Lobe pruning shader optimization #4016
Merged
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
Introduces the lobe pruning concept to MaterialX shaders generated by
either MayaUSD or LookdevX.
We look into shader graph implemnetation and find where we can cut the
graph to simplify it based on the value of the main lobe attributes.
This means a shader with a subsurface weight of zero will completely
prune the evaluation of the subsurface, while a value of 1 will
completely prune the evaluation of the dielectric base.
This provides interesting performance values:
```
MayaUSD: Tumbling speed of a stage with a 11x11x11 set of cubes with a red OpenPBR shader
initial optimized
Default light 434fps 441fps
Dome light 17fps 115fps
LookdevX/MaterialX: Tumbling speed of a heavy geometry scene from MAYA-136105 using OpenPBR surface
initial optimized
Default light 192fps 476fps
Dome light 7.6fps 65fps
```
Technical side: The code looks for nodes in the nodegraph implementation and checks any node that directly connects to the interface on a 0-1 slider:
- Mix nodes: If the mix value if 0 or 1, forward the value/connection of the bg or fg port to the node directly upstream. Delete node.
- Multiply nodes: If any value is zero, write zero to the corresponding port on the upstream node. Delete node.
- PBR nodes: If the weight parameter is zero, substitute with a no-op PBR node. Delete node.
This makes sure the ShaderGen works on a reduced codebase and greatly reduces code size and complexity by disconnecting unused subgraphs and skipping black PBR calls.
| @@ -0,0 +1,606 @@ | |||
| #include "LobePruner.h" | |||
There was a problem hiding this comment.
This class can be copied to USD in HdStorm to allow the same optimizations for usdview or other Storm clients.
Requires the dark closure node as well in the USD shadergen.
frohnej-adsk
approved these changes
Nov 29, 2024
JGamache-autodesk
added
the
ready-for-merge
seando-adsk
reviewed
Dec 3, 2024
| class MAYAUSD_API_PUBLIC TopoNeutralGraph | ||
| { | ||
| public: | ||
| TopoNeutralGraph(const MaterialX::ElementPtr&); | ||
| TopoNeutralGraph(const MaterialX::ElementPtr&, const LobePruner::Ptr& lobePruner); |
There was a problem hiding this comment.
@JGamache-autodesk Can you confirm that adding this overload does not break binary compatibility?
There was a problem hiding this comment.
I confirm it does not break since it is not a virtual function.
| * \return a string vector containing all paths to attibutes that were used to optimize nodes in | ||
| * this graph. | ||
| */ | ||
| const MaterialX::StringVec& getOptimizedAttributes() const; |
There was a problem hiding this comment.
Same here. Not virtual. And since the actual implementation is hidden underneath an Impl pointer, all changes to the core class are unexposed and non-breaking.
seando-adsk
approved these changes
Dec 3, 2024
seando-adsk
deleted the
gamaj/MAYA-136105/optimize_materialx_via_lobe_pruning
branch
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.
Introduces the lobe pruning concept to MaterialX shaders generated by either MayaUSD or LookdevX.
We look into shader graph implemnetation and find where we can cut the graph to simplify it based on the value of the main lobe attributes.
This means a shader with a subsurface weight of zero will completely prune the evaluation of the subsurface, while a value of 1 will completely prune the evaluation of the dielectric base.
This provides interesting performance values:
Technical side: The code looks for nodes in the nodegraph implementation and checks any node that directly connects to the interface on a 0-1 slider:
This makes sure the ShaderGen works on a reduced codebase and greatly reduces code size and complexity by disconnecting unused subgraphs and skipping black PBR calls.