cmuparlay/parlaylib

hugepages

Opened this issue · 0 comments

Some of the benchmarks when run on linux get significant performance improvement by setting /sys/kernel/mm/transparent_hugepage/enabled to "always" instead of the default "madvise". In particular bench_histogram improves by almost a factor of two. Several of the other group_by and reduce_by operations are also affected.

Should add to documentation.