Mablr mablr

Foundry Benchmark Results

Date: 2026-04-24 23:10:24

Repositories Tested

Foundry PR #15070 follow-up — real invariant campaign

Real forge invariant project used to measure end-to-end wall-clock and peak RSS of the corpus-seed worker fan-out (companion to the micro-benchmark harness at https://gist.github.com/mablr/b87b9d75f3723b3de8b5796c72331ae2).

Target.sol exposes a large branch surface (1600 independent branch targets, ~3200 EVM edges, split across 8 functions to stay within solc's parser recursion limit) so coverage-guided invariant fuzzing accumulates a sizeable persisted corpus (~370 entries, depth-2000 sequences).

Decofe Perf PR Benchmark Plan

Static diff triage for deciding which PRs to combine and run through scfuzzbench. The goal is to avoid benchmarking dozens of tiny AI PRs one-by-one.

Run These scfuzzbench Branches

These are the concrete grouped benchmark candidates, in priority order.

1. Invariant campaign steady-state overhead

Foundry v1.8.0

Foundry v1.8.0 is a testing-heavy release. It ships native symbolic testing, mutation testing, and major fuzzing and invariant improvements, while making isolate mode the default after a series of compatibility fixes. This release also ships the Rust foundryup, expands forge lint with a large set of security and gas detectors, and continues the structured CLI output work across Forge and Cast.

Highlights

Testing, Fuzzing, and Symbolic Execution

Testing is the main theme of v1.8.0. The release expands the ways Forge can find failures, make them reproducible, and turn them into smaller actionable counterexamples.

	use std::{env, hint::black_box};

	#[derive(Clone)]
	struct CorpusEntry {
	tx_seq: Vec<u8>,
	cmp_seq: Vec<u8>,
	is_favored: bool,
	}

	#[derive(Clone, Default)]