RE: single software TLB vs. multiple software TLBs (was RE: [Xen-ia64-devel] [PATCH] [Resend]Enable hash vtlb)

[Alex Williamson <alex.williamson@xxxxxx>]

On Thu, 2006-04-13 at 20:20 +0800, Dong, Eddie wrote:
>       Anthony's patch is ready to support all of above as a functionality
> ready solution, and so far I didn't see anybody against multiple
> software TLB support. Can u check in now as a build option? The
> performance difference in 1-2 percent should be a second level
> consideration for now.

Hi Eddie,

   There are still outstanding issues with the Hash vTLB patch.  The
performance loss with the patch has yet to be analyzed.  I still have
trouble believing domU has better performance than bare metal hardware
for anything more than isolated micro benchmarks.  I believe Anthony was
still going to send out some numbers from his system tests with the
memory size increased.  I'm hoping that there might be some profiling
and tuning done at the same time to help minimize or eliminate any
performance differences.  We also have the problem that rebooting or
powering off a guest domain with the Hash vTLB patch causes an MCA.
This is an unacceptable regression IMHO.

   Capturing guest MMIO accesses is an important Xen/ia64 feature, but I
don't think the Hash vTLB patch is the only way we can achieve this
functionality.  Maintenance reduction is certainly a consideration, but
this shouldn't be a reason to rush patches into the tree or accept them
w/o understanding all the issues.  Thanks,

        Alex

-- 
Alex Williamson                             HP Linux & Open Source Lab


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