cpu-loader 0.5.4

A tool to generate CPU load on a system with runtime configuration through a WebUI and REST API

CPU Loader

A tool to generate CPU load on a system with runtime configuration through a WebUI and REST API.

✨ Key Features

• 🎯 Precise Per-Thread Control: Set individual CPU load (0-100%) for each core independently
• ⚡ High Performance: Native C implementation with pthreads ensures accurate load generation
• 🧮 Configurable Algorithms: Choose between 5 different computation types (busy-wait, PI, primes, matrix, fibonacci)
• ⏱️ Time-Controlled Execution: All algorithms respect precise timing for accurate load percentages
• 📊 Real-Time Monitoring: Live WebSocket updates showing actual CPU usage and temperature via psutil
• 🎛️ Interactive WebUI: Beautiful gradient interface with sliders and visual feedback
• 🚀 REST API: Complete programmatic control for automation and testing
• 📱 Responsive Design: Works seamlessly on desktop and mobile devices
• 🔄 Instant Updates: Changes take effect immediately with sub-second response time
• 🌡️ Temperature Monitoring: Optional CPU temperature tracking with sensor auto-detection
• 📡 MQTT Integration: Publish metrics and settings to MQTT broker for IoT/monitoring systems

🖼️ Screenshots

Interface Screenshots

Clean interface showing all CPU threads at idle	Fully responsive mobile interface
Different load levels: 50%, 75%, and 25% on individual threads
Real-time metrics showing actual CPU usage with color-coded bars

Features Shown

• Visual Load Bars: Green progress bars show target load, blue bars display actual CPU usage
• Real-Time Updates: WebSocket connection provides live metrics every second
• Preset Buttons: Quick-set options (0%, 25%, 50%, 75%, 100%) for all threads
• Smooth Gradients: Modern UI with purple-to-blue gradient design

Installation

From Pre-built Wheels (Recommended)

Pre-compiled wheels are available for Linux (x86_64, ARM64) and macOS (x86_64, ARM64):

pip install cpu-loader

From Source

1. Clone the repository:

git clone https://github.com/the78mole/cpu-loader.git
cd cpu-loader

2. Install dependencies and build C extension:

uv pip install -e .

This will compile the high-performance C extension for efficient CPU load generation.

3. (Optional) Set up pre-commit hooks for development:

pre-commit install

4. (Optional) Set up commit message template for semantic versioning:

git config commit.template .gitmessage

See COMMIT_MESSAGE_FORMAT.md for commit message guidelines.

Usage

Starting the Server

uv run src/main.py

The server will start on http://localhost:8000

Command-Line Options

uv run src/main.py --help

Available Options:

• --host HOST: Host to bind the server to (default: 0.0.0.0)
• --port PORT: Port to bind the server to (default: 8000)
• --disable-temperature: Disable CPU temperature monitoring
• --computation-type TYPE: Set computation algorithm (busy-wait, pi, primes, matrix, fibonacci)
• --mqtt-broker-host HOST: MQTT broker hostname
• --mqtt-broker-port PORT: MQTT broker port (default: 1883)
• --mqtt-username USER: MQTT username
• --mqtt-password PASS: MQTT password
• --mqtt-topic-prefix PREFIX: MQTT topic prefix (default: cpu-loader)
• --mqtt-client-id ID: MQTT client ID (default: cpu-loader)

Computation Types

CPU Loader supports different computation algorithms for load generation with precise time control:

• busy-wait (default): Simple busy loop - fastest execution, minimal overhead, most accurate timing
• pi: PI calculation using Leibniz formula - moderate computational intensity, mathematical workload
• primes: Prime number finding - variable computational load, cryptographic-style operations
• matrix: 4x4 matrix multiplication - consistent computational patterns, linear algebra operations
• fibonacci: Lightweight mathematical operations - balanced computational load with micro-pauses

All algorithms are time-controlled to ensure accurate load percentages. The system uses 10ms cycles with frequent timing checks to maintain precise CPU utilization.

Examples:

# Use PI calculation for CPU load
uv run cpu-loader --computation-type pi

# Use prime number calculation
uv run cpu-loader --computation-type primes --port 8001

# Use matrix multiplication
uv run cpu-loader --computation-type matrix

WebUI

Open your browser and navigate to http://localhost:8000

Features:

• Individual Thread Control: Use sliders to set load for each thread (0-100%)
• Preset Buttons: Quick-set all threads to 0%, 10%, 25%, 50%, 80%, 90%, or 100%
• Live Stats: View active thread count and average load
• Real-time Updates: Changes are applied instantly with visual feedback

REST API

Get Thread Status

curl http://localhost:8000/api/threads

Response:

{
  "num_threads": 4,
  "loads": {
    "0": 0.0,
    "1": 0.0,
    "2": 0.0,
    "3": 0.0
  }
}

Set Load for Specific Thread

curl -X PUT http://localhost:8000/api/threads/0/load \
  -H "Content-Type: application/json" \
  -d '{"load_percent": 50.0}'

Set Load for All Threads

curl -X POST http://localhost:8000/api/threads/load/all \
  -H "Content-Type: application/json" \
  -d '{"load_percent": 75.0}'

Change Number of Threads

curl -X POST http://localhost:8000/api/threads \
  -H "Content-Type: application/json" \
  -d '{"num_threads": 8}'

Get Computation Type

curl http://localhost:8000/api/computation-type

Response:

{
  "computation_type": "pi",
  "available_types": ["busy-wait", "pi", "primes", "matrix", "fibonacci"]
}

Set Computation Type

curl -X PUT http://localhost:8000/api/computation-type \
  -H "Content-Type: application/json" \
  -d '{"computation_type": "fibonacci"}'

API Documentation

Once the server is running, visit http://localhost:8000/docs for interactive API documentation powered by Swagger UI.

MQTT Publishing

CPU Loader can publish real-time CPU metrics and load control settings to an MQTT broker for integration with home automation systems, monitoring tools, or custom applications.

MQTT Topics

The application publishes to two topics:

1. {prefix}/cpu_metrics: Published every second with current CPU utilization

   {
     "total_cpu_percent": 25.5,
     "per_cpu_percent": [25.0, 26.0, 25.3, 25.7]
   }
   

2. {prefix}/load_settings: Published when load settings change (retained message)

   {
     "num_threads": 4,
     "loads": {"0": 25.0, "1": 25.0, "2": 25.0, "3": 25.0},
     "average_load": 25.0
   }
   

Configuration

MQTT can be configured using environment variables or command-line arguments. Command-line arguments take precedence over environment variables.

Using Environment Variables

export MQTT_BROKER_HOST=mqtt.example.com
export MQTT_BROKER_PORT=1883
export MQTT_USERNAME=myuser
export MQTT_PASSWORD=mypassword
export MQTT_TOPIC_PREFIX=cpu-loader
export MQTT_CLIENT_ID=cpu-loader-001

uv run src/main.py

Using Command-Line Arguments

uv run src/main.py \
  --mqtt-broker-host mqtt.example.com \
  --mqtt-broker-port 1883 \
  --mqtt-username myuser \
  --mqtt-password mypassword \
  --mqtt-topic-prefix cpu-loader \
  --mqtt-client-id cpu-loader-001

Testing with Mosquitto

To test MQTT publishing locally:

# Install mosquitto
sudo apt-get install mosquitto mosquitto-clients

# Start CPU Loader with MQTT
uv run src/main.py --mqtt-broker-host localhost

# Subscribe to all topics (in another terminal)
mosquitto_sub -h localhost -t "cpu-loader/#" -v

MQTT Settings

Setting	Environment Variable	CLI Argument	Default	Description
Broker Host	MQTT_BROKER_HOST	--mqtt-broker-host	None	MQTT broker hostname or IP
Broker Port	MQTT_BROKER_PORT	--mqtt-broker-port	1883	MQTT broker port
Username	MQTT_USERNAME	--mqtt-username	None	MQTT authentication username
Password	MQTT_PASSWORD	--mqtt-password	None	MQTT authentication password
Topic Prefix	MQTT_TOPIC_PREFIX	--mqtt-topic-prefix	cpu-loader	Prefix for all MQTT topics
Client ID	MQTT_CLIENT_ID	--mqtt-client-id	cpu-loader	MQTT client identifier

Note: If no MQTT broker host is configured, MQTT publishing will be disabled and the application will function normally without it.

Example Script

An example script (src/example.py) is provided to demonstrate programmatic control:

# Start the server first
uv run src/main.py

# In another terminal, run the example
uv run src/example.py

The example demonstrates:

• Getting current status
• Setting all threads to a specific load
• Gradually increasing load
• Individual thread control
• Resetting to idle

Computation Types Demo

A comprehensive demo script is available in examples/computation_types_demo.py to test all computation algorithms:

# Start the server
uv run cpu-loader

# In another terminal, run the demo (tests all computation types)
python examples/computation_types_demo.py

# Test specific computation types only
python examples/computation_types_demo.py --types pi fibonacci

# Custom load and duration
python examples/computation_types_demo.py --load 50 --duration 15

The demo script:

• Tests each computation type systematically
• Monitors actual CPU usage vs. target load
• Provides performance comparisons between algorithms
• Shows timing accuracy for each computation method

Architecture

• src/cpu_loader_core.c: High-performance C implementation using pthreads for CPU load generation
• src/cpu_loader.py: Python wrapper providing a clean API to the C extension
• src/main.py: FastAPI application with REST API and embedded WebUI
• src/mqtt_publisher.py: MQTT client for publishing metrics and settings
• Threading Model: Native pthreads for maximum efficiency and precise timing
• Load Algorithm: High-resolution busy-wait loops with nanosecond precision

Requirements

• Python 3.8+
• FastAPI 0.115.0+
• Uvicorn 0.32.0+
• Pydantic 2.10.0+
• paho-mqtt 2.1.0+ (for MQTT publishing)
• C compiler (gcc or clang) for building the extension

Use Cases

• Performance Testing: Test application behavior under various CPU loads and computation types
• Stress Testing: Validate system stability under high CPU utilization with different algorithms
• Thermal Testing: Use fibonacci or matrix computations for maximum heat generation
• Power Consumption Analysis: Compare power usage across different computation algorithms
• Benchmarking: Create reproducible load scenarios with specific computation patterns
• Algorithm Testing: Test cache performance (matrix), mathematical units (pi), or crypto operations (primes)
• IoT Integration: Integrate with Home Assistant, Node-RED, or other MQTT-based systems
• Monitoring: Feed CPU metrics and temperature data into monitoring dashboards via MQTT

Computation Algorithm Use Cases

• busy-wait: Baseline testing, pure timing validation, minimal computational overhead
• pi: Mathematical processing benchmarks, floating-point unit testing
• primes: Cryptographic algorithm simulation, integer arithmetic testing
• matrix: Linear algebra workloads, cache hierarchy testing, SIMD instruction testing
• fibonacci: Balanced computational load for general stress testing

Development

Pre-commit Hooks

This project uses pre-commit hooks to ensure code quality. The hooks include:

• Code Formatting: Black and isort for consistent Python formatting
• Linting: Flake8 for code quality checks
• Type Checking: Mypy for static type analysis
• General Checks: Trailing whitespace, end-of-file fixes, YAML/JSON/TOML validation

To run pre-commit manually on all files:

pre-commit run --all-files

The hooks will run automatically on git commit after installation.

Building and Publishing Releases

The project uses automatic semantic versioning with GitHub Actions:

Versioning Rules

Versions are automatically determined based on commit messages:

• Patch bump (0.0.X): Every commit to main

• Minor bump (0.X.0): Commits prefixed with feat:

  git commit -m "feat: add new CPU monitoring feature"
  

• Major bump (X.0.0): Commits with major:, breaking:, or BREAKING CHANGE:

  git commit -m "major: redesign API interface"
  git commit -m "breaking: remove deprecated endpoints"
  

Release Process

1. Commit and push to main:

   git add .
   git commit -m "feat: add WebSocket support"
   git push origin main
   

2. Automated workflow:

   • Version is automatically calculated using semantic versioning
   • Wheels are built for Linux (x86_64, ARM64) and macOS (x86_64, ARM64)
   • Python versions: 3.8, 3.9, 3.10, 3.11, 3.12
   • Source distribution (sdist) is created
   • All artifacts are published to PyPI
   • A GitHub Release is created with version tag and artifacts

License

MIT License - see LICENSE file for details.