Vulkan: phase 01

[midwan]

Phase 1: Rendering Abstraction Layer

Objective

The goal of this phase is to untangle the hardcoded #ifdef USE_OPENGL statements and OpenGL/SDL_Render specific code from src/osdep/amiberry_gfx.cpp. We need a clean, uniform abstraction (e.g., IRenderer interface) so that adding Vulkan does not require duplicate if/else logic scattered across the file.

Target Files to Modify

• src/osdep/amiberry_gfx.cpp
• src/osdep/amiberry_gfx.h (or introduce a new renderer.h)

Context and Instructions

Currently, SDL2_renderframe() handles both pure SDL falling-back paths and OpenGL paths.

1. Define the Interface IRenderer

Create a new header file src/osdep/renderer.h containing an abstract base class IRenderer. The interface should cover the following operations:

• bool initialize(...): Set up the API context.
• void destroy(): Clean up resources.
• bool allocate_texture(int w, int h, const char* shader_name): Prepare the backbuffer texture or shader pipelines.
• bool render_frame(...): Execute the main drawing logic.
• void present(...): Present the buffer to the screen.

2. Implement SDLRenderer and OpenGLRenderer

Create src/osdep/sdl_renderer.cpp and src/osdep/gl_renderer.cpp that both inherit from IRenderer.

• Move all SDL texture/surface copying code (SDL_UpdateTexture, SDL_RenderClear, SDL_RenderCopy) into SDLRenderer::render_frame().
• Move all OpenGL specific shader setup (from crt_frame.h, crtemu_shader, glUseProgram) into OpenGLRenderer::render_frame().

3. Wire into amiberry_gfx.cpp

In amiberry_gfx.cpp, instantiate the appropriate renderer subclass based on the build configuration or user preference. Replace calls inside SDL2_renderframe and SDL2_alloctexture with, for example:

if (g_renderer) {
    return g_renderer->render_frame(monid, mode, immediate);
}

Acceptance Criteria

• Code compiles correctly with OpenGL and SDL backends dynamically selected or cleanly isolated.
• The emulator still runs and renders natively or via RTG exactly as before.
• amiberry_gfx.cpp no longer contains direct <GL/gl.h> or shader compilation logic directly mixed with its window event loops.