[llvm] [LVI] Skip self loops in `solveBlockValueNonLocal` (PR #127763)

[Nikita Popov via llvm-commits]

nikic wrote:

> Compile-time impact is high: https://llvm-compile-time-tracker.com/compare.php?from=84ddda58c870681dd12ed765e9d59d5e00567f94&to=af032f1351358f2f5b5d9f4e87c5601c23b9bd37&stat=instructions:u

I think this may look worse than it actually is, due to second order impact. E.g. the bigger regressions are on mafft, which also has 0.1-0.26% code size increase. There is no significant impact on the clang build.

It would be good to confirm comptime with llvm-opt-benchmark as well, but I don't think this is a blocker.

https://github.com/llvm/llvm-project/pull/127763