ttop - Terminal Top

10X Better Than btop - A pure Rust system monitor with GPU support, Apple hardware acceleration, and deterministic rendering.

Installation

# Standard install
cargo install ttop

# With Apple hardware acceleration (macOS)
cargo install ttop --features apple-hardware

Features

• Pure Rust: Zero C dependencies, cross-platform (Linux + macOS)
• 8ms Frame Time: 2X faster than btop's 16ms target
• GPU Monitoring: NVIDIA (via NVML), AMD (via ROCm SMI), Apple Silicon, AMD Radeon (Mac Pro)
• Apple Accelerators: Neural Engine, Afterburner FPGA, Secure Enclave via manzana
• macOS Native: Full support for Apple Silicon, Intel Macs, and Mac Pro with dual AMD GPUs
• Deterministic Mode: Reproducible rendering for testing
• Debug Mode: Verbose logging for troubleshooting collector issues
• CIELAB Colors: Perceptually uniform gradients
• Docker Verified: Tested on Ubuntu 22.04 containers

Platform Support

Panels

Apple Accelerators Panel

With the apple-hardware feature, ttop displays a dedicated Accelerators panel:

┌─ Accelerators │ 3 available ─────────────────────────────────────┐
│ Neural Engine │ 15.8 TOPS │ 16 cores            ████████░░ 78%  │
│ Afterburner   │ 12/23 streams                   ██████░░░░ 52%  │
│ Metal GPU     │ 4.0GB Discrete │ 1024 threads                   │
│ Secure Enclave │ P-256 ECDSA │ Active                           │
│ Unified Memory │ Page size: 4096 bytes                          │
└──────────────────────────────────────────────────────────────────┘

Supported Accelerators

Keyboard Shortcuts

Navigation

• j/k, ↑/↓ - Move up/down
• PgUp/PgDn - Page up/down
• g/G - Go to top/bottom

Sorting & Filtering

• s, Tab - Cycle sort column
• r - Reverse sort order
• f, / - Filter processes
• Del - Clear filter
• t - Toggle tree view

General

• q, Esc - Quit
• ?, F1 - Toggle help
• 0 - Reset all panels

Command Line Options

ttop [OPTIONS]

Options:
  -r, --refresh <MS>     Refresh rate in milliseconds [default: 1000]
      --deterministic    Enable deterministic mode for testing
      --debug            Enable debug logging to stderr
  -c, --config <PATH>    Config file path
      --show-fps         Show frame timing statistics
  -h, --help             Print help
  -V, --version          Print version

Debug Mode

Use --debug to troubleshoot collector initialization:

ttop --debug 2>&1 | head -50

Example output:

[+0000ms] Platform: macos
[+0000ms] CPU: 28 cores
[+0001ms] macOS: Checking for Apple Silicon
[+0002ms] macOS: Not Apple Silicon, checking for AMD GPUs
[+0188ms] macOS: Found 2 AMD GPUs via ioreg
[+0188ms] GPU collector initialized with 2 GPUs
[+0189ms] Apple Accelerators: 3 available
[+0200ms] App initialization complete

Building from Source

# Clone the repository
git clone https://github.com/paiml/trueno-viz
cd trueno-viz/crates/ttop

# Build without Apple hardware (works on Linux)
cargo build --release

# Build with Apple hardware acceleration (macOS only)
cargo build --release --features apple-hardware

# Run
./target/release/ttop

Examples

# Run headless example (programmatic access)
cargo run --example headless

# Run collectors example
cargo run --example collectors

# Run Apple accelerators example (macOS only)
cargo run --example apple_accelerators --features apple-hardware

Docker Testing

# Build and test in Ubuntu container
docker build -t ttop-test -f docker/ttop-test.Dockerfile .
docker run --rm ttop-test

# Interactive shell for debugging
docker run --rm -it ttop-test bash

Feature Flags

GPU Detection

Apple Silicon

Automatically detects M1/M2/M3/M4 chips and their variants (Pro, Max, Ultra).

AMD Radeon (Mac Pro)

Detects discrete AMD GPUs including:

• Radeon Pro W5700X (supports dual GPU configurations)
• Radeon Pro Vega II

NVIDIA (Linux)

Uses NVML for NVIDIA GPU monitoring on Linux systems.

Performance

Part of the Sovereign AI Stack

ttop is part of the trueno-viz visualization library, which integrates with the Sovereign AI Stack:

• manzana - Apple hardware interfaces
• trueno - SIMD/GPU compute primitives
• batuta - Stack orchestration

License

MIT OR Apache-2.0