Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread hpa
On February 20, 2017 2:02:53 PM PST, Paolo Bonzini wrote: > >> > Yes. But 150-200 clock cycles are nothing compared to the cache >misses >> > you get from preemption, so I'd ignore that. Saving 300 clock >cycles on >> > userspace exits from TR+GSBASE would be about 5% on my Haswell. >> >> That'

Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread Paolo Bonzini
> > Yes. But 150-200 clock cycles are nothing compared to the cache misses > > you get from preemption, so I'd ignore that. Saving 300 clock cycles on > > userspace exits from TR+GSBASE would be about 5% on my Haswell. > > That's still 5% :) Yes, 5% on userspace exits is good (though they're r

Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread Andy Lutomirski
On Mon, Feb 20, 2017 at 8:51 AM, Paolo Bonzini wrote: > > > On 20/02/2017 17:46, Andy Lutomirski wrote: >> On Mon, Feb 20, 2017 at 3:05 AM, Paolo Bonzini wrote: >>> >>> >>> On 18/02/2017 04:29, Andy Lutomirski wrote: There's no code here because the patch is trivial, but I want to run t

Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread Paolo Bonzini
On 20/02/2017 17:46, Andy Lutomirski wrote: > On Mon, Feb 20, 2017 at 3:05 AM, Paolo Bonzini wrote: >> >> >> On 18/02/2017 04:29, Andy Lutomirski wrote: >>> There's no code here because the patch is trivial, but I want to run >>> the idea by you all first to see if there are any issues. >>> >>>

Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread Andy Lutomirski
On Mon, Feb 20, 2017 at 3:05 AM, Paolo Bonzini wrote: > > > On 18/02/2017 04:29, Andy Lutomirski wrote: >> There's no code here because the patch is trivial, but I want to run >> the idea by you all first to see if there are any issues. >> >> VMX is silly and forces the TSS limit to the minimum on

Re: RFC: Getting rid of LTR in VMX

2017-02-20 Thread Paolo Bonzini
On 18/02/2017 04:29, Andy Lutomirski wrote: > There's no code here because the patch is trivial, but I want to run > the idea by you all first to see if there are any issues. > > VMX is silly and forces the TSS limit to the minimum on VM exits. KVM > wastes lots of cycles bumping it back up to

Re: RFC: Getting rid of LTR in VMX

2017-02-18 Thread Paolo Bonzini
On 18/02/2017 04:29, Andy Lutomirski wrote: > There's no code here because the patch is trivial, but I want to run > the idea by you all first to see if there are any issues. > > VMX is silly and forces the TSS limit to the minimum on VM exits. KVM > wastes lots of cycles bumping it back up to

RFC: Getting rid of LTR in VMX

2017-02-17 Thread Andy Lutomirski
There's no code here because the patch is trivial, but I want to run the idea by you all first to see if there are any issues. VMX is silly and forces the TSS limit to the minimum on VM exits. KVM wastes lots of cycles bumping it back up to accomodate the io bitmap. I propose that we rework this