287 logger calls analyzed and reclassified (134 debug, 61 info, 66 warning, 26 error)
Existing tag system identified ([ITER X], [CONVERGENCE], [GPU], etc.)
Log bloat root cause fixed — console shows ~66 WARNING lines per run

🎯 Quick Start

Optimization Workflow (Recommended)

# 1. Quick smoke test validation (100 nets, <30s)
.\scripts\optimize_and_validate.ps1 -Compare tests/regression/smoke_metrics.json

# 2. Full backplane test with profiling (512 nets, 11-18 min)
.\scripts\optimize_and_validate.ps1 -ProfileMode -TestBoard backplane -Compare tests/regression/golden_metrics.json

# 3. Analyze results
python scripts/analyze_log.py --compare tests/regression/golden_metrics.json

See: optimization_workflow.md for complete workflow guide

Manual Baseline Test

# Standard performance test
python main.py cli TestBoards/TestBackplane.kicad_pcb

Analyze Performance

# NEW: Standalone log parser (recommended)
python scripts/analyze_log.py
python scripts/analyze_log.py --compare tests/regression/golden_metrics.json

# Legacy manual analysis (for custom queries)

import re, os
log = r"<plugin_dir>/logs/latest.log"
lines = open(log, encoding="utf-8", errors="ignore").readlines()
iters = [l for l in lines if re.search(r"WARNING.*\[ITER\s+\d+\]", l)]
print(f"Iters: {len(iters)}")
vals = [float(re.search(r": ([\d.]+)ms", l).group(1)) for l in lines if "[PROFILE]" in l and re.search(r": ([\d.]+)ms", l)]
print(f"Rebuild total: {sum(vals)/1000:.1f}s avg={sum(vals)/max(1,len(vals)):.0f}ms")
gpu = [float(re.search(r"\(([\d.]+)ms\)", l).group(1)) for l in lines if "Path found in" in l and "ms)" in l]
tot = [float(re.search(r"in ([\d.]+)s", l).group(1)) for l in lines if "SUCCESS! Path found in" in l and re.search(r"in ([\d.]+)s", l)]
print(f"GPU kernel: {sum(gpu)/max(1,len(gpu)):.1f}ms avg,  Total/net: {sum(tot)/max(1,len(tot))*1000:.0f}ms avg")

📈 Optimization Targets — Status Update

Historical Progression

Date	Total Time	Avg Iter	Target achieved	Notes
Apr 3	47.9 min	~39s	Baseline	Multi-launch kernel
Apr 3	25 min	~20s	2× improvement	Persistent kernel enabled
Apr 5	11.96 min	11.0s	4× improvement ✅	Best performance
Apr 8	17.5 min	15.7s	⚠️ 46% regression	Regression investigation needed

Current Investigation Priorities (April 8, 2026)

Priority	Target	Status	Action
🔴 #1	Identify Apr 8 regression root cause	Unknown	Profile with `ORTHO_DEBUG=1`
🔴 #2	Verify GPU persistent kernel usage	Unclear	Check actual kernel mode in logs
🟡 #3	Measure debug screenshot overhead	~5-10%	Test without `ORTHO_SCREENSHOT_FREQ`
🟢 #4	Code diff analysis Apr 5 → Apr 8	Pending	Git log review

Target: Restore 11.96 min performance from April 5 baseline

🚀 Optimization Phases

Phase 1: Instrumentation ✅ COMPLETE

✅ @profile_time decorator in shared/utils/performance_utils.py
✅ Logging reclassified — console shows milestones (WARNING), file captures detail (DEBUG)
✅ GPU timing profiled — MULTI-LAUNCH overhead is the real bottleneck

Phase 2: PERSISTENT KERNEL Activation 🔴 HIGH PRIORITY

Target: 44 min → 3-5 min (10-20× speedup)
File: orthoroute/algorithms/manhattan/pathfinder/cuda_dijkstra.py

Find why MULTI-LAUNCH is selected instead of PERSISTENT KERNEL
Persistent kernel already compiled — routing selection logic needs fixing
Port JITTER + ROUNDROBIN features if not yet in persistent variant
Validate routing quality (zero overuse must be maintained)

Phase 3: Graph Initialization

Target: 20.9s → 15s (30-40% improvement)

Profile initialize_graph() sub-operations
GPU-accelerated CSR construction
Parallel edge array building

Phase 4: Via Rebuild (Deferred)

Status: Incremental implementation deployed. 1% of runtime — not worth further work until Phase 2 is resolved.

Algorithm Architecture: Deep dive into PathFinder algorithm
Performance Profiling Example: @profile_time decorator usage
Contributing Guide: Development practices
Tuning Guide: Routing parameter optimization

🤖 Specialized Agents

Use these agents for optimization work:

# GPU optimization
@GPU Performance Agent - analyze via rebuild performance with CuPy

# Safe refactoring
@Refactoring Agent - extract function from unified_pathfinder.py

# Architecture validation
@Architecture Reviewer - check dependencies in optimization branch

# Test creation
@Testing Agent - create unit tests for via pooling logic

📝 Contribution Guidelines

When adding optimization documentation:

Name baselines with date: optimization_baseline_YYYY-MM-DD.md
Include test board details: Board name, complexity, routing results
Provide reproducible commands: PowerShell/bash commands for verification
Document before/after metrics: Always compare to established baseline
Link to related PRs/commits: Connect documentation to code changes

🔍 Archive

As new baselines are established, previous baselines will be moved to an archive/ subdirectory but kept for historical comparison.

Current Active Baseline: April 3, 2026 (TestBackplane, v0.2.0)

Last Updated: April 3, 2026
Maintainer: Performance Optimization Team

README.md Unescape Escape

OrthoRoute Performance Optimization

📊 Performance Baselines

Baseline: April 8, 2026 ⚠️ CURRENT

Baseline: April 5, 2026 ✅ BEST PERFORMANCE

Baseline: April 3, 2026

🔧 Optimization Tools & Systems

Optimization Workflow ✨ NEW

Scripts (scripts/)

Current Logging Review