WoS Porter — Windows ARM64 Porting Agent

AI-powered agents that automatically port open-source x64 Windows applications to native ARM64. Ships as a GitHub Copilot VS Code extension, a Claude Code plugin, and a Codex CLI plugin — use it wherever you work.

What It Does

Give it a GitHub repository URL or a local path to an x64 project and it will:

1. Analyze the repo for x64-specific code (SIMD intrinsics, inline assembly, architecture guards)
2. Port the build system — adds ARM64 targets to CMake, MSBuild, Meson, Cargo, and 15+ other build systems
3. Port source code — translates SSE/AVX intrinsics to NEON, adds ARM64 preprocessor guards, fixes calling conventions
4. Build the project for ARM64 using MSVC, iteratively fixing compilation errors
5. Validate every output binary with dumpbin to confirm ARM64 architecture
6. Run tests on ARM64 hardware (when available)
7. Generate a report with a ready-to-apply git patch

Requirements

• Visual Studio 2022 with the MSVC v143 - ARM64/ARM64EC build tools and Windows 11 SDK
• GitHub Copilot extension or Claude Code installed and authenticated
• VS Code 1.100+ (for Copilot or Claude Code VS Code extension) or Claude Code CLI
• git on PATH

Get Started

Recommended approach: Run the agent locally using the GitHub Copilot VS Code extension or the Claude Code plugin — local runs give you full visibility and the ability to intervene when needed. For best porting results — whether running locally or in the cloud — use a Claude Sonnet or Opus model. Cloud (GitHub Actions) workflows are also available but are still evolving and may need extra checks.

• GitHub Copilot (VS Code Extension) — Install VSIX, open Copilot Chat, and start porting
• Claude Code (Plugin) — Install and run as a Claude Code plugin
• Codex CLI (Plugin) — Install and run as a Codex plugin
• Run in the Cloud (GitHub Actions) — No local ARM64 machine needed; runs on windows-11-arm runners. Best for small, straightforward projects — use Claude Code or Copilot locally for medium/complex projects.

Skip Optimization (Porting Only)

By default the pipeline runs NEON optimization (Phase 7) after building and testing. Set WOS_SKIP_OPTIMIZE=1 to skip it when you only need a working ARM64 port.

→ Skip Optimization guide

x64 Competitive Analysis

Run the x64-benchmarker agent to capture an x64 baseline, then pass it to the porter via WOS_X64_BENCH to enable gap-driven differential optimization and generate an ARM64 vs x64 performance comparison table.

→ x64 Competitive Analysis guide

ETL Hotspot Optimization

Use the wos-etl-hotspot agent when you have a representative Windows performance trace (.etl). It identifies the top-20 CPU-hottest functions in that workload, walks their transitive call graphs, and applies ARM64 optimizations (NEON/SVE/SVE2/SME vectorization, scalar/branch/memory tuning, build flags) to exactly those hotspots — rather than optimizing speculatively.

→ ETL Hotspot Optimization guide

Documentation

• WoS Porter Details — Agents, instructions, skills, prompts, commands, supported build systems
• ETL Hotspot Optimization — Trace-driven ARM64 optimization using the wos-etl-hotspot agent (Copilot, Claude, Codex)
• Skip Optimization — How to skip Phase 7 when only porting is needed
• x64 Competitive Analysis — Baseline benchmarking and differential optimization guide
• Development — Build, package, and publish commands for contributors

License

Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
SPDX-License-Identifier: BSD-3-Clause-Clear — see LICENSE.txt