[clang] [llvm] [WPD] Enable speculative devirtualization with LTO (PR #233380)

[Hassnaa Hamdi via cfe-commits]

hassnaaHamdi wrote:

> > Hi @KRM7 Did you test the performance for thinLTO ? if the feature is useful for it, could you patch it out separately to simplify this patch ?
> 
> I didn't run benchmarks for thin-LTO separately, only for regular LTO. Is there a reason we would expect some meaningful difference and would want to split this up into full/thin-LTO patches?

I mean if the feature is not useful for thin-LTO, then no reason to support it, right?

https://github.com/llvm/llvm-project/pull/233380