I am still looking for benchmarks comparing ZGC and Shenandoah 2.0 with more than one metric. Pause times are mostly good for both, throughput also seems to be in-line more or less in case you have spare cores that would otherwise be idle (-10% compared to parellel). However what I haven't found at all is a comparison of e.g. total instructions executed / cycles consumed / total energy used for a particular load type - all the benchmarks I saw used spare cores available which might not be the case in the cloud / on high-load servers in clustered environments / in VM environments.
Oh man, this is pure gold. Clears so many areas not inferreable easily by GC docs
It's the clearest presentation about the GC types.
Excellent job, but I'd wish there were more materials on this topic, especially for the general public (a.k.a. "rookies in GC").
Great speech! Very interesting subject
I am still looking for benchmarks comparing ZGC and Shenandoah 2.0 with more than one metric.
Pause times are mostly good for both, throughput also seems to be in-line more or less in case you have spare cores that would otherwise be idle (-10% compared to parellel).
However what I haven't found at all is a comparison of e.g. total instructions executed / cycles consumed / total energy used for a particular load type - all the benchmarks I saw used spare cores available which might not be the case in the cloud / on high-load servers in clustered environments / in VM environments.
Did you manage to find it?
great to learn about gore internals
Excellent speech！
Great talk.
Particular interesting how the 64 bits Platform only use 44 bits, I guess no body has more than 16 tera on a single box.
Hall David Young Daniel Moore Melissa