[WIP] Alternative approach to issue 47202

[BrianBohe]

This PR is intended to measure spmi results. It is another approach to solve issue #47202, in which we don't ask the vm for more memory than what is necessary, but we do scan all variable live ranges before copying them to the vm memory.

The thing is empty variable live ranges are useless. We are reporting them because they can get extended later if the variable becomes alive in the immediately next emitted instruction. If an empty live range is not getting extended, which we can realize after emitting all the code or creating a new live range for the same variable, we can remove it.

Tested running

<path_to_repo>\artifacts\tests\coreclr\windows.x64.Debug\Tests\Core_Root\corerun.exe <path_to_repo>\artifacts\bin\coreclr\windows.x64.Debug\R2RDump\R2RDump.dll --in <path_to_repo>\artifacts\bin\coreclr\windows.x64.Debug\System.Private.CoreLib.dll > output.txt

before and after this PR, checking the differences manually. Files are big and I wasn't able to get a nice diff that I could post here. I am open to any suggestions for more testing.

[ghost]

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

Issue Details

---

null

Author:	BrianBohe
Assignees:	BrianBohe
Labels:	area-CodeGen-coreclr
Milestone:	-

[jakobbotsch]

I'm curious if all this logic can be done inside genSetScopeInfoUsingVariableRanges instead. There is already some existing logic there that extends previous live ranges, can we add some logic that handles this case there too and avoid the complicated emitLocation reasoning?

[BrianBohe]

I'm curious if all this logic can be done inside genSetScopeInfoUsingVariableRanges instead. There is already some existing logic there that extends previous live ranges, can we add some logic that handles this case there too and avoid the complicated emitLocation reasoning?

My guess is this is a little bit faster, but we can measure that with spmi in a new pr after this. Now we are deciding if we want to be faster and a little bit leaky in memory usage or be precise in memory usage but slower. What do you think?

There are cases in which we cannot tell whether two emitLocation are close enough to extend before emitting code, and even after emitting code we can potentially remove instructions, so we cannot avoid checking for extension in genSetScopeInfoUsingVariableRanges. Having said that, computing emitLocation's offset in genSetScopeInfoUsingVariableRanges is expensive, we may need to walk every instruction in its instruction group.

[jakobbotsch]

My guess is this is a little bit faster, but we can measure that with spmi in a new pr after this. Now we are deciding if we want to be faster and a little bit leaky in memory usage or be precise in memory usage but slower. What do you think?

If we can centralize this kind of logic inside genSetScoepInfoUsingVariableRanges I would much prefer that. It is much easier to reason offsets about than comparing emitLocation in IGs given that IGs can be empty and that emitLocation can both point to one-after-the-end and the beginning of (any of the following) IGs.

There are cases in which we cannot tell whether two emitLocation are close enough to extend before emitting code, and even after emitting code we can potentially remove instructions, so we cannot avoid checking for extension in genSetScopeInfoUsingVariableRanges. Having said that, computing emitLocation's offset in genSetScopeInfoUsingVariableRanges is expensive, we may need to walk every instruction in its instruction group.

We are already reporting the offset in genSetScopeInfoUsingVariableRanges so the offset is already being computed and available there. Can we just add a simple case to the reportRange lambda there?

[BrianBohe]

Closing this PR as its alternative performs better.