[Mlir-commits] [mlir] [mlir][linalg] raise generic to named ops. (PR #110421)

[Javed Absar]

javedabsar1 wrote:

> > > "is broadcast" functionality with y
> > 
> > 
> > Hi @rengolin and shahidact. My generic to naming is based on broadcast . Not a coupling of matmul and transpose/broadcast.
> > @rengolin please review my patch on merits of current linalg. `generic to named` has many named ops translation still pending. It is a good feature to have these pass through. If linalg.matmul expands to perm attribute as you propose, the current generic-to-named which translates contractions to matmul variations e.g. linalg.matmul_transpose_a ... I will extend to cover them as well.
> 
> Hey, sorry, I did not mean to review this PR with my tag, just to alert @shahidact to your PR so he can check his own PR against it.

Thanks Renato for reviewing this. yes i was bit confused about your earlier statement. apologies.

https://github.com/llvm/llvm-project/pull/110421