ANARI Course Example Code

This repository contains the example applications for the SIGGRAPH course introducing the ANARI C API. The examples are designed for recorded and self-paced students: each checkpoint is a complete application that can be read and run independently.

The course starts with offline PNG rendering, then moves to a fixed-size window, orbit camera interaction, glTF loading, and finally an ImGui-based viewer that students can extend.

Build

cmake -S . -B build
cmake --build build

The build requires an ANARI-SDK installation. The final ImGui checkpoint fetches SDL 3 and Dear ImGui during CMake configure through anari_sdk_fetch_project(). On Linux, the windowed checkpoints still require platform windowing and graphics development headers for Thirteen and SDL's renderer backends.

Runtime Selection

The examples default to:

library: environment
device:  default

Most checkpoints allow optional runtime overrides. For example:

./build/anari_course_01_triangle_png triangle.png visrtx default
./build/anari_course_05_window_viewer visrtx default

Course Steps

1. Render Your First Triangle

Folder: steps/01_triangle_png

Executable:

./build/anari_course_01_triangle_png triangle.png

Concepts:

• Load an ANARI library and create a device.
• Create triangle geometry with mapped parameter arrays.
• Create a required matte material.
• Connect geometry and material through a surface.
• Attach the surface to a world.
• Configure camera, renderer, frame, and channel.color.
• Render once, map the frame, and write an RGBA PNG.

2. Build A Small Scene

Folder: steps/02_multiple_surfaces

Executable:

./build/anari_course_02_multiple_surfaces multiple_surfaces.png

Concepts:

• Build a box-on-floor scene.
• Use multiple surfaces in world.surface.
• Use separate matte materials for each surface.
• Use per-vertex color for the box and constant material color for the floor.
• Keep rendering offline to PNG.

3. Instance The Cube

Folder: steps/03_instanced_cubes

Executable:

./build/anari_course_03_instanced_cubes instanced_cubes.png

Concepts:

• Keep the floor and cube data from step 2.
• Put the cube surface in an ANARI group.
• Create several transform instances that share that group.
• Set per-instance transform matrices to place cubes over the floor.
• Attach the floor through world.surface and cubes through world.instance.
• Keep rendering offline to PNG.

4. Control The Image And Camera

Folder: steps/04_frame_size_and_camera

Executable:

./build/anari_course_04_frame_size_and_camera frame_size_and_camera.png 1920 1080

Concepts:

• Keep the same box/floor scene while making camera and image size explicit.
• Make output resolution a frame parameter.
• Update camera aspect from the requested image size.
• Keep the view hard-coded before introducing interaction.

Command-line form:

anari_course_04_frame_size_and_camera [output.png] [width height] [library device]

5. Put ANARI Pixels In A Window

Folder: steps/05_window_viewer

Executable:

./build/anari_course_05_window_viewer

Concepts:

• Keep the same scene and camera from step 4.
• Render continuously with blocking ANARI_WAIT.
• Map channel.color each frame.
• Copy mapped pixels into a fixed-size 1200x800 Thirteen window.
• Treat the window as another destination for ANARI pixels.

6. Connect Interaction To The Camera

Folder: steps/06_orbit_camera

Executable:

./build/anari_course_06_orbit_camera

Concepts:

• Use the vendored ANARI-SDK Orbit helper as app-side camera infrastructure.
• Query world.bounds to initialize the orbit camera.
• Convert orbit output into ANARI camera position, direction, and up.
• Commit the camera after mouse interaction changes its parameters.
• Continue using the simple blocking render loop.

Controls:

LMB: orbit
MMB: pan
RMB: dolly
Esc: exit

7. Load A glTF Scene

Folder: steps/07_gltf_loader

Executable:

./build/anari_course_07_gltf_loader scene.gltf

Concepts:

• Keep the orbit-camera viewer shell from step 6.
• Move scene construction into a separate import_gltf() function.
• Introduce vendored tinygltf v3 as the parser used by the loader.
• Return an ANARIWorld from the loader so the viewer remains scene-agnostic.

Command-line form:

anari_course_07_gltf_loader [scene.gltf] [library device]

8. Turn It Into An Extensible Viewer

Folder: steps/08_imgui_viewer

Executable:

./build/anari_course_08_imgui_viewer scene.gltf

Concepts:

• Move to an SDL 3 renderer/ImGui application shell.
• Load the same glTF scene input used by step 7.
• Support framebuffer resizing by updating the ANARI frame size and camera aspect.
• Render continuously using a simple frameInFlight polling model.
• Keep the UI responsive by checking anariFrameReady(..., ANARI_NO_WAIT).
• Expose renderer parameters through ImGui controls.

Command-line form:

anari_course_08_imgui_viewer [scene.gltf] [library device]

Controls:

LMB: orbit
MMB: pan
RMB: dolly
Esc: exit

Supporting Material

• docs/course-design-summary.md summarizes the course design decisions.
• slides/outline.md gives the 90-minute presentation structure.
• slides/*.md contains first-draft slide content for each section.

Third-Party Code

• third_party/thirteen is a small window/pixel display helper.
• third_party/orbit contains the ANARI-SDK orbit camera helper.
• third_party/stb/stb_image_write.h is vendored for PNG output.
• third_party/tinygltf contains the vendored tinygltf v3 parser.
• SDL 3 and Dear ImGui are fetched by CMake for the final checkpoint.