We added a new post-isel CTRLoop pass in D122125. That pass will expand the hardware loop related intrinsic to CTR loop or normal loop based on the loop context.
So we don't need to conservatively check the CTR clobber now on the IR level.
As expected, now, more CTR loops are generated.
Could you please re-generate this test case with complete assembly output, so that we can see why CTR loop is viable on POWER9 now?