[Mlir-commits] [mlir] [MLIR][Linalg] Left over Linalg named ops from previous PR (PR #90405)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Wed May 1 17:03:59 PDT 2024


MaheshRavishankar wrote:

> Adding `erf` as unary and `powf` as binary.
> 
> Same as `max(arg, 0.0)` for `ReLU`, `powf(arg, const)` can be either a generic (with broadcast) or a pair (`linalg.broadcast + linalg.powf`) and then lowered "correctly". Either way, the lower dialects need to know what kind of broadcast anyway, so no materialization of the constant tensors should remain.
> 
> I want to flush the easy ones before we start working on type cast & softmax.

I think I am missing some context here. In a previous PR I saw this was based on some round-table discussion. For people who werent there in person can you please post some notes and provide some more context. Specifically, I want to get some more information about what you are refering to w.r.t `softmax` semantics... 

https://github.com/llvm/llvm-project/pull/90405


More information about the Mlir-commits mailing list