From mboxrd@z Thu Jan  1 00:00:00 1970
From: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Subject: Re: [RFC] Host1x/TegraDRM UAPI
Date: Thu, 25 Jun 2020 02:18:47 +0300
Message-ID: <f85f3fa0-ba18-ddd5-2bac-158dfed9b792@gmail.com>
References: <9b06b7ec-f952-2561-7afb-5653514cd5d3@kapsi.fi>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Return-path: <linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
In-Reply-To: <9b06b7ec-f952-2561-7afb-5653514cd5d3-/1wQRMveznE@public.gmane.org>
Content-Language: en-US
Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
To: Mikko Perttunen <cyndis-/1wQRMveznE@public.gmane.org>, Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>, Jon Hunter <jonathanh-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>, David Airlie <airlied-cv59FeDIM0c@public.gmane.org>, Daniel Vetter <daniel-/w4YWyX8dFk@public.gmane.org>, sumit.semwal-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org, gustavo-THi1TnShQwVAfugRpC6u6w@public.gmane.org
Cc: "linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" <linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, dri-devel <dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>, talho-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org, bhuntsman-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org
List-Id: linux-tegra@vger.kernel.org

23.06.2020 15:09, Mikko Perttunen пишет:
> struct drm_tegra_channel_submit {
>         __u32 channel_id;
>         __u32 flags;
> 
>         /**
>          * [in] Timeout in microseconds after which the kernel may
>          *   consider the job to have hung and may reap it and
>          *   fast-forward its syncpoint increments.
>          *
>          *   The value may be capped by the kernel.
>          */
>         __u32 timeout;
> 
>         __u32 num_syncpt_incrs;
>         __u32 num_relocations;
>         __u32 num_commands;
> 
>         __u64 syncpt_incrs;
>         __u64 relocations;
>         __u64 commands;

Do we really need to retain the multi-gather support? The code-bloat
(both userspace and kernel driver) is very significant just for
preparing and patching of the multi-buffer cmdstreams.

If userspace runs out of a free space within the pushbuffer, then it
should simply reallocate a larger pushbuffer.

I'm suggesting that we should have a single gather per-job, any objections?