Is the max function written in assembly gas optimized

[XuHugo]

I used src/utils/FixedPointMathLib.sol for testing and found that the gas consumption of the max function was not as low as that of the ternary operation.Could you help me clear up my doubts.

Tool:
foundry

Code：
`

function soliditymax(uint256 x, uint256 y) public pure returns (uint256 z) {
    z = x > y ? x : y;
}


function assemblymax(uint256 x, uint256 y) public pure returns (uint256 z) {
    assembly {
        z := xor(x, mul(xor(x, y), gt(y, x)))
    }
}

`
Test result：

`

╭----------------------------------+-----------------+-----+--------+-----+---------╮
| src/MathOpt.sol:MathOpt Contract | | | | | |
+===================================================================================+
| Deployment Cost | Deployment Size | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| 182156 | 581 | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| | | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| Function Name | Min | Avg | Median | Max | # Calls |
|----------------------------------+-----------------+-----+--------+-----+---------|
| assemblymax | 790 | 790 | 790 | 790 | 1 |
╰----------------------------------+-----------------+-----+--------+-----+---------╯

╭----------------------------------+-----------------+-----+--------+-----+---------╮
| src/MathOpt.sol:MathOpt Contract | | | | | |
+===================================================================================+
| Deployment Cost | Deployment Size | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| 182156 | 581 | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| | | | | | |
|----------------------------------+-----------------+-----+--------+-----+---------|
| Function Name | Min | Avg | Median | Max | # Calls |
|----------------------------------+-----------------+-----+--------+-----+---------|
| soliditymax | 776 | 776 | 776 | 776 | 1 |
╰----------------------------------+-----------------+-----+--------+-----+---------╯

`

[Vectorized]

soliditymax(uint256,uint256) has a function selector of 0x38833895.
assemblymax(uint256,uint256) has a function selector of 0x4f02e986.

This puts soliditymax(uint256,uint256) before assemblymax(uint256,uint256) in the function dispatch, which makes it looks like it has lower gas.

To do a proper test for functions with very small difference in gas, you have to make two contracts with the same function selectors.

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.4;

import "./utils/SoladyTest.sol";
import {FixedPointMathLib} from "../src/utils/FixedPointMathLib.sol";

contract MaxSolidityTest is SoladyTest {
    function testMax(uint256 a, uint256 b) public {
        this.max(a, b);
    }

    function max(uint256 a, uint256 b) public pure returns (uint256) {
        return _max(a, b);
    }

    function _max(uint256 a, uint256 b) internal pure returns (uint256) {
        return a > b ? a : b;
    }
}

contract MaxAssemblyTest is SoladyTest {
    function testMax(uint256 a, uint256 b) public {
        this.max(a, b);
    }

    function max(uint256 a, uint256 b) public pure returns (uint256) {
        return FixedPointMathLib.max(a, b);
    }
}

Ran 2 tests for test/FixedPointMathLib.t.sol:MaxAssemblyTest
[PASS] testMax(uint256,uint256) (runs: 299, μ: 1003, ~: 1003)
[PASS] test__codesize() (gas: 1103)
Suite result: ok. 2 passed; 0 failed; 0 skipped; finished in 14.36ms (14.15ms CPU time)

Ran 2 tests for test/FixedPointMathLib.t.sol:MaxSolidityTest
[PASS] testMax(uint256,uint256) (runs: 299, μ: 1076, ~: 1083)
[PASS] test__codesize() (gas: 1138)
Suite result: ok. 2 passed; 0 failed; 0 skipped; finished in 6.39ms (6.30ms CPU time)

[XuHugo]

Cool，Thank you very much for your answer。