Ein Anderssono
6fd2d8d883
- Create profile_detailed.sh script for measuring startup, calculation, and I/O time - Add PROFILING.md documentation explaining the profiling system - Measure startup time (runtime initialization, library loading) - Measure calculation time (algorithm execution) - Estimate I/O time (output formatting) - Support all compiled, JIT, and interpreted languages
3.6 KiB
3.6 KiB
Detailed Profiling System
This document describes the detailed profiling system that breaks down execution time into components.
Overview
The detailed profiling system measures:
- Startup Time - Time to initialize runtime and load libraries
- Calculation Time - Time for the actual π calculation
- I/O Time - Time for output formatting and printing
- Memory Allocation - Memory usage during execution
Usage
Basic Profiling
# Profile all languages with 100 decimals
./profile_detailed.sh 100
# Profile with 1000 decimals
./profile_detailed.sh 1000
# Profile with 20000 decimals
./profile_detailed.sh 20000
Output Format
Language | Startup (ms) | Calculation (ms) | I/O (ms) | Total (ms)
-------------------------------------------------------------------
Assembly | 2 | 7 | 0 | 9
C | 1 | 8 | 0 | 9
Python | 15 | 40 | 2 | 57
Java | 30 | 25 | 2 | 57
Time Components
Startup Time
What it measures:
- Runtime initialization
- Library loading
- JIT compilation (for JIT languages)
- Memory allocation for runtime
Typical values:
- Compiled languages: 1-5 ms (minimal runtime)
- JIT languages: 20-50 ms (JIT compilation overhead)
- Interpreted languages: 10-30 ms (interpreter startup)
Calculation Time
What it measures:
- Algorithm execution
- Numerical operations
- Memory operations for calculation
- Loop iterations
Typical values:
- Compiled languages: 5-30 ms (optimized machine code)
- JIT languages: 20-400 ms (depends on JIT optimization)
- Interpreted languages: 30-900 ms (interpreted execution)
I/O Time
What it measures:
- String formatting
- Output buffering
- Console/terminal output
- Result formatting
Typical values:
- All languages: 0-5 ms (minimal for most languages)
Implementation Details
How It Works
- Startup Measurement: Runs the program with 0 decimals and measures time
- Total Measurement: Runs the program with specified decimals
- Calculation Estimation: Subtracts startup from total time
- I/O Estimation: Approximates I/O time (rough estimate)
Limitations
- I/O time is approximate: Difficult to measure separately
- Startup time varies: First run may be slower due to caching
- Memory measurement: Not yet implemented in detail
Future Enhancements
- Memory Profiling: Track memory allocation over time
- CPU Profiling: Measure CPU usage during execution
- Cache Analysis: Analyze cache hit/miss rates
- Detailed Breakdown: Separate initialization from calculation
Example Results
100 Decimals
| Language | Startup | Calculation | I/O | Total |
|---|---|---|---|---|
| C | 1 ms | 8 ms | 0 ms | 9 ms |
| Python | 15 ms | 40 ms | 2 ms | 57 ms |
| Java | 30 ms | 25 ms | 2 ms | 57 ms |
1000 Decimals
| Language | Startup | Calculation | I/O | Total |
|---|---|---|---|---|
| C | 1 ms | 29 ms | 0 ms | 30 ms |
| Python | 15 ms | 18 ms | 0 ms | 33 ms |
| Java | 30 ms | 46 ms | 0 ms | 76 ms |
Integration with Main Benchmark
The detailed profiling can be run alongside the main benchmark:
# Run main benchmark
./run_all.sh 100
# Run detailed profiling
./profile_detailed.sh 100
Results are stored in timelines/ directory for further analysis.
Generated from Pi Calculation Benchmark - Detailed Profiling System