Re: [Xen-devel] [PATCH RFC v1 4/4] libxc for rt scheduler

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 11/07/14 17:35, Dario Faggioli wrote:
>
>> My concern is: Will the former one has worse performance than the
>> later one? â
>>
> Well, Xen has its tricks, but yes, I think performing e.g., 8 or 16
> hypercalls in a loop is worse than issueing one returning an array.

As far as hypercalls go, the path to Xen is a long one. From userspace,
a hypercall involves a context switch into the kernel (which is a double
context switch through Xen) and a context switch from the kernel into
Xen, and back again. The return from guest kernel to guest userspace
requires a TLB flush.

So, 6 context switches and a TLB flushes per hypercall (assuming a 64bit
dom0).

Use batches wherever possible, even for 2 items. It will be about twice
as fast as not batching.

~Andrew


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