examples/

WebGPU Go Examples

This directory contains examples demonstrating the go-webgpu API.

Prerequisites

1. wgpu_native.dll must be in your PATH or the same directory as the executable
2. Go 1.25+
3. GPU with WebGPU support (most modern GPUs)

Building Examples

# Build all examples
cd examples
go build ./...

# Build specific example
cd adapter_info
go build

Running Examples

Make sure wgpu_native.dll is accessible, then:

cd adapter_info
./adapter_info.exe    # Windows
./adapter_info        # Linux/macOS

Example Categories

🔍 Introspection & Info

adapter_info

Purpose: Query GPU capabilities and adapter information

What it demonstrates:

• Getting adapter limits (max texture size, buffer size, etc.)
• Querying GPU information (vendor, device name, backend type)
• Enumerating supported features
• Checking for specific features

Output:

=== Adapter Information ===
Vendor:       NVIDIA Corporation
Device:       NVIDIA GeForce RTX 3080
Backend Type: Vulkan
Adapter Type: Discrete GPU

=== Key Adapter Limits ===
Max Texture 2D:              16384 x 16384
Max Buffer Size:             4294967296 bytes (4.00 GB)
Max Compute Workgroup Size:  1024 x 1024 x 64

Use cases:

• Check GPU capabilities before creating resources
• Display system information to users
• Debug platform-specific issues
• Validate hardware requirements

buffer_introspection

Purpose: Demonstrate buffer state introspection at runtime

What it demonstrates:

• Querying buffer size
• Checking buffer usage flags
• Getting buffer map state
• Buffer lifecycle management

Output:

=== Buffer Introspection ===
Buffer size: 1048576 bytes (1.00 MB)
Buffer usage: Storage | CopySrc | CopyDst
Buffer map state: Unmapped

=== Mappable Buffer Example ===
Initial map state (MappedAtCreation): Mapped
Map state after Unmap(): Unmapped
Map state after MapAsync(): Mapped

Use cases:

• Debug buffer mapping issues
• Validate buffer state before operations
• GPU memory profiling
• Resource lifecycle tracking

🐛 Debugging & Profiling

render_debug_markers

Purpose: GPU debugging with hierarchical markers

What it demonstrates:

• Inserting single debug markers
• Pushing/popping nested debug groups
• Creating hierarchical GPU timelines
• Integration with GPU profiling tools

Output:

=== Demonstrating RenderPass Debug Markers ===
✓ Inserted debug marker: 'Frame Start'
✓ Pushed debug group: 'Scene Rendering'
  ✓ Pushed nested group: 'Geometry Pass'
    ✓ Inserted marker: 'Draw Opaque Objects'
  ✓ Popped debug group: 'Geometry Pass'

=== Debug Marker Hierarchy ===
Frame Start [marker]
└─ Scene Rendering [group]
   ├─ Geometry Pass [group]
   │  ├─ Draw Opaque Objects [marker]
   │  └─ Draw Alpha-Tested Objects [marker]
   └─ Lighting Pass [group]

GPU Tools Support:

• RenderDoc - Event browser
• PIX (Windows) - Timeline view
• Xcode GPU Debugger (macOS) - Capture hierarchy
• Chrome DevTools - WebGPU profiler
• NVIDIA Nsight - Performance timeline

Use cases:

• Identify GPU bottlenecks
• Debug rendering issues
• Optimize draw call order
• Profile complex render passes
• Create professional GPU captures

Advanced Examples

compute_shader

Demonstrates compute shaders, storage buffers, and GPU computation.

triangle

Basic rendering pipeline with vertex buffers and render passes.

texture

Texture creation, sampling, and rendering to texture.

surface

Window surface creation and presentation (requires windowing library).

Example Structure

All examples follow this pattern:

package main

import (
    "github.com/go-webgpu/webgpu/wgpu"
)

func main() {
    // 1. Create instance
    instance, _ := wgpu.CreateInstance(nil)
    defer instance.Release()

    // 2. Request adapter
    adapter, _ := instance.RequestAdapter(nil)
    defer adapter.Release()

    // 3. Request device
    device, _ := adapter.RequestDevice(nil)
    defer device.Release()

    // 4. Create resources and render
    // ...
}

Important Notes

1. Always defer Release() - Prevents resource leaks
2. Check errors - Production code should handle all errors
3. Init order - Instance → Adapter → Device → Resources
4. GPU polling - Some operations require device.Poll() or instance.ProcessEvents()

Troubleshooting

"Failed to load library"

• Ensure wgpu_native.dll is in PATH or current directory
• Check DLL matches your architecture (x64/x86)

"Failed to create instance"

• Update GPU drivers
• Check system has WebGPU-compatible GPU

"Failed to request adapter"

• Try PowerPreference: wgpu.PowerPreferenceLowPower
• Check GPU supports required features

Example crashes or hangs

• Update to latest wgpu-native build
• Check GPU drivers are up to date
• Try different backend (set via environment variables)

Environment Variables

# Force specific backend (Windows)
set WGPU_BACKEND=dx12    # or vulkan, dx11

# Force specific backend (Linux/macOS)
export WGPU_BACKEND=vulkan  # or metal, gl

# Enable validation (helpful for debugging)
export WGPU_VALIDATION=1

# Enable debug output
export RUST_LOG=wgpu=debug

Performance Tips

1. Reuse resources - Don't create/destroy buffers every frame
2. Batch draw calls - Group by pipeline/bind group
3. Use staging buffers - For frequent CPU→GPU uploads
4. Profile first - Use debug markers to identify bottlenecks
5. Cache limits - Call GetLimits() once, not per frame

Contributing Examples

To add a new example:

1. Create directory: examples/my_example/
2. Add main.go with clear documentation
3. Update this README with description
4. Test on Windows, Linux, and macOS if possible
5. Follow existing code style (see LINTER_RULES.md)

References

• WebGPU Specification
• wgpu-native Documentation
• Project Documentation
• API Implementation Details

Last Updated: 2025-11-28 go-webgpu Version: 0.1.0