Hi,
I am interested in using -XX:+AlwaysPreTouch and I was checking the delay during booting. I am also using TransparentHugePages
THP config[cheva-virtualmachine ~]# cat /sys/kernel/mm/transparent_hugepage/shmem_enabledalways within_size [advise] never deny force[cheva-virtualmachine ~]# cat /sys/kernel/mm/transparent_hugepage/enabled [always] madvise never
I am running a VM with 8 cores and I observed that PreTouch is much faster with G1 that with Gen ZGC. Main reason is that I could check that G1 is using 8 concurrent threads for doing the job while Gen ZGC was using 2 threads (I used top -d 1 and observed the busy threads there)
I made this test with 32G heap, BUT the production enviornment is running with 300G so I expect the figures to be even more different.
# G1 - Using 8 cores (GC Thread #0 ..#8)$> time java -Xmx32G -Xms32G -XX:-UseTransparentHugePages -XX:+AlwaysPreTouch -versionopenjdk version "21.0.4" 2024-07-16 LTSOpenJDK Runtime Environment Zulu21.36+17-CA (build 21.0.4+7-LTS)OpenJDK 64-Bit Server VM Zulu21.36+17-CA (build 21.0.4+7-LTS, mixed mode, sharing)java -Xmx32G -Xms32G -XX:-UseTransparentHugePages -XX:+AlwaysPreTouch -versio 0,44s user 12,28s system 688% cpu 1,848 total
#Gen ZGC - Using 1 thread and at some point switch to 2 threads (ZGCWorker#0 and #1)$> time java -Xmx34G -Xms34G -XX:+UseZGC -XX:+ZGenerational -XX:-UseTransparentHugePages -XX:+AlwaysPreTouch -version ✔
openjdk version "21.0.4" 2024-07-16 LTSOpenJDK Runtime Environment Zulu21.36+17-CA (build 21.0.4+7-LTS)OpenJDK 64-Bit Server VM Zulu21.36+17-CA (build 21.0.4+7-LTS, mixed mode, sharing)java -Xmx34G -Xms34G -XX:+UseZGC -XX:+ZGenerational -XX:+AlwaysPreTouch 1,08s user 11,92s system 136% cpu 9,530 total
Non generational ZGC is even slower.
In this case, GenZGC is 5 times slower than G1 and it is NOT using all available cores to do the job.
Is this somehow expected behaviour? Maybe could be optimized or there is any reason to avoid using more threads?
Thanks in advance,
Evaristo