[llvm] [RISCV] Add correct MachinePointerInfo when putting arguments on the stack. (PR #110140)

[Craig Topper via llvm-commits]

topperc wrote:

> What's the impact if we pass empty MachinePointerInfo? The result of AA might be wrong?

I believe the AA results will be conservative and just assume it can alias anything.

I was hoping adding the MachinePointerInfo would give more information to AA and allow better scheduling on a test I was looking at, but it didn't work. Posted anyway because I thought it was odd that we didn't have proper info here.

https://github.com/llvm/llvm-project/pull/110140