feat: draw meshes with instanced multi-draw indirect

destrum/assets_src/shaders/mesh.vert

@@ -8,6 +8,7 @@ layout (location = 0) out vec3 outPos;
 layout (location = 1) out vec2 outUV;
 layout (location = 2) out vec3 outNormal;
 layout (location = 3) out vec4 outTangent;
+layout (location = 4) flat out uint outMaterialID;
 
 // The depth pre-pass reuses this shader, positions have to match bit for bit
 // so the main pass can depth test against it with LESS_OR_EQUAL
@@ -15,9 +16,11 @@ invariant gl_Position;
 
 void main()
 {
-    Vertex v = pcs.vertexBuffer.vertices[gl_VertexIndex];
+    // gl_InstanceIndex already includes the indirect command's firstInstance
+    InstanceData instance = pcs.instanceBuffer.instances[gl_InstanceIndex];
+    Vertex v = instance.vertexBuffer.vertices[gl_VertexIndex];
 
-    vec4 worldPos = pcs.transform * vec4(v.position, 1.0f);
+    vec4 worldPos = instance.transform * vec4(v.position, 1.0f);
 
     gl_Position = pcs.sceneData.viewProj * worldPos;
     outPos = worldPos.xyz;
@@ -25,9 +28,10 @@ void main()
     // A bit inefficient, but okay - this is needed for non-uniform scale
     // models. See: http://www.lighthouse3d.com/tutorials/glsl-12-tutorial/the-normal-matrix/
     // Simpler case, when everything is uniform
-    // outNormal = (pcs.transform * vec4(v.normal, 0.0)).xyz;
-    outNormal = mat3(transpose(inverse(pcs.transform))) * v.normal;
+    // outNormal = (instance.transform * vec4(v.normal, 0.0)).xyz;
+    outNormal = mat3(transpose(inverse(instance.transform))) * v.normal;
 
     // w is the bitangent sign, the fragment shader builds the TBN from the interpolated vectors
-    outTangent = vec4(mat3(pcs.transform) * v.tangent.xyz, v.tangent.w);
+    outTangent = vec4(mat3(instance.transform) * v.tangent.xyz, v.tangent.w);
+    outMaterialID = instance.materialID;
 }