Follows #37

Summary

Function-level profiling with yappi and SnakeViz visualization to complement existing line_profiler for pipeline performance analysis.

Changes

• Add profiling utilities (src/indeca/utils/profiling.py)

  • Minimal yappi_profile context manager for pipeline-level profiling
  • Supports wall-clock and CPU time measurements
  • Outputs pstat format compatible with snakeviz

• Add benchmark scripts (benchmarks/profile/)

  • profile_pipeline_small.py - 10 cells × 1K frames (quick iterations)
  • profile_pipeline_medium.py - 50 cells × 5K frames (realistic workload)
  • profile_pipeline_large.py - 100 cells × 10K frames (comprehensive)

• Add dependencies

  • yappi>=1.6.0 (main dependencies)
  • snakeviz>=2.2.0 (dev dependencies)

• Update .gitignore

  • Ignore *.prof files

Usage

from indeca.utils.profiling import yappi_profile

with yappi_profile("pipeline.prof"):
    pipeline_bin_new(Y, config=config, spawn_dashboard=False)

# Visualize: snakeviz pipeline.prof

# Run benchmark
python benchmarks/profile/profile_pipeline_small.py --profile

# Visualize
snakeviz benchmarks/profile/output/profile_pipeline_small.prof

Testing

• Verified yappi profiling generates valid pstat files
• Benchmark scripts run successfully with profiling enabled
• snakeviz can visualize generated profiles