{"data":{"kind":"file","path":"README.md","version_id":"yjin3w5yra9tm31sr4w9bd6b","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1579,"modified_at":"2026-08-10T13:44:16.584000","content_hash":"094a6d7fbe8aece15460c028f84324ad42c9246d40cb7d3334745745ea3047ad"},"entries":[],"content":"# Performance Profiler & Bottleneck Finder\n\nEvaluate an LLM's ability to analyze code for performance issues, identify computational bottlenecks, and suggest concrete optimizations.\n\n## What It Tests\n\n- **CPU Profiling**: Identify O(n^2) loops, unnecessary recomputations, suboptimal algorithms\n- **Memory Profiling**: Detect memory leaks, excessive allocations, unbounded caches\n- **I/O Bottlenecks**: Find synchronous blocking on I/O, N+1 queries, missing connection pooling\n- **Algorithmic Complexity**: Recognize when better data structures or algorithms apply\n- **Optimization Suggestions**: Propose concrete, correct refactors with measurable impact\n\n## Task Types\n\n1. **Bottleneck Identification**: Given code with hidden performance issues, identify the primary bottleneck\n2. **Complexity Analysis**: Determine time/space complexity and identify improvement opportunities\n3. **Code Profiling**: Analyze profiling output (callgrind, cProfile) and interpret results\n4. **Optimization Propose**: Suggest and validate optimization strategies\n5. **Multi-dimensional Analysis**: Evaluate code across CPU, memory, and I/O simultaneously\n\n## Scoring\n\n- **correctness** (weight 0.40): Accurately identifies the real bottleneck\n- **specificity** (weight 0.25): Provides concrete line references and complexity analysis\n- **optimization_quality** (weight 0.20): Suggests correct, implementable optimizations\n- **completeness** (weight 0.15): Covers all bottleneck dimensions (CPU/memory/IO)\n\n## Usage\n\n```python\nimport verifiers as vf\nenv = vf.load_environment(\"perf-profiler\")\n```\n","encoding":"utf-8","truncated":false,"total_bytes":1579},"status":null}