DYN1805 Validate the performance tracking system by using validated improvements Part II

[reddyashish]

Purpose

This PR is a copy of the one that Craig made: #9620

Instead of removing the TraceUtils class from from Protocore.Lang namespace, we just obsolete it for now.

Since this trace data is checked for in the Execute function, this part of the code is executed for all the graphs. A small percentage decrease or increase in the run times for those graphs, that take less absolute run times can be ignored.

Performance results comparison with the base line data(for model tests):

Declarations

Check these if you believe they are true

• The code base is in a better state after this PR
• Is documented according to the standards
• The level of testing this PR includes is appropriate
• User facing strings, if any, are extracted into *.resx files
• All tests pass using the self-service CI.
• Snapshot of UI changes, if any.
• Changes to the API follow Semantic Versioning, and are documented in the API Changes document.

Reviewers

@mjkkirschner @aparajit-pratap @QilongTang @saintentropy

[QilongTang]

Since this trace data is checked for in the Execute function, this part of the code is executed for all the graphs.

Can you explain a little bit more or point to the actual caller to help understand?

[QilongTang]

LGTM @reddyashish Is there anything we are still waiting in this PR?

[reddyashish]

No Aaron, this is good to go. Merging it.

[junmendoza]

Very impressive performance uplift! @aparajit-pratap I'm no longer up to date with the VM but is this improvement attributed to fixing a fundamental issue in the callsites or an iterative process?

[aparajit-pratap]

@junmendoza glad to know you are still tracking progress of Dynamo :)
This perf improvement was a result of work by @saintentropy : #9620 to replace the backing implementation of TLS from a thread data slot to a thread relative static field per Microsoft recommendations: https://docs.microsoft.com/en-us/dotnet/standard/threading/thread-local-storage-thread-relative-static-fields-and-data-slots
Since the TLS is checked for each VM interaction in the callsites it basically affects performance globally, not only for trace/element binding related calls.

[junmendoza]

@aparajit-pratap, @saintentropy Interesting! and thanks for the background. Btw did this arise due to performance profiling? Curious how much effort is still put in making the VM faster at this stage :)

[aparajit-pratap]

@saintentropy has been slowly chipping away at the VM trying to make performance improvements and he has made some along the way. We have more ongoing work in this area. One of the things I was looking at is at optimizing execution for modified inputs in graphs: #9485. This was based on one of your design proposals some time ago.
This work is yet to be fully completed and merged.