolves, the messaging will evolve with it to continue reflecting current capabilities.

1. https://mojolang.org/docs/roadmap/

Mojo and Modular's Max platform would do to heterogeneous compute what LLVM did to programming language development. People who dismiss the real value offering here know nothing. Modular have already raised $350m+ from industry giants (including Nvidia and Google) to solve this, and I believe they will.

Of course, this won't be necessary in most cases if you're building on top of abstractions provided by Modular.

You don't get this choice using vendor-specific libraries; you're locked into this or that.

Comments URL: https://news.ycombinator.com/item?id=48053091

Points: 17

# Comments: 1