Skip to content

Are there any matmul kernels designed for MxK lhs and NxK rhs?

I’m just getting started with kernel optimization, and I’ve got two questions I’d like to ask:

  1. May I ask whether KleidiAI currently supports matmul with MxK LHS and NxK RHS matrices?
  2. I noticed that most kernels require packing the RHS — are they specifically designed for linear operations? For general matrix multiplication like MxK LHS and KxN RHS, would packing the RHS every time be inefficient?
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information