r0ller <r0ller%freemail.hu@localhost> writes: > Yepp, that I know, that's why I asked it:) However, as the > question is rather about the overhead of running code in one or the > other way I asked jym at NetBSD who told that -besides the fact that > only measurement can tell- based on his gut feeling, usermode should > have less overhead if we don't take into account I/O. I suspect that for CPU-bound jobs that they are close enough not to matter. But indeed, measuring beats guessing and if you are inclined to test both and post (on the same hardware) that would be great.
Attachment:
signature.asc
Description: PGP signature