glsl

GLSL Shader Programming Skill

File Organization: This skill uses split structure. See references/ for advanced shader patterns.

1. Overview

This skill provides GLSL shader expertise for creating holographic visual effects in the JARVIS AI Assistant HUD. It focuses on efficient GPU programming for real-time rendering.

Risk Level: LOW - GPU-side code with limited attack surface, but can cause performance issues

Primary Use Cases:

• Holographic panel effects with scanlines
• Animated energy fields and particle systems
• Data visualization with custom rendering
• Post-processing effects (bloom, glitch, chromatic aberration)

2. Core Responsibilities

2.1 Fundamental Principles

1. TDD First: Write visual regression tests and shader unit tests before implementation
2. Performance Aware: Profile GPU performance, optimize for 60 FPS target
3. Precision Matters: Use appropriate precision qualifiers for performance
4. Avoid Branching: Minimize conditionals in shaders for GPU efficiency
5. Optimize Math: Use built-in functions, avoid expensive operations
6. Uniform Safety: Validate uniform inputs before sending to GPU
7. Loop Bounds: Always use constant loop bounds to prevent GPU hangs
8. Memory Access: Optimize texture lookups and varying interpolation

3. Implementation Workflow (TDD)

3.1 Step 1: Write Failing Test First

// tests/shaders/holographic-panel.test.ts
import { describe, it, expect, beforeEach } from 'vitest'
import { WebGLTestContext, captureFramebuffer, compareImages } from '../utils/webgl-test'

describe('HolographicPanelShader', () => {
  let ctx: WebGLTestContext

  beforeEach(() => {
    ctx = new WebGLTestContext(256, 256)
  })

  // Unit test: Shader compiles
  it('should compile without errors', () => {
    const shader = ctx.compileShader(holoFragSource, ctx.gl.FRAGMENT_SHADER)
    expect(shader).not.toBeNull()
    expect(ctx.getShaderErrors()).toEqual([])
  })

  // Unit test: Uniforms are accessible
  it('should have required uniforms', () => {
    const program = ctx.createProgram(vertSource, holoFragSource)
    expect(ctx.getUniformLocation(program, 'uTime')).not.toBeNull()
    expect(ctx.getUniformLocation(program, 'uColor')).not.toBeNull()
    expect(ctx.getUniformLocation(program, 'uOpacity')).not.toBeNull()
  })

  // Visual regression test
  it('should render scanlines correctly', async () => {
    ctx.renderShader(holoFragSource, { uTime: 0, uColor: [0, 0.5, 1], uOpacity: 1 })
    const result = captureFramebuffer(ctx)
    const baseline = await loadBaseline('holographic-scanlines.png')
    expect(compareImages(result, baseline, { threshold: 0.01 })).toBeLessThan(0.01)
  })

  // Edge case test
  it('should handle extreme UV values', () => {
    const testCases = [
      { uv: [0, 0], expected: 'no crash' },
      { uv: [1, 1], expected: 'no crash' },
      { uv: [0.5, 0.5], expected: 'no crash' }
    ]
    testCases.forEach(({ uv }) => {
      expect(() => ctx.renderAtUV(holoFragSource, uv)).not.toThrow()
    })
  })
})

3.2 Step 2: Implement Minimum to Pass

// Start with minimal shader that passes tests
#version 300 es
precision highp float;

uniform float uTime;
uniform vec3 uColor;
uniform float uOpacity;

in vec2 vUv;
out vec4 fragColor;

void main() {
  // Minimal implementation to pass compilation test
  fragColor = vec4(uColor, uOpacity);
}

3.3 Step 3: Refactor with Full Implementation

// Expand to full implementation after tests pass
void main() {
  vec2 uv = vUv;
  float scanline = sin(uv.y * 100.0) * 0.1 + 0.9;
  float pulse = sin(uTime * 2.0) * 0.1 + 0.9;
  vec3 color = uColor * scanline * pulse;
  fragColor = vec4(color, uOpacity);
}

3.4 Step 4: Run Full Verification

# Run all shader tests
npm run test:shaders

# Visual regression tests
npm run test:visual -- --update-snapshots  # First time only
npm run test:visual

# Performance benchmark
npm run bench:shaders

# Cross-browser compilation check
npm run test:webgl-compat

4. Technology Stack & Versions

4.1 GLSL Versions

Version	Context	Features
GLSL ES 3.00	WebGL 2.0	Modern features, better precision
GLSL ES 1.00	WebGL 1.0	Legacy support

4.2 Shader Setup

#version 300 es
precision highp float;
precision highp int;

// WebGL 2.0 shader header

5. Performance Patterns

5.1 Avoid Branching - Use Mix/Step

// ❌ BAD - GPU branch divergence
vec3 getColor(float value) {
  if (value < 0.3) {
    return vec3(1.0, 0.0, 0.0);  // Red
  } else if (value < 0.7) {
    return vec3(1.0, 1.0, 0.0);  // Yellow
  } else {
    return vec3(0.0, 1.0, 0.0);  // Green
  }
}

// ✅ GOOD - Branchless with mix/step
vec3 getColor(float value) {
  vec3 red = vec3(1.0, 0.0, 0.0);
  vec3 yellow = vec3(1.0, 1.0, 0.0);
  vec3 green = vec3(0.0, 1.0, 0.0);

  vec3 color = mix(red, yellow, smoothstep(0.3, 0.31, value));
  color = mix(color, green, smoothstep(0.7, 0.71, value));
  return color;
}

5.2 Texture Atlases - Reduce Draw Calls

// ❌ BAD - Multiple texture bindings
uniform sampler2D uIcon1;
uniform sampler2D uIcon2;
uniform sampler2D uIcon3;

vec4 getIcon(int id) {
  if (id == 0) return texture(uIcon1, vUv);
  if (id == 1) return texture(uIcon2, vUv);
  return texture(uIcon3, vUv);
}

// ✅ GOOD - Single atlas texture
uniform sampler2D uIconAtlas;
uniform vec4 uAtlasOffsets[3];  // [x, y, width, height] for each icon

vec4 getIcon(int id) {
  vec4 offset = uAtlasOffsets[id];
  vec2 atlasUV = offset.xy + vUv * offset.zw;
  return texture(uIconAtlas, atlasUV);
}

5.3 Level of Detail (LOD) - Distance-Based Quality

// ❌ BAD - Same quality regardless of distance
const int NOISE_OCTAVES = 8;

float noise(vec3 p) {
  float result = 0.0;
  for (int i = 0; i < NOISE_OCTAVES; i++) {
    result += snoise(p * pow(2.0, float(i)));
  }
  return result;
}

// ✅ GOOD - Reduce octaves based on distance
uniform float uCameraDistance;

float noise(vec3 p) {
  // Fewer octaves when far away (detail not visible)
  int octaves = int(mix(2.0, 8.0, 1.0 - smoothstep(10.0, 100.0, uCameraDistance)));
  float result = 0.0;
  for (int i = 0; i < 8; i++) {
    if (i >= octaves) break;
    result += snoise(p * pow(2.0, float(i)));
  }
  return result;
}

5.4 Uniform Batching - Minimize CPU-GPU Transfers

// ❌ BAD - Many individual uniforms
uniform float uPosX;
uniform float uPosY;
uniform float uPosZ;
uniform float uRotX;
uniform float uRotY;
uniform float uRotZ;
uniform float uScaleX;
uniform float uScaleY;
uniform float uScaleZ;

// ✅ GOOD - Packed into vectors/matrices
uniform vec3 uPosition;
uniform vec3 uRotation;
uniform vec3 uScale;
// Or even better:
uniform mat4 uTransform;

5.5 Precision Optimization - Use Appropriate Precision

// ❌ BAD - Everything highp (wastes GPU cycles)
precision highp float;

highp vec3 color;
highp float alpha;
highp vec2 uv;

// ✅ GOOD - Match precision to data needs
precision highp float;  // Default for calculations

mediump vec3 color;     // 0-1 range, mediump sufficient
mediump float alpha;    // 0-1 range
highp vec2 uv;          // Need precision for texture coords
lowp int flags;         // Boolean-like values

5.6 Cache Texture Lookups

// ❌ BAD - Redundant texture fetches
void main() {
  vec3 diffuse = texture(uTexture, vUv).rgb;
  // ... some code ...
  float alpha = texture(uTexture, vUv).a;  // Same lookup!
  // ... more code ...
  vec3 doubled = texture(uTexture, vUv).rgb * 2.0;  // Again!
}

// ✅ GOOD - Cache the result
void main() {
  vec4 texSample = texture(uTexture, vUv);
  vec3 diffuse = texSample.rgb;
  float alpha = texSample.a;
  vec3 doubled = texSample.rgb * 2.0;
}

6. Implementation Patterns

6.1 Holographic Panel Shader

// shaders/holographic-panel.frag
#version 300 es
precision highp float;

uniform float uTime;
uniform vec3 uColor;
uniform float uOpacity;
uniform vec2 uResolution;

in vec2 vUv;
out vec4 fragColor;

const int SCANLINE_COUNT = 50;

void main() {
  vec2 uv = vUv;

  // Scanline effect
  float scanline = 0.0;
  for (int i = 0; i < SCANLINE_COUNT; i++) {
    float y = float(i) / float(SCANLINE_COUNT);
    scanline += smoothstep(0.0, 0.002, abs(uv.y - y));
  }
  scanline = 1.0 - scanline * 0.3;

  // Edge glow
  float edge = 1.0 - smoothstep(0.0, 0.05, min(
    min(uv.x, 1.0 - uv.x),
    min(uv.y, 1.0 - uv.y)
  ));

  // Animated pulse
  float pulse = sin(uTime * 2.0) * 0.1 + 0.9;

  vec3 color = uColor * scanline * pulse;
  color += vec3(0.0, 0.5, 1.0) * edge * 0.5;

  fragColor = vec4(color, uOpacity);
}

6.2 Energy Field Shader

// shaders/energy-field.frag
#version 300 es
precision highp float;

uniform float uTime;
uniform vec3 uColor;

in vec2 vUv;
in vec3 vNormal;
in vec3 vViewPosition;
out vec4 fragColor;

float snoise(vec3 v) {
  return fract(sin(dot(v, vec3(12.9898, 78.233, 45.543))) * 43758.5453);
}

void main() {
  vec3 viewDir = normalize(-vViewPosition);
  float fresnel = pow(1.0 - abs(dot(viewDir, vNormal)), 3.0);
  float noise = snoise(vec3(vUv * 5.0, uTime * 0.5));

  vec3 color = uColor * fresnel;
  color += uColor * noise * 0.2;
  float alpha = fresnel * 0.8 + noise * 0.1;

  fragColor = vec4(color, alpha);
}

6.3 Data Visualization Shader

// shaders/data-bar.frag
#version 300 es
precision highp float;

uniform float uValue;
uniform float uThreshold;
uniform vec3 uColorLow;
uniform vec3 uColorHigh;
uniform vec3 uColorWarning;

in vec2 vUv;
out vec4 fragColor;

void main() {
  float fill = step(vUv.x, uValue);
  vec3 color = mix(uColorLow, uColorHigh, uValue);
  color = mix(color, uColorWarning, step(uThreshold, uValue));
  float gradient = vUv.y * 0.3 + 0.7;
  fragColor = vec4(color * gradient * fill, fill);
}

7. Security & Performance Standards

7.1 GPU Safety

Risk	Mitigation
Infinite loops	Always use constant loop bounds
GPU hangs	Test shaders with small datasets first
Memory exhaustion	Limit texture sizes

7.2 Loop Safety Pattern

// ❌ BAD - Dynamic loop bound
for (int i = 0; i < int(uCount); i++) { }

// ✅ GOOD - Constant loop bound
const int MAX_ITERATIONS = 100;
for (int i = 0; i < MAX_ITERATIONS; i++) {
  if (i >= int(uCount)) break;
}

8. Common Mistakes & Anti-Patterns

8.1 Never: Use Dynamic Loop Bounds

// ❌ DANGEROUS - May cause GPU hang
for (int i = 0; i < uniformValue; i++) { }

// ✅ SAFE - Constant bound with early exit
const int MAX = 100;
for (int i = 0; i < MAX; i++) {
  if (i >= uniformValue) break;
}

8.2 Never: Divide Without Checking Zero

// ❌ DANGEROUS - Division by zero
float result = value / divisor;

// ✅ SAFE - Guard against zero
float result = value / max(divisor, 0.0001);

9. Pre-Implementation Checklist

Phase 1: Before Writing Code

• Write shader compilation test
• Write uniform accessibility test
• Create baseline images for visual regression tests
• Define performance targets (FPS, draw calls)
• Review existing shaders for reusable patterns

Phase 2: During Implementation

• All loops have constant bounds
• No division by zero possible
• Using branchless patterns (mix/step)
• Appropriate precision qualifiers
• Texture lookups cached
• Uniforms batched into vectors/matrices

Phase 3: Before Committing

• All shader tests pass: npm run test:shaders
• Visual regression tests pass: npm run test:visual
• Performance benchmark meets targets: npm run bench:shaders
• Cross-browser compatibility verified
• No artifacts at edge cases (UV 0,0 and 1,1)
• Smooth animation timing verified

10. Summary

GLSL shaders power the visual effects in JARVIS HUD:

1. TDD First: Write tests before shaders - compilation, uniforms, visual regression
2. Performance: Use branchless patterns, texture atlases, LOD, precision optimization
3. Safety: Constant loop bounds, guard divisions
4. Testing: Verify across target browsers, benchmark GPU performance

Remember: Shaders run on GPU - a single bad shader can freeze the entire system.

---

References:

• references/advanced-patterns.md - Complex shader techniques