Re: [Xen-ia64-devel] [PATCH 0/10] [RFC] performance tuning TAKE 3

[Tristan Gingold <Tristan.Gingold@xxxxxxxx>]

Le Mercredi 09 Août 2006 13:27, Isaku Yamahata a écrit :
> Hi all
Hi

some comments enclosed.

> These patches are for performance tuning TAKE 3.
> They are for comment, review and evaluation.
>
> PATCHES:
> - performace counter
>   NEW
> - per vcpu vhpt
> - tlb tracking
> - deferred page freeing
> - p2m exposure
>   NEW
>
> CHANGES:
> - various bug fixes.
>   especially deferred page freeing patch bug.
>   It spoiled performance much.
>
> - performance counter
>   added performance counter.
>   I added these for evaluation of the patches.
>   But I think some of them are usefull for normal use.
>   Is there any objection for commit?
I think they can be safely commited.

> - P2M table exposure
>   I also included test module for p2m exposure module.
>   It showed that p2m conversion with p2m exposure is faster than
>   the hypercall conversion at about order 2.
>   So I think p2m exposure patch is acceptable.
>
>
> PATCH DETAIL:
> - per vcpu vhpt
>   It focuses on vcpu migration between physical cpus.
>   With credit scheduler, vcpu is heavily migrated.
>   This patch tries to reduce vTLB flush when vcpu is migrated.
Although I have no objection for integrating this patch I have some doubts 
about the gain.
If vcpus are heavily migrated, this is a real loss due to caches.
If vcpus are not heavily migrated the gain should be in the noise.
(Of course it depends on the weight!)

> - tlb tracking
>   It forcuses on grant table mapping.
>   When page is unmapped, full vTLB flush is necessary.
>   By tracking tlb insert on grant mapped page, full vTLB flush
>   can be avoided.
>   Especially vbd does only DMA, so dom0 doesn't insert tlb entry
>   on the grant mapped page. In such case any vTLB flush isn't needed.
>
> - deferred page freeing
>   When the page in which tlb insert isn't tracked is unmapped/zapped from
>   domain, full vTLB flush is necessary again.
>   Balloon driver and grant table page transfer is the case.
>   This patch focuses on it.
>   It tries to batch freeing/zapping page from domain in order
>   to reduce full vTLB flush.
>
> - p2m exposure
>   DMA paravirtualization requires the conversion from pseudo physical
> address to machine address. Currently it is done by hypercall.
>   This patch tries to reduce the conversion overhead by read-only
>   mapping the xen p2m table to domain.

Thank you for the optimization job.

Tristan.

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel