Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refine dot node elision and remove unused code #1522

Conversation

JGamache-autodesk
Copy link
Contributor

No description provided.

{
bypass(context, node, 0);
++numEdits;
}
}
else if (node->hasClassification(ShaderNode::Classification::IFELSE))
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This used to target a node called "compare" which was removed in 1.37.

++numEdits;
}
}
else if (node->hasClassification(ShaderNode::Classification::SWITCH))
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This worked in 1.37 where the "which" input was a "parameter" instead of an "input". Since this input is not marked as "uniform", it should not be elided.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this should be restored as it checks to see if indeed a uniform unconnected value. I have log an new issue and can discuss there, but now new shaders will always add more code in. (#1526)

ShaderInput* in = node->getInput("in");
if (in->getChannels().empty())
if (in->getChannels().empty() && in->getType() == Type::FILENAME)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Still solves the issue with extra samplers, but preserves the distinction between "dot" nodes and "constant" nodes.

@jstone-lucasfilm
Copy link
Member

This looks very promising, thanks @JGamache-autodesk. Would it be reasonable to consider this proposal for after today's release of MaterialX 1.38.8, so that we have sufficient time to think through all of the potential edge cases?

@JGamache-autodesk
Copy link
Contributor Author

Hi @jstone-lucasfilm. I would consider the automatic elision of dot nodes regardless of type a showstopper. I can provide a PR with only a one line fix to restrict to filenames if that can help acceptance.

@jstone-lucasfilm
Copy link
Member

Don't worry about simplifying the changelist, @JGamache-autodesk, and let's start testing the proposal as it currently stands. Thoughts from @kwokcb would be appreciated as well!

@JGamache-autodesk
Copy link
Contributor Author

Dot nodes were elided in #1152 to prevent creating extra samplers in rasterized shadergen.
This preserves the change written by @kwokcb while keeping the distinction between a topological constant node and a non-topological dot node.
In scenarios where you want to feed a single color to multiple shader nodes in a complex material, using a constant node will force recompilation every time the color is tweaked, we want users to be able to use a dot node to feed the graph, which should allow tweaking the color without having to recompile the shader. Also allows re-using the shader code for material instances differing on this input color.

@jstone-lucasfilm jstone-lucasfilm changed the title Refined the dot elision code. Removed dead code. Refine dot node elision and remove unused code Sep 8, 2023
<tiledimage name="image_color" type="color3">
<input name="file" type="filename" value="wood_color.jpg" colorspace="srgb_texture" />
<input name="file" type="filename" nodename="color_filename" colorspace="srgb_texture" />
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sufficient to trigger the issue fixed in #1152. This will not generate valid GLSL code unless dot elision is active for filenames.

Copy link
Member

@jstone-lucasfilm jstone-lucasfilm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the fix, @JGamache-autodesk! Once all tests have passed, I'll move forward and include this in the release.

@jstone-lucasfilm jstone-lucasfilm merged commit f69e4c2 into AcademySoftwareFoundation:main Sep 8, 2023
29 checks passed
@JGamache-autodesk JGamache-autodesk deleted the gamaj/refine_dot_elision branch September 8, 2023 19:18
@kwokcb
Copy link
Contributor

kwokcb commented Sep 8, 2023

Just a comment -- after the fact now since this is merged.

Since it is possible to not perform an upgrade on a document, Thus, I'm unsure what it means to remove support for older nodes. if someone loads in a document and tries to code-gen. Could be things will not work at all, but just wanted to add this note.

{
newNode->_classification = Classification::TEXTURE | Classification::CONDITIONAL | Classification::IFELSE;
}
else if (nodeDef.getNodeString() == SWITCH)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure way it means to remove the texture classification here. I think nodes like try to perform filtered sample upstream will no longer work. But it could be that it does not work anyways.

JGamache-autodesk added a commit to Autodesk/maya-usd that referenced this pull request Sep 11, 2023
The MaterialX topological node list was in need of an update:

For MaterialX 1.38.7:

"dot" nodes are now topological.
    Via AcademySoftwareFoundation/MaterialX#1152

For upcoming MaterialX 1.38.8:

"dot" node no longer elided (except for filenames)
"switch" node no longer elided
    Via AcademySoftwareFoundation/MaterialX#1522
Michaelredaa pushed a commit to Michaelredaa/MaterialX that referenced this pull request Oct 21, 2023
…tion#1522)

Dot nodes were elided in AcademySoftwareFoundation#1152 to prevent creating extra samplers in rasterized shadergen.

This preserves the change written by @kwokcb while keeping the distinction between a topological constant node and a non-topological dot node.

In scenarios where you want to feed a single color to multiple shader nodes in a complex material, using a constant node will force recompilation every time the color is tweaked, we want users to be able to use a dot node to feed the graph, which should allow tweaking the color without having to recompile the shader. Also allows re-using the shader code for material instances differing on this input color.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants