[llvm] [SLP]Transform stores + reverse to strided stores with stride -1, if profitable. (PR #90464)

[Roger Ferrer Ibáñez via llvm-commits]

rofirrim wrote:

Hi @alexey-bataev.

I'm confused by the code generation of this GCC torture after this change:

https://www.godbolt.org/z/frnz9K1n4

Check line 5. I would expect the pointer of a negative strided store be moved at the address of the last element of the store. Instead we store at the original place (i.e. I'd expect something like `addi a2, a0, 4` and then `vsse32.v  v8, (a2), a1`)

https://github.com/llvm/llvm-project/pull/90464