[llvm-branch-commits] [llvm] [Passes] Disable code sinking in InstCombine early on. (PR #72567)

[Nikita Popov via llvm-branch-commits]

nikic wrote:

This looks plausible to me, but let me play devil's advocate here: The flip side of "code sinking prevents if conversion" is "if conversion prevents code sinking". Looking at the first test case `test_find_min`, we are now unconditionally executing a fairly large number of instructions that were previously behind a branch. I'm assuming that this happens to be great for your motivating case because this happens to be hot branch -- but couldn't this just as easily be a regression if this were the cold branch instead? Is the idea here that if the original code executed these unconditionally, then it's more likely than not that unconditionally executing them is beneficial?

https://github.com/llvm/llvm-project/pull/72567