# Thrust: Code at the speed of light Thrust is a C++ parallel programming library which resembles the C++ Standard Library. Thrust's **high-level** interface greatly enhances programmer **productivity** while enabling performance portability between GPUs and multicore CPUs. **Interoperability** with established technologies (such as CUDA, TBB, and OpenMP) facilitates integration with existing software. Develop **high-performance** applications rapidly with Thrust! Thrust is included in the NVIDIA HPC SDK and the CUDA Toolkit. ## Quick Start ### Getting the Thrust Source Code The CUDA Toolkit provides a recent release of the Thrust source code in `include/thrust`. This will be suitable for most users. Users that wish to contribute to Thrust or try out newer features should recursively clone the Thrust Github repository: ``` git clone --recursive https://github.com/NVIDIA/thrust.git ``` ### Using Thrust From Your Project Thrust is a header-only library; there is no need to build or install the project unless you want to run the Thrust unit tests. For CMake-based projects, we provide a CMake package for use with `find_package`. See the [CMake README](thrust/cmake/README.md) for more information. Thrust can also be added via `add_subdirectory` or tools like the [CMake Package Manager](https://github.com/cpm-cmake/CPM.cmake). For non-CMake projects, compile with: - The Thrust include path (`-I`) - The CUB include path, if using the CUDA device system (`-I/dependencies/cub/`) - By default, the CPP host system and CUDA device system are used. These can be changed using compiler definitions: - `-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_XXX`, where `XXX` is `CPP` (serial, default), `OMP` (OpenMP), or `TBB` (Intel TBB) - `-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_XXX`, where `XXX` is `CPP`, `OMP`, `TBB`, or `CUDA` (default). ### Examples Thrust is best explained through examples. The following source code generates random numbers serially and then transfers them to a parallel device where they are sorted. ```c++ #include #include #include #include #include #include #include int main(void) { // generate 32M random numbers serially thrust::host_vector h_vec(32 << 20); std::generate(h_vec.begin(), h_vec.end(), rand); // transfer data to the device thrust::device_vector d_vec = h_vec; // sort data on the device (846M keys per second on GeForce GTX 480) thrust::sort(d_vec.begin(), d_vec.end()); // transfer data back to host thrust::copy(d_vec.begin(), d_vec.end(), h_vec.begin()); return 0; } ``` This code sample computes the sum of 100 random numbers in parallel: ```c++ #include #include #include #include #include #include #include int main(void) { // generate random data serially thrust::host_vector h_vec(100); std::generate(h_vec.begin(), h_vec.end(), rand); // transfer to device and compute sum thrust::device_vector d_vec = h_vec; int x = thrust::reduce(d_vec.begin(), d_vec.end(), 0, thrust::plus()); return 0; } ``` Additional usage examples can be found in the [`examples/`](examples/) and [`testing/`](testing/) directories of the Github repo. ## Documentation Resources - [API Reference](https://thrust.github.io/doc/modules.html) - [Examples](https://github.com/NVIDIA/thrust/tree/main/examples) - [User Support](https://github.com/NVIDIA/thrust/discussions) ## CI Status ## Supported Compilers Thrust is regularly tested using the specified versions of the following compilers. Unsupported versions may emit deprecation warnings, which can be silenced by defining THRUST_IGNORE_DEPRECATED_COMPILER during compilation. - NVCC 11.0+ - NVC++ 20.9+ - GCC 5+ - Clang 7+ - MSVC 2019+ (19.20/16.0/14.20) ## Releases Thrust is distributed with the NVIDIA HPC SDK and the CUDA Toolkit in addition to GitHub. See the [changelog](CHANGELOG.md) for details about specific releases. | Thrust Release | Included In | | ----------------- | --------------------------------------- | | 1.16.0 | TBD | | 1.15.0 | NVIDIA HPC SDK 22.1 & CUDA Toolkit 11.6 | | 1.14.0 | NVIDIA HPC SDK 21.9 | | 1.13.1 | CUDA Toolkit 11.5 | | 1.13.0 | NVIDIA HPC SDK 21.7 | | 1.12.1 | CUDA Toolkit 11.4 | | 1.12.0 | NVIDIA HPC SDK 21.3 | | 1.11.0 | CUDA Toolkit 11.3 | | 1.10.0 | NVIDIA HPC SDK 20.9 & CUDA Toolkit 11.2 | | 1.9.10-1 | NVIDIA HPC SDK 20.7 & CUDA Toolkit 11.1 | | 1.9.10 | NVIDIA HPC SDK 20.5 | | 1.9.9 | CUDA Toolkit 11.0 | | 1.9.8-1 | NVIDIA HPC SDK 20.3 | | 1.9.8 | CUDA Toolkit 11.0 Early Access | | 1.9.7-1 | CUDA Toolkit 10.2 for Tegra | | 1.9.7 | CUDA Toolkit 10.2 | | 1.9.6-1 | NVIDIA HPC SDK 20.3 | | 1.9.6 | CUDA Toolkit 10.1 Update 2 | | 1.9.5 | CUDA Toolkit 10.1 Update 1 | | 1.9.4 | CUDA Toolkit 10.1 | | 1.9.3 | CUDA Toolkit 10.0 | | 1.9.2 | CUDA Toolkit 9.2 | | 1.9.1-2 | CUDA Toolkit 9.1 | | 1.9.0-5 | CUDA Toolkit 9.0 | | 1.8.3 | CUDA Toolkit 8.0 | | 1.8.2 | CUDA Toolkit 7.5 | | 1.8.1 | CUDA Toolkit 7.0 | | 1.8.0 | | | 1.7.2 | CUDA Toolkit 6.5 | | 1.7.1 | CUDA Toolkit 6.0 | | 1.7.0 | CUDA Toolkit 5.5 | | 1.6.0 | | | 1.5.3 | CUDA Toolkit 5.0 | | 1.5.2 | CUDA Toolkit 4.2 | | 1.5.1 | CUDA Toolkit 4.1 | | 1.5.0 | | | 1.4.0 | CUDA Toolkit 4.0 | | 1.3.0 | | | 1.2.1 | | | 1.2.0 | | | 1.1.1 | | | 1.1.0 | | | 1.0.0 | | ## Development Process Thrust uses the [CMake build system](https://cmake.org/) to build unit tests, examples, and header tests. To build Thrust as a developer, the following recipe should be followed: ``` # Clone Thrust and CUB repos recursively: git clone --recursive https://github.com/NVIDIA/thrust.git cd thrust # Create build directory: mkdir build cd build # Configure -- use one of the following: cmake .. # Command line interface. ccmake .. # ncurses GUI (Linux only) cmake-gui # Graphical UI, set source/build directories in the app # Build: cmake --build . -j # invokes make (or ninja, etc) # Run tests and examples: ctest ``` By default, a serial `CPP` host system, `CUDA` accelerated device system, and C++14 standard are used. This can be changed during configuration -- see [CMake Options](CONTRIBUTING.md#cmake-options) for details. More information on configuring your Thrust build and creating a pull request can be found in [CONTRIBUTING.md](CONTRIBUTING.md).