performance

name	performance
version	3.0.0
description	Production-grade skill for C++ performance optimization. Covers profiling, benchmarking, cache optimization, SIMD vectorization, multithreading, and lock-free programming techniques.
sasmp_version	1.3.0
skill_version	3.0.0
bonded_agent	05-performance-optimizer
bond_type	PRIMARY_BOND
category	development
parameters	{"optimization_target":{"type":"string","required":false,"enum":["throughput","latency","memory","cpu","all"],"default":"all","description":"Primary optimization target"},"profiling_tool":{"type":"string","required":false,"enum":["perf","vtune","valgrind","tracy","instruments"],"description":"Profiling tool to use"},"optimization_level":{"type":"string","required":false,"enum":["quick_wins","moderate","aggressive"],"default":"moderate","description":"Depth of optimization effort"},"maintain_readability":{"type":"boolean","required":false,"default":true,"description":"Whether to prioritize code readability"}}
error_handling	{"retry_logic":{"max_attempts":3,"backoff":"exponential","initial_delay_ms":1000,"max_delay_ms":16000,"jitter":true},"fallback":{"on_benchmark_unstable":"increase_iterations","on_profiling_fail":"use_alternative_tool","on_no_improvement":"try_different_approach","on_regression":"rollback_and_analyze"},"validation":{"verify_no_regression":true,"statistical_significance":true,"test_multiple_inputs":true}}

Performance Skill

Production-Grade Development Skill | C++ Performance Engineering

Optimize C++ code for maximum performance through profiling, analysis, and targeted optimization.

---

Golden Rules

┌─────────────────────────────────────────────────────────────────┐
│  1. MEASURE first - never optimize without profiling data       │
│  2. OPTIMIZE hotspots - focus on the 20% that takes 80% time   │
│  3. VERIFY improvements - benchmark before and after            │
│  4. MAINTAIN readability - premature optimization is evil       │
└─────────────────────────────────────────────────────────────────┘

---

Profiling Tools

Linux perf

# Record CPU profile
perf record -g ./program
perf report

# Flamegraph generation
perf script | stackcollapse-perf.pl | flamegraph.pl > flame.svg

# Hardware counters
perf stat -e cache-misses,cache-references,instructions,cycles ./program

# Specific function profiling
perf record -g -e cycles:u --call-graph dwarf ./program

Valgrind Callgrind

# Instruction-level profiling
valgrind --tool=callgrind ./program
kcachegrind callgrind.out.*

# Cache simulation
valgrind --tool=cachegrind ./program
cg_annotate cachegrind.out.*

Google Benchmark

#include <benchmark/benchmark.h>

static void BM_VectorPushBack(benchmark::State& state) {
    for (auto _ : state) {
        std::vector<int> v;
        v.reserve(state.range(0));  // Fair comparison
        for (int i = 0; i < state.range(0); ++i) {
            v.push_back(i);
        }
        benchmark::(v.());
        benchmark::();
    }
    state.(state.());
}
(BM_VectorPushBack)
    ->(,  << )
    ->(benchmark::oN);

();