Revert "Reland "Move Dawn Ozone Image Representation over to Shared Texture Memory"" [chromium/src : main]
Peter McNeeley (Gerrit)
Peter McNeeley has uploaded the change for review![]( "Open in Gerrit")
Peter McNeeley would like Chromium LUCI CQ, Vasiliy Telezhnikov and Michael Spang to review this change.
Commit message
Revert "Reland "Move Dawn Ozone Image Representation over to Shared Texture Memory""
This reverts commit ccff0a208cb087dc006d881143b47197d63c6a30.
Reason for revert: May be failing chromeos generic mailbox tests
Original change's description:
> Reland "Move Dawn Ozone Image Representation over to Shared Texture Memory"
>
> Attempt #4. Resolving Cast build failures.
>
> This is a reland of commit e36bb082f72c0049cc8a54af29e70e0f756434e1
>
> Original change's description:
> > Move Dawn Ozone Image Representation over to Shared Texture Memory
> >
> > (Attempt #3)
> >
> > TESTED again on chromeos device using webgpu meets.
> >
> > This is done as far as ozone is concerned but the issue with opaque fences still remains. Basically ozone backing works with sync fences
> > and the external_vk_image_backing doesnt not.
> >
> > We are able to make this change only because webGPU for LINUX is not
> > enabled and will not be for the foreseeable future. WebGPU on linux
> > will work with this change if it uses the external_vk_image_backing
> > and associated representations.
> >
> > This is dependent on:
> > https://dawn-review.googlesource.com/c/dawn/+/194440
> >
> > This means that we must use single copy for linux video import as we
> > cannot use zero copy through ozone image backing (again due to
> > the fence sync vs opaque issue)
> >
> > See: 'CanImportNativePixmapToWebGPU' function
> >
> > Bug: 330385376
> > Change-Id: Ifaf2b8de9e9186a9f9e44ab0a74f2774de9035f6
> > Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6173060
> > Reviewed-by: Michael Spang <sp...@chromium.org>
> > Commit-Queue: Peter McNeeley <peterm...@google.com>
> > Reviewed-by: Vasiliy Telezhnikov <vas...@chromium.org>
> > Cr-Commit-Position: refs/heads/main@{#1589395}
>
> Bug: 330385376
> Change-Id: I240b8d2c46b36b38e13f7e0f65044dd48b13e568
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/7603194
> Reviewed-by: Vasiliy Telezhnikov <vas...@chromium.org>
> Reviewed-by: Michael Spang <sp...@chromium.org>
> Commit-Queue: Peter McNeeley <peterm...@google.com>
> Cr-Commit-Position: refs/heads/main@{#1599166}
Change diff
Change information
- M gpu/command_buffer/service/BUILD.gn
- M gpu/command_buffer/service/shared_image/dawn_ozone_image_representation.cc
- M gpu/command_buffer/service/shared_image/dawn_ozone_image_representation.h
- M gpu/command_buffer/service/webgpu_decoder_impl.cc
- M ui/gfx/linux/BUILD.gn
Related details
- Michael Spang
- Vasiliy Telezhnikov
Code-Coverage
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Rubber Stamper (Gerrit)
Rubber Stamper voted Bot-Commit+1![]( "Open in Gerrit")
| Bot-Commit | +1 |
Related details
- Michael Spang
- Vasiliy Telezhnikov
Code-Coverage
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Peter McNeeley (Gerrit)
Peter McNeeley added 1 comment![]( "Open in Gerrit")
I am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
Related details
- Michael Spang
- Vasiliy Telezhnikov
Code-Coverage
Code-Owners
Code-Review
No-Unresolved-Comments
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Vasiliy Telezhnikov (Gerrit)
Vasiliy Telezhnikov added 1 comment![]( "Open in Gerrit")
I am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
From logs it fails [here](https://source.chromium.org/chromium/chromium/src/+/main:third_party/angle/third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp;drc=a5a289f9476fa269f837d7daef8bb12af6e464ca;l=239) and then crashes [here](https://source.chromium.org/chromium/chromium/src/+/main:gpu/command_buffer/tests/webgpu_mailbox_unittest.cc;drc=cad5903484d19fe15c828ee76c0a507ddc42184e;l=1492) because adapter is null.
Related details
- Michael Spang
- Peter McNeeley
Peter McNeeley (Gerrit)
Peter McNeeley added 1 comment![]( "Open in Gerrit")
Vasiliy TelezhnikovI am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
From logs it fails [here](https://source.chromium.org/chromium/chromium/src/+/main:third_party/angle/third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp;drc=a5a289f9476fa269f837d7daef8bb12af6e464ca;l=239) and then crashes [here](https://source.chromium.org/chromium/chromium/src/+/main:gpu/command_buffer/tests/webgpu_mailbox_unittest.cc;drc=cad5903484d19fe15c828ee76c0a507ddc42184e;l=1492) because adapter is null.
Yes but I have yet to draw the connection to my change.
Also I cannot seem to replicate these failures.
Related details
- Michael Spang
- Vasiliy Telezhnikov
Vasiliy Telezhnikov (Gerrit)
Vasiliy Telezhnikov added 1 comment![]( "Open in Gerrit")
Vasiliy TelezhnikovI am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
Peter McNeeleyFrom logs it fails [here](https://source.chromium.org/chromium/chromium/src/+/main:third_party/angle/third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp;drc=a5a289f9476fa269f837d7daef8bb12af6e464ca;l=239) and then crashes [here](https://source.chromium.org/chromium/chromium/src/+/main:gpu/command_buffer/tests/webgpu_mailbox_unittest.cc;drc=cad5903484d19fe15c828ee76c0a507ddc42184e;l=1492) because adapter is null.
Yes but I have yet to draw the connection to my change.
Also I cannot seem to replicate these failures.
The only change that might have effect is in `gpu/command_buffer/service/webgpu_decoder_impl.cc`, rest doesn't run in crashing tests.
I guess it now fails to create adapter if we don't have `SharedFenceSyncFD`, which we probably don't on VM?
Related details
- Michael Spang
- Peter McNeeley
Peter McNeeley (Gerrit)
Peter McNeeley added 1 comment![]( "Open in Gerrit")
Vasiliy TelezhnikovI am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
Peter McNeeleyFrom logs it fails [here](https://source.chromium.org/chromium/chromium/src/+/main:third_party/angle/third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp;drc=a5a289f9476fa269f837d7daef8bb12af6e464ca;l=239) and then crashes [here](https://source.chromium.org/chromium/chromium/src/+/main:gpu/command_buffer/tests/webgpu_mailbox_unittest.cc;drc=cad5903484d19fe15c828ee76c0a507ddc42184e;l=1492) because adapter is null.
Vasiliy TelezhnikovYes but I have yet to draw the connection to my change.
Also I cannot seem to replicate these failures.
The only change that might have effect is in `gpu/command_buffer/service/webgpu_decoder_impl.cc`, rest doesn't run in crashing tests.
I guess it now fails to create adapter if we don't have `SharedFenceSyncFD`, which we probably don't on VM?
Right this was the original path I was thinking. But the log you mentioned also seems critical. I also noticed a change in vulkan-deps. Is it possible that that cause a change which results in that early log failure?
Since this is all 'vm' I should be able to repo locally on my cloudtop.
I also in all honesty dont know why the SyncFD wouldnt work even if virtualized.
Related details
- Michael Spang
- Vasiliy Telezhnikov
Vasiliy Telezhnikov (Gerrit)
Vasiliy Telezhnikov added 1 comment![]( "Open in Gerrit")
Vasiliy TelezhnikovI am just testing the mailbox failures that exist on the farm right now
https://ci.chromium.org/ui/p/chromium/builders/ci/ChromeOS%20FYI%20Release%20(amd64-generic)/17782/overview
Peter McNeeleyFrom logs it fails [here](https://source.chromium.org/chromium/chromium/src/+/main:third_party/angle/third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp;drc=a5a289f9476fa269f837d7daef8bb12af6e464ca;l=239) and then crashes [here](https://source.chromium.org/chromium/chromium/src/+/main:gpu/command_buffer/tests/webgpu_mailbox_unittest.cc;drc=cad5903484d19fe15c828ee76c0a507ddc42184e;l=1492) because adapter is null.
Vasiliy TelezhnikovYes but I have yet to draw the connection to my change.
Also I cannot seem to replicate these failures.
Peter McNeeleyThe only change that might have effect is in `gpu/command_buffer/service/webgpu_decoder_impl.cc`, rest doesn't run in crashing tests.
I guess it now fails to create adapter if we don't have `SharedFenceSyncFD`, which we probably don't on VM?
Right this was the original path I was thinking. But the log you mentioned also seems critical. I also noticed a change in vulkan-deps. Is it possible that that cause a change which results in that early log failure?
Since this is all 'vm' I should be able to repo locally on my cloudtop.
I also in all honesty dont know why the SyncFD wouldnt work even if virtualized.
The log is present in a passing run too:
```
Warning: Vulkan shaderUniform*ArrayDynamicIndexing required.
- While initializing adapter (backend=BackendType::Vulkan)
at InitializeImpl (third_party/dawn/src/dawn/native/vulkan/PhysicalDeviceVk.cpp:240)
gpu/command_buffer/tests/webgpu_mailbox_unittest.cc:1494: Skipped
info.backendType != wgpu::BackendType::D3D12
```
But it is able to get adapter (maybe GL compat?) and skip the test, as the test is windows only anyway.
Related details
- Michael Spang
- Peter McNeeley