qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <philmd@linaro.org>
To: "Daniel Henrique Barboza" <dbarboza@ventanamicro.com>,
	"Daniel P. Berrangé" <berrange@redhat.com>
Cc: qemu-devel@nongnu.org, peter.maydell@linaro.org,
	eduardo@habkost.net, marcel.apfelbaum@gmail.com,
	wangyanan55@huawei.com, zhao1.liu@intel.com, eblake@redhat.com,
	armbru@redhat.com, ajones@ventanamicro.com
Subject: Re: [PATCH] qapi, machine-qmp-cmds.c: query-accelerator support
Date: Thu, 3 Oct 2024 22:59:23 +0200	[thread overview]
Message-ID: <e898c191-23f8-4e5d-9361-0ae7e8c5b78b@linaro.org> (raw)
In-Reply-To: <046df1e9-5e4c-4f1c-8edb-1ee1244fe340@ventanamicro.com>

On 2/10/24 20:06, Daniel Henrique Barboza wrote:
> 
> 
> On 9/27/24 7:50 AM, Daniel P. Berrangé wrote:
>> Markus: QAPI design Qs for you at the bottom
>>
>> On Wed, Sep 25, 2024 at 10:19:33AM -0300, Daniel Henrique Barboza wrote:
>>>
>>>
>>> On 9/19/24 9:22 AM, Daniel P. Berrangé wrote:
>>>> On Thu, Sep 19, 2024 at 08:20:56AM -0300, Daniel Henrique Barboza 
>>>> wrote:
>>>>> Add a QMP command that shows all specific properties of the current
>>>>> accelerator in use.
>>>>
>>>> Why do we need to expose /everything/ ?
>>>
>>> I wouldn't mind pick and choose advertised properties for the 
>>> accelerators
>>> like we do with other APIs.
>>>
>>> This would mean that each arch should choose what to advertise or 
>>> not, given that
>>> some accelerator properties might be relevant just for some archs. 
>>> The API would
>>> be implemented by each arch individually.
>>
>> Well with qemu-system-any we might get multiple arches reporting
>> info in the same binary, so we'll need to fan out to fill in the
>> per-arch info, after doing a common base.
>>
>> Hmmm, i wonder if qemu-system-any will support mixing KVM and TCG ?
>> ie KVM for the host native accelerator, combined with TCG for the
>> foreign archs ??? Hopefully not !
> 
> If you're talking about Phil's patches it seems that it'll be TCG only:

Orthogonaly to the single binary series (mentioned below with the
'any'-architecture target), we plan to add a "Dual HW/SW accelerator".
It will be implemented using the same AccelOps structure we currently
have, and internally it will dispatch between HW/SW. First prototype
plan to provide HVF+TCG.

IMHO query-accelerator should expose accelerator generic properties,
not the target-specific ones. At least by default, we could add an
optional flag to include target-specific stuff. At least that'd keep
the core accel core simpler.

> https://lore.kernel.org/qemu-devel/20240305220938.85410-1-philmd@linaro.org/
> 
> Patch 2 commit msg states:
> 
> ------
> Add the 'any'-architecture target.
> 
> - Only consider 64-bit targets
> - Do not use any hardware accelerator (except qtest)
> - For architecture constants, use:
>    . max of supported targets phys/virt address space
>    . max of supported targets MMU modes
>    . min of supported targets variable page bits
> ------
> 
> 
> Thanks,
> 
> Daniel



  reply	other threads:[~2024-10-03 21:00 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-19 11:20 [PATCH] qapi, machine-qmp-cmds.c: query-accelerator support Daniel Henrique Barboza
2024-09-19 12:22 ` Daniel P. Berrangé
2024-09-25 13:19   ` Daniel Henrique Barboza
2024-09-27 10:50     ` Daniel P. Berrangé
2024-10-02 18:06       ` Daniel Henrique Barboza
2024-10-03 20:59         ` Philippe Mathieu-Daudé [this message]
2024-10-07  9:53       ` Markus Armbruster
2024-10-11 12:26         ` Daniel Henrique Barboza

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=e898c191-23f8-4e5d-9361-0ae7e8c5b78b@linaro.org \
    --to=philmd@linaro.org \
    --cc=ajones@ventanamicro.com \
    --cc=armbru@redhat.com \
    --cc=berrange@redhat.com \
    --cc=dbarboza@ventanamicro.com \
    --cc=eblake@redhat.com \
    --cc=eduardo@habkost.net \
    --cc=marcel.apfelbaum@gmail.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=wangyanan55@huawei.com \
    --cc=zhao1.liu@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).