get rid of saved_allocated

[Maoni0]

I thought incorrectly when I made the change to use saved_allocated to calculate the frag ratio - this frag ratio actually should use the allocated that's adjusted by the plan phase, not the original allocated for the heap_segment. this can affect microbenchmarks dramatically when the last live object is way smaller than allocated, instead of doing a sweeping GC (which is the right choice) we ended up doing a compacting GC because we'd be calculating a very large frag ratio when we call decide_on_compacting.

[ghost]

Tagging subscribers to this area: @dotnet/gc
See info in area-owners.md if you want to be subscribed.

Issue Details

---

I thought incorrectly when I made the change to use saved_allocated to calculate the frag ratio - this frag ratio actually should use the allocated that's adjusted by the plan phase, not the original allocated for the heap_segment. this can affect microbenchmarks dramatically when the last live object is way smaller than allocated, instead of doing a sweeping GC (which is the right choice) we ended up doing a compacting GC because we'd be calculating a very large frag ratio when we call decide_on_compacting.

Author:	Maoni0
Assignees:	-
Labels:	area-GC-coreclr
Milestone:	-

[mangod9]

Would this help reduce impact on String.Replace microbenchmark which was observed with regions?

[Maoni0]

@mangod9 yes, I ran the simple string test and verified that it fixes the remaining regression. we'll be running more microbenchmarks.

[mangod9]

@Maoni0, is this PR still relevant?

[Maoni0]

it is. I just didn't have time to take care of it. will do so soon.

[stephentoub]

@Maoni0, looking at old PRs... should this be closed?

[Maoni0]

I'm closing this for now. I've added this to my list of items to keep track of.