SP1

Documentation for SP1 users and developers.

SP1 is a performant, open-source zero-knowledge virtual machine (zkVM) that verifies the execution of arbitrary Rust (or any LLVM-compiled language) programs.

SP1 has undergone multiple audits from leading ZK security firms and is currently used in production by many top blockchain teams.

The future of ZK is writing normal code

Zero-knowledge proofs (ZKPs) are one of the most critical technologies to blockchain scaling, interoperability and privacy. But, historically building ZKP systems was extrememly complicated--requiring large teams with specialized cryptography expertise and taking years to go to production.

SP1 provides a performant, general-purpose zkVM that enables any developer to use ZKPs by writing normal code (in Rust), and get cheap and fast proofs. SP1 will enable ZKPs to become mainstream, introducing a a new era of verifiability for all of blockchain infrastructure and beyond.

SP1 enables a diversity of use-cases

ZKPs enable a diversity of use-cases in blockchain and beyond, including:

• Rollups: Use SP1 to generate a ZKP for the state transition function of your rollup and connect to Ethereum, Bitcoin or other chains with full validity proofs or ZK fraud proofs.
• Interoperability: Use SP1 for fast-finality, cross rollup interoperability
• Bridges: Use SP1 to generate a ZKP for verifying consensus of L1s, including Tendermint, Ethereum’s Light Client protocol and more, for bridging between chains.
• Oracles: Use SP1 for large scale computations with onchain state, including consensus data and storage data.
• Aggregation: Use SP1 to aggregate and verify other ZKPs for reduced onchain verification costs.
• Privacy: Use SP1 for onchain privacy, including private transactions and private state.

Why use SP1?

Use-Cases

Zero-knowledge proofs (ZKPs) are a powerful primitive that enable verifiable computation. With ZKPs, anyone can verify a cryptographic proof that a program has executed correctly, without needing to trust the prover, re-execute the program or even know the inputs to the program.

Historically, building ZKP systems has been extremely complicated, requiring large teams with specialized cryptography expertise and taking years to go to production. SP1 is a performant, general-purpose zkVM that solves this problem and creates a future where all blockchain infrastructure, including rollups, bridges, coprocessors, and more, utilize ZKPs via maintainable software written in Rust.

SP1 is especially powerful in blockchain contexts which rely on verifiable computation. Example applications include:

• Rollups: SP1 can be used in combination with existing node infrastructure like Reth to build rollups with ZKP validity proofs or ZK fraud proofs.
• Coprocessors: SP1 can be used to outsource onchain computation to offchain provers to enable use cases such as large-scale computation over historical state and onchain machine learning, dramatically reducing gas costs.
• Light Clients: SP1 can be used to build light clients that can verify the state of other chains, facilitating interoperability between different blockchains without relying on any trusted third parties.

SP1 has already been integrated in many of these applications, including but not limited to:

• SP1 Tendermint: An example of a ZK Tendermint light client on Ethereum powered by SP1.
• SP1 Reth: A performant, type-1 zkEVM written in Rust & SP1 using Reth.
• and many more!

100x developer productivity

SP1 enables teams to use ZKPs in production with minimal overhead and fast timelines.

Maintainable: With SP1, you can reuse existing Rust crates, like revm, reth, tendermint-rs, serde and more, to write your ZKP logic in maintainable, Rust code.

Go to market faster: By reusing existng crates and expressing ZKP logic in regular code, SP1 significantly reduces audit surface area and complexity, enabling teams to go to market with ZKPs faster.

Blazing Fast Performance

SP1 is the fastest zkVM and has blazing fast performance on a variety of realistic blockchain workloads, including light clients and rollups. With SP1, ZKP proving costs are an order of magnitude less than alternative zkVMs or even circuits, making it cost-effective and fast for practical use.

Read more about our benchmarking results here.

Open Source

SP1 is 100% open-source (MIT / Apache 2.0) with no code obfuscation and built to be contributor friendly, with all development done in the open. Unlike existing zkVMs whose constraint logic is closed-source and impossible to audit or modify, SP1 is modularly architected and designed to be customizable from day one. This customizability (unique to SP1) allows for users to add “precompiles” to the core zkVM logic that yield substantial performance gains, making SP1’s performance not only SOTA vs. existing zkVMs, but also competitive with circuits in a variety of use-cases.

What is a zkVM?

A zero-knowledge virtual machine (zkVM) is zero-knowledge proof system that allows developers to prove the execution of arbitrary Rust (or other LLVM-compiled language) programs.

Conceptually, you can think of the SP1 zkVM as proving the evaluation of a function f(x) = y by following the steps below:

• Define f using normal Rust code and compile it to an ELF (covered in the writing programs section).
• Setup a proving key (pk) and verifying key (vk) for the program given the ELF.
• Generate a proof π using the SP1 zkVM that f(x) = y with prove(pk, x).
• Verify the proof π using verify(vk, x, y, π).

As a practical example, f could be a simple Fibonacci program. The process of generating a proof and verifying it can be seen here.

For blockchain applications, the verification usually happens inside of a smart contract.

How does SP1 Work?

At a high level, SP1 works with the following steps:

• Write a program in Rust that defines the logic of your computation for which you want to generate a ZKP.
• Compile the program to the RISC-V ISA (a standard Rust compilation target) using the cargo prove CLI tool (installation instructions here) and generate a RISC-V ELF file.
• SP1 will prove the correct execution of arbitrary RISC-V programs by generating a STARK proof of execution.
• Developers can leverage the sp1-sdk crate to generate proofs with their ELF and input data. Under the hood the sp1-sdk will either generate proofs locally or use a beta version of Succinct's prover network to generate proofs.

SP1 leverages performant STARK recursion that allows us to prove the execution of arbitrarily long programs and also has a STARK -> SNARK "wrapping system" that allows us to generate small SNARK proofs that can be efficiently verified on EVM chains.

Proof System

For more technical details, check out the SP1 technical note that explains our proof system in detail. In short, we use:

• STARKs + FRI over the Baby Bear field
• We use performant STARK recursion that allows us to prove the execution of arbitrarily long programs
• We have a system of performant precompiles that accelerate hash functions and cryptographic signature verification that allow us to get substantial performance gains on blockchain workloads

Installation

SP1 currently runs on Linux and macOS. You can either use prebuilt binaries through sp1up or build the Succinct Rust toolchain and CLI from source.

Requirements

• Git
• Rust (Nightly)
• Docker

Option 1: Prebuilt Binaries (Recommended)

sp1up is the SP1 toolchain installer. Open your terminal and run the following command and follow the instructions:

curl -L https://sp1.succinct.xyz | bash

Then simply follow the instructions on-screen, which will make the sp1up command available in your CLI.

After following the instructions, you can run sp1up to install the toolchain and the cargo prove CLI:

sp1up

This will install two things:

1. The succinct Rust toolchain which has support for the riscv32im-succinct-zkvm-elf compilation target.
2. cargo prove CLI tool that provides convenient commands for compiling SP1 programs and other helper functionality.

You can verify the installation of the CLI by running cargo prove --version:

cargo prove --version

You can check the version of the Succinct Rust toolchain by running:

RUSTUP_TOOLCHAIN=succinct cargo --version

or equivalently:

cargo +succinct --version

If this works, go to the next section to compile and prove a simple zkVM program.

Troubleshooting

Rate-limiting

If you experience rate-limiting when using the sp1up command, you can resolve this by using the --token flag and providing your GitHub token. To create a Github token, follow the instructions here.

Unsupported OS Architectures

Currently our prebuilt binaries are built on Ubuntu 20.04 (22.04 on ARM) and macOS. If your OS uses an older GLIBC version, it's possible these may not work and you will need to build the toolchain from source.

Conflicting cargo-prove installations

If you have installed cargo-prove from source, it may conflict with sp1up's cargo-prove installation or vice versa. You can remove the cargo-prove that was installed from source with the following command:

rm ~/.cargo/bin/cargo-prove

Or, you can remove the cargo-prove that was installed through sp1up:

rm ~/.sp1/bin/cargo-prove

Option 2: Building from Source

Make sure you have installed the dependencies needed to build the rust toolchain from source.

Clone the sp1 repository and navigate to the root directory.

git clone git@github.com:succinctlabs/sp1.git
cd sp1
cd crates
cd cli
cargo install --locked --path .
cd ~
cargo prove build-toolchain

Building the toolchain can take a while, ranging from 30 mins to an hour depending on your machine. If you're on a machine that we have prebuilt binaries for (ARM Mac or x86 or ARM Linux), you can use the following to download a prebuilt version.

cargo prove install-toolchain

To verify the installation of the toolchain, run and make sure you see succinct:

rustup toolchain list

You can delete your existing installation of the toolchain with:

rustup toolchain remove succinct

Quickstart

In this section, we will show you how to create a simple Fibonacci program using the SP1 zkVM.

Create an SP1 Project

Option 1: Cargo Prove New CLI (Recommended)

You can use the cargo prove CLI to create a new project using the cargo prove new <name> command. This command will create a new folder in your current directory.

cargo prove new fibonacci
cd fibonacci

Option 2: Project Template (Solidity Contracts for Onchain Verification)

If you want to use SP1 to generate proofs that will eventually be verified on an EVM chain, you should use the SP1 project template. This Github template is scaffolded with a SP1 program, a script to generate proofs, and also a contracts folder that contains a Solidity contract that can verify SP1 proofs on any EVM chain.

Either fork the project template repository or clone it:

git clone https://github.com/succinctlabs/sp1-project-template.git

Project Overview

Your new project will have the following structure (ignoring the contracts folder, if you are using the project template):

.
├── program
│   ├── Cargo.lock
│   ├── Cargo.toml
│   ├── elf
│   │   └── riscv32im-succinct-zkvm-elf
│   └── src
│       └── main.rs
├── rust-toolchain
└── script
    ├── Cargo.lock
    ├── Cargo.toml
    ├── build.rs
    └── src
        └── bin
            ├── prove.rs
            └── vkey.rs

6 directories, 4 files

There are 2 directories (each a crate) in the project:

• program: the source code that will be proven inside the zkVM.
• script: code that contains proof generation and verification code.

We recommend you install the rust-analyzer extension. Note that if you use cargo prove new inside a monorepo, you will need to add the manifest file to rust-analyzer.linkedProjects to get full IDE support.

Build

Before we can run the program inside the zkVM, it must be compiled to a RISC-V executable using the succinct Rust toolchain. This is called an ELF (Executable and Linkable Format). To compile the program, you can run the following command:

cd program && cargo prove build

which will output the compiled ELF to the file program/elf/riscv32im-succinct-zkvm-elf.

Note: the build.rs file in the script directory will use run the above command automatically to build the ELF, meaning you don't have to manually run cargo prove build every time you make a change to the program!

Execute

To test your program, you can first execute your program without generating a proof. In general this is helpful for iterating on your program and verifying that it is correct.

cd script
RUST_LOG=info cargo run --release -- --execute

Prove

When you are ready to generate a proof, you should run the script with the --prove flag that will generate a proof.

cd script
RUST_LOG=info cargo run --release -- --prove

The output should show something like this:

n: 20
2024-07-23T17:07:07.874856Z  INFO prove_core:collect_checkpoints: clk = 0 pc = 0x2017e8
2024-07-23T17:07:07.876264Z  INFO prove_core:collect_checkpoints: close time.busy=2.00ms time.idle=1.50µs
2024-07-23T17:07:07.913304Z  INFO prove_core:shard: close time.busy=32.2ms time.idle=791ns
2024-07-23T17:07:10.724280Z  INFO prove_core:commit: close time.busy=2.81s time.idle=1.25µs
2024-07-23T17:07:10.725923Z  INFO prove_core:prove_checkpoint: clk = 0 pc = 0x2017e8     num=0
2024-07-23T17:07:10.729130Z  INFO prove_core:prove_checkpoint: close time.busy=3.68ms time.idle=1.17µs num=0
2024-07-23T17:07:14.648146Z  INFO prove_core: execution report (totals): total_cycles=9329, total_syscall_cycles=20
2024-07-23T17:07:14.648180Z  INFO prove_core: execution report (opcode counts):
2024-07-23T17:07:14.648197Z  INFO prove_core:   1948 add
...
2024-07-23T17:07:14.648277Z  INFO prove_core: execution report (syscall counts):
2024-07-23T17:07:14.648408Z  INFO prove_core:   8 commit
...
2024-07-23T17:07:14.648858Z  INFO prove_core: summary: cycles=9329, e2e=9.193968459, khz=1014.69, proofSize=1419780
2024-07-23T17:07:14.653193Z  INFO prove_core: close time.busy=9.20s time.idle=12.2µs
Successfully generated proof!
fib(n): 10946

The program by default is quite small, so proof generation will only take a few seconds locally. After it generates, the proof will be verified for correctness.

Note: When benchmarking proof generation times locally, it is important to note that there is a fixed overhead for proving, which means that the proof generation time for programs with a small number of cycles is not representative of the performance of larger programs (which often have better performance characteristics as the overhead is amortized across many cycles).

Recommended Workflow

Please see the Recommended Workflow section for more details on how to develop your SP1 program and generate proofs.

We strongly recommend that developers who want to use SP1 for non-trivial programs generate proofs on the beta version of our Prover Network. The prover network generates SP1 proofs across multiple machines, reducing latency and also runs SP1 on optimized hardware instances that result in faster + cheaper proof generation times.

We recommend that for any production benchmarking, you use the prover network to estimate latency and costs of proof generation. We also would love to chat with your team directly to help you get started with the prover network--please fill out this form.

Project Template

Another option for getting started with SP1 is to use the SP1 Project Template.

You can use this as a Github template to create a new repository that has a SP1 program, a script to generate proofs, and also a contracts folder that contains a Solidity contract that can verify SP1 proofs on any EVM chain.

Proof Generation Requirements

Local Proving

If you want to generate SP1 proofs locally, this section has an overview of the hardware requirements required. These requires depend on which types of proofs you want to generate and can also change over time as the design of the zKVM evolves.

The most important requirement is CPU for performance/latency and RAM to prevent running out of memory.

CPU

The execution & trace generation of the zkVM is mostly CPU bound, having a high single-core performance is recommended to accelerate these steps. The rest of the prover is mostly bound by hashing/field operations which can be parallelized with multiple cores.

Memory

Our prover requires keeping large matrices (i.e., traces) in memory to generate the proofs. Certain steps of the prover have a minimum memory requirement, meaning that if you have less than this amount of memory, the process will OOM.

This effect is most noticeable when using the PLONK prover, which requires around 128GB of RAM to generate a proof. We use PLONK to avoid having to perform a trusted setup, which other SNARKs like Groth16 require. We have future optimizations planned to reduce the memory requirements of the PLONK prover substantially.

Disk

Disk is required to install the SP1 zkVM toolchain and to install the trused setup artifacts, if you plan to locally build the PLONK prover.

Furthermore, disk is used to checkpoint the state of the program execution, which is required to generate the proofs.

Writing Programs: Setup

In this section, we will teach you how to setup a self-contained crate which can be compiled as a program that can be executed inside the zkVM.

Create Project with CLI (Recommended)

The recommended way to setup your first program to prove inside the zkVM is using the method described in Quickstart which will create a program folder.

cargo prove new <name>
cd program

Manual Project Setup

You can also manually setup a project. First create a new Rust project using cargo:

cargo new program
cd program

Cargo Manifest

Inside this crate, add the sp1-zkvm crate as a dependency. Your Cargo.toml should look like the following:

[workspace]
[package]
version = "0.1.0"
name = "program"
edition = "2021"

[dependencies]
sp1-zkvm = "1.1.0"

The sp1-zkvm crate includes necessary utilities for your program, including handling inputs and outputs, precompiles, patches, and more.

main.rs

Inside the src/main.rs file, you must make sure to include these two lines to ensure that your program properly compiles to a valid SP1 program.

#![no_main]
sp1_zkvm::entrypoint!(main);

These two lines of code wrap your main function with some additional logic to ensure that your program compiles correctly with the RISC-V target.

Writing Programs: Basics

The easiest way to understand how to write programs for the SP1 zkVM is to look at some examples.

Example: Fibonacci

This program is from the examples directory in the SP1 repo which contains several example programs of varying complexity.

//! A simple program that takes a number `n` as input, and writes the `n-1`th and `n`th fibonacci
//! number as an output.

// These two lines are necessary for the program to properly compile.
//
// Under the hood, we wrap your main function with some extra code so that it behaves properly
// inside the zkVM.
#![no_main]
sp1_zkvm::entrypoint!(main);

pub fn main() {
    // Read an input to the program.
    //
    // Behind the scenes, this compiles down to a system call which handles reading inputs
    // from the prover.
    let n = sp1_zkvm::io::read::<u32>();

    // Write n to public input
    sp1_zkvm::io::commit(&n);

    // Compute the n'th fibonacci number, using normal Rust code.
    let mut a = 0;
    let mut b = 1;
    for _ in 0..n {
        let mut c = a + b;
        c %= 7919; // Modulus to prevent overflow.
        a = b;
        b = c;
    }

    // Write the output of the program.
    //
    // Behind the scenes, this also compiles down to a system call which handles writing
    // outputs to the prover.
    sp1_zkvm::io::commit(&a);
    sp1_zkvm::io::commit(&b);
}

As you can see, writing programs is as simple as writing normal Rust.

After you've written your program, you must compile it to an ELF that the SP1 zkVM can prove. To read more about compiling programs, refer to the section on Compiling Programs. To read more about how inputs and outputs work, refer to the section on Inputs & Outputs.

Compiling Programs

Once you have written an SP1 program, you must compile it to an ELF file that can be executed in the zkVM. The cargo prove CLI tool (downloaded during installation) provides convenient commands for compiling SP1 programs.

Development Builds

WARNING: This may not generate a reproducible ELF which is necessary for verifying that your binary corresponds to given source code.

Use the reproducible build system for production builds.

To build a program while developing, simply run the following command in the crate that contains your SP1 program:

cargo prove build

This will compile the ELF that can be executed in the zkVM and put it in the file elf/riscv32im-succinct-zkvm-elf. The output from the command will look something like this:

[sp1]     Compiling version_check v0.9.4
[sp1]     Compiling proc-macro2 v1.0.86
[sp1]     Compiling unicode-ident v1.0.12
[sp1]     Compiling cfg-if v1.0.0
...
[sp1]     Compiling sp1-lib v1.0.1
[sp1]     Compiling sp1-zkvm v1.0.1
[sp1]     Compiling fibonacci-program v0.1.0 (/Users/username/Documents/fibonacci/program)
[sp1]      Finished `release` profile [optimized] target(s) in 8.33s

Under the hood, this CLI command calls cargo build with the riscv32im-succinct-zkvm-elf target and other required environment variables and flags. The logic for this command is defined in the sp1-build crate.

Advanced Build Options

You can pass additional arguments to the cargo prove build command to customize the build process, like configuring what features are enabled, customizing the output directory and more. To see all available options, run cargo prove build --help. Many of these options mirror the options available in the cargo build command.

Production Builds

For production builds of programs, you can build your program inside a Docker container which will generate a reproducible ELF on all platforms. To do so, just use the --docker flag and optionally the --tag flag with the release version you want to use (defaults to latest). For example:

cargo prove build --docker --tag v1.0.1

To verify that your build is reproducible, you can compute the SHA-512 hash of the ELF on different platforms and systems with:

$ shasum -a 512 elf/riscv32im-succinct-zkvm-elf
f9afb8caaef10de9a8aad484c4dd3bfa54ba7218f3fc245a20e8a03ed40b38c617e175328515968aecbd3c38c47b2ca034a99e6dbc928512894f20105b03a203

Build Script

If you want your program crate to be built automatically whenever you build/run your script crate, you can add a build.rs file inside of script/ (at the same level as Cargo.toml of your script crate) that utilizes the sp1-build crate:

fn main() {
    sp1_build::build_program(&format!("{}/../program", env!("CARGO_MANIFEST_DIR")));
}

The path passed in to build_program should point to the directory containing the Cargo.toml file for your program. Make sure to also add sp1-build as a build dependency in script/Cargo.toml:

[build-dependencies]
sp1-build = "1.2.0"

You will see output like the following from the build script if the program has changed, indicating that the program was rebuilt:

[fibonacci-script 0.1.0] cargo:rerun-if-changed=../program/src
[fibonacci-script 0.1.0] cargo:rerun-if-changed=../program/Cargo.toml
[fibonacci-script 0.1.0] cargo:rerun-if-changed=../program/Cargo.lock
[fibonacci-script 0.1.0] cargo:warning=fibonacci-program built at 2024-03-02 22:01:26
[fibonacci-script 0.1.0] [sp1]    Compiling fibonacci-program v0.1.0 (/Users/umaroy/Documents/fibonacci/program)
[fibonacci-script 0.1.0] [sp1]     Finished release [optimized] target(s) in 0.15s
warning: fibonacci-script@0.1.0: fibonacci-program built at 2024-03-02 22:01:26

The above output was generated by running RUST_LOG=info cargo run --release -vv for the script folder of the Fibonacci example.

Advanced Build Options

To configure the build process when using the sp1-build crate, you can pass a BuildArgs struct to to the build_program_with_args function. The build arguments are the same as the ones available from the cargo prove build command.

As an example, you could use the following code to build the Fibonacci example with the docker flag set to true and a custom output directory for the generated ELF:

use sp1_build::{build_program_with_args, BuildArgs};

fn main() {
    let args = BuildArgs {
        docker: true,
        output_directory: "./fibonacci-program".to_string(),
        ..Default::default()
    };
    build_program_with_args("../program", &args);
}

Note: If you want reproducible builds with the build.rs approach, you should use the docker flag and the build_program_with_args function, as shown in the example above.

Inputs and Outputs

In real world applications of zero-knowledge proofs, you almost always want to verify your proof in the context of some inputs and outputs. For example:

• Rollups: Given a list of transactions, prove the new state of the blockchain.
• Coprocessors: Given a block header, prove the historical state of some storage slot inside a smart contract.
• Attested Images: Given a signed image, prove that you made a restricted set of image transformations.

In this section, we cover how you pass inputs and outputs to the zkVM and create new types that support serialization.

Reading Data

Data that is read is not public to the verifier by default. Use the sp1_zkvm::io::read::<T> method:

let a = sp1_zkvm::io::read::<u32>();
let b = sp1_zkvm::io::read::<u64>();
let c = sp1_zkvm::io::read::<String>();

Note that T must implement the serde::Serialize and serde::Deserialize trait. If you want to read bytes directly, you can also use the sp1_zkvm::io::read_vec method.

let my_vec = sp1_zkvm::io::read_vec();

Committing Data

Committing to data makes the data public to the verifier. Use the sp1_zkvm::io::commit::<T> method:

sp1_zkvm::io::commit::<u32>(&a);
sp1_zkvm::io::commit::<u64>(&b);
sp1_zkvm::io::commit::<String>(&c);

Note that T must implement the Serialize and Deserialize trait. If you want to write bytes directly, you can also use sp1_zkvm::io::write_slice method:

let mut my_slice = [0_u8; 32];
sp1_zkvm::io::commit_slice(&my_slice);

Creating Serializable Types

Typically, you can implement the Serialize and Deserialize traits using a simple derive macro on a struct.

use serde::{Serialize, de::Deserialize};

#[derive(Serialize, Deserialize)]
struct MyStruct {
    a: u32,
    b: u64,
    c: String
}

For more complex usecases, refer to the Serde docs.

Example

Here is a basic example of using inputs and outputs with more complex types.

#![no_main]
sp1_zkvm::entrypoint!(main);

use serde::{Deserialize, Serialize};

#[derive(Serialize, Deserialize, Debug, PartialEq)]
struct MyPointUnaligned {
    pub x: usize,
    pub y: usize,
    pub b: bool,
}

pub fn main() {
    let p1 = sp1_zkvm::io::read::<MyPointUnaligned>();
    println!("Read point: {:?}", p1);

    let p2 = sp1_zkvm::io::read::<MyPointUnaligned>();
    println!("Read point: {:?}", p2);

    let p3: MyPointUnaligned = MyPointUnaligned {
        x: p1.x + p2.x,
        y: p1.y + p2.y,
        b: p1.b && p2.b,
    };
    println!("Addition of 2 points: {:?}", p3);
    sp1_zkvm::io::commit(&p3);
}

Patched Crates

We maintain forks of commonly used libraries in blockchain infrastructure to significantly accelerate the execution of certain operations. Under the hood, we use precompiles to achieve tremendous performance improvements in proof generation time.

If you know of a library or library version that you think should be patched, please open an issue or a pull request!

Supported Libraries

Using Patched Crates

To use the patched libraries, you can use corresponding patch entries in your program's Cargo.toml such as:

[patch.crates-io]
sha2-v0-9-8 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha2", branch = "patch-sha2-v0.9.8" }
sha2-v0-10-6 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha2", branch = "patch-sha2-v0.10.6" }
sha2-v0-10-8 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha2", branch = "patch-sha2-v0.10.8" }
sha3-v0-9-8 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha3", branch = "patch-sha3-v0.9.8" }
sha3-v0-10-6 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha3", branch = "patch-sha3-v0.10.6" }
sha3-v0-10-8 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha3", branch = "patch-sha3-v0.10.8" }
crypto-bigint = { git = "https://github.com/sp1-patches/RustCrypto-bigint", branch = "patch-v0.5.5" }
tiny-keccak = { git = "https://github.com/sp1-patches/tiny-keccak", branch = "patch-v2.0.2" }
curve25519-dalek = { git = "https://github.com/sp1-patches/curve25519-dalek", branch = "patch-curve25519-v4.1.3" }
curve25519-dalek-ng = { git = "https://github.com/sp1-patches/curve25519-dalek-ng", branch = "patch-v4.1.1" }
ed25519-consensus = { git = "https://github.com/sp1-patches/ed25519-consensus", branch = "patch-v2.1.0" }
ecdsa-core = { git = "https://github.com/sp1-patches/signatures", package = "ecdsa", branch = "patch-ecdsa-v0.16.9" }
secp256k1 = { git = "https://github.com/sp1-patches/rust-secp256k1", branch = "patch-secp256k1-v0.29.0" }
substrate-bn = { git = "https://github.com/sp1-patches/bn", branch = "patch-v0.6.0" }
bls12_381 = { git = "https://github.com/sp1-patches/bls12_381", branch = "patch-v0.8.8" }

If you are patching a crate from Github instead of from crates.io, you need to specify the repository in the patch section. For example:

[patch."https://github.com/RustCrypto/hashes"]
sha3 = { git = "https://github.com/sp1-patches/RustCrypto-hashes", package = "sha3", branch = "patch-sha3-v0.10.8" }

An example of using patched crates is available in our Tendermint Example.

Ed25519 Acceleration

To accelerate Ed25519 operations, you'll need to patch crates depending on if you're using the ed25519-consensus or ed25519-dalek library in your program or dependencies.

Generally, ed25519-consensus has better performance than ed25519-dalek by a factor of 2.

Patches

Apply the following patches based on what crates are in your dependencies.

• ed25519-consensus

  ed25519-consensus = { git = "https://github.com/sp1-patches/ed25519-consensus", branch = "patch-v2.1.0" }
  

  Note: The curve operations for Ed25519 occur mainly inside of curve25519-dalek-ng, but the crate also exposes a u32_backend feature flag which accelerates signature recovery by 10% over the default u64_backend, which is why ed25519-consensus is patched rather than ed25519-dalek.

• ed25519-dalek

  curve25519-dalek = { git = "https://github.com/sp1-patches/curve25519-dalek", branch = "patch-curve25519-v4.1.3" }
  

  Note: The curve operations occur inside of the curve25519-dalek crate.

• curve25519-dalek

  curve25519-dalek = { git = "https://github.com/sp1-patches/curve25519-dalek-ng", branch = "patch-v4.1.3" }
  

Secp256k1 Acceleration

To accelerate Secp256k1 operations, you'll need to patch k256 or secp256k1 depending on your usage.

Generally, if a crate you're using (ex. revm) has support for using k256 instead of secp256k1, you should use k256.

Patches

Apply the following patches based on what crates are in your dependencies.

• k256

  ecdsa-core = { git = "https://github.com/sp1-patches/signatures", package = "ecdsa", branch = "patch-ecdsa-v0.16.9" }
  

  Note: The curve operations for k256 are inside of the ecdsa-core crate, so you don't need to patch k256 itself, and just patching ecdsa-core is enough.

• secp256k1

  secp256k1 = { git = "https://github.com/sp1-patches/rust-secp256k1", branch = "patch-v0.29.0" }
  

BN254 Acceleration

To accelerate BN254 (Also known as BN128 and Alt-BN128), you will need to patch the substrate-bn crate.

Patches

Apply the patch by adding the following to your list of dependencies:

#![allow(unused)]
fn main() {
substrate-bn = { git = "https://github.com/sp1-patches/bn", branch = "patch-v0.6.0" }
}

Performance Benchmarks for Patched substrate-bn in revm

Note: The operations run-add, run-mul, and run-pair are from the revm crate, specifically from the file crates/precompile/src/bn128.rs on GitHub. In the patched version of the substrate-bn crate, these functions utilize SP1's BN254 Fp precompiles.

To accelerate revm in SP1 using the BN254 patched crate, replace the substrate-bn crate with the patched crate by adding the following to crates/precompile/Cargo.toml:

#![allow(unused)]
fn main() {
bn = { git = "https://github.com/sp1-patches/bn", package = "substrate-bn", branch = "patch-v0.6.0" }
}

BLS12-381 Acceleration

To accelerate BLS12-381 operations, you'll need to patch the bls12_381 crate. Apply the following patch by adding the following to your list of dependencies:

blst = { git = "https://github.com/sp1-patches/bls12_381", branch = "patch-v0.8.0" }

This patch significantly improves the performance of BLS12-381 operations, making it essential for applications that rely heavily on these cryptographic primitives.

Performance Benchmarks for Patched bls12_381 in kzg-rs

Troubleshooting

Verifying Patch Usage: Cargo

You can check if the patch was applied by using cargo's tree command to print the dependencies of the crate you patched.

cargo tree -p sha2@0.9.8

Next to the package name, it should have a link to the Github repository that you patched with.

Ex.

sha2 v0.9.8 (https://github.com/sp1-patches/RustCrypto-hashes?branch=patch-sha2-v0.9.8#afdbfb09)
├── ...

Verifying Patch Usage: SP1

To check if a precompile is used by your program, you can view SP1's ExecutionReport, which is returned when executing a program with execute. In ExecutionReport you can view the syscall_counts map to view if a specific syscall was used.

For example, if you wanted to check sha256 was used, you would look for SHA_EXTEND and SHA_COMPRESS in syscall_counts.

An example of this is available in our Patch Testing Example.

Cargo Version Issues

If you encounter issues with version commits on your patches, you should try updating the patched crate manually.

cargo update -p <patch-crate-name>

If you encounter issues relating to cargo / git, you can try setting CARGO_NET_GIT_FETCH_WITH_CLI:

CARGO_NET_GIT_FETCH_WITH_CLI=true cargo update -p <patch-crate-name>

You can permanently set this value in ~/.cargo/config:

[net]
git-fetch-with-cli = true

Precompiles

Precompiles are built into the SP1 zkVM and accelerate commonly used operations such as elliptic curve arithmetic and hashing. Under the hood, precompiles are implemented as custom STARK tables dedicated to proving one or few operations. They typically improve the performance of executing expensive operations in SP1 by a few orders of magnitude.

Inside the zkVM, precompiles are exposed as system calls executed through the ecall RISC-V instruction. Each precompile has a unique system call number and implements an interface for the computation.

SP1 also has been designed specifically to make it easy for external contributors to create and extend the zkVM with their own precompiles. To learn more about this, you can look at implementations of existing precompiles in the precompiles folder. More documentation on this will be coming soon.

To use precompiles, we typically recommend you interact with them through patches, which are crates modified to use these precompiles under the hood, without requiring you to call system calls directly.

Specification

If you are an advanced user you can interact with the precompiles directly using external system calls.

Here is a list of all available system calls & precompiles.

{{#include ../../zkvm/lib/src/lib.rs}}

Cycle Tracking

When writing a program, it is useful to know how many RISC-V cycles a portion of the program takes to identify potential performance bottlenecks. SP1 provides a way to track the number of cycles spent in a portion of the program.

Tracking Cycles with Annotations

To track the number of cycles spent in a portion of the program, you can either put println!("cycle-tracker-start: block name") + println!("cycle-tracker-end: block name") statements (block name must be same between start and end) around the portion of your program you want to profile or use the #[sp1_derive::cycle_tracker] macro on a function. An example is shown below:

#![no_main]
sp1_zkvm::entrypoint!(main);

#[sp1_derive::cycle_tracker]
pub fn expensive_function(x: usize) -> usize {
    let mut y = 1;
    for _ in 0..100 {
        y *= x;
        y %= 7919;
    }
    y
}

pub fn main() {
    let mut nums = vec![1, 1];

    // Setup a large vector with Fibonacci-esque numbers.
    println!("cycle-tracker-start: setup");
    for _ in 0..100 {
        let mut c = nums[nums.len() - 1] + nums[nums.len() - 2];
        c %= 7919;
        nums.push(c);
    }
    println!("cycle-tracker-end: setup");

    println!("cycle-tracker-start: main-body");
    for i in 0..2 {
        let result = expensive_function(nums[nums.len() - i - 1]);
        println!("result: {}", result);
    }
    println!("cycle-tracker-end: main-body");
}

Note that to use the macro, you must add the sp1-derive crate to your dependencies for your program.

[dependencies]
sp1-derive = "1.1.0"

In the script for proof generation, setup the logger with utils::setup_logger() and run the script with RUST_LOG=info cargo run --release. You should see the following output:

$ RUST_LOG=info cargo run --release
    Finished release [optimized] target(s) in 0.21s
     Running `target/release/cycle-tracking-script`
2024-03-13T02:03:40.567500Z  INFO execute: loading memory image
2024-03-13T02:03:40.567751Z  INFO execute: starting execution
2024-03-13T02:03:40.567760Z  INFO execute: clk = 0 pc = 0x2013b8
2024-03-13T02:03:40.567822Z  INFO execute: ┌╴setup
2024-03-13T02:03:40.568095Z  INFO execute: └╴4,398 cycles
2024-03-13T02:03:40.568122Z  INFO execute: ┌╴main-body
2024-03-13T02:03:40.568149Z  INFO execute: │ ┌╴expensive_function
2024-03-13T02:03:40.568250Z  INFO execute: │ └╴1,368 cycles
stdout: result: 5561
2024-03-13T02:03:40.568373Z  INFO execute: │ ┌╴expensive_function
2024-03-13T02:03:40.568470Z  INFO execute: │ └╴1,368 cycles
stdout: result: 2940
2024-03-13T02:03:40.568556Z  INFO execute: └╴5,766 cycles
2024-03-13T02:03:40.568566Z  INFO execute: finished execution clk = 11127 pc = 0x0
2024-03-13T02:03:40.569251Z  INFO execute: close time.busy=1.78ms time.idle=21.1µs

Note that we elegantly handle nested cycle tracking, as you can see above.

Get Tracked Cycle Counts

To include tracked cycle counts in the ExecutionReport when using ProverClient::execute, use the following annotations:

fn main() {
  println!("cycle-tracker-report-start: block name");
  // ...
  println!("cycle-tracker-report-end: block name");
}

This will log the cycle count for block name and include it in the ExecutionReport in the cycle_tracker map.

Tracking Cycles with Tracing

The cycle-tracker annotation is a convenient way to track cycles for specific sections of code. However, sometimes it can also be useful to track what functions are taking the most cycles across the entire program, without having to annotate every function individually.

First, we need to generate a trace file of the program counter at each cycle while the program is executing. This can be done by simply setting the TRACE_FILE environment variable with the path of the file you want to write the trace to. For example, you can run the following command in the script directory for any example program:

TRACE_FILE=trace.log RUST_LOG=info cargo run --release

When the TRACE_FILE environment variable is set, as SP1's RISC-V runtime is executing, it will write a log of the program counter to the file specified by TRACE_FILE.

Next, we can use the cargo prove CLI with the trace command to analyze the trace file and generate a table of instruction counts. This can be done with the following command:

cargo prove trace --elf <path_to_program_elf> --trace <path_to_trace_file>

The trace command will generate a table of instruction counts, sorted by the number of cycles spent in each function. The output will look something like this:

  [00:00:00] [########################################] 17053/17053 (0s)

Total instructions in trace: 17053


 Instruction counts considering call graph
+----------------------------------------+-------------------+
| Function Name                          | Instruction Count |
| __start                                | 17045             |
| main                                   | 12492             |
| sp1_zkvm::syscalls::halt::syscall_halt | 4445              |
| sha2::sha256::compress256              | 4072              |
| sp1_lib::io::commit                    | 258               |
| sp1_lib::io::SyscallWriter::write      | 255               |
| syscall_write                          | 195               |
| memcpy                                 | 176               |
| memset                                 | 109               |
| sp1_lib::io::read_vec                  | 71                |
| __rust_alloc                           | 29                |
| sp1_zkvm::heap::SimpleAlloc::alloc     | 22                |
| syscall_hint_len                       | 3                 |
| syscall_hint_read                      | 2                 |
+----------------------------------------+-------------------+


 Instruction counts ignoring call graph
+----------------------------------------+-------------------+
| Function Name                          | Instruction Count |
| main                                   | 12075             |
| sha2::sha256::compress256              | 4073              |
| sp1_zkvm::syscalls::halt::syscall_halt | 219               |
| memcpy                                 | 180               |
| syscall_write                          | 123               |
| memset                                 | 111               |
| sp1_lib::io::commit                    | 88                |
| sp1_lib::io::SyscallWriter::write      | 60                |
| __start                                | 45                |
| sp1_lib::io::read_vec                  | 35                |
| sp1_zkvm::heap::SimpleAlloc::alloc     | 23                |
| anonymous                              | 7                 |
| __rust_alloc                           | 7                 |
| syscall_hint_len                       | 4                 |
| syscall_hint_read                      | 3                 |
+----------------------------------------+-------------------+

Proof Aggregation

SP1 supports proof aggregation and recursion, which allows you to verify an SP1 proof within SP1. Use cases include:

• Reducing on-chain verification costs by aggregating multiple SP1 proofs into a single SP1 proof.
• Proving logic that is split into multiple proofs, such as proving a statement about a rollup's state transition function by proving each block individually and aggregating these proofs to produce a final proof of a range of blocks.

For an example of how to use proof aggregation and recursion in SP1, refer to the aggregation example.

Note that to verify an SP1 proof inside SP1, you must generate a "compressed" SP1 proof (see Proof Types for more details).

When to use aggregation

Note that by itself, SP1 can already prove arbitarily large programs by chunking the program's execution into multiple "shards" (contiguous batches of cycles) and generating proofs for each shard in parallel, and then recursively aggregating the proofs. Thus, aggregation is generally not necessary for most use-cases, as SP1's proving for large programs is already parallelized. However, aggregation can be useful for aggregating computations that require more than the zkVM's limited (~2GB) memory or for aggregating multiple SP1 proofs from different parties into a single proof to save on onchain verification costs.

Verifying Proofs inside the zkVM

To verify a proof inside the zkVM, you can use the sp1_zkvm::lib::verify::verify_proof function.

sp1_zkvm::lib::verify::verify_proof(vkey, public_values_digest);

You do not need to pass in the proof as input into the syscall, as the proof will automatically be read for the proof input stream by the prover.

Note that you must include the verify feature in your Cargo.toml for sp1-zkvm to be able to use the verify_proof function (like this).

Generating Proofs with Aggregation

To provide an existing proof as input to the SP1 zkVM, you can use the existing SP1Stdin object which is already used for all inputs to the zkVM.

Generating proving key and verifying key.
let (input_pk, input_vk) = client.setup(PROOF_INPUT_ELF);
let (aggregation_pk, aggregation_vk) = client.setup(AGGREGATION_ELF);

// Generate a proof that will be recursively verified / aggregated. Note that we use the "compressed"
// proof type, which is necessary for aggregation.
let mut stdin = SP1Stdin::new();
let input_proof = client
    .prove(&input_pk, stdin)
    .compressed()
    .run()
    .expect("proving failed");

// Create a new stdin object to write the proof and the corresponding verifying key to.
let mut stdin = SP1Stdin::new();
stdin.write_proof(input_proof, input_vk);

// Generate a proof that will recusively verify / aggregate the input proof.
let aggregation_proof = client
    .prove(&aggregation_pk, stdin)
    .compressed()
    .run()
    .expect("proving failed");

Generating Proofs: Setup

In this section, we will teach you how to setup a self-contained crate which can generate proofs of programs that have been compiled with the SP1 toolchain inside the SP1 zkVM, using the sp1-sdk crate.

CLI (Recommended)

The recommended way to setup your first program to prove inside the zkVM is using the method described in Quickstart which will create a script folder.

cargo prove new <name>
cd script

Manual

You can also manually setup a project. First create a new cargo project:

cargo new script
cd script

Cargo Manifest

Inside this crate, add the sp1-sdk crate as a dependency. Your Cargo.toml should look like as follows:

[workspace]
[package]
version = "0.1.0"
name = "script"
edition = "2021"

[dependencies]
sp1-sdk = "1.1.0"

The sp1-sdk crate includes the necessary utilities to generate, save, and verify proofs.

Generating Proofs: Basics

All the methods you'll need for generating proofs are included in the sp1_sdk crate. Most importantly, you'll need to use the ProverClient to setup a proving key and verifying key for your program and then use the execute, prove and verify methods to execute your program, and generate and verify proofs.

To make this more concrete, let's walk through a simple example of generating a proof for a Fiboancci program inside the zkVM.

Example: Fibonacci

use sp1_sdk::{utils, ProverClient, SP1ProofWithPublicValues, SP1Stdin};

/// The ELF we want to execute inside the zkVM.
const ELF: &[u8] = include_bytes!("../../program/elf/riscv32im-succinct-zkvm-elf");

fn main() {
    // Setup logging.
    utils::setup_logger();

    // Create an input stream and write '500' to it.
    let n = 1000u32;

    // The input stream that the program will read from using `sp1_zkvm::io::read`. Note that the
    // types of the elements in the input stream must match the types being read in the program.
    let mut stdin = SP1Stdin::new();
    stdin.write(&n);

    // Create a `ProverClient` method.
    let client = ProverClient::new();

    // Execute the program using the `ProverClient.execute` method, without generating a proof.
    let (_, report) = client.execute(ELF, stdin.clone()).run().unwrap();
    println!(
        "executed program with {} cycles",
        report.total_instruction_count()
    );

    // Generate the proof for the given program and input.
    let (pk, vk) = client.setup(ELF);
    let mut proof = client.prove(&pk, stdin).run().unwrap();

    println!("generated proof");

    // Read and verify the output.
    //
    // Note that this output is read from values commited to in the program using
    // `sp1_zkvm::io::commit`.
    let _ = proof.public_values.read::<u32>();
    let a = proof.public_values.read::<u32>();
    let b = proof.public_values.read::<u32>();

    println!("a: {}", a);
    println!("b: {}", b);

    // Verify proof and public values
    client.verify(&proof, &vk).expect("verification failed");

    // Test a round trip of proof serialization and deserialization.
    proof
        .save("proof-with-pis.bin")
        .expect("saving proof failed");
    let deserialized_proof =
        SP1ProofWithPublicValues::load("proof-with-pis.bin").expect("loading proof failed");

    // Verify the deserialized proof.
    client
        .verify(&deserialized_proof, &vk)
        .expect("verification failed");

    println!("successfully generated and verified proof for the program!")
}

You can run the above script in the script directory with RUST_LOG=info cargo run --release. Note that running the above script will generate a proof locally.

We recommend using the prover network to generate proofs. Read more about the recommended workflow for developing with SP1.

Proof Types

There are a few different types of proofs that can be generated by the SP1 zkVM. Each proof type has its own tradeoffs in terms of proof generation time, verification cost, and proof size.

The ProverClient follows a "builder" pattern that allows you to configure the proof type and other options after creating a ProverClient and calling prove on it.

For a full list of options, see the following docs.

Core (Default)

The default prover mode generates a list of STARK proofs that in aggregate have size proportional to the size of the execution. Use this in settings where you don't care about verification cost / proof size.

let client = ProverClient::new();
client.prove(&pk, stdin).run().unwrap();

Compressed

The compressed prover mode generates STARK proofs that have constant size. Use this in settings where you care about verification cost / proof size. This is useful for applications where you want to recursively verify SP1 proofs within SP1 (see the proof aggregation section).

let client = ProverClient::new();
client.prove(&pk, stdin).compressed().run().unwrap();

PLONK

The PLONK prover mode generates a SNARK proof with extremely small proof size and low verification cost. This mode is necessary for generating proofs that can be verified onchain for around ~300k gas.

let client = ProverClient::new();
client.prove(&pk, stdin).plonk().run().unwrap();

Recommended Workflow for Developing with SP1

We recommend the following workflow for developing with SP1.

Step 1: Iterate on your program with execution only

While iterating on your SP1 program, you should only execute the program with the RISC-V runtime. This will allow you to verify the correctness of your program and test the SP1Stdin as well as the SP1PublicValues that are returned, without having to generate a proof (which can be slow and/or expensive). If the execution of your program succeeds, then proof generation should succeed as well!

use sp1_sdk::{utils, ProverClient, SP1Stdin};

/// The ELF we want to execute inside the zkVM.
const ELF: &[u8] = include_bytes!("../../program/elf/riscv32im-succinct-zkvm-elf");

fn main() {
    // Setup logging.
    utils::setup_logger();

    // Create an input stream and write '500' to it.
    let n = 500u32;

    let mut stdin = SP1Stdin::new();
    stdin.write(&n);

    // Only execute the program and get a `SP1PublicValues` object.
    let client = ProverClient::new();
    let (mut public_values, execution_report) = client.execute(ELF, stdin).run().unwrap();

    // Print the total number of cycles executed and the full execution report with a breakdown of
    // the RISC-V opcode and syscall counts.
    println!(
        "Executed program with {} cycles",
        execution_report.total_instruction_count() + execution_report.total_syscall_count()
    );
    println!("Full execution report:\n{:?}", execution_report);

    // Read and verify the output.
    let _ = public_values.read::<u32>();
    let a = public_values.read::<u32>();
    let b = public_values.read::<u32>();

    println!("a: {}", a);
    println!("b: {}", b);
}

Note that printing out the total number of executed cycles and the full execution report provides helpful insight into proof generation latency and cost either for local proving or when using the prover network.

Crate Setup: We recommend that your program crate that defines the main function (around which you wrap the sp1_zkvm::entrypoint! macro) should be kept minimal. Most of your business logic should be in a separate crate (in the same repo/workspace) that can be tested independently and that is not tied to the SP1 zkVM. This will allow you to unit test your program logic without having to worry about the zkvm compilation target. This will also allow you to efficient reuse types between your program crate and your crate that generates proofs.

Step 2: Generate proofs

After you have iterated on your program and finalized that it works correctly, you can generate proofs for your program for final end to end testing or production use.

Generating proofs on the prover network (recommended)

Using Succinct's prover prover network will generally be faster and cheaper than local proving, as it parallelizes proof generation amongst multiple machines and also uses SP1's GPU prover that is not yet available for local proving. Follow the setup instructions to get started with the prover network. Using the prover network only requires adding 1 environment variable from a regular SP1 proof generation script with the ProverClient.

There are a few things to keep in mind when using the prover network.

Prover Network FAQ

Benchmarking latency on the prover network

The prover network currently parallelizes proof generation across multiple machines. This means the latency of proof generation does not scale linearly with the number of cycles of your program, but rather with the number of cycles of your program divided by the number of currently available machines on the prover network.

Our prover network currently has limited capacity because it is still in beta. If you have an extremely latency sensitive use-case and you want to figure out the minimal latency possible for your program, you should reach out to us and we can onboard you to our reserved capacity cluster that has a dedicated instances that can significantly reduce latency.

Costs on the prover network

The cost of proof generation on the prover network scales approximately linearly with the number of cycles of your program (along with the number of syscalls that your program makes). For larger workloads with regular proof frequency (like rollups and light clients), we can offer discounted pricing. To figure out how much your program will cost to prove, you can get in touch with us to discuss pricing options.

Note that latency is not the same as cost, because we parallelize proof generation across multiple machines, so two proofs with the same latency can be using a different number of machines, impacting the cost.

Benchmarking on small vs. large programs

In SP1, there is a fixed overhead for proving that is independent of your program's cycle count. This means that benchmarking on small programs is not representative of the performance of larger programs. To get an idea of the scale of programs for real-world workloads, you can refer to our benchmarking blog post and also some numbers below:

• An average Ethereum block can be between 100-500M cycles (including merkle proof verification for storage and execution of transactions) with our keccak and secp256k1 precompiles.
• For a Tendermint light client, the average cycle count can be between 10M and 50M cycles (including our ed25519 precompiles).
• We consider programs with <2M cycles to be "small" and by default, the fixed overhead of proving will dominate the proof latency. If latency is incredibly important for your use-case, we can specialize the prover network for your program if you reach out to us.

Note that if you generate PLONK proofs on the prover network, you will encounter a fixed overhead of 90 seconds for the STARK -> SNARK wrapping step. We're actively working on reducing this overhead in our next release.

On-Demand vs. Reserved Capacity

The prover network is currently in beta and has limited capacity. For high volume use-cases, we can offer discounted pricing and a reserved capacity cluster that has a dedicated instances that can significantly reduce latency and have higher throughput and guaranteed SLAs.

Generating proofs locally

If you want to generate proofs locally, you can use the sp1_sdk crate to generate proofs locally as outlined in the Basics section. By default, the ProverClient will generate proofs locally using your CPU. Check out the hardware requirements for locally proving here.

Prover Network Beta

Currently, the supported version of SP1 on the prover network is v1.1.0.

Succinct has been building the Succinct Prover Network, a distributed network of provers that can generate proofs of any size quickly and reliably. It's currently in private beta, but you can get access by following the steps below.

To get started, FILL OUT THIS FORM to gain access to the Succinct Network. Completing this form requires you to complete the key setup steps below.

Prover Network: Key Setup

The prover network uses Secp256k1 keypairs for authentication, similar to Ethereum wallets. You may generate a new keypair explicitly for use with the prover network, or use an existing keypair.

You do not need to hold any funds in this account, it is used solely for access control.

Generate a Public Key

Prover network keypair credentials can be generated using the cast CLI tool.

First install Foundry:

curl -L https://foundry.paradigm.xyz | bash

Upon running this command, you will be prompted to source your shell profile and run foundryup. Afterwards you should have access to the cast command.

Use cast to generate a new keypair:

cast wallet new

which will give you an output similar to this:

The "Address" what you should submit in the form, in the example above this is 0x552f0FC6D736ed965CE07a3D71aA639De15B627b. The "Private key" should be kept safe and secure. When interacting with the network, you will set your SP1_PRIVATE_KEY environment variable to this value.

Retrieve an Existing Key

If you already have an existing key you would like to use, you can also use cast retrieve your address:

cast wallet address --private-key $PRIVATE_KEY

Prover Network: Usage

Currently, the supported version of SP1 on the prover network is v1.1.0.

Sending a proof request

To use the prover network to generate a proof, you can run your script that uses sp1_sdk::ProverClient as you would normally but with additional environment variables set:

// Generate the proof for the given program.
let client = ProverClient::new();
let (pk, vk) = client.setup(ELF);
let mut proof = client.prove(&pk, stdin).run().unwrap();

SP1_PROVER=network SP1_PRIVATE_KEY=... RUST_LOG=info cargo run --release

• SP1_PROVER should be set to network when using the prover network.

• SP1_PRIVATE_KEY should be set to your private key. You will need to be using a whitelisted key to use the network.

When you call any of the prove functions in ProverClient now, it will first simulate your program, then wait for it to be proven through the network and finally return the proof.

View the status of your proof

You can view your proof and other running proofs on the explorer. The page for your proof will show details such as the stage of your proof and the cycles used. It also shows the program hash which is the keccak256 of the program bytes.

Advanced Usage

Skip simulation

To skip the simulation step and directly submit the program for proof generation, you can set the SKIP_SIMULATION environment variable to true. This will save some time if you are sure that your program is correct. If your program panics, the proof will fail and ProverClient will panic.

Use NetworkProver directly

By using the sp1_sdk::NetworkProver struct directly, you can call async functions directly and have programmatic access to the proof ID.

impl NetworkProver {
    /// Creates a new [NetworkProver] with the private key set in `SP1_PRIVATE_KEY`.
    pub fn new() -> Self;

    /// Creates a new [NetworkProver] with the given private key.
    pub fn new_from_key(private_key: &str) -> Self;

    /// Requests a proof from the prover network, returning the proof ID.
    pub async fn request_proof(
        &self,
        elf: &[u8],
        stdin: SP1Stdin,
        mode: ProofMode,
    ) -> Result<String>;

    /// Waits for a proof to be generated and returns the proof.
    pub async fn wait_proof<P: DeserializeOwned>(&self, proof_id: &str) -> Result<P>;

    /// Requests a proof from the prover network and waits for it to be generated.
    pub async fn prove<P: ProofType>(&self, elf: &[u8], stdin: SP1Stdin) -> Result<P>;
}

Supported Versions

The prover network currently only supports specific versions of SP1:

Where X denotes that any patch version is supported (e.g. v1.2.0, v1.2.1).

If you submit a proof request to the prover network and your are not using the supported version, you will receive an error message.

Changing versions

You must switch to a supported version before submitting a proof. To do so, replace the sp1-zkvm version in your progam's Cargo.toml:

[dependencies]
sp1-zkvm = "1.1.0"

replace the sp1-sdk version in your script's Cargo.toml:

[dependencies]
sp1-sdk = "1.1.0"

Re-build your program and script, and then try again.

Hardware Acceleration

SP1 supports hardware acceleration on the following platforms:

• AVX256/AVX512 on x86 CPUs
• CUDA on Nvidia GPUs

To enable hardware acceleration, please refer to the platform specific instructions available in this section.

AVX

SP1 supports both AVX256 and AVX512 acceleration on x86 CPUs due to support in Plonky3. Whenever possible, we recommend using AVX512 acceleration as it provides better performance.

Checking for AVX

To check if your CPU supports AVX, you can run the following command:

grep avx /proc/cpuinfo

Look for the flags avx2 and avx512.

Enabling AVX256

To enable AVX256 acceleration, you can set the RUSTFLAGS environment variable to include the following flags:

RUSTFLAGS="-C target-cpu=native" cargo run --release

Enabling AVX512

To enable AVX512 acceleration, you can set the RUSTFLAGS environment variable to include the following flags:

RUSTFLAGS="-C target-cpu=native -C target-feature=+avx512f" cargo run --release

Note that the +avx512f flag is required to enable AVX512 acceleration.

CUDA

SP1 supports CUDA acceleration, which can provide dramatically better latency and cost performance compared to using the CPU prover, even with AVX acceleration.

Software Requirements

Please make sure you have the following installed before using the CUDA prover:

• CUDA 12
• CUDA Container Toolkit

Hardware Requirements

• CPU: We recommend having at least 8 CPU cores with 32GB of RAM available to fully utilize the GPU.
• GPU: 24GB or more for core/compressed proofs, 40GB or more for shrink/wrap proofs

Usage

To use the CUDA prover, you can compile the sp1-sdk crate with the cuda feature enabled. You can use the normal methods on the ProverClient to generate proofs.

FAQ

Logging and Tracing Information

You can use sp1_sdk::utils::setup_logger() to enable logging information respectively. You can set the logging level with the RUST_LOG environment variable.

sp1_sdk::utils::setup_logger();

Example of setting the logging level to info (other options are debug, trace, and warn):

RUST_LOG=info cargo run --release

Onchain Verification: Setup

The best way to get started with verifying SP1 proofs on-chain is to refer to the SP1 Project Template.

• The template program shows how to write outputs that can be decoded in Solidity.
• The template script shows how to generate the proof using the SDK and save it to a file.
• The template contract shows how to verify the proof onchain using Solidity.

Refer to the section on Contract Addresses for the addresses of the deployed verifiers.

Generating SP1 Proofs for Onchain Verification

By default, the proofs generated by SP1 are not verifiable onchain, as they are non-constant size and STARK verification on Ethereum is very expensive. To generate a proof that can be verified onchain, we use performant STARK recursion to combine SP1 shard proofs into a single STARK proof and then wrap that in a SNARK proof. Our ProverClient has a prover option for this called plonk. Behind the scenes, this function will first generate a normal SP1 proof, then recursively combine all of them into a single proof using the STARK recursion protocol. Finally, the proof is wrapped in a SNARK proof using PLONK.

WARNING: The PLONK prover is only guaranteed to work on official releases of SP1. To use PLONK proving & verification locally, ensure that you have Docker installed and have at least 128GB of RAM.

Example

use sp1_sdk::{utils, ProverClient, SP1Stdin};

/// The ELF we want to execute inside the zkVM.
const ELF: &[u8] = include_bytes!("../../program/elf/riscv32im-succinct-zkvm-elf");

fn main() {
    // Setup logging.
    utils::setup_logger();

    // Create an input stream and write '500' to it.
    let n = 500u32;

    let mut stdin = SP1Stdin::new();
    stdin.write(&n);

    // Generate the proof for the given program and input.
    let client = ProverClient::new();
    let (pk, vk) = client.setup(ELF);
    let proof = client.prove(&pk, stdin).plonk().run().unwrap();

    println!("generated proof");

    // Get the public values as bytes.
    let public_values = proof.public_values.raw();
    println!("public values: {:?}", public_values);

    // Get the proof as bytes.
    let solidity_proof = proof.raw();
    println!("proof: {:?}", solidity_proof);

    // Verify proof and public values
    client.verify(&proof, &vk).expect("verification failed");

    // Save the proof.
    proof
        .save("proof-with-pis.bin")
        .expect("saving proof failed");

    println!("successfully generated and verified proof for the program!")
}

You can run the above script with RUST_LOG=info cargo run --bin plonk_bn254 --release in examples/fibonacci/script.

Using PLONK without Docker (Advanced)

If you would like to run the PLONK prover directly without Docker, you must have Go 1.22 installed and enable the native-plonk feature in sp1-sdk. This path is not recommended and may require additional native dependencies.

sp1-sdk = { version = "1.1.0", features = ["native-plonk"] }

Solidity Verifier

We maintain a suite of contracts used for verifying SP1 proofs onchain. We highly recommend using Foundry.

Installation

To install the latest release version:

forge install succinctlabs/sp1-contracts

To install a specific version:

forge install succinctlabs/sp1-contracts@<version>

Finally, add @sp1-contracts/=lib/sp1-contracts/contracts/src/ in remappings.txt.

Usage

Once installed, you can use the contracts in the library by importing them:

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.20;

import {ISP1Verifier} from "@sp1-contracts/ISP1Verifier.sol";

/// @title Fibonacci.
/// @author Succinct Labs
/// @notice This contract implements a simple example of verifying the proof of a computing a
///         fibonacci number.
contract Fibonacci {
    /// @notice The address of the SP1 verifier contract.
    /// @dev This can either be a specific SP1Verifier for a specific version, or the
    ///      SP1VerifierGateway which can be used to verify proofs for any version of SP1.
    ///      For the list of supported verifiers on each chain, see:
    ///      https://github.com/succinctlabs/sp1-contracts/tree/main/contracts/deployments
    address public verifier;

    /// @notice The verification key for the fibonacci program.
    bytes32 public fibonacciProgramVKey;

    constructor(address _verifier, bytes32 _fibonacciProgramVKey) {
        verifier = _verifier;
        fibonacciProgramVKey = _fibonacciProgramVKey;
    }

    /// @notice The entrypoint for verifying the proof of a fibonacci number.
    /// @param _proofBytes The encoded proof.
    /// @param _publicValues The encoded public values.
    function verifyFibonacciProof(bytes calldata _publicValues, bytes calldata _proofBytes)
        public
        view
        returns (uint32, uint32, uint32)
    {
        ISP1Verifier(verifier).verifyProof(fibonacciProgramVKey, _publicValues, _proofBytes);
        (uint32 n, uint32 a, uint32 b) = abi.decode(_publicValues, (uint32, uint32, uint32));
        return (n, a, b);
    }
}

For more details on the contracts, refer to the sp1-contracts repo.

Testing

To test the contract, we recommend setting up Foundry Tests. We have an example of such a test in the SP1 Project Template.

Solidity Versions

The officially deployed contracts are built using Solidity 0.8.20 and exist on the sp1-contracts main branch.

If you need to use different versions that are compatible with your contracts, there are also other branches you can install that contain different versions. For example for branch main-0.8.15 contains the contracts with:

pragma solidity ^0.8.15;

and you can install it with:

forge install succinctlabs/sp1-contracts@main-0.8.15

If there is different versions that you need but there aren't branches for them yet, please ask in the SP1 Telegram.

Contract Addresses

When using SP1, we recommend using our deployed verifiers. Each contract is a SP1VerifierGateway which can automatically routes your SP1 proof to the correct verifier based on the prover version.

Currently officially supported version of SP1 is v1.1.0. If you'd like official support for a verifier on a different chain, please ask in the SP1 Telegram.

ISP1Verifier Interface

All verifiers implement the ISP1Verifier interface.

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.20;

/// @title SP1 Verifier Interface
/// @author Succinct Labs
/// @notice This contract is the interface for the SP1 Verifier.
interface ISP1Verifier {
    /// @notice Verifies a proof with given public values and vkey.
    /// @dev It is expected that the first 4 bytes of proofBytes must match the first 4 bytes of
    /// target verifier's VERIFIER_HASH.
    /// @param programVKey The verification key for the RISC-V program.
    /// @param publicValues The public values encoded as bytes.
    /// @param proofBytes The proof of the program execution the SP1 zkVM encoded as bytes.
    function verifyProof(
        bytes32 programVKey,
        bytes calldata publicValues,
        bytes calldata proofBytes
    ) external view;
}

Common Issues

Bus Error

If you are running a executable that uses the sp1-sdk crate, you may encounter a bus error like this:

zsh: bus error

This is fixed by running with the --release flag, as the sp1-sdk crate only supports release builds as of right now.

Alloy Errors

If you are using a library that depends on alloy_sol_types, and encounter an error like this:

perhaps two different versions of crate `alloy_sol_types` are being used?

This is likely due to two different versions of alloy_sol_types being used. To fix this, you can set default-features to false for the sp1-sdk dependency in your Cargo.toml.

[dependencies]
sp1-sdk = { version = "1.1.0", default-features = false }

This will configure out the network feature which will remove the dependency on alloy_sol_types and configure out the NetworkProver.

Rust Version Errors

If you are using a library that has an MSRV (minimum supported rust version) of 1.76.0 or higher, you may encounter an error like this when building your program.

package `alloy v0.1.1 cannot be built because it requires rustc 1.76 or newer, while the currently active rustc version is 1.75.0-nightly`

This is due to the fact that the Succinct Rust toolchain might be built with a lower version than the MSRV of the crates you are using. You can check the version of the Succinct Rust toolchain by running cargo +succinct --version. The Succinct Rust toolchain's latest version is 1.79, and you can update to the latest version by running sp1up.

If that doesn't work (i.e. the MSRV of the crates you are using is still higher than the version of the Succinct Rust toolchain), you can try the following:

• If you're using cargo prove build directly, pass the --ignore-rust-version flag:

  cargo prove build --ignore-rust-version
  

• If using build_program in an build.rs file with the sp1-build crate, set ignore_rust_version to true inside the BuildArgs struct and use build_program_with_args:

  
  #![allow(unused)]
  fn main() {
  let args = BuildArgs {
      ignore_rust_version: true,
      ..Default::default()
  };
  build_program_with_args("path/to/program", args);
  }
  

Stack Overflow Errors

If you encounter the following in a script using sp1-sdk:

thread 'main' has overflowed its stack
fatal runtime error: stack overflow

Segmentation fault (core dumped)

Re-run your script with --release.

Note that the core sp1-core library and sp1-recursion require being compiled with the release profile.

C Binding Errors

If you are building a program that uses C bindings or has dependencies that use C bindings, you may encounter the following errors:

cc did not execute successfully

Failed to find tool. Is `riscv32-unknown-elf-gcc` installed?

To resolve this, re-install sp1 with the --c-toolchain flag:

sp1up --c-toolchain

This will install the C++ toolchain for RISC-V and set the CC_riscv32im_succinct_zkvm_elf environment variable to the path of the installed riscv32-unknown-elf-gcc binary. You can also use your own C++ toolchain be setting this variable manually:

export CC_riscv32im_succinct_zkvm_elf=/path/to/toolchain

Compilation Errors with sp1-lib::syscall_verify_sp1_proof

If you are using the sp1-lib::syscall_verify_sp1_proof function, you may encounter compilation errors when building your program.

  [sp1]    = note: rust-lld: error: undefined symbol: syscall_verify_sp1_proof
  [sp1]            >>> referenced by sp1_lib.b593533d149f0f6e-cgu.0
  [sp1]            >>>               sp1_lib-8f5deb4c47d01871.sp1_lib.b593533d149f0f6e-cgu.0.rcgu.o:(sp1_lib::verify::verify_sp1_proof::h5c1bb38f11b3fe71) in ...
  [sp1]            
  [sp1]  
  [sp1]  error: could not compile `package-name` (bin "package-name") due to 1 previous error

To resolve this, ensure that you're importing both sp1-lib and sp1-zkvm with the verify feature enabled.

[dependencies]
sp1-lib = { version = "<VERSION>", features = ["verify"] }
sp1-zkvm = { version = "<VERSION>", features = ["verify"] }

Usage in CI

You may want to use SP1 in your Github Actions CI workflow.

You first need have Rust installed, and you can use actions-rs/toolchain for this:

- name: Install Rust Toolchain
  uses: actions-rs/toolchain@v1
  with:
    toolchain: 1.79.0
    profile: default
    override: true
    default: true
    components: llvm-tools, rustc-dev

And then you can install the SP1 toolchain:

- name: Install SP1 toolchain
  run: |
    curl -L https://sp1.succinct.xyz | bash
    ~/.sp1/bin/sp1up 
    ~/.sp1/bin/cargo-prove prove --version

RV32IM Specification

SP1 implements the RISC-V RV32IM instruction set with some implementation details that make it more suitable for proving.

• LW/SW memory access must be word aligned.
• LH/LHU/SH memory access must be half-word aligned.
• Memory access is only valid for addresses 0x20, 0x78000000. Accessing addresses outside of this range will result in undefined behavior. The global heap allocator in sp1_zkvm will panic if memory exceeds this range.
• The ECALL instruction is used for system calls and precompiles. Only valid syscall IDs should be called, and only using the specific convention of loading the ID into register T0 and arguments into registers A0 and A1. If the arguments are addresses, they must be word-aligned. Failure to follow this convention can result in undefined behavior. Correct usages can be found in the sp1_zkvm and sp1_lib crates.

Building Circuit Artifacts

To build the production PLONK and Groth16 Bn254 artifacts from scratch, you can use the Makefile inside the prover directory.

cd prover
RUST_LOG=info make build-circuits