perf

name	perf
description	Performance profiling -- CPU, memory, concurrency, benchmarking with statistical rigor.

Activate When

• /godmode:perf, "profile this", "find bottleneck"
• "memory leak", "race condition", "deadlock"
• "benchmark", "flame graph", "why is this slow"

Workflow

1. Profiling Strategy

# Node.js CPU profile
node --cpu-prof --cpu-prof-dir=./profiles app.js
# Go CPU profile
go tool pprof http://localhost:6060/debug/pprof/profile
# Python memory
python -c "import tracemalloc; tracemalloc.start()"

Target: <application/function>
Symptom: <slow response, high CPU, OOM, hang>
Language: <Node.js|Python|Go|Rust|Java>

2. CPU Profiling & Flame Graphs

For each hotspot found:

Function: <name> at <file>:<line>
CPU share: <N>% of total
Root cause: <why slow>
Remediation: <optimized code>
Expected improvement: <N>% reduction

IF hotspot >10% CPU: must address. IF hotspot <5% CPU: diminishing returns.

3. Memory Leak Detection

Tools: Node.js (--heap-prof), Python (tracemalloc), Go (pprof), Rust (DHAT), Java (JFR+JMC).

Pattern: baseline -> load -> analyze growth.

• Constant: OK
• Linear: leak
• Stepped: periodic leak

For each finding:

Type: LEAK | EXCESSIVE ALLOCATION | UNBOUNDED CACHE
Source: <file>:<line>
Growth rate: <MB per hour/request>
Retention chain: <root -> ... -> leaked object>

IF growth is linear under load: confirmed leak. IF cache grows without bound: add max size + eviction.

4. Concurrency Bug Detection

go  -race ./...

gcc -fsanitize=thread -o app app.c && ./app