[Mlir-commits] [mlir] [MLIR] Add apply_patterns.vector.arm_neon.contraction_to_i8mm TD Op (PR #140251)

Momchil Velikov llvmlistbot at llvm.org
Wed May 21 04:24:03 PDT 2025


momchil-velikov wrote:

> * `apply_patterns.{dialect}.{meta-name-for-patterns}`
> 
> This will help us avoid "polluting" the "Vector" namespace with things specific to hardware targets.

In that case the TD op has to be simply `apply_patterns.vector.lower_contraction` (which already exists).
But then how do we route to a specific target-dependent rewrite?

IMHO, having target dependent components in pattern names (like we have `GPU` or `NVVM`) is not an issue and
namespace pollution can be avoided by employing a hierarchical naming convention.


https://github.com/llvm/llvm-project/pull/140251


More information about the Mlir-commits mailing list