macos-perf

macos-perf

Purpose

This skill enables monitoring and profiling macOS system performance using native tools, focusing on CPU, memory, thermal, and power metrics to identify bottlenecks and optimize applications.

When to Use

Use this skill for diagnosing high CPU usage in apps, investigating memory leaks, profiling resource-intensive processes, checking thermal states during heavy workloads, or analyzing power consumption on macOS devices.

Key Capabilities

• Monitor real-time system metrics via Activity Monitor or top/htop for CPU, memory, and network usage.
• Profile applications with Instruments, capturing detailed traces for CPU, memory, and energy.
• Check memory pressure using vm_stat to detect low-memory conditions.
• Query thermal state via powermetrics for CPU/GPU temperatures and fan speeds.
• Analyze power metrics with powermetrics to measure battery drain and efficiency.
• Use command-line tools like top -o cpu for sorted process lists or instruments -s devices for available templates.

Usage Patterns

Invoke this skill in scripts for automated monitoring or integrate into AI workflows for real-time analysis. For example, run periodic checks in a loop for server-like macOS setups, or trigger profiling when an app exceeds resource thresholds. Always specify exact tools and flags based on the task; e.g., use top for quick views and Instruments for deep dives. If integrating with automation, export data to JSON for parsing.

Common Commands/API

• Use top -o cpu -s 1 to display processes sorted by CPU usage, updating every second; pipe output to grep for filtering, e.g., top -o cpu | grep "MyApp".
• Check memory pressure with vm_stat | grep "Pages" to get free pages; sample code:

  output = subprocess.run(['vm_stat'], capture_output=True).stdout
  free_pages = int(re.search(r'free:\s+(\d+)', output.decode()).group(1))
  

• Run powermetrics --samplers cpu,thermal -i 1000 for CPU and thermal data every 1 second.
• Launch Instruments via CLI: instruments -w <device> -t Time Profiler -D output.trace /path/to/app; use exported .trace files for analysis.
• For thermal state, use ioreg -l | grep "Ambient" to parse ambient temperature from system registry.
• No API keys required for these tools; they run natively on macOS.

Integration Notes

Integrate by wrapping commands in Python or shell scripts; set environment variables for custom paths, e.g., export INSTRUMENTS_PATH=/Applications/Xcode.app/Contents/Developer/usr/bin/instruments. For data export, use plutil to convert .plist outputs to JSON. If combining with other skills, ensure macOS version compatibility (e.g., Instruments requires Xcode); check via sw_vers -productVersion. Avoid running multiple intensive tools simultaneously to prevent resource contention.

Error Handling

Check command exit codes; for example, if top fails, verify with echo $? and handle by logging errors or retrying. For Instruments, parse stderr for messages like "Template not found" and fallback to alternatives. Use try-except in scripts, e.g.:

try:
    result = subprocess.run(['instruments', '-t', 'Time Profiler'], check=True)
except subprocess.CalledProcessError as e:
    print(f"Error: {e.returncode} - {e.output}")

Common issues include permission errors (run with sudo) or missing dependencies (install Xcode command line tools via xcode-select --install).

Concrete Usage Examples

1. To monitor CPU usage of a specific process: Run top -pid <processID> -o cpu in a loop script, then analyze output to alert if usage > 80%; example script line: while true; do top -pid 1234 -l 1 | grep CPU; sleep 5; done.
2. For profiling an app's memory: Use instruments -t Allocations -D profile.trace /Applications/MyApp.app, then open the trace in Instruments GUI to identify leaks; integrate by scripting: instruments ... && open profile.trace.

Graph Relationships

• Related to: macos cluster (e.g., macos-core for basic system ops), performance tag (links to general profiling skills), profiling tag (connects to app optimization tools).
• Dependencies: Requires macos cluster skills for foundational access; no direct API links.