Re: -cpu Questions

[Thomas Huth]

On 09/12/2022 16.23, Harold Grovesteen wrote:
> I am brand new to qemu for s390x.  Installed it from the Ubuntu
> package.  As output from qemu-system-s390x executable:
>
> QEMU emulator version 6.2.0 (Debian 1:6.2+dfsg-2ubuntu6.5)
>
> Just trying to familiarize myself with Qemu.  Did a few helps (qemu -h,
> machine help, cpu help).
>
> Tried to run -cpu z13.2.  Produced a long list of unavailable features.
> Tried to run -cpu z14.2.  Produced another long list, but not
> identical, list of unavailable features.

These CPU models correspond to real hardware CPUs. They are currently only 
usable with KVM on a real mainframe host.

The emulation in QEMU (TCG) is not advanced enough yet to be able to fully 
emulate all features of a real CPU (except the oldest models like the z900), 
that's why you are getting these messages about unavailable features.

> Tried -cpu max and all of the unavailable features disappeared,
> apparently now available to -cpu max, although two warnings resulted.
>
> Four questions...
>
> Why is a full list of features supported by -cpu max but not -cpu z13.2
> or -cpu z14.2?

The "max" CPU model just enables as many CPU features as possible, not all 
CPU features. So in fact, when using emulation,  "max" less features enable 
than the z13, for example.

> Is this an artifact of using the Ubuntu package version of qemu s390x
> vs. a build from scratch?

QEMU 6.2.0 is still quite recent. There have been some additions in the past 
year, so you will get more emulated features with QEMU 7.2, but it's still 
not enough to emulate a full z13, for example.

> Additionally, most cpu's would apparently correspond to individual
> hardware models.  What is the significance of genxx cpu models?

gen15 = z15, gen16 = z16. The CPU models have just been introduced before 
IBM announced the official names, that's why they are named differently. We 
should maybe introduce aliases for those now that we know the official names.

 HTH,
  Thomas