A blazingly fast Rust CLI tool that analyzes x86-64 binaries to detect and classify SIMD instructions by their ISA extension. Perfect for performance analysis, compiler optimization verification, and understanding what SIMD features your binaries actually use.

# Install from crates.io
cargo install simdscan

# Analyze a binary
simdscan ./my_program

# Get detailed breakdown in YAML
simdscan -f yaml --show-insts ./my_program

• Fast Analysis - Written in Rust for maximum performance
• Comprehensive Detection - Supports SSE, SSE2, SSE3, SSSE3, SSE4, AVX, AVX2, and AVX-512
• Multiple Formats - Output in JSON or YAML
• Detailed Breakdowns - See which specific instructions are used most
• Cross-Platform - Works on Linux, macOS, and Windows
• Easy Installation - Single command install via cargo

• objdump (from GNU binutils) must be available in your PATH
• Rust 1.70+ (for building from source)

cargo install simdscan

git clone https://github.com/yourusername/simdscan
cd simdscan
cargo install --path .

Download pre-built binaries from the releases page.

# Analyze any x86-64 binary (ELF, Mach-O, PE)
simdscan path/to/binary

# YAML output with instruction details
simdscan -f yaml --show-insts my_program

# JSON output (default)
simdscan -f json my_program

# Help
simdscan --help

• SSE - Streaming SIMD Extensions (Pentium III)
• SSE2 - Streaming SIMD Extensions 2 (Pentium 4)
• SSE3 - Streaming SIMD Extensions 3 (Pentium 4)
• SSSE3 - Supplemental Streaming SIMD Extensions 3 (Core 2)
• SSE4 - SSE4.1, SSE4.2, plus POPCNT, LZCNT, CRC32 (Core i7)
• AVX - Advanced Vector Extensions, including AVX2 (Sandy Bridge+)
• AVX-512 - 512-bit Advanced Vector Extensions (Skylake-X+)

{
  "binary": "./my_program",
  "has_simd": true,
  "isa_summary": {
    "AVX": 156,
    "SSE2": 43,
    "SSE4": 12
  },
  "total_simd_insts": 211
}

{
  "binary": "./my_program",
  "has_simd": true,
  "isa_summary": {
    "AVX": 156,
    "SSE2": 43,
    "SSE4": 12
  },
  "total_simd_insts": 211,
  "isa_details": {
    "AVX": {
      "unique_mnemonics": 8,
      "occurrences": {
        "vmovaps": 45,
        "vaddps": 32,
        "vmulps": 28,
        "vsubps": 21,
        "vxorps": 15,
        "vdivps": 10,
        "vmovups": 3,
        "vzeroupper": 2
      }
    },
    "SSE2": {
      "unique_mnemonics": 5,
      "occurrences": {
        "movdqa": 18,
        "movdqu": 12,
        "paddq": 8,
        "pshufd": 3,
        "pxor": 2
      }
    }
  }
}

1. Disassembly - Uses objdump -d to disassemble the target binary
2. Parsing - Efficiently parses assembly output using compiled regex patterns
3. Classification - Matches instruction mnemonics against comprehensive ISA tables
4. Reporting - Aggregates statistics and generates structured output

• Performance Analysis - Verify your compiler is generating the SIMD code you expect
• Binary Auditing - Understand what instruction sets a binary requires
• Optimization Verification - Confirm auto-vectorization is working
• Compatibility Checking - Ensure binaries will run on target hardware
• Research - Analyze SIMD usage patterns across different codebases

Simdscan is designed for speed:

• Rust Performance - Zero-overhead abstractions and memory safety
• Efficient Parsing - Pre-compiled regex patterns and optimized string processing
• Minimal Dependencies - Fast startup and low memory usage
• Streaming Processing - Handles large binaries efficiently

Contributions are welcome! Here's how you can help:

1. Report Issues - Found a bug or have a feature request? Open an issue
2. Submit PRs - Fix bugs, add features, or improve documentation
3. Add ISA Support - Help expand coverage of instruction sets
4. Performance - Optimize parsing or add new output formats

git clone https://github.com/vimkim/simdscan
cd simdscan
cargo build

• ARM NEON instruction support
• Intel CET (Control-flow Enforcement Technology) detection
• Function-level SIMD analysis
• Integration with popular build systems
• Web assembly support
• Instruction frequency heatmaps

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)

• Inspired by the original Python version
• Built with the amazing Rust ecosystem
• Thanks to the objdump maintainers for the reliable disassembly foundation

Made with ❤️ and ⚡ by the Rust community