Triton: Open-Source GPU Programming for Neural Networks

#1

Post by ag8 » Wed, Jul 28, 2021, 4:18 PM UTC

Triton: Open-Source GPU Programming for Neural Networks
openai.com

Re: Triton: Open-Source GPU Programming for Neural Networks

#2

Post by riyadparvez » Wed, Jul 28, 2021, 4:55 PM UTC

I am confused. Is it another competitor of Tensorflow, JAX, and Pytorch? Or something else?

Re: Triton: Open-Source GPU Programming for Neural Networks

#3

Post by peytoncasper » Wed, Jul 28, 2021, 5:00 PM UTC

riyadparvez wrote: post #2

I am confused. Is it another competitor of Tensorflow, JAX, and Pytorch? Or something else?

I believe this is more of an optimization layer to be utilized by libraries like Tensorflow and JAX. More of a simplification of the interaction with traditional CUDA instructions.

I imagine these libraries and possibly some users would implement libraries on top of this language and reap some of the optimization benefit without having to maintain low-level CUDA specific code.

Re: Triton: Open-Source GPU Programming for Neural Networks

#4

Post by croes » Wed, Jul 28, 2021, 5:03 PM UTC

Too bad it's CUDA Sooner or later this will become a problem because you are depending on the benevolence of a single manufacturer.

Re: Triton: Open-Source GPU Programming for Neural Networks

#5

Post by boulos » Wed, Jul 28, 2021, 5:04 PM UTC

Folks might find the author’s research paper [1] while at Harvard more informative. This is a great high-level description, but if you want more detail, I recommend the paper.

[1] https://dl.acm.org/doi/abs/10.1145/3315508.3329973

Re: Triton: Open-Source GPU Programming for Neural Networks

#6

Post by blueblisters » Wed, Jul 28, 2021, 5:05 PM UTC

peytoncasper wrote: post #3

riyadparvez wrote: post #2

I am confused. Is it another competitor of Tensorflow, JAX, and Pytorch? Or something else?

I believe this is more of an optimization layer to be utilized by libraries like Tensorflow and JAX. More of a simplification of the interaction with traditional CUDA instructions. I imagine these libraries and possibly some users would implement libraries on top of this language and reap some of the optimization benefit without having to maintain low-level CUDA specific code.

So is this similar to XLA?

Re: Triton: Open-Source GPU Programming for Neural Networks

#7

Post by yumraj » Wed, Jul 28, 2021, 5:05 PM UTC

croes wrote: post #4

Too bad it's CUDA Sooner or later this will become a problem because you are depending on the benevolence of a single manufacturer.

Sure, but if this abstraction layer becomes popular, then it becomes much easier to support other GPUs without requiring client libraries to change, which is a much harder problem.

Re: Triton: Open-Source GPU Programming for Neural Networks

#8

Post by shubuZ » Wed, Jul 28, 2021, 5:07 PM UTC

croes wrote: post #4

Too bad it's CUDA Sooner or later this will become a problem because you are depending on the benevolence of a single manufacturer.

Which other hardware vendor provides the level of performance that Nvidia's GPU provide? Wasnt the benevolence on single (or couple) manufacturer(s) true in 90s, 2020s?

Re: Triton: Open-Source GPU Programming for Neural Networks

#9

Post by shubuZ » Wed, Jul 28, 2021, 5:09 PM UTC

I have found writing CUDA code is much simpler than writing correct multi-threaded AVX2/AVX-512 code.

Re: Triton: Open-Source GPU Programming for Neural Networks

#10

Post by mvanaltvorst » Wed, Jul 28, 2021, 5:12 PM UTC

As a sidenote, these SVG graphs are absolutely beautiful. The flow diagram is even responsive on mobile!