Intel is introducing big.little (or hybrid) core distribution in their upcoming desktop series of processors (Alder Lake), having "P"erformance and "E"fficiency cores.

However current thread sizing in the Java VM assumes that all cpu cores are of the same kind (performance), e.g. with it's

  ParallelGCThreads = 8 + (#cores - 8) * 5/8 (or so)

default thread sizing policy formula for #cores >= 8.

Another complication is that in Intel's hybrid design, the performance cores have hyperthreading, while the efficiency cores do not, skewing the calculation even more.

It is also likely that at least the low-end server (Xeon-E) line will also get this addition in the future, and other vendors may also follow.

Investigate the impact of this processor design on performance, maybe there is need to improve the various thread sizing calculations.

This is an umbrella CR just stating that the VM will not use the available resources optimally with some performance impact.