1552590 - [Wayland] Implement basic Wayland dmabuf surface support

Status: RESOLVED FIXED

(Core :: Graphics, enhancement, P2)

Description

[Martin Stránský [:stransky] (ni? me)]

Wayland dmabuf surface are located in GPU and can be attached as EGLImage or wl_buffer. It allows direct rendering to GPU and share the HW buffer across processes.

Comment 1

[Martin Stránský [:stransky] (ni? me)]

Attachment: WIP patch

WIP patch

Comment 2

[Francois Guerraz]

For me build fails with

15:31.86 mozilla-unified/obj-x86_64-pc-linux-gnu/dist/include/mozilla/widget/nsWaylandDisplay.h:19:12: fatal error: 'drm/drm_fourcc.h' file not found
15:31.86 #  include <drm/drm_fourcc.h>
15:31.86            ^~~~~~~~~~~~~~~~~~

Same for mozilla-unified/gfx/2d/WaylandSurface.cpp

On Arch Linux.

Replacing drm with libdrm fixes the compilation issue (result not tested yet though).

Comment 3

[Martin Stránský [:stransky] (ni? me)]

(In reply to Francois Guerraz from comment #2)

For me build fails with

mozilla-unified/gfx/2d/WaylandSurface.cpp:10:
15:31.86
mozilla-unified/obj-x86_64-pc-linux-gnu/dist/include/mozilla/widget/
nsWaylandDisplay.h:19:12: fatal error: 'drm/drm_fourcc.h' file not found
15:31.86 #  include <drm/drm_fourcc.h>
15:31.86            ^~~~~~~~~~~~~~~~~~

It's really wip and don't expect anything from it right now.

Comment 4

[Francois Guerraz]

always happy to test new cool features / optimizations :-)

I've been running it for a few hours, I don't know if the code path is used anywhere yet, but I didn't notice any problem. Let us know when there is something to expect from it.

Comment 5

[Martin Stránský [:stransky] (ni? me)]

Attachment: WIP v.2

Use DMAbuf as a Wayland backend. Not finished/polished yet.

Comment 6

[Francois Guerraz]

Not sure this is useful feedback because you probably know about it already. It's working but it's really really slow and I get a lot of

Hard copy 3200 x 1746 -> 22348800```

Comment 7

[Martin Stránský [:stransky] (ni? me)]

Attachment: patch

Working patch, uses DMABuf as Wayland backend if gfx.wayland_dmabuf_backend.enabled is set to true.

Comment 8

[Francois Guerraz]

I have not tried the individual patches, but the latest unified one with gfx.wayland_dmabuf_backend.enabled and no odd behaviour so far, slowdown is gone.

Comment 9

[Francois Guerraz]

I don't know if it's related, but I got this: bp-17143081-1ae6-4a48-8f2f-d79820190528 ... probably not super helpful backtrace.

Comment 10

[Francois Guerraz]

When the feature is enabled, is there any visibility in about:support ? How do we know it's working/enabled?

Comment 11

[Martin Stránský [:stransky] (ni? me)]

(In reply to Francois Guerraz from comment #10)

When the feature is enabled, is there any visibility in about:support ? How
do we know it's working/enabled?

No, there's no evidence is enabled right now :) We should address that.

Comment 12

[Martin Stránský [:stransky] (ni? me)]

(In reply to Francois Guerraz from comment #10)

When the feature is enabled, is there any visibility in about:support ? How
do we know it's working/enabled?

Filed as Bug 1561859.

Comment 13

[Robert Mader [:rmader]]

Is this also required (or beneficial) for hardware video decoding (bug 1210726)? In that case, should we mark it as blocker?

Comment 14

[mthw0]

Could this help with terrible performance in situations like fishbowl benchmark? (W10: FF and Chrome ~60fps, Linux: Chrome 60fps, FF 7fps)

Comment 15

[Martin Stránský [:stransky] (ni? me)]

(In reply to mthw0 from comment #14)

Could this help with terrible performance in situations like fishbowl benchmark? (W10: FF and Chrome ~60fps, Linux: Chrome 60fps, FF 7fps)

This is Bug 1572697.
This one can be closed as basic dmabuf support landed.