Expose more information on compiled entry points

Problem

Right now the entry_points in CompileResult contains names only. This limits limits our/the users ability to generate code.

Proposal

Exposing some additional information on them could help with generating code, without prior knowledge of how each entry point works (like having a dictionary of graphics and compute shaders).

Expose SpirV representation

Right now I am thinking of putting a version of the generated EntryPoint spirv lang struct into rustc_codegen_spirv_types because of how easy it is to implement.

pub struct EntryPoint {
    pub execution_model: spirv::ExecutionModel,
    pub entry_point: spirv::Word,
    pub name: String,
    pub interface: Vec<spirv::Word>,
}

Use naga

Another idea I had was using naga::ir::EntryPoint

pub struct EntryPoint {
    pub name: String,
    pub stage: ShaderStage,
    pub early_depth_test: Option<EarlyDepthTest>,
    pub workgroup_size: [u32; 3],
    pub workgroup_size_overrides: Option<[Option<Handle<Expression>>; 3]>,
    pub function: Function,
}

but this is a bit detached from the underlying spirv (harder to parse from spirv code + less shader stages etc)

Unique struct

The best version would probably be a combination of the 2.

Even if it doesn't end up getting merged, I'll do it for my own project. All insights are welcome.

Commencez par CompileResult.entry_points et examinez l’EntryPoint proposé dans rustc_codegen_spirv_types en regard de naga::ir::EntryPoint. Déterminez d’abord quelle représentation et quelles métadonnées le projet devrait exposer ; la tâche est terminée lorsque CompileResult fournit des informations convenues suffisantes sur les points d’entrée pour la génération de code sans connaissance préalable du shader.