[mlir] [MLIR][Linalg] Introduce broadcast/transpose semantic to 'linalg.batc… (PR #122275)

Andrzej Warzyński llvmlistbot at llvm.org
Mon Jan 13 04:08:50 PST 2025


banach-space wrote:

> > Please, could you provide references to relevant RFCs where this was discussed/agreed-on? And include in the summary. Thanks :)
> 
> Added :)

Thanks! @MaheshRavishankar , relevant links have been added in the summary at the top:

> RFCs and related PR:
> https://discourse.llvm.org/t/rfc-linalg-opdsl-constant-list-attribute-definition/80149
> https://discourse.llvm.org/t/rfc-op-explosion-in-linalg/82863
> https://discourse.llvm.org/t/rfc-mlir-linalg-operation-tree/83586
> #115319

Looks like the syntax introduced here matches #115319 (unless I missed sth), so this is consistent with the previous patches in the series 👍🏻 

One high level (kind) request:

> Goals:
> 
> To add syntax and semantic to 'batch_matmul' without changing any of the existing syntax expectations for current usage. batch_matmul is still just batch_matmul.
> 
> Move the definition of batch_matmul from linalg OpDsl to tablegen ODS infra.

Why not split this into two patches? Those two goals seem orthogonal to me and reviewing small patches is easier (at least on GitHub).

https://github.com/llvm/llvm-project/pull/122275


More information about the Mlir-commits mailing list