tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: RFC: NUMA support



Hi,

On Mon, Nov 10, 2008 at 05:11:37PM +0200, Christoph Egger wrote:

> I started to work on NUMA support. First step is to set up the
> topology.

Very cool!

> NUMA: SRAT table found
> NUMA: SLIT table not found
> numa0: memory: 0x0 - 0xa0000 (0xa0000, physical, raw, raw)
> numa0: memory: 0x100000 - 0x40000000 (0x3ff00000, physical, raw, raw)

Could these be emitted via aprint_debug()?

> http://www.netbsd.org/~cegger/numa.diff

- Is it possible to eliminate MAX_NUMA_NODES and make it dynamic?

- I wonder how much more can be made MI, for example some of the lookup
  functions. We could assume both CPUs and numa nodes nodes have global
  IDs.

- It may or may not be be worthwhile to use cpu_lock instead of a new lock,
  in order to save future pain. I'm not sure.

Thanks,
Andrew


Home | Main Index | Thread Index | Old Index