On Wed, 3 Jun 2026 08:32:43 GMT, Stefan Karlsson <[email protected]> wrote:

> I  think that changing the naming to use the word resident makes sense. I 
> still think it is highly odd, and maybe bug-prone in the future, to use 
> first_resident_in_range to fill in the values for the NMT function 
> next_committed

Yes I agree, it is not ideal that NMT actually reports resident size rather 
than committed size for thread stacks. However, this is pre-existing as we're 
limited by Linux since it can only check resident status of an address (using 
mincore), not committed status. Windows can do both. The alternative (for 
linux) is to do something like parse /proc/PID/smaps to get committed status, 
but I think this will be slower and add also complexity. 

Additionally, NMT may be moving toward reporting RSS in the future: 
[JDK-8249666](https://bugs.openjdk.org/browse/JDK-8249666).

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

PR Comment: https://git.openjdk.org/jdk/pull/31124#issuecomment-4612976365

Reply via email to