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

Re: [Xen-devel] [PATCH v3 4/7] libxl: vNUMA supporting interface



On mar, 2013-11-26 at 17:14 -0500, Elena Ufimtseva wrote:
> On Thu, Nov 21, 2013 at 4:59 AM, Li Yechen <lccycc123@xxxxxxxxx> wrote:
> > Hi Elena and Dario,
> >> +    /* Determine the best nodes to fit vNUMA nodes */
> >> +    /* TODO: change algorithm. The current just fits the nodes
> >> +     * Will be nice to have them also sorted by size
> >> +     * If no p-node found, will be set to NUMA_NO_NODE
> >> +     */
> >> +    claim = calloc(info->nr_vnodes, sizeof(*claim));
> >> +    if (claim == NULL)
> >> +        return rc;
> >> +
> >> +    libxl_for_each_set_bit(n, info->nodemap)
> >> +    {
> >> +        for (i = 0; i < info->nr_vnodes; i++)
> >> +        {
> >> +            if (((claim[n] + (mems[i] << 20)) <= ninfo[n].free) &&
> >> +                 /*vnode was not set yet */
> >> +                 (info->vnode_to_pnode[i] == VNUMA_NO_NODE ) )
> >> +            {
> >> +                info->vnode_to_pnode[i] = n;
> >> +                claim[n] += (mems[i] << 20);
> >> +            }
> >> +        }
> >> +    }
> >> +
> >
> > So in the initial part, each vnode is set to EXACTLY ONE pnode, right?
> 
> Hi Yechen
> 
> Yes, one node or VNUMA_NO_NODE (any node);
> 
Yes, but, at the same time, this particular chunk of code is most likely
going to change, in order to better integrate with automatic NUMA
placement.

In fact, if I'm reading this right (as I said when reviewing this
patch), this kind of try to compact more vnodes in the same pnode, if
they fit.

Although I see the reason why you ma want to do that, I'm still of the
opinion that, if info->nodemap has some info coming from the placement
algorithm, we should just honor them... That thing (the placement
algorithm) is too complicated to just trow away the results! :-P

I'm sorry I haven't got to write some code about this, I hope to find
some time to do that soon enough.

Thanks and Regards,
Dario

-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.