RFR(s): 8234086: VM operation can be simplified

[Man Cao]

Sorry for being late to this thread.
I also found it quite unsettling to change several async VM ops to
synchronous.
Has anyone done any rigorous performance benchmarking to ensure this
change does not cause noticeable slowdowns in the JVM?

Currently I'm working on a synchronization protocol for G1's refinement
threads
for JDK-8226731, where I added an async VM op. Surely I can try finding a
workaround
to make it synchronous, but I'm not quite convinced on the cost-benefit
for removing async VM op.

>>> Perhaps the ability to execute an async-safepoint VM operation needs
>>> to remain, for simplicity (compared to working around the issues).
>>
>> I'm hoping not :(
> To be quite upfront I don't think support for async VM ops really adds
> any complexity, whereas ensuring the existing async ops can safely
> become sync ops has been a bit of effort :) But it is a one-time effort
> so ...

I totally agree with David. In addition I also found the behavior change
to Thread::send_async_exception() concerning, although I'm no expert
on this area.

-Man