On Fri, 27 Oct 2023 14:43:56 GMT, Ben Perez <d...@openjdk.org> wrote:

> Modified `getService` method to prevent caching of `ServiceKey`, which was 
> negatively impacting multithreaded performance

I have a few questions:
- The change listed in the bug is not the same as this PR.  The bug removes the 
section and leaves only the new ServiceKey().  The condition statement left 
will always be false.
- This seems like a large performance difference for just removing a condition 
for the existing key.  Is the problem that is volatile?  Might removing 
volatile help?
- This change is done for vCPU machines.  What is the impact on non vCPU?  The 
reason for the caches was most likely to help single threaded performance.  How 
have those been situations been effected?

-------------

PR Comment: https://git.openjdk.org/jdk/pull/16403#issuecomment-1783239321

Reply via email to