-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Revert "cpu-kvm: Support perf counters on hybrid host architectures" #1127
Conversation
Wouldn't it be better to just disable this feature by default instead? That way, we can treat the feature as experimental and we don't end up reverting a large amount of code (less disruptive). It's presumably working on some systems when enabled. |
Related: #1138 |
Change-Id: Ieff05ae342cdc9e3b23718e365ff86b83784d41f
38fbd7f
to
537dc69
Compare
Sorry for the late reply. I agree with @andysan — I think it would be better to simply disable hybrid performance counters by default and treat them as an experimental feature rather than just revert the entire patch. Reverting this patch has once again made the KVM CPU non-functional on the E-cores of my Alder Lake system. Can I just resubmit this patch with this line changed from |
I agree that it would be good to have support for hybrid performance counters. However, we were not able to get #1065 to work on our systems (see discussion here: #1126 (comment) apologies for the information being spread around). I am positive that there are some problems with the current KVM support. We are also having issues here: #1119 (comment) I'm also wary of adding parameters to SimObjects that are required for different hosts. I don't like the idea of tying simulation parameters to a host. That said, if we can 1) document the problems and which hosts are causing the issue and 2) explain the needed configurations for different hosts, then we can resubmit the patch. However, we need to do a lot more testing this time before committing 🙂 |
Reverts #1065
Reverting this change because this PR breaks X86 kvm as mentioned in the issue #1126.