Re:AI Is Writing Its Own Kernels, and They Are 17x Faster
by pakt1 on Thursday November 20, 2025 @05:42PM
(#45998889)
Trainium has always been a black box to me compared to GPUs. Seeing an automated tool reverse-engineer the best way to use the VectorEngine vs the TensorEngine is fascinating. It reveals just how much performance is left on the table by standard compilers.
[ Parent
| on HN ]
There are no comments on this comment yet.