[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v4 8/8] x86: avoid double CR3 reload when switching to guest user mode



On 19/03/18 14:41, Jan Beulich wrote:
> When XPTI is active, the CR3 load in restore_all_guest is sufficient
> when switching to user mode, improving in particular system call and
> page fault exit paths for the guest.
> 
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> Tested-by: Juergen Gross <jgross@xxxxxxxx>
> Reviewed-by: Juergen Gross <jgross@xxxxxxxx>

I've done some more simple performance tests with some of the patches
with xpti=false and xpti=true. Data is always elapsed time, system time
and user time in seconds for a make -j 4 in dom0 with 4 vcpus, stddev
in braces, based on 5 runs each (I tried 30 runs, but the result didn't
really change):

xpti=false
no patch:  89.96 ( 2.84)   97.05 ( 2.39)  178.64 ( 1.55)
Jan p1:    90.65 ( 2.57)   99.50 ( 3.85)  180.35 ( 2.44)
Jan p5:    91.33 ( 0.89)   99.72 ( 2.56)  180.97 ( 1.71)
Jan p6:    90.86 ( 2.59)   97.09 ( 2.59)  177.85 ( 2.35)
Jan p7:    90.72 ( 2.84)  100.10 ( 4.60)  179.85 ( 2.61)
Jan p8:    88.59 ( 0.71)   96.31 ( 2.14)  178.47 ( 0.86)

xpti=true
no patch: 113.42 ( 3.13)  165.99 ( 3.89)  180.99 ( 1.63)
Jan p1:   111.69 ( 3.15)  163.63 ( 3.61)  181.22 ( 1.93)
Jan p5:   114.76 ( 2.28)  167.15 ( 4.67)  181.13 ( 1.75)
Jan p6:   116.85 ( 2.35)  168.73 ( 3.68)  181.27 ( 1.98)
Jan p7:   115.37 ( 2.71)  166.96 ( 4.41)  180.82 ( 1.98)
Jan p8:   114.85 ( 2.83)  167.08 ( 5.00)  181.27 ( 1.85)

Summing it up: performance isn't really changing for any of the patches.


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.