[NDCOslo2024] Intro to 3D Graphics – Chris Ryan
In the intricate interplay of pixels and polygons, where digital dimensions dance, Chris Ryan, a self-professed enthusiast of orthogonal artistry, unveils the underpinnings of 3D graphics. Eschewing the opacity of engines like Unity, Chris, an all-around engineer, constructs a C++ crucible to expose the mechanics—points, matrices, transforms—driving virtual vistas. His pedagogy, a bottom-up ballet, bridges novices to nuanced techniques, spotlighting sequential subtleties and multi-threaded musings.
Chris confesses his amateur allure: no expert, but an explorer of Euclidean elegance. His canvas: a C++ program, peeling back the pipeline—points plotted, matrices multiplied, perspectives projected—to illuminate 3D’s inner workings.
Points and Matrices: The Geometry Genesis
The journey begins with coordinates: 2D dots evolve into 3D vertices, vectors venturing through virtual voids. Chris clarifies: matrices mold movements—rotation, scaling, translation—mathematical maestros orchestrating object odysseys.
His demo: a cube, corners computed, transformed through matrix multiplications—row-major rigor rendering rotations. Chris’s counsel: master matrices, for they maneuver the mesh’s march.
Transforms and Coordinate Spaces: From Model to Screen
Transforms traverse terrains: model to world, world to view, view to screen—a cascade of coordinate conversions. Chris charts: model space molds objects, world space weaves scenes, view space aligns eyes, screen space flattens for display.
Rasterization resolves: 3D depths distilled to 2D displays, depth buffers dictating dominance. Chris cautions: affine errors—interpolation inaccuracies—mar mappings, demanding derivative diligence.
Rasterization Realities: Rendering the Raster
Rasterization reigns: pixels painted, triangles traced, interpolation iterating intensities. Chris’s code: scanlines sweep surfaces, z-buffers zapping overlaps—ensuring foreground fidelity.
His highlight: rasterization’s rigor, consuming cycles—60 million pixels per second, full HD faltering at 30fps on modest machines. Chris’s clarity: optimize judiciously, for pixel-pushing predominates.
Multi-Threading Musings: Parallelizing Pixels
The pipeline’s sequential soul—single-threaded—spurs scrutiny. Chris explores: multi-threading matrices, a minor marvel; rasterization’s richness resists parallel promises. GPU glances gleam, yet data transfers deter—his demo, lean with points, sidesteps silicon speedups.
His horizon: simplicity suffices for starters, but game engines’ grandeur—point profusion—demands GPU gusto.