-
Story
-
Resolution: Unresolved
-
Normal
-
rhel-9.2.0
-
libvirt-10.9.0-1.el9
-
rhel-sst-virtualization
-
ssg_virtualization
-
2
-
Dev ack
-
False
-
-
None
-
None
-
Pass
-
None
-
Enhancement
-
-
Proposed
-
-
Unspecified
-
10.9.0
-
None
+++ This bug was initially created as a clone of Bug #1643887 +++
(In reply to Vitaly Kuznetsov from comment #6)
> The most up-to-date list of enlightenments in QEMU upstream is:
>
> hv-relaxed
> hv-vapic
> hv-spinlocks
> hv-vpindex
> hv-runtime
> hv-crash
> hv-time
> hv-synic
> hv-stimer
> hv-tlbflush
> hv-ipi
> hv-reset
> hv-frequencies
> hv-reenlightenment
> hv-evmcs (Intel only)
> hv-stimer-direct
These ^^ are supported.
> hv-avic (== hv-apicv)
> hv-emsr-bitmap
> hv-xmm-input
> hv-no-nonarch-coresharing = on/off/auto (careful! Can only be used with
> proper vCPU pinning!)
And these ^^ need to be implemented yet. What's the priority? I haven't found any libvirt RFE for these, should I create one?
— Additional comment from Vitaly Kuznetsov on 2022-09-05 15:34:17 CEST —
(In reply to Michal Privoznik from comment #8)
> > hv-avic (== hv-apicv)
> > hv-emsr-bitmap
> > hv-xmm-input
> > hv-no-nonarch-coresharing = on/off/auto (careful! Can only be used with
> > proper vCPU pinning!)
>
> And these ^^ need to be implemented yet. What's the priority? I haven't
> found any libvirt RFE for these, should I create one?
The priority should be 'normal' I believe, all these give moderate performance
improvement for certain scenarios.
'hv-avic' is supported since QEMU-6.2, hv-xmm-input and hv-emsr-bitmap are new
(QEMU-7.1). 'hv-no-nonarch-coresharing' is old (QEMU-4.2) but it's very specific
and hard to use properly.