OpenGL Backend

OpenGL Backend

Release boundary: This page documents an experimental implementation outside the blocked and uncertified stable-v1 product scope (Windows 11 x64/MSVC v143 with D3D11 or Windows NullRHI and C++ modules). It is not a supported or release-certified backend.

Experimental OpenGL RHI development implementation. The direct EGL/GLX device path requests a 4.5 context, while the SDL runtime/editor hosts request 3.3; source-level DSA and SPIR-V paths still depend on the context and driver that is actually created. It is not a generic or certified fallback.

Overview

• Namespace: Spark::RHI::OpenGL
• Files: Graphics/RHI/OpenGL/OpenGLDevice.{h,cpp} (~450 LOC header).
• Guard: #ifdef SPARK_OPENGL_SUPPORT (CMake toggle ENABLE_OPENGL=ON).
• Linkage: GLAD loader (bundled in ThirdParty/) + SDL2 for context creation.
• Context/feature boundary: EGL/GLX requests OpenGL 4.5; SDL hosts request 3.3. DSA and SPIR-V binary-shader calls exist in source but are not a blanket OpenGL 4.6 runtime guarantee.

Architecture

Class	Purpose	RHI interface
GLDevice	Context owner, resource factory	IRHIDevice
GLBuffer	glCreateBuffers + DSA upload	IRHIBuffer
GLTexture	glCreateTextures + optional framebuffer attachment	IRHITexture
GLShader	Program or separable shader object	IRHIShader
GLSampler	glCreateSamplers state	IRHISampler
GLPipelineState	VAO + pipeline composition	IRHIPipelineState
GLSwapChain	SDL2-backed double-buffer swap	IRHISwapChain
GLCommandList	Immediate-mode translation (no real GL command buffer)	IRHICommandList

OpenGL is single-threaded — command recording is serialized on the render thread.

Features

• Direct state access (DSA) — all resource creation uses the DSA variants (glCreateTextures, glNamedBufferStorage, etc.). Cleaner than bind-to-modify.
• SPIR-V ingest — shader loading accepts compiled SPIR-V binaries via glShaderBinary(GL_SHADER_BINARY_FORMAT_SPIR_V), so the offline shader cook path produces one binary format for both OpenGL and Vulkan.
• Mesa llvmpipe development route — requires explicit Mesa/driver and headless-display configuration; it is not a generic CI or release-certified fallback. See the RHI fallback chain.
• KHR_debug callback — validation messages routed through Spark::Logger in debug builds.
• Persistent-mapped buffers — GL_MAP_PERSISTENT_BIT used for per-frame transient UI / debug-draw streams.
• Framebuffer object (FBO) abstraction — GLTexture::SetFramebuffer caches the FBO that owns this texture as a color/depth attachment, avoiding per-draw-call FBO churn.

Capability detection

auto& caps = device->GetCapabilities();
caps.maxTextureSize           // GL_MAX_TEXTURE_SIZE
caps.computeShaderSupport     // true on 4.3+
caps.bindlessResourceSupport  // GL_ARB_bindless_texture (NVIDIA)
caps.meshShaderSupport        // GL_NV_mesh_shader (experimental)

Factory registration

// RHIFactory.cpp
case GraphicsBackend::OpenGL:
    device = std::make_unique<OpenGL::GLDevice>();
    break;

RHIFactory::GetRecommendedBackend() selects a preferred compiled and available backend; it does not retry device startup. RHIBridge::Initialize tries available GPU candidates after device or swap-chain failure and creates NullRHIDevice only if they all fail. OpenGL is one optional candidate, not a NullRHI fallback tier.

Limitations

• No timeline semaphores — CPU/GPU sync via glFenceSync / glClientWaitSync.
• No bindless (standard) — vendor-specific extensions only.
• No ray tracing, no mesh shaders (portable) — upgrade to Vulkan or D3D12 for those.
• Single thread — render commands must run on the thread that created the context.

Related

• RHI Abstraction Layer
• Vulkan Backend — experimental Linux development alternative when compiled.
• D3D11 Backend — Windows equivalent tier.
• Shader Pipeline — how shaders are built into the SPIR-V blobs that OpenGL consumes.