Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
Date:   Fri, 27 Oct 2023 09:32:38 +0200
From:   Boris Brezillon <boris.brezillon@collabora.com>
To:     Christian =?UTF-8?B?S8O2bmln?= <christian.koenig@amd.com>
Cc:     Danilo Krummrich <dakr@redhat.com>, airlied@gmail.com,
        daniel@ffwll.ch, matthew.brost@intel.com, faith@gfxstrand.net,
        luben.tuikov@amd.com, dri-devel@lists.freedesktop.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH drm-misc-next v3] drm/sched: implement dynamic job-flow
 control
Message-ID: <20231027093238.2ff8172e@collabora.com>
In-Reply-To: <0bc79ae3-04fe-4e85-9fd0-e8b281148390@amd.com>
References: <20231026161431.5934-1-dakr@redhat.com>
        <0bc79ae3-04fe-4e85-9fd0-e8b281148390@amd.com>
Organization: Collabora
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, 27 Oct 2023 09:22:12 +0200
Christian K=C3=B6nig <christian.koenig@amd.com> wrote:

> > +
> > +	/**
> > +	 * @update_job_credits: Called once the scheduler is considering this
> > +	 * job for execution.
> > +	 *
> > +	 * Drivers may use this to update the job's submission credits, which=
 is
> > +	 * useful to e.g. deduct the number of native fences which have been
> > +	 * signaled meanwhile.
> > +	 *
> > +	 * The callback must either return the new number of submission credi=
ts
> > +	 * for the given job, or zero if no update is required.
> > +	 *
> > +	 * This callback is optional.
> > +	 */
> > +	u32 (*update_job_credits)(struct drm_sched_job *sched_job); =20
>=20
> Why do we need an extra callback for this?
>=20
> Just document that prepare_job() is allowed to reduce the number of=20
> credits the job might need.

->prepare_job() is called only once if the returned fence is NULL, but
we need this credit-update to happen every time a job is considered for
execution by the scheduler. If you're saying this control-flow should
be implemented with a dma_fence that's signaled when enough space is
available, I fear Danilo's work won't be that useful to the PowerVR
driver, unfortunately.