[RFC] Add Comprehensive Performance Benchmark Suite for Intelligent Contract Execution

[riyannode]

Proposal

Add a dedicated BenchmarkRunner class and gltest bench CLI command to the GenLayer Testing Suite. This delivers production-grade performance benchmarks for Intelligent Contracts (deterministic + non-deterministic paths) as outlined in the attached analysis.

Motivation

GenLayer Intelligent Contracts introduce LLM calls, web fetches, and equivalence-based consensus. Current testing supports statistical analysis via .analyze(runs=100), but lacks standardized wall-clock, resource, and throughput benchmarks. This feature enables:

• Regression testing
• LLM provider comparison
• Gas-equivalent metering validation
• Studio-mode network simulation under load

Proposed Implementation

• New module: gltest/bench/benchmark_runner.py
• CLI command: gltest bench --mode direct|studio --workload llm-heavy --validators 8 --iterations 1000
• Core metrics: mean/p95 latency (ms), CPU/memory peak, TPS, consensus overhead
• Reuse existing fixtures (direct_vm, studio_network) and psutil + statistics

Reference implementation (ready to add):

from genlayer.test import ContractFactory
import time, psutil, statistics, json
from pathlib import Path

class BenchmarkRunner:
    def __init__(self, mode: str = "direct"):
        self.factory = ContractFactory()
        self.mode = mode

    def run(self, contract_code: str, method: str, inputs: list, iterations: int = 1000):
        contract = self.factory.deploy(contract_code)
        times = []
        resources = []
        for _ in range(iterations):
            start = time.perf_counter_ns()
            result = contract.call(method, *inputs)
            duration_ms = (time.perf_counter_ns() - start) / 1e6
            times.append(duration_ms)
            resources.append(psutil.Process().memory_info().rss / 1024**2)
        return {
            "mean_ms": statistics.mean(times),
            "p95_ms": statistics.quantiles(times, n=20)[18],
            "memory_mb_peak": max(resources),
            "results": json.dumps({"times": times})
        }