Hi Keir and Alex,
A Virtual Hash Page Table(VHPT) is a special feature of IA64.
It's an extension of TLB that resides in memory and can be
automatically searched by the processor.
This patch allows user to configure VHPT size per HVM domain.
We have an interesting result.
We tried some large
On Tue, Oct 16, 2007 at 10:29:53AM +0900, Akio Takebe wrote:
Hi, Alex, Horms
Sorry for my late responce.
What do you think about it?
The panic does seem a bit much, but shouldn't we return rather than
continue on with a NULL_VECTOR? Thanks,
Yes, I also thik so.
I'll post the new
Hi, Horms
On Tue, Oct 16, 2007 at 10:29:53AM +0900, Akio Takebe wrote:
Hi, Alex, Horms
Sorry for my late responce.
What do you think about it?
The panic does seem a bit much, but shouldn't we return rather than
continue on with a NULL_VECTOR? Thanks,
Yes, I also thik so.
I'll post
Hi Kouya.
I guess Linux PV domains (especially dom0) possibly can gain similar
performance improvement by increasing VHPT size too.
So would it be easy to make vhpt parameter non hvm specific?
I know your concern is hvm domain, though.
On Tue, Oct 16, 2007 at 03:21:39PM +0900, Kouya Shimura
I've cleaned up the xc.hvm_build() interface in changeset 16116. Please
merge against that. The main change will be that you will set the new HVM
parameter in IA64_HVM_ImageHandler.buildDomain(), rather than passing the
new parameter down to xc.hvm_build().
-- Keir
On 16/10/07 07:21, Kouya
Quoting Kouya Shimura [EMAIL PROTECTED]:
Hi Keir and Alex,
A Virtual Hash Page Table(VHPT) is a special feature of IA64.
It's an extension of TLB that resides in memory and can be
automatically searched by the processor.
This patch allows user to configure VHPT size per HVM domain.
Good
On Tue, Oct 16, 2007 at 01:51:25PM +0200, [EMAIL PROTECTED] wrote:
And if you want to go farther, you may also want to specify vtlb size!
Yes. The results showed that reducing vTLB flush overhead even
with smaller vTLB size is important for Windows performance.
So possibly eliminating whole the
Hi Yamahata-san,
I'm now thinking about PV domain. Please stay tuned.
As I mentioned before, vTLB is already minimum size(16KB)
and no collision for windows.
http://lists.xensource.com/archives/html/xen-ia64-devel/2007-08/msg00108.html
So I feel it's enough about vTLB. If smaller vTLB size is
On Tue, 2007-10-16 at 11:06 +0900, Atsushi SAKAI wrote:
Hi, Alex
Thank you for your comments.
I added xen0 and xenU patch.
Sorry to bother you for a trivial cleanup patch, but I need a
Sign-off for this new patch. Thanks,
Alex
--
Alex Williamson HP
On Mon, Oct 15, 2007 at 08:09:27PM +0900, Kouya Shimura wrote:
Content-Description: message body text
Hi Tristan,
Here is a patch to support windows 2008 (aka Longhorn).
I don't know EFI spec well and this patch might do something wrong.
But I succeed in installing and running windows 2008
Hi,
Edge triggered interrupts don't need an EOI but EOI-ing them is harmless
according to the IOSAPIC documentation.
Do not emit a warning in such case.
Tristan.
# HG changeset patch
# User Tristan Gingold [EMAIL PROTECTED]
# Date 1192594387 -7200
# Node ID
Hi,
 This is clean up patch for linux-defconfig(xen, xen0, xenU)
Signed-off-by: Atsushi SAKAI [EMAIL PROTECTED]
Thanks
Atsushi SAKAI
Alex Williamson [EMAIL PROTECTED] wrote:
On Tue, 2007-10-16 at 11:06 +0900, Atsushi SAKAI wrote:
Hi, Alex
Thank you for your comments.
I added
Hi,
mini-cleanup found when trying to compile with gcc 4.2
Tristan.
# HG changeset patch
# User Tristan Gingold [EMAIL PROTECTED]
# Date 1192596414 -7200
# Node ID 65468c0c6c64a234a203f3c8c223196e916ca6a4
# Parent cad488592f731bc59b2bd101ff265f8b1654ef9d
Cleanup: remove unused definitions.
Hi,
M28 has now a field x (for fc vs fc.i).
Tristan.
# HG changeset patch
# User Tristan Gingold [EMAIL PROTECTED]
# Date 1192596324 -7200
# Node ID cad488592f731bc59b2bd101ff265f8b1654ef9d
# Parent 701a3e8f8c64adc32e8c49631558da506cbfd8bc
Declare field x of M28 (fc vs fc.i).
Signed-off-by:
On Fri, Sep 28, 2007 at 05:01:44PM -0600, Alex Williamson wrote:
Hi Simon,
I'm sure you're off on holiday by now, but I've applied most of your
patches for kexec. I left out the EFI remapping and the ACPI table
save/restore. As we discussed in a separate email, this should allow UP
15 matches
Mail list logo