[llvm] [DAGCombiner] Allow freeze(fmul) + fadd/fsub → FMA when using contract (PR #142250)

[Harrison Hao via llvm-commits]

harrisonGPU wrote:

Hi, I want to implement support for folding freeze(fmul) + fadd/fsub into FMA when using the contract flag. This is important for graphics GPU compilers because we set nnan by default. Could you please give me some suggestions? I have already updated the PR and lit tests. @jayfoad @nikic @RKSimon @arsenm  :-)

https://github.com/llvm/llvm-project/pull/142250