It's a 16-core processor. I've attached the diagnostics.
If I run
GLIBC_TUNABLES=glibc.cpu.x86_rep_movsb_threshold=2114 ./memcpy_loop -b 2113 -p 1000000 -t mmap -S 0 -D 1 0
then I hit the high performance again (>100GB/s).
It's a 16-core processor. I've attached the diagnostics.
If I run
GLIBC_TUNABLES= glibc.cpu. x86_rep_ movsb_threshold =2114 ./memcpy_loop -b 2113 -p 1000000 -t mmap -S 0 -D 1 0
then I hit the high performance again (>100GB/s).