Mark cloned loops as cold

[EgorBo]

diffs

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Copilot reviewed 1 out of 1 changed files in this pull request and generated no comments.

[AndyAyersMS]

Would be good for you to describe why you think this is the right thing to do.

If the cloning conditions include GDV then I think the slow loop likelihood needs to reflect the GDV likelihoods. If there's no GDV then I'm less sure.

[EgorBo]

@AndyAyersMS Mainly, for three reasons:

1. We do other clonings (e.g. finally or range-check-clone) with 0 weight for the fallback
2. SPMI reports huge PerfScore winnings (mainly, because the slow path now gets 0 perfscore mostly)
   
3. My upcoming "loop cloning for Spans" shows mixed PerfScore results due to non-zero weight of the fallback, I want to know the perfscore of the fast path

But looks you're right, we need to use GDV's fallback weight (if GDV is involved), today we don't do it either. And GDV is not involved if its fallback's probability is >5%

[EgorBo]

Anyway, I was mostly curious about the diffs for now. I'll take a look at changing the weight to match GDV likelihoods