Render WebVR frames to OVR VR surfaces

[daoshengmu]

I put my Gecko work at https://bugzilla.mozilla.org/show_bug.cgi?id=1562134.

[MortimerGoro]

Thanks for the updates! This is getting closer to be ready!

Other than the mentioned issues WebVR samples demo looks zoomed compared to the master build. It seems something related to bad sizes or bad eye offsets. Could you check that?

[daoshengmu]

Thanks for the updates! This is getting closer to be ready!

Other than the mentioned issues WebVR samples demo looks zoomed compared to the master build. It seems something related to bad sizes or bad eye offsets. Could you check that?

ExternalVRSurface

EGLTexture

@MortimerGoro
Regarding to the immersive texture size, I have checked they both are 1024 x 1024, and their difference is not so obvious as my attachments.

Besides, I notice that if I release my surfaces at LeaveVR(), the surfaces will not be presented properly after re-entering the immersive mode. However, if I don't release my surfaces, it will still be good after exiting FxR and re-enter.

[kennardconsulting]

Looking forward to this being merged! Any progress on resolving conflicts?

[daoshengmu]

Looking forward to this being merged! Any progress on resolving conflicts?

I have solved conflicts in local and am working on other new issues after merging with master branch.

[kennardconsulting]

Awesome. Thanks for the update. Can you comment on whether it helps #1408 in the WebVR samples demo app you show above?

Specifically, whether the moving cube (and the stationary cubes if moving your head) appear to judder every few seconds

[daoshengmu]

I am going to give another commit to provide an UI switch in the developer setting to turn on/off external surface feature.

[kennardconsulting]

Awesome. Thanks again for all your hard work on this. I really appreciate it and look forward to seeing the end result.

If you get time, could you please comment on how this feature affects the 'judder' on the moving cubes (and the stationary cubes if moving your head) in the WebVR samples demo on Oculus Quest?

request a new review again

[daoshengmu]

It seems like I am using wrong surface size when adopting Swapchain surfaces to do immersive render at Sketchfab.com

[daoshengmu]

Per #304

[kennardconsulting]

I'm unclear if this issue is now closed?

[daoshengmu]

I'm unclear if this issue is now closed?

Not yet. We are blocked by Bugzilla review. Sorry about that.

[daoshengmu]

Now, doing profiling experiments from webgfx-tests, and run it via webgfx-tests run webvr_cubes -a "Serial Number" -b fxr -c ./examples/tests/webgfx-tests.config.json -o result.json -n 8

We can see we can get a little higher FPS if we adopt this approach (enable). The first time to draw or page loading time would be more than the original one (close) is because we need to ask OVR to resize our external surface buffer from the content side, the only way I can get is asking the content thread sleep for 80 ms to make sure we can get the resized surface.

However, if we choose to use two external surface buffers. The FPS will be low. It seems like the framebuffer memory pressure (we have a couple of screen size framebuffers) causes the FPS goes down. I have checked disabling all extra JNI bindings and WebGLContext::BlitBackbufferToCurDriverFB for this experiment, and we still get the same result. So, I pretty sure this would be our bottleneck.