Benchmark Baseline - 2025-11-15

Pre-Optimization Results

This baseline was established before implementing the optimization improvements outlined in the 2025-11-15 optimization plan. All benchmarks were run using Criterion with the pre-optimization baseline saved.

Parse Benchmarks

Low-level parsing operations:

• Format Detection: 5.99 ns (5.89 - 6.08 ns range)
• JPEG Segment Parsing: 52.99 ns (52.31 - 53.62 ns range)
• TIFF IFD Parsing: 167.84 ns (164.38 - 170.90 ns range)
• Full Read Metadata: 7.49 ms (7.26 - 7.72 ms range)

Integration Benchmarks

End-to-end performance across file formats:

Single File Extraction

• JPEG Simple: 12.58 ms (12.09 - 13.21 ms range)
• JPEG Complex: 6.52 ms (6.41 - 6.64 ms range)
• PNG Simple: 16.58 ms (15.05 - 18.20 ms range)
• PNG Complex: 6.36 ms (6.29 - 6.43 ms range)
• TIFF Simple: 6.94 ms (6.82 - 7.07 ms range)

Batch Processing

• Batch 16 JPEGs: 147.23 ms (138.91 - 156.21 ms range)
  • Average per file: ~9.20 ms

Note: The batch benchmark used 16 JPEG files (not 100) as that was the number available in the test fixtures directory.

Large File Handling

• TIFF Multipage: 7.78 ms (7.63 - 7.94 ms range)
• JPEG Large Dimension: 6.96 ms (6.81 - 7.13 ms range)

Format Comparison

• JPEG: 5.89 ms (5.77 - 6.01 ms range)
• PNG: 21.59 ms (19.27 - 23.95 ms range)
• TIFF: 27.18 ms (26.37 - 28.08 ms range)
• PDF: 25.46 ms (24.68 - 26.29 ms range)
• MP4: 25.28 ms (24.74 - 25.83 ms range)

GPS Coordinate Extraction

• GPS Coordinate Extraction (3 files): 77.61 ms (73.27 - 82.29 ms range)
  • Average per file: ~25.87 ms

Binary Size

• Release build: 3.6 MB (with LTO enabled, stripped)

Flamegraph Analysis

A flamegraph was successfully generated for baseline profiling analysis. The flamegraph is saved as flamegraph-baseline.svg in the project root directory.

Key observations:

• Flamegraph generated using cargo flamegraph on macOS
• Profiled against JPEG test fixture: tests/fixtures/jpeg/simple/sample_with_exif.jpg
• Generated without requiring sudo privileges (macOS Instruments used as backend)
• 47 KB SVG file containing detailed call stack visualization

Note: For detailed hotspot analysis, open flamegraph-baseline.svg in a web browser. The interactive visualization shows function call hierarchy with time percentages.

Performance Characteristics

Strengths

1. Very fast format detection: Sub-6ns indicates excellent optimization
2. Good JPEG performance: JPEG files process faster than other formats
3. Consistent small-file performance: Most operations in the 6-7ms range

Areas for Improvement

1. PNG processing: ~21ms average is significantly slower than JPEG (~6ms)
2. TIFF processing: ~27ms average shows room for optimization
3. Batch processing: At 9.20ms per file, we're not meeting the 5ms target for bulk operations
4. GPS extraction: At 25.87ms per file, GPS coordinate extraction is slower than expected

Target Improvements

Based on the optimization plan:

• Conservative goal: 10-20% faster runtime

  • Parse benchmarks: Target 5-6ms for full read metadata (currently 7.49ms)
  • Integration: Target 5-7ms for JPEG, 15-18ms for PNG

• Optimistic goal: 20-35% with allocation optimizations

  • Parse benchmarks: Target 4-5ms for full read metadata
  • Integration: Target 4-5ms for JPEG, 12-15ms for PNG

• Binary size: 3-10% reduction

  • Current: 3.6 MB
  • Target: 3.2-3.5 MB

Methodology

Benchmark Execution

# Parse benchmarks
cargo bench --bench parse_benchmarks -- --save-baseline pre-optimization

# Integration benchmarks
cargo bench --bench integration_benchmarks -- --save-baseline pre-optimization

Flamegraph Generation

# Build release binary
cargo build --release

# Generate flamegraph (macOS - no sudo required)
cargo flamegraph --bin oxidex -- tests/fixtures/jpeg/simple/sample_with_exif.jpg

Environment

• Platform: macOS (Darwin 25.1.0)
• Rust toolchain: As configured in project
• Optimization level: 3 (release profile)
• LTO: Enabled (thin for benchmarks, full for release)
• Code generation units: 1
• Strip symbols: Yes (release)

Next Steps

1. Implement Task 5: Zero-allocation format detection optimization
2. Implement Task 6: SIMD-based byte scanning for segment parsing
3. Compare against this baseline using: cargo bench -- --baseline pre-optimization
4. Document improvements and verify targets are met