Messages in this thread | | | Date | Tue, 14 Dec 1999 02:10:54 +0100 | From | Andi Kleen <> | Subject | Re: Per-Processor Data Paget |
| |
On Mon, Dec 13, 1999 at 04:47:36PM -0800, Scott Lurndal wrote: > > slurn@griffin.engr.sgi.com (Scott Lurndal) writes: > > > > > > > > And what percentage of the time does a thread in the same address space > > > get scheduled on a single processor. > > > > Also lots of apps simply rely on kernel thread for IO context. Aio is > > implemented in glibc using cloned threads. Other programs use them e.g. > > There are additional forms of aio (much lower overhead) that don't rely > on threads.
... which assumes that all programs are rewriten to use them first (a lot of programs on Linux don't use the aio_* API, but use their threads to eat mmap latencies etc.)
> > > to tolerate paging latencies on mmap'ed files or for sendfile. For that it > > makes sense to have a lot of threads per process on a single CPU, even on > > a big multi CPU machine. > > Of course, but on a big multiCPU machine, it is not probable that two > consecutively dispatched threads will be from the same address space.
WIth the current Linux scheduler it definitely is.
Do you have numbers on Irix that show otherwise?
-Andi
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |