[Mlir-commits] [mlir] [mlir][gpu] Skip address space checks for memrefs between launchFuncOp and kernel func (PR #102925)

[Petr Kurapov]

kurapov-peter wrote:

> > The problem is that it is unclear what should happen to a kernel that has two pointers as arguments: one with an attribute and one without.
> > 
> You know that you can attach attributes to individual arguments?

Sure, I mean it is unclear whether the user's intention actually was to have one of the arguments without the attribute for whatever reason. I suspect this shouldn't be a problem in practice though, since you need `addrpspace(1)` in any case.

> One option is adding: #gpu.address_space<unified>.

That's interesting. Probably not in gpu too? Like, for all the devices on a platform?

https://github.com/llvm/llvm-project/pull/102925