[Mlir-commits] [mlir] [linalg] Add quantized version of `conv_3d_ncdhw_fcdhw` (PR #113953)

[Renato Golin]

rengolin wrote:

> It's just that the `linalg` convolution Ops are probably among the Ops with the most downstream uses, so such an op would likely need much more careful and elaborate design than I could provide.

This doesn't scale. If no downstream user is "able to provide" a reasonable implementation upstream and just start dumping all variations of the same op to suit their needs, the dialect will be bloated and unusable.

https://github.com/llvm/llvm-project/pull/113953