deal() performance

[frontier159]

I've noticed the performance of StdCheats::deal() is orders of magnitude slower than if there were a mint function directly on the ERC20.

This is particularly exposed within invariant tests which had a large speedup after I updated mock tokens to add a direct mint() function on it. This wouldn't always be possible however, I think it could be a good hunting ground for performance for various projects.

Example fuzz test below:

pragma solidity ^0.8.19;
// SPDX-License-Identifier: UNLICENSED

import {Test} from "forge-std/Test.sol";
import {MockERC20} from "forge-std/mocks/MockERC20.sol";

contract MintableERC20 is MockERC20 {   
    function mint(address to, uint256 amount) external {
        _mint(to, amount);
    }
}

contract DealPerformanceTest is Test {
    MintableERC20 internal token;

    function setUp() public {
        token = new MintableERC20();
        token.initialize("TOKEN", "TOKEN", 18);
    }

    /// forge-config: default.fuzz.runs = 100000
    function test_mint(address to, uint256 amount) public {
        token.mint(to, amount);
    }

    /// forge-config: default.fuzz.runs = 100000
    function test_deal(address to, uint256 amount) public {
        deal(address(token), to, amount, true);
    }
}

Ran 1 test for test/foundry/unit/DealPerformance.t.sol:DealPerformanceTest
[PASS] test_deal(address,uint256) (runs: 100000, μ: 297655, ~: 297734)
Suite result: ok. 1 passed; 0 failed; 0 skipped; finished in 47.83s (47.82s CPU time)

Ran 1 test for test/foundry/unit/DealPerformance.t.sol:DealPerformanceTest
[PASS] test_mint(address,uint256) (runs: 100000, μ: 51723, ~: 52197)
Suite result: ok. 1 passed; 0 failed; 0 skipped; finished in 2.23s (2.22s CPU time)

So mint is > 20x faster

$ forge --version
forge Version: 1.0.0-nightly
Commit SHA: 8da73730b4033553589aa67ef404e527149c2e92
Build Timestamp: 2025-04-02T06:02:24.621284000Z (1743573744)
Build Profile: maxperf

[frontier159]

NB: I wasn't sure if this should be reported here or in the foundry repo. @mattsse you may be able to guide me on that

[DaniPopes]

they're not really comparable, deal is a lot more complex because it uses stdStorage and many calls to vm

here's the flamegraph: https://share.firefox.dev/4leDn8X

it's about 40% in the EVM, 60% in Foundry inspectors; not much specific to 'deal' other than vm.record which takes about 6%

if this is a problem for you, you can try compiling with optimizations and see if it's faster overall

[frontier159]

Thanks - i raised in case there may be a more efficient way to implement the deal at as a lower primative rather than than many VM calls.

If nothing can be done then can be closed and I'll either use a custom mock or etch where possible. It's worth the effort in this case