Re: [PATCH 09/32] cpu_map: Group models in index.xml
On Wed, Nov 20, 2024 at 12:11:19 +0000, Daniel P. Berrangé wrote:
On Tue, Nov 19, 2024 at 07:49:45PM +0100, Jiri Denemark wrote:
> We already visually group the included models according to vendor using
> comments. This patch introduces a new <group> element for doing it
> properly in a machine friendly way.
AFAICT the <group> has no functional effect
I added it so the following commit can automatically add new CPU models
in the right place. So yes, no functional effect for the CPU map itself.
>
> Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com>
> ---
> src/cpu/cpu_map.c | 2 +-
> src/cpu_map/index.xml | 226 ++++++++++++++++++++++--------------------
> 2 files changed, 121 insertions(+), 107 deletions(-)
> diff --git a/src/cpu_map/index.xml b/src/cpu_map/index.xml
> index 15cb63afe5..2cb97a83ba 100644
> --- a/src/cpu_map/index.xml
> +++ b/src/cpu_map/index.xml
> @@ -3,122 +3,136 @@
> <include filename='x86_vendors.xml'/>
> <include filename='x86_features.xml'/>
> +
> + <group name='Intel-based QEMU generic CPU models'>
> + <include filename='x86_pentium.xml'/>
> + <include filename='x86_pentium2.xml'/>
> + <include filename='x86_pentium3.xml'/>
> + <include filename='x86_pentiumpro.xml'/>
> + <include filename='x86_coreduo.xml'/>
> + <include filename='x86_n270.xml'/>
> + <include filename='x86_core2duo.xml'/>
> + </group>
> +
> + <group name='Generic QEMU CPU models'>
> + <include filename='x86_qemu32.xml'/>
> + <include filename='x86_kvm32.xml'/>
> + <include filename='x86_cpu64-rhel5.xml'/>
> + <include filename='x86_cpu64-rhel6.xml'/>
> + <include filename='x86_qemu64.xml'/>
> + <include filename='x86_kvm64.xml'/>
> + </group>
> +
> + <group vendor='Intel'>
In some cases youve used 'name' and some 'vendor', which seems
fairly arbitrary, but I guess we're ignoring this attrbiute
entirely.
Still all the CPUs above have either 'Intel' or 'AMD' set
as their vendor in QEMU, so separating them off feels a
bit odd.
Yes, it's completely arbitrary. I basically just changed the comments
into <group name='the text of the comment'/>, with the exception of a
few groups to which the script is going to add new models. I agree it's
odd, but after playing with XPath and trying to come up with a way to
use the existing comments to detect where new models should be added
this solution looked very clean and nice :-) I guess I could also go
with just <group name='the text of the comment'/> in all cases for
consistency.
BTW, the XPath way almost works as you can really reference a comment in
the document and even match its content. I just didn't find a way to
select a node that is after a specific comment, but before another
comment. Which I guess is a good thing as it was pretty disgusting :-)
If we're just going to group everything based on vendor,
why not just call the tag <vendor name=...> ?
We can't use just vendor because some models (the old ones) don't have a
vendor. We could perhaps use something like vendor='generic',
vendor='QEMU' or something similar, although using the complete text
from the comment makes reading a bit easier for people.
Jirka