[Mlir-commits] [mlir] [polynomial] distribute add/sub through ntt to reduce ntts (PR #93132)

[Jeremy Kun]

j2kun wrote:

> That looks good from an MLIR point of view, I don't understand NTT enough to know if the patterns are correct though. So LG if you're owning the maths correctness :)

Speaking of which, should I solicit someone who can review the math aspects of these PRs? For now they are mainly coming from out-of-tree ports, but some of the larger parts (e.g., lowering `ntt` to an optimized CPU implementation) are more sophisticated.

If you know of folks who already have commit access who would be interested in this, please let me know, otherwise I can ask some of my colleagues to request membership.

https://github.com/llvm/llvm-project/pull/93132