[libvirt] [PATCH] cpu: Add new EPYC CPU model

[Jiri Denemark]

On Wed, Aug 23, 2017 at 13:49:41 -0500, Brijesh Singh wrote:
> Add a new CPU model called 'EPYC' to model processors from AMD EPYC
> family (which includes EPYC 76xx,75xx,74xx, 73xx and 72xx).
> 
> The following features bits have been added/removed compare to Opteron_G5
> 
> Added: monitor, movbe, rdrand, mmxext, ffxsr, rdtscp, cr8legacy, osvw,
>        fsgsbase, bmi1, avx2, smep, bmi2, rdseed, adx, smap, clfshopt, sha
>        xsaveopt, xsavec, xgetbv1, arat
> 
> Removed: xop, fma4, tbm
> 
> The patch is depend on EPYC CPU model supported introduced in qemu [1]
> 
> [1] https://patchwork.kernel.org/patch/9902205/
> 
> Cc: Tom Lendacky <Thomas.Lendacky at amd.com>
> Signed-off-by: Brijesh Singh <brijesh.singh at amd.com>
> ---
>  src/cpu/cpu_map.xml | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 74 insertions(+)
> 
> diff --git a/src/cpu/cpu_map.xml b/src/cpu/cpu_map.xml
> index 8e7ac49..522d66b 100644
> --- a/src/cpu/cpu_map.xml
> +++ b/src/cpu/cpu_map.xml
> @@ -251,6 +251,9 @@
>      <feature name='clflushopt'>
>        <cpuid eax_in='0x07' ecx_in='0x00' ebx='0x00800000'/>
>      </feature>
> +    <feature name='sha_ni'>

QEMU calls this feature "sha-ni".

> +      <cpuid eax_in='0x07' ecx_in='0x00' ebx='0x20000000'/>
> +    </feature>

The CPU features in cpu_map.xml are sorted by CPUID bits, which means
this new feature should go between avx512cd and avx512bw.

>      <feature name='avx512pf'>
>        <cpuid eax_in='0x07' ecx_in='0x00' ebx='0x04000000'/>
>      </feature>
> @@ -1545,6 +1548,77 @@
>        <feature name='xop'/>
>        <feature name='xsave'/>
>      </model>
> +
> +    <model name='EPYC'>
> +      <signature family='23' model='1'/>
> +      <vendor name='AMD'/>
> +      <feature name='sse2'/>
> +      <feature name='sse'/>
> +      <feature name='fxsr'/>
...

We list model's features sorted by name.

ACK with the issues fixed.

However, it would be nice to add some CPUID data to our test suite.
Luckily enough I have such data so I'll just resend [1] a fixed version
of this patch with the test data added.

Jirka

[1] https://www.redhat.com/archives/libvir-list/2017-September/msg00111.html