[llvm] [AggressiveInstCombine] Implement store merge optimization (PR #147540)

[Nikita Popov via llvm-commits]

nikic wrote:

> I'm wondering if it would be worth having a general memory optimization pass to do such transforms, possibly using MSSA?

Possibly, though I don't think MSSA is particularly useful in this context. I think the only thing MSSA would really give us is a slightly more efficient way to iterate over the stores in a block, we wouldn't benefit from the actual analysis capabilities.

https://github.com/llvm/llvm-project/pull/147540