Today, for each emulated MMIO operation, there is a long path from
page fault to kvmppc_hv_emulate_mmio(), include looking up Hash Page
Table, locking HPTE twice and searching mmio gfn from memslots.
This series introduced a cache mechanism for MMIO HPTEs to shorten
the path, which will improve the IO performance of emulated device,
especially virtio device. Because currently almost all of virtio
nofications(Port IO access) will hit the same HPTE.
In this series, patch 1 fixed a bug on key field of HPTE; patch 2
implemented the per vcpu cache for recently page faulted MMIO
entries.
The performance test result:
FIO(null_blk, iops)
===========================
Before After
1st 32095 32616
2nd 32032 32635
3rd 32057 32638
4th 32005 32598
5th 32069 32628
NetPerf(TCP_RR, Trans.Rate per sec)
===========================
Before After
1st 30734.66 31328.63
2nd 30749.26 31337.77
3rd 30825.99 31278.19
4th 30708.94 31350.81
5th 30832.35 31484.80
Changelog v2:
- Fix a bug of the function is_mmio_hpte()
- Add some performance test results
Yongji Xie (2):
KVM: PPC: Book3S HV: Clear the key field of HPTE when the page is paged out
KVM: PPC: Book3S HV: Add a per vcpu cache for recently page faulted MMIO entries
arch/powerpc/include/asm/kvm_host.h | 21 ++++++
arch/powerpc/kvm/book3s_64_mmu_hv.c | 20 ++++++-
arch/powerpc/kvm/book3s_hv.c | 9 +++
arch/powerpc/kvm/book3s_hv_rm_mmu.c | 122 +++++++++++++++++++++++++++++------
4 files changed, 151 insertions(+), 21 deletions(-)
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
