1 |
Just to point out, amd was calling the opterons and such more of a |
2 |
SUMO configuration (Sufficiently Uniform Memory Organization, not |
3 |
joking here), instead of NUMA. Whereas technically, it clearly is a |
4 |
NUMA system, the differences in latency when accessing memory from a |
5 |
bank attached to another processors memory controller is very small. |
6 |
Small enough to be largely ignored, and treated like uniform memory |
7 |
access latencies in a SMP system. Sorta in between SMP unified style |
8 |
memory access and NUMA. This holds for up to 3 hypertranport link |
9 |
hops, or up to 8 chips/sockets. You add hypertransport switches to |
10 |
scale over 8 chips/sockets, it'll most likely be a different story... |
11 |
|
12 |
What I've always wondered is, the NUMA code in the linux kernel, is |
13 |
this for handling traditional NUMA, like in a large computer system |
14 |
(big iron) where NUMA memory access latencies will vary greatly, or is |
15 |
it simply for optimizing the memory usage across the memory banks. |
16 |
Keeping data in the memory of the processor using it, etc, etc. Of |
17 |
course none of this matters for single chip/socket amd systems, as |
18 |
dual cores as well as single cores share a memory controller. Hmm, |
19 |
maybe I should drink some coffee and shutup until I'm awake... |
20 |
|
21 |
On 7/27/05, Duncan <1i5t5.duncan@×××.net> wrote: |
22 |
> Dulmandakh Sukhbaatar posted <20050727075012.79549.qmail@××××××××.mn>, |
23 |
> excerpted below, on Wed, 27 Jul 2005 15:50:12 +0800: |
24 |
> |
25 |
> > Thanks. How can I enable hypertransport in kernel or somewhere? Anyone |
26 |
> > knows about NUMA? I read about it, and it seems technology for |
27 |
> > multiprocessor systems. Thus I have single CPU, I don't need it. Right? |
28 |
> |
29 |
> NUMA is indeed for multi-processor systems. NUMA is Non-Uniform Memory |
30 |
> Architecture. With AMD CPUs that have the memory controller on the same |
31 |
> chip as the CPU, that means that each CPU can control it's own memory. If |
32 |
> you run NUMA mode in this case (and if your BIOS is set up accordingly), |
33 |
> the kernel will try to keep the memory for each task in the memory handled |
34 |
> by, local to, that CPU. If either the kernel or BIOS is set to unified |
35 |
> memory, or if you only have memory sticks in the slots for one of the |
36 |
> CPUs, then you won't get NUMA mode and the kernel won't care what memory |
37 |
> addresses the memory for each process lives at. |
38 |
> |
39 |
> AFAIK, hypertransport is automatically handled by your choice of chipset. |
40 |
> If the chipset you configure has it, it will be enabled, if not, it won't. |
41 |
> I was therefore a bit puzzled when you mentioned hypertransport |
42 |
> specifically in the previous post, since I don't believe there's a |
43 |
> specific kernel option for it. (It's possible, however, that there is and |
44 |
> I've just forgotten about it, since it's been awhile since I reviewed |
45 |
> the settings for the entire kernel -- I just run make oldconfig and deal |
46 |
> with any new options in each newer kernel, and additionally do any |
47 |
> specific tweaking I might want to try.) |
48 |
> |
49 |
> -- |
50 |
> Duncan - List replies preferred. No HTML msgs. |
51 |
> "Every nonfree program has a lord, a master -- |
52 |
> and if you use the program, he is your master." Richard Stallman in |
53 |
> http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
54 |
> |
55 |
> |
56 |
> -- |
57 |
> gentoo-amd64@g.o mailing list |
58 |
> |
59 |
> |
60 |
|
61 |
-- |
62 |
gentoo-amd64@g.o mailing list |