[flang] [llvm] [mlir] [llvm][mlir][flang][OpenMP] Emit __atomic_load and __atomic_compare_exchange libcalls for complex types in atomic update (PR #92364)

[Tom Eccles via llvm-commits]

tblah wrote:

I think the problem might have been that `flang-new -fc1 -emit-llvm` will run an optimization pipeline on the LLVM IR before printing. Even without specifying `-O` I don't think the result will always be the same as taking the output of `mlir-translate`.

https://github.com/llvm/llvm-project/pull/92364