I don’t understand the use of “-disable-loop-unrolling” option in the optimizer. I assumed it would disable unrolling but while compiling the test in “llvm/test/Transforms/LoopUnroll/disable-loop-unrolling_forced.ll” with and without this option I see no difference in output IR as well as the order in which passes are run.
Ohh! I missed that this option is overridden when unroll metadata is specified.