[Mlir-commits] [mlir] [MLIR][XeGPU] Allow some nd ops to have argument shapes mismatch for … (PR #120566)

[Petr Kurapov]

kurapov-peter wrote:

> It looks to me the current implementation is following option 2, am I right?

Yes, these validation changes unblock option 2

>  Can we drop the sgMap in future and derive it from kernel attributes? We may use the presence of VectorComputeINTEL attribute to indicate the kernel is for SIMD, which will trigger the SIMD logic, otherwise it is a SIMT kernel and trigger the SIMT logic. Is it safe?

Deriving from an attribute likely means the map would be essentially the same for all operations. I suspect this might be too restrictive, in presence of packing for example. Other than that, I don't see any technical problems with getting this information not from the type attribute but from somewhere else like a kernel attribute (which would probably be the same sg map but attached to a kernel?).

https://github.com/llvm/llvm-project/pull/120566