GPU acceleration

This guide explains how to update your existing program to leverage GPU acceleration, or to start a new program using GPU.

TFHE-rs now supports a GPU backend with CUDA implementation, enabling integer arithmetic operations on encrypted data.

Prerequisites

Cuda version >= 10
Compute Capability >= 3.0
gcc >= 8.0 - check this page for more details about nvcc/gcc compatible versions
cmake >= 3.24
Rust version - check this page

Importing to your project

To use the TFHE-rs GPU backend in your project, add the following dependency in your Cargo.toml.

If you are using an x86 machine:

tfhe = { version = "0.7.5", features = [ "boolean", "shortint", "integer", "x86_64-unix", "gpu" ] }

If you are using an ARM machine:

tfhe = { version = "0.7.5", features = [ "boolean", "shortint", "integer", "aarch64-unix", "gpu" ] }

For optimal performance when using TFHE-rs, run your code in release mode with the --release flag.

Supported platforms

TFHE-rs GPU backend is supported on Linux (x86, aarch64).

x86

aarch64

A first example

Configuring and creating keys.

Comparing to the CPU example, GPU set up differs in the key creation, as detailed here

Here is a full example (combining the client and server parts):

use tfhe::{ConfigBuilder, set_server_key, FheUint8, ClientKey, CompressedServerKey};
use tfhe::prelude::*;

fn main() {

    let config = ConfigBuilder::default().build();

    let client_key= ClientKey::generate(config);
    let compressed_server_key = CompressedServerKey::new(&client_key);

    let gpu_key = compressed_server_key.decompress_to_gpu();

    let clear_a = 27u8;
    let clear_b = 128u8;

    let a = FheUint8::encrypt(clear_a, &client_key);
    let b = FheUint8::encrypt(clear_b, &client_key);

    //Server-side

    set_server_key(gpu_key);
    let result = a + b;

    //Client-side
    let decrypted_result: u8 = result.decrypt(&client_key);

    let clear_result = clear_a + clear_b;

    assert_eq!(decrypted_result, clear_result);
}

Setting the keys

The configuration of the key is different from the CPU. More precisely, if both client and server keys are still generated by the client (which is assumed to run on a CPU), the server key has then to be decompressed by the server to be converted into the right format. To do so, the server should run this function: decompressed_to_gpu().

Once decompressed, the operations between CPU and GPU are identical.

Encryption

On the client-side, the method to encrypt the data is exactly the same than the CPU one, as shown in the following example:

    let clear_a = 27u8;
    let clear_b = 128u8;
    
    let a = FheUint8::encrypt(clear_a, &client_key);
    let b = FheUint8::encrypt(clear_b, &client_key);

Computation

The server first need to set up its keys with set_server_key(gpu_key).

Then, homomorphic computations are performed using the same approach as the CPU operations.

    //Server-side
    set_server_key(gpu_key);
    let result = a + b;

    //Client-side
    let decrypted_result: u8 = result.decrypt(&client_key);

    let clear_result = clear_a + clear_b;

    assert_eq!(decrypted_result, clear_result);

Decryption

Finally, the client decrypts the results using:

    let decrypted_result: u8 = result.decrypt(&client_key);

Improving performance

TFHE-rs allows to leverage the high number of threads given by a GPU. To maximize the number of GPU threads, update your configuration accordingly:

let config = ConfigBuilder::with_custom_parameters(PARAM_GPU_MULTI_BIT_MESSAGE_2_CARRY_2_GROUP_3_KS_PBS, None).build();

Here's the complete example:

use tfhe::{ConfigBuilder, set_server_key, FheUint8, ClientKey, CompressedServerKey};
use tfhe::prelude::*;
use tfhe::shortint::parameters::PARAM_GPU_MULTI_BIT_MESSAGE_2_CARRY_2_GROUP_3_KS_PBS;

fn main() {

    let config = ConfigBuilder::with_custom_parameters(PARAM_GPU_MULTI_BIT_MESSAGE_2_CARRY_2_GROUP_3_KS_PBS, None).build();

    let client_key= ClientKey::generate(config);
    let compressed_server_key = CompressedServerKey::new(&client_key);

    let gpu_key = compressed_server_key.decompress_to_gpu();

    let clear_a = 27u8;
    let clear_b = 128u8;

    let a = FheUint8::encrypt(clear_a, &client_key);
    let b = FheUint8::encrypt(clear_b, &client_key);

    //Server-side

    set_server_key(gpu_key);
    let result = a + b;

    //Client-side
    let decrypted_result: u8 = result.decrypt(&client_key);

    let clear_result = clear_a + clear_b;

    assert_eq!(decrypted_result, clear_result);
}

List of available operations

The GPU backend includes the following operations for both signed and unsigned encrypted integers:

All operations follow the same syntax than the one described in here.

Multi-GPU support

TFHE-rs supports platforms with multiple GPUs with some restrictions at the moment: the platform should have NVLink support, and only GPUs that have peer access to GPU 0 via NVLink will be used for the computation. Depending on the platform, this can restrict the number of GPUs used to perform the computation.

There is nothing to change in the code to execute on multiple GPUs, when they are available and have peer access to GPU 0 via NVLink. To keep the API as user-friendly as possible, the configuration is automatically set, i.e., the user has no fine-grained control over the number of GPUs to be used.

Benchmarks

All GPU benchmarks presented here were obtained on H100 GPUs, and rely on the multithreaded PBS algorithm. The cryptographic parameters PARAM_GPU_MULTI_BIT_MESSAGE_2_CARRY_2_GROUP_3_KS_PBS were used.

1xH100

Below come the results for the execution on a single H100. The following table shows the performance when the inputs of the benchmarked operation are encrypted:

The following table shows the performance when the left input of the benchmarked operation is encrypted and the other is a clear scalar of the same size:

2xH100

Below come the results for the execution on two H100's. The following table shows the performance when the inputs of the benchmarked operation are encrypted:

The following table shows the performance when the left input of the benchmarked operation is encrypted and the other is a clear scalar of the same size:

PreviousRust configuration NextOverflow detection

Last updated 21 days ago