What problem does it solve? Writing procedural shaders that are both visually rich and fast on the GPU is difficult: naive GLSL code wastes ALU cycles, uses expensive transcendental functions, branches unpredictably, and produces numerical artifacts. This Skill provides senior-level GPU engineering guidance to make shaders faster, cleaner, and more stable without sacrificing artistic intent. ## Core Features & Use Cases - Branchless & Vectorized GLSL: Replaces if-statements with step, smoothstep, mix, and clamp, and restructures scalar math into vector operations for better GPU throughput. - Raymarch & SDF Optimization: Applies sphere tracing, adaptive step sizes, bounding volumes, hierarchical distance fields, and early-exit conditions to minimize distance evaluations. - Numerical Stability & Precision Control: Guards against NaN, division by zero, and precision drift using epsilon values, highp/mediump selection, and periodic normalization. - Use Case: You have a raymarched fractal shader running at low frame rates on mobile. Use this Skill to restructure the march loop with bounding volumes, reduce iterations from 256 to 96, cache repeated trigonometric evaluations, and replace nested FBM with a cheaper noise model. ## Quick Start Ask the agent to review and optimize your GLSL raymarching shader for performance and numerical stability while preserving its visual appearance.
