Hacker News Reader

Comment by cbrogrammer | Hacker News Reader

> is **like** writing x86 assembly by hand

yes absolutely, way too granular and low level

What? CUDA is intentionally granular and low level, why do you feel that is a negative thing at this level of abstraction? Are you suggesting the tools should be better so that doesn't have to be the case? I can't figure out what you're actually trying to express here.

CUDA is absolutely fine just like x86 assembly is absolutely fine. They are well thought out. What I am saying is that from my perspective they are as granular and difficult, and programming directly in them for even small projects is not a good idea. GPU programming is, at the moment, stuck in 1960's way of thinking, as there are multiple people writing CUDA code by hand.

There's nothing 1960s about it, that's just not well reasoned. These computation constructs/tools just don't currently have better abstractions while maintaining the desired computational performance.

It's a strange intersection of needs where one wants or needs something like CUDA, but doesn't care to ensure their computation is actually running optimally. If you don't want to be bothered with control and granularity, why are you trying to write high performance CUDA code in the first place?

Would you mind elaborating as to what your hobby project was?

Now C++17 with NVidia parallel libraries is too granular and low level, what a joke.

Yes one can make use of the Assembly like low level APIs, which unlike OpenCL, are opt-in, not forced upon everyone.

Clever devs use the high level C++ and Fortran NVidia libraries.

Reply on news.ycombinator.com