
Engine Performance Profiling & Optimization
Analyze profiling data and generate engine optimization recommendations
What You Can Do
You can upload profiling data, flame graphs, or performance metrics from your game engine and receive systematic analysis of bottlenecks. Claude identifies root causes by cross-referencing metrics against common engine patterns, reveals cascading performance issues, and delivers production-ready code solutions with clear trade-offs for each optimization approach.
Features
decode CPU/GPU frame time breakdowns, memory snapshots, and call stack hierarchies into actionable insights
identify whether bottlenecks stem from algorithm inefficiency, memory pressure, synchronization issues, or architectural patterns
receive specific optimization techniques with before/after code examples tailored to your engine architecture
detect memory leaks, unexpected allocations, and fragmentation patterns with fixes targeting the source
compare profiling data across engine versions or platforms to pinpoint what changed
analyze optimization approaches with clear cost-benefit analysis (speed vs. complexity vs. memory vs. maintainability)
focus on the 10% of code consuming 90% of frame time with prioritized fix strategies
Example Output
Input: GPU frame time breakdown showing 22ms in shadow map rendering (16.7ms budget for 60fps)
Output:
- Root Cause: Redundant shadow map updates each frame for static geometry; cascading re-renders from missed frustum culling
- Recommendation 1: Cache shadow maps for static meshes, update only when lights move (estimated 8ms recovery)
- Code Fix: Implement shadow cache with dirty-flag system; add frustum culling pre-pass
- Trade-off: +2KB per shadow-casting object, minimal CPU overhead, high frame time savings
Input: Memory profile showing 450MB allocations in physics step
Output:
- Root Cause: Physics broadphase creates temporary pair lists without pooling; GC pressure from 10k+ small allocations per frame
- Recommendation: Replace allocation with object pool + bitset pair tracking
- Code Example: Pool-based broadphase reducing allocations from 8,000/frame to 0
- Memory Saved: ~380MB peak, eliminates GC stalls
What's Included
- SKILL.md instruction file with core workflow and best practices:
- Profiling Data Template: structured format for submitting CPU/GPU metrics, memory snapshots, and flame graphs
- Analysis Checklist: step-by-step guide for root cause diagnosis (check for algorithm inefficiency, memory pressure, synchronization, etc.)
- Optimization Recommendation Framework: template for code fixes with trade-off evaluation matrix
- Common Engine Bottleneck Patterns: reference document mapping symptoms to typical causes (draw call overhead, physics broadphase, animation systems, etc.)
Who It's For
- Engine programmers optimizing frame rate and memory consumption on shipped or in-development titles
- Performance engineers analyzing profiling data across multiple platforms and engine versions
- Technical leads preparing optimization roadmaps and stakeholder recommendations
- Graphics programmers diagnosing GPU bottlenecks and shader optimization opportunities
- Physics/animation programmers reducing hot path overhead in engine systems
Best For
- Interpreting profiler output (CPU/GPU frame time breakdowns, call hierarchies, memory snapshots)
- Diagnosing performance regressions between engine builds or platform ports
- Optimizing hot paths identified in profiling sessions with code-level solutions
- Memory leak investigation and unexpected allocation pattern analysis
- Evaluating trade-offs between competing optimization strategies







