[Bug 50542] New: get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

Bug ID: 50542
           Summary: get_gpu_properties_from_vulkan() assumes a 1-1 mapping
                    of XRRProvider to VkPhysicalDevice
           Product: Wine
           Version: unspecified
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: winex11.drv
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

The function currently takes in a single XRRProvider, and will stop execution
after it finds the first eligible device. This causes an issue for setups where
multiple Vulkan ICDs may expose devices which have valid RROutputs on a single
XRRProvider.

I believe a minimal test-case for this would be the LUID part of the vulkan-1
test already in the WINE tree, as it enumerates LUID information for every
VkPhysicalDevice it can find.

See link to mesa bug for more information:
https://gitlab.freedesktop.org/mesa/mesa/-/issues/4112#note_774347

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

Zhiyi Zhang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

--- Comment #1 from Zhiyi Zhang  ---
I am not sure what should be done in Wine for this bug. Wait for the device
selection layer like
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8551 to land? We
could also sort the Vulkan devices in Wine.

This is probably the same bug as 49764.

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

--- Comment #2 from Zhiyi Zhang  ---
Created attachment 69242
  --> https://bugs.winehq.org/attachment.cgi?id=69242
patch

Similar to the device selection layer patch proposed
at https://gitlab.freedesktop.org/mesa/mesa/-/issues/4112#note_774347

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

--- Comment #3 from Liam Middlebrook  ---
I think that logic makes sense. Although maybe this is a staging-level patch?

A Vulkan Device selection layer should be affecting this section of code in
winex11.drv also, not just the application+winevulkan. If users aren't seeing
any changes from enabling it, I'd think there may be something in the newer
Proton versions which make use of a container-based runtime that just isn't
importing the device selection layer properly.

In upstream I think it's important to respect what order the VkPhysicalDevice
entries are being supplied to us in.

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

Zhiyi Zhang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |DUPLICATE
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #4 from Zhiyi Zhang  ---
Marking this as a duplicate of 49764

*** This bug has been marked as a duplicate of bug 49764 ***

[Bug 50542] get_gpu_properties_from_vulkan() assumes a 1-1 mapping of XRRProvider to VkPhysicalDevice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=50542

Alistair Leslie-Hughes  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #5 from Alistair Leslie-Hughes  ---
Closing Duplicate.