[Qemu-devel] question about VM kernel parameter idle=poll/mwait/halt/nomwait

2013-11-20 Thread Zhanghaoyu (A) 
Hi, all

What's the difference of the linux guest kernel parameter 
idle=poll/mwait/halt/nomwait, especially in performance?

Taking the performance into account, which one is best?

In my opinion, if the number of all VMs' vcpus is far more than that of pcpus, 
e.g. SPECVirt test, idle=halt is better for server's total throughput,
otherwise, e.g. in some CT scenario, the number of total vcpus is not greater 
than that of pcpus, idle=poll is better for server's total throughput,
because of less latency and VMEXIT.

linux-3.9 and above, idle=mwait is not recommended.

Thanks,
Zhang Haoyu

Re: [Qemu-devel] question about VM kernel parameter idle=poll/mwait/halt/nomwait

2013-11-20 Thread Michael S. Tsirkin 
On Thu, Nov 21, 2013 at 03:45:28AM +, Zhanghaoyu (A) wrote:
 Hi, all
 
 What's the difference of the linux guest kernel parameter 
 idle=poll/mwait/halt/nomwait, especially in performance?
 
 Taking the performance into account, which one is best?
 
 In my opinion, if the number of all VMs' vcpus is far more than that of 
 pcpus, e.g. SPECVirt test, idle=halt is better for server's total throughput,
 otherwise, e.g. in some CT scenario, the number of total vcpus is not greater 
 than that of pcpus, idle=poll is better for server's total throughput,
 because of less latency and VMEXIT.

Makes sense overall.

 linux-3.9 and above, idle=mwait is not recommended.
 
 Thanks,
 Zhang Haoyu

Does it actually have effect? I didn't think it would.

Re: [Qemu-devel] question about VM kernel parameter idle=poll/mwait/halt/nomwait

2013-11-20 Thread Gleb Natapov 
On Thu, Nov 21, 2013 at 09:01:39AM +0200, Michael S. Tsirkin wrote:
 On Thu, Nov 21, 2013 at 03:45:28AM +, Zhanghaoyu (A) wrote:
  Hi, all
  
  What's the difference of the linux guest kernel parameter 
  idle=poll/mwait/halt/nomwait, especially in performance?
  
  Taking the performance into account, which one is best?
  
  In my opinion, if the number of all VMs' vcpus is far more than that of 
  pcpus, e.g. SPECVirt test, idle=halt is better for server's total 
  throughput,
  otherwise, e.g. in some CT scenario, the number of total vcpus is not 
  greater than that of pcpus, idle=poll is better for server's total 
  throughput,
  because of less latency and VMEXIT.
 
 Makes sense overall.
 
  linux-3.9 and above, idle=mwait is not recommended.
  
  Thanks,
  Zhang Haoyu
 
 Does it actually have effect? I didn't think it would.
mwait instruction is never exposed to a guest. With idle=mwait is will
likely fall back to halt silently.

--
Gleb.