The One Billion Row Challenge in CUDA: from 17 minutes to 17 seconds

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by jbochi on Sunday April 14, 2024 @11:20AM (#40031743)

Pretty cool. Shameless plug: My team at Anthropic is hiring people that can write accelerator kernels. Please reach out to <my username>@gmail.com if you want to make state of the art models faster :)

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by petermcneeley on Sunday April 14, 2024 @11:31AM (#40031834)

qq: Would this be in CUDA?

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by jbochi on Sunday April 14, 2024 @12:27PM (#40032266)

Not sure what I can share publicly, but we use TPUs as well: https://cloud.google.com/blog/products/compute/announcing-cl...

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by fpgamlirfanboy on Sunday April 14, 2024 @01:26PM (#40032693)

It's funny - former Google devs (whom I just maintain a good relationship with their former employer) are ideally positioned to profitably take advantage of the arb of TPU over GPU.

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by bee_rider on Sunday April 14, 2024 @01:31PM (#40032741)

Google should improve their abstractions until there isn’t room for that anymore, haha.

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by trashtensor on Sunday April 14, 2024 @02:08PM (#40033060)

not looking for a job at this time but i do this kind of work - what is the name of the team that you are working on / typically works on this?

[ Parent | on HN ]

Re:The One Billion Row Challenge in CUDA: from 17 minutes to... by winwang on Monday April 15, 2024 @12:14PM (#40042475)

Curious, do you guys use Triton? I was surprised to find out PyTorch slowdown is palpable even compared to basic CUDA kernels.

[ Parent | on HN ]