[release/7.0-rc2] [Mono] Restore old code to solve the recent SpanHelpers regressions

[github-actions]

Backport of #75917 to release/7.0-rc2

/cc @radical@adamsitnik

Customer Impact

Fixes large performance regression #75709 in library code when running on mono (AOT, interpreter) introduced in rc1 (#74086)

Testing

Automated benchmark is running (will add results)

Risk

Medium/Low - The patch restores the previous code path for mono

[ghost]

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

Issue Details

---

Backport of #75917 to release/7.0-rc2

/cc @radical@adamsitnik

Customer Impact

Testing

Risk

IMPORTANT: Is this backport for a servicing release? If so and this change touches code that ships in a NuGet package, please make certain that you have added any necessary package authoring and gotten it explicitly reviewed.

Author:	github-actions[bot]
Assignees:	-
Labels:	area-System.Memory
Milestone:	-

[carlossanlop]

@adamsitnik@radical please fill out the template, add the servicing-consider label, and email Tactics today, so in case it gets approved, I merge it tonight.

cc @mmitche, we have one more.

[lewing]

@adamsitnik@radical please fill out the template, add the servicing-consider label, and email Tactics today, so in case it gets approved, I merge it tonight.

cc @mmitche, we have one more.

we're going to let the perf pipeline run finish first but Steve is aware of this one

[radical]

The helix jobs for wasm/perf finally started running after being queued up for 2hrs. Build - 20220920.13 on dotnet-runtime-perf. And this will now take another ~1.5hours to complete :/

[radical]

In the earlier runs 62db96b was missing because it was merged after the last changes here.

New runs:

• For this PR
• For baseline - release/7.0-rc262db96b

[mmitche]

@carlossanlop How close on this are we? Can we get this in ASAP?

[adamsitnik]

How close on this are we? Can we get this in ASAP?

I am waiting for local benchmarks run to compare the results and verify that the PR solves the regressions. According to the tool it needs two more hours.

[carlossanlop]

The single failure is networking, unrelated (infra, timeout): #58628

[mmitche]

How close on this are we? Can we get this in ASAP?

I am waiting for local benchmarks run to compare the results and verify that the PR solves the regressions. According to the tool it needs two more hours.

@adamsitnik How are we looking here?

[adamsitnik]

How are we looking here?

The results I got don't prove that all the regressions are gone: #75917 (comment)

@radekdoulik has shared offline results that show that the fix actually helps. He has run the benchmarks in different way that I did, using perf pipeline developed by @radical.

I leave the decision up to @radekdoulik and @radical who own WASM and have more experience in benchmarking it.

[jeffhandley]

I was worried I'd get the backport merge update wrong and I did. I'm going to close this PR and do a fresh backport from #75917 now that it's merged.