Re: Tegra and Vulkan 1.3 Inquiry

--000000000000057d850657a2c1c6
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Aaron Kling (added to the CC since I'm not sure if he's on mesa-dev) has
been working on this precisely for the switch.  There are some issues with
some of some parts (IIRC, the Tegra X2 is sketchy) but most of it does
work. Mostly, we're missing some kernel work to fix issues that Vulkan
seems to hammer where GL doesn't.  You can find the Tegra ToDo list here:

https://gitlab.freedesktop.org/mesa/mesa/-/work_items/14105

~Faith

On Fri, Jul 24, 2026 at 4:17=E2=80=AFAM Gabriele Melluso <
[email protected]> wrote:

> Hi,
> noticed NVK's docs state Tegra/integrated GPUs aren't currently supported=
,
> but there's already a debug option for forcing coherent memory maps
> specifically on Tegra devices. Has bringing up NVK on Tegra X1/GM20B ever
> been scoped out (e.g. for nouveau-based Linux on Jetson TX1 or Nintendo
> Switch homebrew Linux distros)? Since the class-engine version logic
> already treats Maxwell as capable of 1.3, I'm trying to understand what's
> actually missing beyond memory coherency, class-engine enumeration for th=
e
> integrated part, host1x/syncpoint integration, something else entirely? A=
nd
> is this an area the project would welcome outside contribution on, or is
> there a reason it's been deprioritized I should know before looking into =
it?
>
> Regards,
>
> Gabriel
>

--000000000000057d850657a2c1c6
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Aaron Kling (added to the CC since I&#39;m not sure i=
f he&#39;s on mesa-dev) has been working on this precisely for the switch.=
=C2=A0 There are some issues with some of some parts (IIRC, the Tegra X2 is=
 sketchy) but most of it does work. Mostly, we&#39;re missing some kernel w=
ork to fix issues that Vulkan seems to hammer where GL doesn&#39;t.=C2=A0 Y=
ou can find the Tegra ToDo list here:</div><div><br></div><div><a href=3D"h=
ttps://gitlab.freedesktop.org/mesa/mesa/-/work_items/14105">https://gitlab.=
freedesktop.org/mesa/mesa/-/work_items/14105</a></div><div><br></div><div>~=
Faith</div></div><br><div class=3D"gmail_quote gmail_quote_container"><div =
dir=3D"ltr" class=3D"gmail_attr">On Fri, Jul 24, 2026 at 4:17=E2=80=AFAM Ga=
briele Melluso &lt;<a href=3D"mailto:[email protected]">noski=
[email protected]</a>&gt; wrote:<br></div><blockquote class=3D"gma=
il_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,2=
04,204);padding-left:1ex"><div dir=3D"auto">Hi,<div dir=3D"auto">noticed NV=
K&#39;s docs state Tegra/integrated GPUs aren&#39;t currently supported, bu=
t there&#39;s already a debug option for forcing coherent memory maps speci=
fically on Tegra devices. Has bringing up NVK on Tegra X1/GM20B ever been s=
coped out (e.g. for nouveau-based Linux on Jetson TX1 or Nintendo Switch ho=
mebrew Linux distros)? Since the class-engine version logic already treats =
Maxwell as capable of 1.3, I&#39;m trying to understand what&#39;s actually=
 missing beyond memory coherency, class-engine enumeration for the integrat=
ed part, host1x/syncpoint integration, something else entirely? And is this=
 an area the project would welcome outside contribution on, or is there a r=
eason it&#39;s been deprioritized I should know before looking into it?</di=
v><div dir=3D"auto"><br></div><div dir=3D"auto">Regards,</div><div dir=3D"a=
uto"><br></div><div dir=3D"auto">Gabriel=C2=A0</div></div>
</blockquote></div>

--000000000000057d850657a2c1c6--