Thread by @RobertOsfield

I have checked in improvements to #VulkanScneneGraph shader composition support so that now a range of shader variants can be pre-compiled to SPIR-V and built as part of the library itself:

github.com/vsg-dev/Vulkan…

This can mean no more loading and compiling shaders at runtime!

Shader composition might at first seem like another layer of complexity on top of GLSL shaders, but it actually makes shader and C++ code management easier once you understand how to use it.

To explain the improvements I've started a thread on vsg-users:
groups.google.com/g/vsg-users/c/…

The #VulkanSceneGraph uses the #pragma(tic) shader composition approach that I developed as part of the #OpenScenegraph project, allow apps to inject #define into shader compilation

This text.vert shader illustrates use of 3 different #defines:
github.com/vsg-dev/vsgExa…

The #VulkanSceneGraph takes the concept further adding a ShaderSet class that provides introspection interface so that C++ programs can query the supported arrays and uniforms and their layouts:

github.com/vsg-dev/Vulkan…

When constructing a scene graph applications can use the vsg::ShaderSet to guide the arrays, uniforms, textures and optional code path in the shaders they can use and select the appropriate pre-compield SPIR-V shader variant required to achieve the desired rendering.

In the case of Text rendering we need just 4 variants, but to render all the #glTF sample dataset with Physics Based Rendering we need 28 different variants!

Here's the vsgdynamicload example loadin all the .glb's in the sample dataset, all handled by one "uber" PBR ShaderSet: