The attached file is the super page patch which supports both 2MB and 4MB (depending on the paging modes of hypervisor) under hardware assisted paging. The idea is to allocate super pages when guests are being created. Whenever such requests cannot be satisfied, it falls back to normal 4KB allocation. Also it splits large pages into normal 4KB pages whenever necessary. This patch is applicable on the latest xen-unstable tree. Any comment/suggestion is welcomed! Since it also touches a small portion of IA64 and PowerPC code, I would appreciate the information of whether this patch breaks IA64 or PowerPC compilation. Thanks, -Wei _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> The attached file is the super page patch which supports both 2MB and > 4MB (depending on the paging modes of hypervisor) under hardware > assisted paging. The idea is to allocate super pages when guests are > being created. Whenever such requests cannot be satisfied, it falls > back to normal 4KB allocation. Also it splits large pages into normal > 4KB pages whenever necessary. This patch is applicable on the latest > xen-unstable tree. > > Any comment/suggestion is welcomed! Since it also touches a small > portion of IA64 and PowerPC code, I would appreciate the informationof> whether this patch breaks IA64 or PowerPC compilation.It''s great to see this patch, thanks. Could you give a summary of what testing its been subjected to? Also, do you have any performance results? I''m particularly interested in results with a 64b hypervisor. We''ll have to update the balloon driver to try and hand back 2MB aligned slabs to avoid fragmentation. Thanks, Ian _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian, Super page has many advantages by reducing TLB misses and the nested page table walk. I am currently collecting the performance results of commonly used benchmarks. The results will be posted next time along with a formal patch (probably this week). Thanks, -Wei -----Original Message----- From: Ian Pratt [mailto:Ian.Pratt@eu.citrix.com] Sent: Monday, April 21, 2008 6:52 PM To: Huang2, Wei; xen-devel@lists.xensource.com Cc: Ian Pratt Subject: RE: [Xen-devel] [Patch][RFC] Super Page Patch> The attached file is the super page patch which supports both 2MB and > 4MB (depending on the paging modes of hypervisor) under hardware > assisted paging. The idea is to allocate super pages when guests are > being created. Whenever such requests cannot be satisfied, it falls > back to normal 4KB allocation. Also it splits large pages into normal > 4KB pages whenever necessary. This patch is applicable on the latest > xen-unstable tree. > > Any comment/suggestion is welcomed! Since it also touches a small > portion of IA64 and PowerPC code, I would appreciate the informationof> whether this patch breaks IA64 or PowerPC compilation.It''s great to see this patch, thanks. Could you give a summary of what testing its been subjected to? Also, do you have any performance results? I''m particularly interested in results with a 64b hypervisor. We''ll have to update the balloon driver to try and hand back 2MB aligned slabs to avoid fragmentation. Thanks, Ian _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Reasonably Related Threads
- [PATCH] patch to support super page (2M) with EPT
- Hardware Assisted Paging Param and Message
- [PATCH] HAP: Add global enable/disable command line option
- [PATCH][HAP][2/2] fix CR4 initialization when hap is on
- Question regarding the number of P2M l3e entries