.gitignore - ignore __pycache__ and .pyc #3
Conversation
This file contains hidden or 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
Minor developer quality of life PR...
pmolodo
pushed a commit
that referenced
this pull request
Jan 26, 2026
dirty expression prior to removing it from the collection tables.
Reported by the testLightLinkingSceneIndex test under asan, as:
=1343067==ERROR: AddressSanitizer: heap-use-after-free on address 0x60c0009bd598 at pc 0x7f2b478c6b23 bp 0x7ffff1bf6350 sp 0x7ffff1bf6340
READ of size 8 at 0x60c0009bd598 thread T0
#0 0x7f2b478c6b22 in std::vector<SdfPathExpression::Op, std::allocator<SdfPathExpression::Op> >::size() const /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_vector.h:919
#1 0x7f2b478c6b22 in std::vector<SdfPathExpression::Op, std::allocator<SdfPathExpression::Op> >::vector(std::vector<SdfPathExpression::Op, std::allocator<SdfPathExpression::Op> > const&) /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_vector.h:555
#2 0x7f2b478c6b22 in SdfPathExpression::SdfPathExpression(SdfPathExpression const&) /depts/tools/build_archive/dev/builds/2379295/inst/fedora-gcc64-opt-asan/pxr/include/pxr/usd/sdf/pathExpression.h:54
#3 0x7f2b478cf89d in std::pair<SdfPathExpression, std::optional<std::pair<SdfPath, TfToken> > >::pair<std::pair<SdfPath, TfToken> const&, true>(SdfPathExpression const&, std::pair<SdfPath, TfToken> const&) /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_pair.h:337
#4 0x7f2b478cf89d in HdsiLightLinkingSceneIndex_Impl::_Cache::ProcessCollection(SdfPath const&, TfToken const&, SdfPathExpression const&) imaging/hdsi/lightLinkingSceneIndex.cpp:156
(Internal change: 2379722)
pmolodo
pushed a commit
that referenced
this pull request
Jan 26, 2026
pxr/usdImaging. This change introduces a prototype implementation for supporting UsdSkel native instancing in Storm, utilizing vertex shader-based skinning. In Hydra 1.0, when UsdSkel was natively instanced, USD scene composition would deduplicate the scene graph. However, the SkelRootAdapter would de-instance itself and all of its descendants during imaging due to architectural limitations. Hydra 2.0's native instancing model is based on binding signatures, meaning that UsdSkel instances can only share a prototype if they have identical skel binding signatures. In practice, this becomes problematic because a common UsdSkel use case involves overriding skel:animation per instance to apply unique animations - such as in a crowd of 1,000 HumanFemale_Group instances, each playing a different animation clip. This variation causes the NiPrototypePropagatingSceneIndex to treat them as distinct and thus de-instance them. Our proposal addresses this by relocating skel:animation to a primvar before NiInstanceAggregation. This ensures all instances share the same skel binding signature, allowing proper aggregation. The per-instance animation data is then aggregated automatically as an instance primvar on the instancer, which can be accessed downstream with minimal code changes. During the SkeletonResolvingSceneIndex stage, we restore the per-instance skel:animation data and aggregate it into tensored skinningTransforms and blendShapeWeights, without requiring significant schema changes to ResolvedSkeleton. A new "range" attribute is added to handle cases where blend shapes vary in size between animations, ensuring correct reconstruction downstream. In PointsResolvingSceneIndex, we disable the skinning extComputation but reuse its input generation logic. These inputs are now emitted as primvars to feed vertex shaders directly, instead of compute shaders. Note that currently all the tensored per-vertex and per-instance primvars are all concatenated and passed to the vertex shader as constant primvars (see TODO #3 and #4). This is also potentially the reason why the vertex shader is quite a bit slower than the compute shader even though it's a direct port. We have to do two indirect index lookups from primvar inputs and Nsight profiling shows that we're spending 30% of the time waiting for memory fetch. On the Storm side, the existing skinning compute shaders have been ported to shared skinning methods in skinning.glsl, with added instance support (current shared between the two skinnable point-based prims: mesh and basisCurves). This entire feature is gated behind the HD_ENABLE_DEFERRED_SKINNING environment variable or the HdSkinningSettings::IsSkinningDeferred() API call (his also requires HGI_ENABLE_VULKAN=1 for more complex test scenes like HumanFemale_Group). Added imaging test coverage for this implementation. TODO: 1. fix DQS (skinningMethod is getting aggregated currently) 2. testUsdImagingGLUsdSkelInstancing is very slow for deferred skinning in test.usda, the majority of the time is spent in shader compilation, we'll file a separate ticket for that. 3. use existing skel:jointIndices and skel:jointWeights vertex primvars on the mesh instead of influences constant primvar from extComputation. This will require adding tensor value support to vertex primvars. 4. convert skinningTransforms and blendShapeWeights from constant primvars to instance primvars so we don't need the instance indexing logic in the vertex shader. This will require add tensor value support to instance primvars. (Internal change: 2390250)
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.
Description of Change(s)
.gitignore - ignore pycache and .pyc
Minor developer quality of life PR...
Fixes Issue(s)
git status, might accidentally get committed