[llvm] Decompose gep of complex type struct to its element type (PR #107848)

[Nikita Popov via llvm-commits]

https://github.com/nikic requested changes to this pull request.

This patch doesn't make sense to me. It *partially* decomposes the GEP, so that we both have an explicit multiplication in the offset, and a GEP scale multiply. We should have only one or the other. 

On top of that, this is driven by a GEP source element type based heuristic, which is not allowed.

https://github.com/llvm/llvm-project/pull/107848