[pull] dev from PixarAnimationStudios:dev #5
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
![pull[bot]](https://avatars.githubusercontent.com/in/12910?s=60&v=4){kind=small}
…in a temporary location instead of in the same directory as the original file. This allows users to run usdedit --noeffect in read-only directories. (Internal change: 2355902)
The cause of the crash in XPU is because:
CreateFramebufferAndRenderViewFromAovs() ends up calling
Riley::CreateDisplay().
This CreateDisplay() to the d_qn then goes to DspyImageOpen
where in we try to acquire the same lock that
CreateFramebufferAndRenderViewFromAovs() is already holding on to.
This only happens in XPU because Dspy calls in RIS are deferred.
In RIS, Dspy calls happen in the Render() call.
The fix here is decouple the lock's scope
in CreateFramebufferAndRenderViewFromAovs,
so that modifying member variables of that HdPrmanFramebuffer
does not also lock the CreateDisplay() call.
Also discussed the addition of the lock around buffer->Resize() in XPU's
Display API i.e., Rebind().
(Internal change: 2355916)
…ne Index std::optional In Hydra 2.0 the render settings path from the stage's metadata is stored on the UsdImaging Stage Scene Index. Then in the the Scene Globals Scene Index this was overwitten by the value set via that Scene Indices SetActiveRenderSettingsPrimPath(). This means that if this function was not called, the value from the metadata will be overwritten by an empty SdfPath. This changes the render settings path stored by the Scene Globals Scene Index to be std::optional so that we make sure to fallback to the upstream scene index's opion. Fixes #3367 (Internal change: 2355965)
Instead of crashing, silently skip the terminal. This allows an easy way for a scene index to disonnect e.g. displacement, by overriding material.terminals.displacement.upstreamNodePath. (Internal change: 2355968)
Add a second env setting, PCP_ENABLE_PARALLEL_NON_USD_LAYER_PREFETCH, that controls whether parallel layer prefetch is enabled for non-USD layers. Note that PCP_ENABLE_PARALLEL_LAYER_PREFETCH only affects whether parallel prefetch is enabled in USD mode, even though its name suggests a global effect. (Internal change: 2355978)
There was an errant availability macro that was causing the path to be taken on only non-macOS platforms. This caused issues where the background would retain the previous background, causing a Solitaire win like effect. This patch makes the color pass use the same pattern as the depth pass, and adequately clear the background. I believe this resolves to the same LoadAction, but something about the availability macro was causing an issue here. Either way, at least it cleans it up. Tested on iOS 18.3 Closes #3505 (Internal change: 2355986)
…ed by Solaris; use renderProgressAnnotation to print the render resolution in the viewport, and keep track of the totalClockTime of the render. (Internal change: 2355996)
This fixes the warnings these tests have been emitting for a long time, and gets them passing again... (Internal change: 2356024)
Production denoisers generally require configuration attributes that are best provided by an applied API schema. It makes more sense to provide all denoise controls as part of such a schema, specific to a particular technique, rather than an implied universal denoise ability on the UsdRenderPass that is enabled solely with a boolean. This changelist comprises an edit to schema.usda files along with associated code-gen output. (Internal change: 2356065)
Adds the env setting HD_PRMAN_DISABLE_ADAPTIVE_SAMPLING. When this is set, hdPrman forces Ri:PixelVariance to 0, disabling adaptive sampling. Intended for use in testing scenarios only. (Internal change: 2356126)
UsdShade semantics for material binding resolution factors the origin path of the binding in addition to its strength, so this information needs to be transported. (Internal change: 2356146)
Closes #3522 (Internal change: 2356158)
…path. This was missed in the previous checkin. (Internal change: 2356173)
(Internal change: 2356186)
(Internal change: 2356187)
(Internal change: 2356188)
(Internal change: 2356189)
…no longer automatically sets fixedSampleCount on lights. Use the environment setting HD_PRMAN_ALL_LIGHTS_FIXED_SAMPLE_COUNT instead. (Internal change: 2356193)
… effects of jitter and variable count light sampling. Perceptual diff is removed. (Internal change: 2356194)
(Internal change: 2356203)
…GetNodeDiscoveryResults and GetShaderProperties in favor of GetDiscoveryResults and GetProperties, respectively (Internal change: 2356231)
(Internal change: 2356232)
…g VFX2023. I lean towards the reported error (array bounds) being a compiler bug for a few reasons. 1) It appeared after upgrading to VFX 2023. 2) As far as I can tell, the error only happens for the first for-loop over highlight modes and not the second, even though they use the same for-loop conditions, array indices, and syntax. 3) I haven't been able to recreate the error in the compiler explorer (although it doesn't currently support the exact setup from what I can tell i.e. gcc 11.2.1/vfx2023). Simple workarounds like unrolling for-loops or replacing the problematic for-loop entirely with hard-coded statements also fixed the issue. I'm guessing the size_t typing is allowing the compiler to handle bounds checking somewhat differently and avoid the spurious error. One theory is that since the first for-loop is setting all the elements to the same value unconditionally while the second is not, the compiler might be optimizing the first for-loop into a memset (__builtin_memset) that isn't working correctly with the 32-bit index values. (Internal change: 2356294)
CMake 3.30 removed support for FindBoost, so we switched to using the BoostConfig.cmake file. This change is backwards compatible down to Boost 1.70 and much older CMake versions.
Apply the following changes:
- extraArgs.append('-DBoost_NO_BOOST_CMAKE=OFF')
- extraArgs.append('-DBoost_NO_SYSTEM_PATHS=ON')
Closes #3485
(Internal change: 2356335)
Now that we have moved to boost version (1.76+) which has stabilize boost cmake config workflow, we can safely take out Boost_NO_BOOST_CMAKE from OpenUSD build configurations. This is a followup change for 2356335 (Internal change: 2356337)
(Internal change: 2356388)
…omment. (Internal change: 2356389)
Fixes failure to compile with GCC 15 Closes #3487 (Internal change: 2356412)
Baking should be backing (I believe). Can be confusing as baking is used when talking about pre-rendered shadows. Closes #3465 (Internal change: 2356415)
….0 prim adapters. (Internal change: 2356416)
(Internal change: 2358508)
…lders -- changes Tf_HasAttribute to remove the reparse point attribute on a directory if TfReadLink returns the original path or an empty string (i.e. we couldn't follow the link). This means TfIsLink returns false in these cases, preventing recursive calls to TfReadLink because we always call TfIsLink before calling TfReadLink. Closes #3415 (Internal change: 2358571)
Addressing SkelBinding inheritance and adding a data source to indicate whether a prim is under a skel root. This required adding UsdImagingSceneIndexPlugin::FlattenedDataSourceProviders so that usdSkelImaging can customize the flattening behavior. (Internal change: 2358619)
…y GfCamera::[...]UNIT. Divided we fail. (Internal change: 2358620)
…ransforms. (Internal change: 2358649)
(Internal change: 2358687)
…binding resolution. - Fix bug in the adapter to return a null data source only we don't have any direct or collection bindings for that purpose. - Update scene index to implement the material binding resolution semantics of UsdShadeMaterialBindingAPI. - Add unit test. (Internal change: 2358714)
target. Fixes #2844 (Internal change: 2358718)
…ing variants. (Internal change: 2358721)
build_usd.py would skip installing zlib if the script were run on Linux or if the user explicitly requested it via "--no-zlib" at the command line. This was broken in the case where more that one dependency required zlib (e.g. Alembic/HDF5 and OpenImageIO). Closes #3551 (Internal change: 2358725)
This change removes build support for using boost::python instead of the internal pxr_boost::python library for Python bindings. Follow-up changes will clean up remnants of this support in the rest of the codebase. (Internal change: 2358757)
(Internal change: 2358773) (Internal change: 2358781) (Internal change: 2358783)
…ng. Now any Vulkan validation errors reported by the test will cause the test to fail. A couple of usdImagingGL tests were purposefully creating errors, so add functionality to handle that, in the form of test argument -numErrorsAllowed. (Internal change: 2358795)
Renames the existing _assetPath member and meaning of first parameter to the constructor to now represent the raw authored path in the layerr. Added explicit getters for this authored path. It is important to note that when authoring scene description, this value will be used for serilization, all other fields will be ignored. Adds a new evaluatedPath field to SdfAssetPath. This field is meant to hold the original asset path as authored in the layer with any variable expressions evaluated. This field is only populated if passed in at construction time. Changes the behavior of SdfAssetPath::GetAssetPath to return the evaulated path if present, otherwise returns the raw authored path. This is to preserve compatibility with existing callers who were relying on this method to return them the unresolved asset path containing evaulated expression variables. Additionally, _MakeResolvedAssetPathsImpl in UsdStage has been augmented to provide the unresolved, evaluated path when creating SdfAssetPath objects. This function services calls to UsdAttribute::Get(). (Internal change: 2358824)
"bare predicates" following stretch, like `//{pred}...`.
For example, the pattern `/foo//{pred}/bar` should match the path
`/foo/bar` if `/foo` passes {pred}. The way that matching proceeds is
to split patterns into components by //, and match the components. The
previous code assumed that components could not overlap, which is true
*except* in just this one case. So now for segments that start with
bare predicates (i.e. //{pred}/...) we look for matches that overlap by
one with the prior component.
(Internal change: 2358827)
Closes #3205 (Internal change: 2358834)
`mx_microfacet.glsl` uses functions from `mx_math.metal`, so the latter must be included first. Closes #3519 (Internal change: 2358835)
…ving relocates (Internal change: 2358836)
…nvolve conflicts from sibling composition arcs. (Internal change: 2358863)
…ing inherits and specializes (Internal change: 2358864)
…kel bindings. (Internal change: 2358926)
… single bone. (Internal change: 2358927)
Properly procesing relationship targets on USD native instances that point to namespace descendants of the native instance. Semantically, these point to a prim inside the respective prototype. This requires exposing the rerooting data source from the UsdImaging_NiInstanceAggregationSceneIndex. (Internal change: 2358959)
a new module. It generates the CMakeLists.txt, __init__.py, module.cpp, schema.usda, and schemaUserDoc.usda. (Internal change: 2358968)
- fixes compile issue due to recent function Shader Generator signature changes and the addition of TypeSystem (Internal change: 2359028)
…schema locally. The definition is tied to and only used by the blend shape, so moving it out out schemaTypeDefs.h which can now be deleted. (Internal change: 2359093)
Now that build support for using boost::python has been removed, we can clean up the headers in pxr_boost::python. This change removes all boost::python includes that had previously been used when the PXR_USE_INTERNAL_BOOST_PYTHON preprocessor symbol was not defined. (Internal change: 2359145)
AcademySoftwareFoundation/openvdb#1977 fixed a syntax issue that caused errors when building with Xcode 16.3. This change updates build_usd.py to apply the (simple) patch to the version of OpenVDB it builds. We manually back-port this fix for a few reasons: - The PR was merged in December 2024, but is currently not part of an official release. - build_usd.py currently uses OpenVDB 9.1.0, which is several major versions behind the current latest version 12.0.0. Even if the PR were released in a new 12.x version, we wouldn't want to jump all the way there since it's so far ahead of what we're building and testing against internally. If OpenVDB officially ports the fix back to older versions, we can revert this change and just update build_usd.py to pull in the official release instead. (Internal change: 2359154)
brought in from a header, rather than implicitly used, as it allows delineation between different declarations of size_t. This change includes cstddef explicitly in gf. Closes #3535 (Internal change: 2359159)
…Aggregation Scene Index and including case addressed by change list 2358959. (Internal change: 2359253)
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.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.1)
Can you help keep this open source service alive? 💖 Please sponsor : )