Annotation with debug_utils example

[timoore]

This example shows how to insert nodes in the scene graph with annotations that can be used by debugging programs like RenderDoc. A screenshot of the RenderDoc display is included.

[robertosfield]

I have merged as a timoore-annotation branch:

https://github.com/vsg-dev/vsgExamples/tree/timoore-annotation

I'll now get things compiling and have a bash at use RenderDoc with it. I did wonder about naming the example vsgrendrdoc if it's mainly about enabling high quality RenderDoc usage.

[robertosfield]

I have got it working with qrenderdoc, found out that ~ doesn't work for paths in RenderDoc which lots me few minutes trying to figure out why things weren't starting up...

I can see the annotation, it's a bit limited though, so I'm thinking that it might be nice to have a visitor that goes through the scene graph and insert annotation nodes. Can this extension handle nesting?

I am also thinking about whether we need the EXT at the end of the function names. I presume at some point Vulkan will move these from EXT.

[timoore]

I have got it working with qrenderdoc, found out that ~ doesn't work for paths in RenderDoc which lots me few minutes trying to figure out why things weren't starting up...

I admit that I've never used an animation path, which I presume you mean?

Renderdoc does define an in-application interface, so an application could make a capture on a specific frame, among other things.

I can see the annotation, it's a bit limited though, so I'm thinking that it might be nice to have a visitor that goes through the scene graph and insert annotation nodes. Can this extension handle nesting?

Yes, my screenshot shows the output from a nested use of annotation nodes. It also shows a limitation; the transparent lenses on the flight helmet end up without any annotation because they are rendered in a bin.

I am also thinking about whether we need the EXT at the end of the function names. I presume at some point Vulkan will move these from EXT.

I don't know.

[robertosfield]

Oooh, just tried:

 vsgannotation models/lz.vsgt models/teapot.vsgt -a

And if I search for the EXT I find:

Thread 0, Frame 0:
vkCmdBeginDebugUtilsLabelEXT(commandBuffer, pLabelInfo) returns void:
    commandBuffer:                  VkCommandBuffer = 0x564788659c10
    pLabelInfo:                     const VkDebugUtilsLabelEXT* = 0x7ffe5cb102b0:
        sType:                          VkStructureType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT (1000128002)
        pNext:                          const void* = NULL
        pLabelName:                     const char* = "Scene"
        color:                          float[4] = 0x7ffe5cb102c8
            color[0]:                       float = 0.8
            color[1]:                       float = 0.8
            color[2]:                       float = 0.8
            color[3]:                       float = 1

Thread 0, Frame 0:
vkCmdBeginDebugUtilsLabelEXT(commandBuffer, pLabelInfo) returns void:
    commandBuffer:                  VkCommandBuffer = 0x564788659c10
    pLabelInfo:                     const VkDebugUtilsLabelEXT* = 0x7ffe5cb10230:
        sType:                          VkStructureType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT (1000128002)
        pNext:                          const void* = NULL
        pLabelName:                     const char* = "/home/robert/Dev/vsgExamples/data/models/lz.vsgt"
        color:                          float[4] = 0x7ffe5cb10248
            color[0]:                       float = 1
            color[1]:                       float = 1
            color[2]:                       float = 1
            color[3]:                       float = 1

Which is cool :-)

[robertosfield]

@timoore I am curious why you added the following definitions as I don't think they are necessary, and if they are the a using should be sufficient to make sure they are still available. Just commenting them out things still work fine for me.

#if 0
        // ParentClass::accept() is hidden by overriding accept(RecordTraversal&), so we need to
        // reimplement the other variations too in order to be able to call them directly.
        void accept(vsg::Visitor& visitor) override
        {
            ParentClass::accept(visitor);
        }

        void accept(vsg::ConstVisitor& visitor) const override
        {
            ParentClass::accept(visitor);
        }
#endif

[robertosfield]

I have got it working with qrenderdoc, found out that ~ doesn't work for paths in RenderDoc which lots me few minutes trying to figure out why things weren't starting up...

I admit that I've never used an animation path, which I presume you mean?

No, it was the command line paths to the files I was trying to load, I copy and pasted command line parameters that worked in bash but I didn't think about the ~ that bash automatically expands. Unfortunately RenderDoc was quietly failing without information when nothing was happening so I looked in all the wrong places to fix things.

Renderdoc does define an in-application interface, so an application could make a capture on a specific frame, among other things.

I have only started tinkering with RenderDoc and am pretty amazed by it.

Yes, my screenshot shows the output from a nested use of annotation nodes. It also shows a limitation; the transparent lenses on the flight helmet end up without any annotation because they are rendered in a bin.

I noticed the the Scene is coming from the annotation too, but didn't realize this at first.

I am also thinking about whether we need the EXT at the end of the function names. I presume at some point Vulkan will move these from EXT.

I don't know.

I will do another code review of the PR to the VSG and make decision. Without EXT is more readable so I'm drawn in that direction.

[robertosfield]

If I'm reading this correct it looks like this extension is part of Vulkan 1.3:

https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_EXT_tooling_info.html

[timoore]

@timoore I am curious why you added the following definitions as I don't think they are necessary, and if they are the a using should be sufficient to make sure they are still available. Just commenting them out things still work fine for me.

#if 0
        // ParentClass::accept() is hidden by overriding accept(RecordTraversal&), so we need to
        // reimplement the other variations too in order to be able to call them directly.
        void accept(vsg::Visitor& visitor) override
        {
            ParentClass::accept(visitor);
        }

        void accept(vsg::ConstVisitor& visitor) const override
        {
            ParentClass::accept(visitor);
        }
#endif

It doesn't make any difference in the example. I put it in there as just as for good pratice. You are write that using Annotation::accept; works well and is probably better style. I was stuck in C++98 land.

[timoore]

If I'm reading this correct it looks like this extension is part of Vulkan 1.3:

https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_EXT_tooling_info.html

The tooling info extension has been promoted, but I don't think that debug_utils has been.

[robertosfield]

On the tooling info page I says "Interacts with VK_EXT_debug_utils == Deprecation State" which makes me wonder if the VK_EXT_debug_utils is being subsumed into something else.

[robertosfield]

I have replaced the extra void accept(..) methods with using: 7f05b1a

Save for the decision about EXT suffix I think we are good to go.

[robertosfield]

Reviewing the code, I find the line:

std::copy(&debugColor.value[0], &debugColor.value[4], &markerInfo.color[0]);

rather more cryptic than I'd like. I have tried a couple variations of help functions but not come up with anything that I'm happy with yet. Perhaps a helper function in include/vsg/maths/color,h? A copy member function of vsg::vec4?

I'm open to suggestions about what might be more expressive than a std::copy() that you have closely read to know for sure it's doing something intended.

This needn't hold up merging this branch though.

[robertosfield]

Forgot to mention, if we want to annotate stuff that goes into the transpent bin then subclassing from vsg::Bin and then assigning this to the View at the appropriate slot would be the way to do it.