On Thu, 30 Jan 2020 09:41:08 -0800
Michael Hinton <[email protected]> wrote:

> On Monday, January 27, 2020 at 12:16:08 AM UTC-7, Henning Schild
> wrote:
> >
> > Ok, so we are just looking for differences between the inmate and
> > the linux as non-root cell, because the jailhouse/virtualization
> > overhead is acceptable or known. 
> >  
> I think so, yeah.
>  
> 
> > In that case a memory bound workload boils down to the mapping and
> > the tlb misses or CAT. And cpu bound could be an issue with the
> > FPU. If your binary uses FPU instructions but is able to fall back
> > to soft-fpu, you should check which path it takes in the inmate. 
> >  
> Let's see: CAT isn't supported by my chip, so that won't help, 
> unfortunately. But the Linux workload is mostly idle, so I'm not sure
> how much that would have helped anyways.
> 
> My inmates don't use FPU instructions, and it's not even set up, so I
> don't think that will cause a problem.
> 
> I will investigate TLB misses and page mappings and see what I can
> find.

To align the mapping on the hypervisor side, you could just run that
inmate in the same cell configuration you are running that guest-linux
in. For the inmate itself the pagetable is constructed by the mapping
library. The code looks like it tries to do huge pages, make sure the
call map_range just once with your full memory range. Aligned and maybe
more than you actually need. Consider putting a few printfs into the
mapping code to see which path (page-size) it goes.

That way you might get lucky without having to find how to read
tlb-miss counters from the inmate.

Henning

> Thanks,
> Michael 
> 

-- 
You received this message because you are subscribed to the Google Groups 
"Jailhouse" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jailhouse-dev/20200131100917.3ad1b337%40md1za8fc.ad001.siemens.net.

Reply via email to