Understand EVM bytecode – Part 4

In previous section:

Understand EVM bytecode – Part 1

Understand EVM bytecode – Part 2

Understand EVM bytecode – Part 3

We have talked about how different Solidity data types are implemented in storage. For this section we will dig more about memory and its usage in external calls in this section.

We have learned some basics about memory from previous sections. We know memory is designed for HASH calculation or interactions with external calls or returns. The memory structure has reserved 0x0 and 0x20 for HASH calculation. At address 0x40 it will store a free memory pointer for future use. When some memory need to be allocated, the pointer can be adjusted accordingly so the allocated memory won’t be re-visited again. Also, Memory only accessible during contract execution. Once execution is finished, its contents are discarded. Comparing to storage, it is more like the RAM in computer.

Let’s look at all the opcodes which depends on memory besides MLOAD and MSTORE:


We can see besides the SHA3 is used for calculate HASH value, most of the rest opcodes are related to interactions with EVM. It includes loading data from EVM data payload or code , or arranging data for external calls and returns.

First, let’s look at opcode CALLDATACOPY. From Solidity document, “calldatacopy(t, f, s)” is defined as “copy s bytes from calldata at position f to mem at position t”. If you got experience to analyze bytecode level contracts, you may observe that another similar opcode CALLDATALOAD is more popular than this one. But the difference of these 2 opcodes is that CALLDATALOAD only load 32-bytes data into the stack instead of memory. If the contract public function only uses integers in their arguments, then CALLDATALOAD is good enough for the calls. The format of the data payload will be as follows:

0x00: <function signature hash>
0x04: <first integer argument if any>
0x24: <second integer argument if any>


However, there are exceptions. Functions can support more data types in Solidity other than pure integers. For example of a function with struct or fixed size arrays:

contract Data7 {
address a;
address b;

function test(address [2] addresses) public returns (bool) {
a = addresses[0];
b = addresses[1];
return true;

When you look at the bytecode generated from above code, you can see :

temp0 = mload(0x40);
mstore(0x40,(0x40 + temp0));

Apparently, this time CALLDATACOPY is used to copy the whole argument into the memory for future use. It is not just fixed size of arrays. For any parameter which has fixed size like struct. CALLDATACOPY will be the one for the work.

However, things can be even more complicated when there is dynamic arrays. For example:

contract Data8 {
address a;
address b;

function test(address [] addresses) public returns (bool) {
a = addresses[0];
b = addresses[1];
return true;

We can see there is only one function test() uses an address array as the argument. So how the data will be arranged inside the data payload? Let’s still go into the bytecode for the truth. Here is the code snippet before calling the test() function:

temp0 = mload(0x40);
temp1 = msg.data(0x4)
mstore(0x40,(0x20 + (temp0 + (msg.data((0x4 + temp1)) * 0x20))));
mstore(temp0,msg.data((0x4 + temp1)));
calldatacopy((temp0 + 0x20),(0x24 + temp1),(msg.data((0x4 + temp1)) * 0x20));
var1 = test(temp0);

It might not be that obvious to get the logic of the data payload structure. But don’t worry. Let’s go through it together. The first line “temp0 = mload(0x40);

” is a very popular one. It gets the free memory pointer from memory address 0x40 into variable temp0. Then, temp1 will be assigned with the value get from data payload at offset 0x4, which regularly hold the first parameter when the type is integer. However, apparently it is not finished yet in this case. This value in temp1 will be used as an offset to locate the data starting from 0x4. This value can be shown as “msg.data((0x4 + temp1))”. From above code snippet, this value is the length of the array. The size of each item in the array is 0x20. So “mstore(0x40,(0x20 + (temp0 + (msg.data((0x4 + temp1)) * 0x20))));” will adjust the free memory pointer to save a piece of memory for this argument. Then, the array length will be copied into the old free memory pointer and items of array will be copied too by CALLDATACOPY. Finally, the memory pointer will be transferred to test() function for operation.
After we have some basic idea how EVM data payload was arranged, we can see how CALL related opcodes work and how memory is involved in. Solidity document states “call(g, a, v, in, insize, out, outsize) – call contract at address a with input mem[in..(in+insize)) providing g gas and v wei and output area mem[out..(out+outsize)) returning 0 on error (eg. out of gas) and 1 on success”. So when you use this opcode to call other smart contract you need to supply all of the information. Let’s look at a real example of external contract reference:

pragma solidity ^0.4.18;
contract Deployed {
function a() public pure returns (uint) {}
contract Existing {
Deployed dc;
function Existing(address _t) public {
dc = Deployed(_t);
function getA() public view returns (uint result) {
return dc.a();

Here we defined 2 smart contracts. The getA() function in the Existing contract will call the external function a() in Deployed. Here is the compiled bytecode:

temp0 = mload(0x40);
var11 =uint160(sload(0x0));
var6 = var11.gas(gasleft).value(0).call(temp0,4,temp0,0x20);

As always, the free memory pointer was loaded into variable temp0. Then the function hash 0xDBE671F is saved into the free memory. Then var11 will hold the first storage variable which is dc in this case. The require line will check if the address in var11 holds a contract address. Finally, this address will be used to make an external call “var11.gas(gasleft).value(0).call(temp0,4,temp0,0x20)”. The parameters (temp0 and 4) in this call is the data payload it wants to send to the external contract. In this case it sends 4 bytes from the address temp0.  From early code we know the 4 bytes are the function signature hash value for a(). Because there is no other arguments in the function, only 4 bytes function hash is needed to make this call. If the function you want to call does have arguments, then the compiler will arrange the memory in the way we have discussed early for the call. The parameters (temp0 and 0x20) will have the return data form the external call. EVM will get the data returned from the external call and put it into the memory address opcode CALL specified.

There is one thing worth mention about external call is that there is some hard-coded address for builtin function in Solidity. If you look at some real world EVM bytecode you will always find some external calls using addresses 1,2,3 and 4. Apparently they are not normal smart contract addresses. I have searched Internet, and not a lot of information about them can be found. But I do find some clue from Solidity documents:

Expressions that might have a side-effect on memory allocation are allowed, but those that might have a side-effect on other memory objects are not. The built-in functions keccak256, sha256, ripemd160, ecrecover, addmod and mulmod are allowed (even though they do call external contracts).

Check what it says in the bracket. They call external contracts for these built-in functions. Then I just wrote a smart contract with all these functions inside, then I found following mapping for the hard-coded addresses:

1 - ecrecover
2 - sha256
3 - ripemd160
4 - sha3

For recent version of Solidity compiler, SHA3 has its own opcode, so no external call is needed for this calculation. But you can still see some online smart contracts are using 0x4 to send external call for this functionality.

So far, we have discussed how memory plays an important role in the EVM environment, especially when making external calls to other smart contracts. This is would be the last section for this series. We have talked most of the things you might encounter when you want to analyze the EVM bytecode. Hope it can help you a little bit to understand how it works.

One thought on “Understand EVM bytecode – Part 4

  1. Pingback: Smart Contract Insight – an online EVM decompiler – Trustlook

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s