Add more prime sizes to HashMap

[jkotas]

We have seen cases where the current sizes are not sufficient and end up with pathological behavior.

[Copilot]

Pull Request Overview

This PR extends the list of prime numbers used by the hash table to address potential performance issues when the current sizes are insufficient.

• Added additional larger primes to support cases where the hash table may require more buckets.
• Enhanced the available prime range for better distribution in high-load scenarios.

Comments suppressed due to low confidence (1)

src/coreclr/vm/hash.cpp:33

• Consider adding unit tests to verify that the new extended prime list maintains the expected hash table performance under high load scenarios.

4999559, 5999471, 7199369, 8639249, 10367101, 12440521, 14928637, 17914367,

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[mangod9]

is there a specific benchmark this will improve?

[jkotas]

is there a specific benchmark this will improve?

I have cced you on the internal email thread. The 7M limit was hit on .NET 8 by a workload that kept creating marshalled delegates on 50 worker threads and the finalizer thread fell behind cleaning them up. Once the 7M limit was hit, it went downhill even more rapidly due this hashtable behaving pathologically.

Marshalled delegates do not use this hashtable in .NET 10 anymore. This change is to avoid pathological behaviors in case there are other scenario that can store more than 7M elements in the runtime hashtables.

[jkotas]

/ba-g test failure without a log