Awesome
cairo_vm.c
This is a work in progress implementation of the Cairo VM in C
. The reasons for doing this include:
- Having a diversity of implementations helps find bugs and make the whole ecosystem more resilient.
- It's a good opportunity to extensively document the VM in general, as currently the documentation on its internals is very scarce and mostly lives on the minds of a few people.
Requirements
- A working C compiler with
C11
support. clang-format
installed in the system.- Rust
- cmake
- pkg-config
- cpputest
- Docker (for valgrind on MacOS)
Local development
To build the vm, run:
make
This will compile the code and store the cairo_vm
binary under the build
directory, so you can run it like so:
./build/cairo_vm
To format the code, do
make fmt
To remove all compilation objects:
make clean
Note: When building outside of macos, you may have to run this command after building the collections lib to make the system's runtime aware of the location of the new library:
sudo ldconfig
Valgrind/Asan on MacOS
To run valgrind
on MacOS, first run:
make docker_build
This will build a Linux Docker image with all the dependencies needed to build the vm and run valgrind
. Then run:
make docker_run
This will run a new container from the built image and will execute bash by using the repo root as the working directory.
Finally, run:
make SANITIZER_FLAGS=-fno-omit-frame-pointer docker_valgrind
Tests
Tests are located in the test
directory. To run them:
make test
Project Guidelines
Because C
as a language leaves everything to the programmer, a lot of things can go wrong; discipline is required. This repo has strict rules to address this. They are:
- PRs addressing performance are forbidden. We are currently concerned with making it work without bugs and nothing more.
- All PRs must contain tests. Code coverage has to be above 98%.
- To check for security and other types of bugs, the code will be fuzzed extensively.
- To catch memory errors, all code is compiled by default with AddressSanitizer. The CI pipeline runs Valgrind as well.
- PRs must be accompanied by its corresponding documentation. A book will be written documenting the entire inner workings of it, so anyone can dive in to a Cairo VM codebase and follow it along.
Documentation
High Level Overview
The Cairo virtual machine is meant to be used in the context of STARK validity proofs. What this means is that the point of Cairo is not just to execute some code and get a result, but to prove to someone else that said execution was done correctly, without them having to re-execute the entire thing. The rough flow for it looks like this:
- A user writes a Cairo program.
- The program is compiled into Cairo's VM bytecode.
- The VM executes said code and provides a trace of execution, i.e. a record of the state of the machine and its memory at every step of the computation.
- This trace is passed on to a STARK prover, which creates a cryptographic proof from it, attesting to the correct execution of the program.
- The proof is passed to a verifier, who checks that the proof is valid in a fraction of a second, without re-executing.
The main three components of this flow are:
- A Cairo compiler to turn a program written in the Cairo programming language into bytecode.
- A Cairo VM to then execute it and generate a trace.
- A STARK prover and verifier so one party can prove correct execution, while another can verify it.
While this repo is only concerned with the second component, it's important to keep in mind the other two; especially important are the prover and verifier that this VM feeds its trace to, as a lot of its design decisions come from them. This virtual machine is designed to make proving and verifying both feasible and fast, and that makes it quite different from most other VMs you are probably used to.
Basic VM flow
Our virtual machine has a very simple flow:
- Take a compiled cairo program as input. You can check out an example program here, and its corresponding compiled version here.
- Run the bytecode from the compiled program, doing the usual
fetch->decode->execute
loop, running until program termination. - On every step of the execution, record the values of each register.
- Take the register values and memory at every step and write them to a file, called the
execution trace
.
Barring some simplifications we made, this is all the Cairo VM does. The two main things that stand out as radically different are the memory model and the use of Field Elements
to perform arithmetic. Below we go into more detail on each step, and in the process explain the ommisions we made.
Architecture
The Cairo virtual machine uses a Von Neumann architecture with a Non-deterministic read-only memory. What this means, roughly, is that memory is immutable after you've written to it (i.e. you can only write to it once); this is to make the STARK proving easier, but we won't go into that here.
Memory Segments and Relocation
The process of memory allocation in a contiguous write-once memory region can get pretty complicated. Imagine you want to have a regular call stack, with a stack pointer pointing to the top of it and allocation and deallocation of stack frames and local variables happening throughout execution. Because memory is immutable, this cannot be done the usual way; once you allocate a new stack frame that memory is set, it can't be reused for another one later on.
Because of this, memory in Cairo is divided into segments
. This is just a way of organizing memory more conveniently for this write-once model. Each segment is nothing more than a contiguous memory region. Segments are identified by an index
, an integer value that uniquely identifies them.
Memory cells
(i.e. values in memory) are identified by the index of the segment they belong to and an offset
into said segment. Thus, the memory cell {2,0}
is the first cell of segment number 2
.
Even though this segment model is extremely convenient for the VM's execution, the STARK prover needs to have the memory as just one contiguous region. Because of this, once execution of a Cairo program finishes, all the memory segments are collapsed into one; this process is called Relocation
. We will go into more detail on all of this below.
Registers
There are only three registers in the Cairo VM:
- The program counter
pc
, which points to the next instruction to be executed. - The allocation pointer
ap
, pointing to the next unused memory cell. - The frame pointer
fp
, pointing to the base of the current stack frame. When a new function is called,fp
is set to the currentap
. When the function returns,fp
goes back to its previous value. The VM creates new segments whenever dynamic allocation is needed, so for example the cairo analog to a RustVec
will have its own segment. Relocation at the end meshes everything together.
Instruction Decoding/Execution
TODO: explain the components of an instruction (dst_reg
, op0_reg
, etc), what each one is used for and how they're encoded/decoded.
Felts
Felts, or Field Elements, are cairo's basic integer type. Every variable in a cairo vm that is not a pointer is a felt. From our point of view we could say a felt in cairo is an unsigned integer in the range [0, CAIRO_PRIME). This means that all operations are done modulo CAIRO_PRIME. The CAIRO_PRIME is 0x800000000000011000000000000000000000000000000000000000000000001, which means felts can be quite big (up to 252 bits), luckily, we have the Lambdaworks library to help with handling these big integer values and providing fast and efficient modular arithmetic.
More on memory
The cairo memory is made up of contiguous segments of variable length identified by their index. The first segment (index 0) is the program segment, which stores the instructions of a cairo program. The following segment (index 1) is the execution segment, which holds the values that are created along the execution of the vm, for example, when we call a function, a pointer to the next instruction after the call instruction will be stored in the execution segment which will then be used to find the next instruction after the function returns. The following group of segments are the builtin segments, one for each builtin used by the program, and which hold values used by the builtin runners. The last group of segments are the user segments, which represent data structures created by the user, for example, when creating an array on a cairo program, that array will be represented in memory as its own segment.
An address (or pointer) in cairo is represented as a relocatable
value, which is made up of a segment_index
and an offset
, the segment_index
tells us which segment the value is stored in and the offset
tells us how many values exist between the start of the segment and the value.
As the cairo memory can hold both felts and pointers, the basic memory unit is a maybe_relocatable
, a variable that can be either a relocatable
or a felt
While memory is continous, some gaps may be present. These gaps can be created on purpose by the user, for example by running:
[ap + 1] = 2;
Where a gap is created at ap. But they may also be created indireclty by diverging branches, as for example one branch may declare a variable that the other branch doesn't, as memory needs to be allocated for both cases if the second case is ran then a gap is left where the variable should have been written.
Memory API
The memory can perform the following basic operations:
-
memory_add_segment
: Creates a new, empty segment in memory and returns a pointer to its start. Values cannot be inserted into a memory segment that hasn't been previously created. -
memory_insert
: Inserts amaybe_relocatable
value at an address indicated by arelocatable
pointer. For this operation to succeed, the pointer's segment_index must be an existing segment (created usingmemory_add_segment
), and there mustn't be a value stored at that address, as the memory is immutable after its been written once. If there is a value already stored at that address but it is equal to the value to be inserted then the operation will be successful. -
memory_get
: Fetches amaybe_relocatable
value from a memory address indicated by arelocatable
pointer.
Other operations:
memory_load_data
: This is a convenience method, which takes an array ofmaybe_relocatable
and inserts them contiguosuly in memory by callingmemory_insert
and advancing the pointer by one after each insertion. Returns a pointer to the next free memory slot after the inserted data.
Memory Relocation
During execution, the memory consists of segments of varying length, and they can be accessed by indicating their segment index, and the offset within that segment. When the run is finished, a relocation process takes place, which transforms this segmented memory into a contiguous list of values. The relocation process works as follows:
1- The size of each segment is calculated (The size is equal to the highest offset within the segment + 1, and not the amount of maybe_relocatable
values, as there can be gaps)
2- A base is assigned to each segment by accumulating the size of the previous segment. The first segment's base is set to 1.
3- All relocatable
values are converted into a single integer by adding their offset
value to their segment's base calculated in the previous step
For example, if we have this memory represented by address, value pairs:
0:0 -> 1
0:1 -> 4
0:2 -> 7
1:0 -> 8
1:1 -> 0:2
1:4 -> 0:1
2:0 -> 1
Step 1: Calculate segment sizes:
0 -> 3
1 -> 5
2 -> 1
Step 2: Assign a base to each segment:
0 -> 1
1 -> 4 (1 + 3)
2 -> 9 (4 + 5)
Step 3: Convert relocatables to integers
1 (base[0] + 0) -> 1
2 (base[0] + 1) -> 4
3 (base[0] + 2) -> 7
4 (base[1] + 0) -> 8
5 (base[1] + 1) -> 3 (base[0] + 2)
.... (memory gaps)
8 (base[1] + 4) -> 2 (base[0] + 1)
9 (base[2] + 0) -> 1
Program parsing
The input of the Virtual Machine is a compiled Cairo program in Json format. The main part of the file are listed below:
-
data: List of hexadecimal values that represent the instructions and immediate values defined in the cairo program. Each hexadecimal value is stored as a maybe_relocatable element in memory, but they can only be felts because the decoder has to be able to get the instruction fields in its bit representation.
-
debug_info: This field provides information about the instructions defined in the data list. Each one is identified with its index inside the data list. For each one it contains information about the cairo variables in scope, the hints executed before that instruction if any, and its location inside the cairo program.
-
hints: All the hints used in the program, ordered by the pc offset at which they should be executed.
-
identifiers: User-defined symbols in the Cairo code representing variables, functions, classes, etc. with unique names. The expected offset, type and its corresponding information is provided for each identifier
For example, the identifier representing the main function (usually the entrypoint of the program) is of
function
type, and a list of decorators wrappers (if there are any) are provided as additional information. Another example is a user defined struct, is ofstruct
type, it provides its size, the members it contains (with its information) and more. -
main_scope: Usually something like main. All the identifiers associated with main function will be identified as main.identifier_name. Useful to identify the entrypoint of the program.
-
prime: The cairo prime in hexadecimal format. As explained above, all arithmetic operations are done over a base field, modulo this primer number.
-
reference_manager: Contains information about cairo variables. This information is useful to access to variables when executing cairo hints.
In this project, we use a C++ library called simdjson, the json is stored in a custom structure which the vm can use to run the program and create a trace of its execution.
Code walkthrough/Write your own Cairo VM
Lets begin by creating the basic types and structures for our VM:
Felt
As anyone who has ever written a cairo program will know, everything in cairo is a Felt. We can think of it as our unsigned integer. In this project, we use the Lambdaworks
library to abstract ourselves from modular arithmetic.
TODO: Instructions on how to use Lambdaworks felt from C
Relocatable
This is how cairo represents pointers, they are made up of segment_index
, which segment the variable is in, and offset
, how many values exist between the start of a segment and the variable. We represent them like this:
typedef struct relocatable {
unsigned int segment_index;
unsigned int offset;
} relocatable;
MaybeRelocatable
As the cairo memory can hold both felts and relocatables, we need a data type that can represent both in order to represent a basic memory unit, therefore:
union maybe_relocatable_value {
struct relocatable relocatable;
felt_t felt;
};
typedef struct maybe_relocatable {
union maybe_relocatable_value value;
bool is_felt;
} maybe_relocatable;
We use two structs to represent it as we need to be able to distinguish between the two union types during execution.
Memory
As we previously described, the memory is made up of a series of segments of variable length, each containing a continuous sequence of maybe_relocatable
elements. Memory is also immutable, which means that once we have written a value into memory, it can't be changed.
There are multiple valid ways to represent this memory structure, but the simples way to represent it is by using a hashmap, maping a relocatable
address to a maybe_relocatable
value.
As we don't have an actual representation of segments, we have to keep track of the number of segments.
In this project we decided to use the Collections-C library for our data structures, but you can choose any other library (or implement your own!).
typedef struct memory {
unsigned int num_segments;
CC_HashTable *data;
} memory;
Now we can define the basic memory operations:
Add Segment
As we are using a hashmap, we dont have to allocate memory for the new segment, so we only have to raise our segment counter and return the first address of the new segment:
relocatable memory_add_segment(memory *memory) {
relocatable rel = {memory->num_segments, 0};
memory->num_segments += 1;
return rel;
}
Insert
Here we need to make perform some checks to make sure that the memory remains consistent with its rules:
-
We must check that insertions are performed on previously-allocated segments, by checking that the address's segment_index is lower than our segment counter
-
We must check that we are not mutating memory we have previously written, by checking that the memory doesn't already contain a value at that address that is not equal to the one we are inserting
ResultMemory memory_insert(memory *mem, relocatable ptr, maybe_relocatable value) {
// Guard out of bounds writes
if (ptr.segment_index >= mem->num_segments) {
ResultMemory error = {.is_error = true, .value = {.error = Insert}};
return error;
}
// Guard overwrites
maybe_relocatable *prev_value = NULL;
if (cc_hashtable_get(mem->data, &ptr, (void *)&prev_value) == CC_OK) {
if (maybe_relocatable_equal(prev_value, &value)) {
ResultMemory ok = {.is_error = false, .value = {.none = 0}};
return ok;
} else {
ResultMemory error = {.is_error = true, .value = {.error = Insert}};
return error;
}
}
// Write new value
// Allocate new values
relocatable *ptr_alloc = malloc(sizeof(relocatable));
*ptr_alloc = ptr;
maybe_relocatable *value_alloc = malloc(sizeof(maybe_relocatable));
*value_alloc = value;
if (cc_hashtable_add(mem->data, ptr_alloc, value_alloc) == CC_OK) {
ResultMemory ok = {.is_error = false, .value = {.none = 0}};
return ok;
}
ResultMemory error = {.is_error = true, .value = {.error = Insert}};
return error;
}
Get
This is the easiest operation, as we only need to fetch the value from our hashmap:
ResultMemory memory_get(memory *mem, relocatable ptr) {
maybe_relocatable *value = NULL;
if (cc_hashtable_get(mem->data, &ptr, (void *)&value) == CC_OK) {
ResultMemory ok = {.is_error = false, .value = {.memory_value = *value}};
return ok;
}
ResultMemory error = {.is_error = true, .value = {.error = Get}};
return error;
}
Then we have some convenience methods that make specific functions of the vm more readable:
Load Data
This method inserts a contiguous array of values starting from a certain addres in memory, and returns the next address after the inserted values. This is useful when inserting the program's instructions in memory.
In order to perform this operation, we only need to iterate over the array, inserting each value at the address indicated by ptr
while advancing the ptr with each iteration and then return the final ptr.
ResultMemory memory_load_data(memory *mem, relocatable ptr, CC_Array *data) {
// Load each value sequentially
CC_ArrayIter data_iter;
cc_array_iter_init(&data_iter, data);
maybe_relocatable *value = NULL;
while (cc_array_iter_next(&data_iter, (void *)&value) != CC_ITER_END) {
// Insert Value
if (memory_insert(mem, ptr, *value).is_error) {
ResultMemory error = {.is_error = true, .value = {.error = LoadData}};
return error;
}
// Advance ptr
ptr.offset += 1;
}
ResultMemory ok = {.is_error = false, .value = {.ptr = ptr}};
return ok;
}
Builtins
TODO
Hints
TODO