[flang-commits] [flang] [flang] Add nsw to DO loop parameters (PR #113854)

[Tom Eccles via flang-commits]

https://github.com/tblah commented:

The changes look okay to me. Perhaps in another patch you could investigate if there is any benefit in doing the same for fir.do loops generated through other paths through lowering (e.g. bufferization of a hlfir.elemental). 

Please could you update the description describing what performance regression testing have you done on this patch and what speedups you expect (if there are any beyond the one testcase in the RFC).

https://github.com/llvm/llvm-project/pull/113854