[Mesa-dev] [Mesa-stable] i915: Enable EGL_KHR_{fence, wait}_sync

[Mauro Rossi]

2016-07-15 21:18 GMT+02:00 Emil Velikov <emil.l.velikov at gmail.com>:

> Hi Mauro,
>
> On 14 July 2016 at 04:33, Mauro Rossi <issor.oruam at gmail.com> wrote:
> > Sending patches to implement DRI2_Fence extension for i915,
> > to enable EGL_KHR_fence_sync and EGL_KHR_wait_sync.
> >
> > It is the step-by-step porting to i915 of i965/sync,
> > plus patch to support {server,client}_wait_sync without bound context.
> >
> > [PATCH 1/3] i915/sync: Replace prefix 'intel_sync' -> 'intel_gl_sync'
> > [PATCH 2/3] i915/sync: Implement DRI2_Fence extension
> These two are
> Acked-by: Emil Velikov <emil.l.velikov at gmail.com>
>
> > [PATCH 3/3] i915: store reference to the context within struct
> >
> And this one needs a re-spin to include locking, as done my Tomasz.
>

I'm already doing the re-spin, I will submit v2 by step by step replication
of Tomasz patch for i915.


> > Motivation is to support (web)chromium which requires
> EGL_KHR_{fence,wait}_sync
> > Tested with marshmallow-x86 build on Asus Eee PC 1015PEM
> >
> Did you perform any tests as stated in the i965 "Implement DRI2_Fence
> extension" patch ? Afaict the new functionality is nicely separated so
> there's little chance for regressions but giving it greater testing
> would be appreciated.
>

Tests were performed by Paulo and they consisted in checking chromium
browser that was crashing without those features.

I will test on intel G33 in the weekend and  perform Android CTS deqp EGL
and report back.


>
> > Spotted by Paulo Travaglia <pstglia at gmail.com>
> >
> > Cc: "12.0" <mesa-stable at lists.freedesktop.org>
> I'm not sold that these are stable material - they add feature and
> don't fix bug(s). If anything Android could use some updates to honour
> the lack of the EGL (EGL_KHR_fence_sync and EGL_KHR_wait_sync)
> extensions.
>

Thanks, I understand your point about additional features for mesa-stable.

It depends if we consider that these are step-by-step cloned from i965 ones,
with the only difference that 'brw' was renamed as 'intel' as in all other
similar functions in i915

Also these capabilities are independent from underlying HW, if I understood
correctly



> That said if the Intel devs are OK with having these in stable I won't
> object too much ;-)
>
> Thanks
> Emil
>

Merging in mesa-dev will be ok, If patches not accepted for mesa-stable it
is not a problem,
because they can be used as they are in android-x86 marshmallow-x86 branch.

M.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160715/9c582d12/attachment-0001.html>