Forced the test to wait for the GPU after rendering #12
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Not waiting was causing some bad behavior in the array and wide swapchain tests on D3D12.
|
Does this need discussion on the call? |
|
Opinions about whether to discuss this one differ, which I think means yes. |
rpavlik
approved these changes
Jan 14, 2021
rpavlik
added a commit
that referenced
this pull request
Jan 28, 2021
Note that the procedure for generating your conformance submission has
changed slightly.
- Registry
- All changes found in 1.0.13, and 1.0.14.
- Conformance Tests
- Build: Initial setup of CMake for conformance build on Android.
(Not complete - no gradle part.) (internal MR 1910)
- Fix: Properly apply function attributes to fix build of
conformance layer on Android for ARM. (internal MR 1910,
OpenXR-CTS/#3, internal issue 1479)
- Fix: Use android.app.NativeActivity correctly in place of
NativeActivity subclass for the conformance tests. (internal MR
1965, internal MR 1976)
- Fix: The D3D12, OpenGL, and Vulkan graphics plugins sometimes
did not update their swapchain image context maps due to rare
key collisions. (OpenXR-CTS/#4)
- Fix: Removed extra check that would fail if the CTS was compiled
against a version of openxr.h that included extensions that
added results that were newer than the results present in the
version of openxr.h that the runtime was compiled against.
(OpenXR-CTS/#8)
- Fix: Fixed conformance failures on runtimes where x and y
components can have different last changed times (OpenXR-CTS/#9)
- Fix: “Projection Mutable Field-of-View” was an old, broken
version of the test. It is reset to the intended version of the
code now. (OpenXR-CTS/#10)
- Fix: CopyRGBAImage was using the wrong array slice when setting
image barriers. This broke the “Subimage Tests” case on some
hardware/drivers. (OpenXR-CTS/#11)
- Fix: Added WaitForGpu call at the end of RenderView in
D3D12GraphicsPlugin. Without this the array and wide swapchain
tests failed on some hardware/driver versions. This is not ideal
behavior, but it fixes the test for now, and has been noted for
future fixing in a better way. (OpenXR-CTS/#12)
- Fix: Allow negated quaternion to be equivalent to the expected
orientation value. This test only cares about orientation, not
which path the rotation took. (OpenXR-CTS/#13)
- Fix: The test assumed that X and Y components of a vector2
action would have exactly the same timestamp. Changed that to
check that the vector2 action would have the most recent of
those two timestamps instead. (OpenXR-CTS/#14, internal issue
1490)
- Fix: The test checked for “float value == X” and then checked
for its binary thresholded value after a subsequent call to
xrSyncActions. Changed the test so both checks happen in the
same input frame. (OpenXR-CTS/#14)
- Fix: Added support for GL_SRGB8 textures to OpenGL tests.
(OpenXR-CTS/#15)
- Fix: Increased allowed pipeline overhead threshold to 50%. The
purpose of the test is to ensure the runtime isn’t serializing
(100% overhead) so this is a safe increase. (OpenXR-CTS/#16)
- Fix: Use int64_t for YieldSleep calculations, to not overflow.
(OpenXR-CTS/#17)
- Improvement: Modify the test instructions to change the output
format and increase its content, and require the console output
to be submitted as well. This makes it easier to review and
provides a more complete look at both passed and failed tests.
(internal MR 1953)
- Lower the amount of time that the renderer blocks. The CTS is
not a highly optimized application and due to thread scheduling
and extra GPU waits 90% CPU wait makes it fairly tight to fit
everything inside of a single display period. (OpenXR-CTS/#18)
- New test: Verify that triangles returned from
XR_KHR_visibility_mask are counter-clockwise. (internal MR 1943)
- New test: Add tests for xrBeginFrame call order violations
(OpenXR-CTS/#7)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Not waiting was causing some bad behavior in the array and wide swapchain tests on D3D12. I think this was caused by the textures being destroyed and recreated on subsequent views, but I'm not sure.
This is actually @brycehutchings 's fix. I'm just making a PR for it.