Skip to main content

Quickstart

Get Wafer installed and start using it in your GPU development workflow.

Installation

  1. Open VS Code or Cursor
  2. Go to Extensions (Cmd+Shift+X on Mac, Ctrl+Shift+X on Windows/Linux)
  3. Search for “Wafer”
  4. Click Install

Getting Started

1

Open Wafer

Click the Wafer icon in the VS Code activity bar (left sidebar) to open the Wafer panel.
2

Sign In

Sign in with GitHub to access all features. Click the sign-in button in the Wafer panel.
3

Choose a Tool

Select a tool from the dropdown in the top bar:
  • NCU Profiler — Analyze Nsight Compute reports
  • Compiler Explorer — View PTX/SASS assembly
  • GPU Docs — Chat with CUDA documentation

Quick Examples

Analyze an NCU Report

1

Select NCU Profiler

Choose NCU Profiler from the tool dropdown.
2

Open a Report

Click Select .ncu-rep file and choose an existing Nsight Compute report.
3

View Results

Browse kernel metrics, performance diagnostics, and optimization recommendations.

Compile and View Assembly

1

Select Compiler Explorer

Choose Compiler Explorer from the tool dropdown.
2

Select a CUDA File

Click the file selector and choose a .cu file from your workspace.
3

Compile

Click Compile to generate PTX and SASS output. Use the tabs to switch between views.

Ask the GPU Docs Agent

1

Open GPU Docs

Choose GPU Docs from the tool dropdown (or select “Wevin” for the agent mode).
2

Ask a Question

Type your question in the chat input. For example:
  • “What is TMEM?”
  • “How do I use tcgen05 for matrix multiplication?”
  • “Explain the memory hierarchy in Blackwell”
3

Get Answers with Citations

View the response with inline citations linking to official documentation.

Prerequisites

Before using certain features, make sure you have the required tools installed:
NVIDIA Nsight Compute must be installed and the ncu CLI available on your PATH.
CUDA Toolkit must be installed with nvcc on your PATH.For SASS output, nvdisasm must also be available (included with CUDA Toolkit).
No local installation required—GPU Docs works out of the box.

Next Steps