ADDITIONAL SYSTEM INFORMATION :
# java -version
openjdk version "17.0.6" 2023-01-17
OpenJDK Runtime Environment (Red_Hat-17.0.6.0.10-1.fc37) (build 17.0.6+10)
OpenJDK 64-Bit Server VM (Red_Hat-17.0.6.0.10-1.fc37) (build 17.0.6+10, mixed mode, sharing)

# uname -a
Linux fedora 6.1.9-200.fc37.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Feb 2 00:21:48 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

# cat /etc/fedora-release
Fedora release 37 (Thirty Seven)


A DESCRIPTION OF THE PROBLEM :
Adding more than two identical things inside a busy loop in very heavy multithreaded application causes performance and CPU usage to plummet. In my testing environment and application -DmsgCount=2 causes 1200% cpu load (6c12t machine) where -DmsgCount=3 howers around 20%-30%. Using -Xint increases performance significantly in in -DmsgCount=3 case but also increases CPU usage.

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
All of the source code and required steps are in https://github.com/StrongestNumber9/threadbug repository.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
CPU usage and performance to scale linearly with available CPU cores and processing threads.

ACTUAL -
CompilerThreads keeps on trying to optimize code, causing significantly reduced performance and low CPU usage.

---------- BEGIN SOURCE ----------
All of the source code and required steps are in https://github.com/StrongestNumber9/threadbug
---------- END SOURCE ----------

CUSTOMER SUBMITTED WORKAROUND :
Partial workaround is to use -Xint but it is still significant performance loss as other optimizations are lost.

FREQUENCY : always