[llvm] [DAGCombiner] ignore loads when finding store in store forwarding optimization (PR #87465)

[Chen Zheng via llvm-commits]

chenzheng1030 wrote:

I am going to close this patch as it does not show compile time improvement in the tested benchmarks although in theory it should speed up LLVM by saving the time to optimize the chain of load instructions.

https://github.com/llvm/llvm-project/pull/87465