You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I was wondering if there were any plans to add padding support to global loads and store. This would bring the functionality of Thunderkittens a lot closer to Triton but with the flexibility of using Cuda. Without this, it's rather hard to justify using Thunderkittens as it doesn't make any sense to pad the sequence in HBM before passing it to a thunder kittens kernel.
The text was updated successfully, but these errors were encountered:
Hi, I was wondering if there were any plans to add padding support to global loads and store. This would bring the functionality of Thunderkittens a lot closer to Triton but with the flexibility of using Cuda. Without this, it's rather hard to justify using Thunderkittens as it doesn't make any sense to pad the sequence in HBM before passing it to a thunder kittens kernel.
The text was updated successfully, but these errors were encountered: