Frequently Asked Questions

General

• What is Thrust?
  • Thrust is a C++ template library of parallel algorithms. Thrust allows you to program parallel architectures using an interface similar to the C++ Standard Template Library (STL).
• What is a C++ template library?
  • C++ templates are a way to write generic algorithms and data structures. A template library is simply a cohesive collection of such algorithms and data structures in a single package.
• Do I need to build Thrust?
  • No. Since Thrust is a template library you just #include the appropriate header files into your .cu (or .cpp) file and compile with nvcc (c++ compiler).
• What data structures does Thrust provide?
  • Currently Thrust provides vector data structures (e.g., thrust::device_vector) which are analogous to std::vector in the STL. These vector data structures simplify memory management and data transfer.
• What algorithms does Thrust provide?
  • sorting: thrust::sort and thrust::sort_by_key
  • transformations: thrust::transform
  • reductions: thrust::reduce and thrust::transform_reduce
  • scans: thrust::inclusive_scan, thrust::exclusive_scan, thrust::transform_inclusive_scan, etc.
  • Refer to the documentation for a complete listing.
• What version of CUDA do I need to develop GPU applications with Thrust?
  • The latest version of Thrust requires CUDA 4.0 or newer.
• What platforms does Thrust support?
  • Thrust has been tested extensively on Linux, Windows, and OSX systems.
• When will Thrust support OpenCL?
  • The primary barrier to OpenCL support is the lack of an OpenCL compiler and runtime with support for C++ templates (e.g., something similar to nvcc and the CUDA Runtime). These features are necessary to achieve close coupling of host and device codes.
• Does Thrust depend on any other libraries?
  • No, Thrust is self-contained and requires no additional libraries.
• Can I distribute Thrust with my application?
  • Yes! Thrust is open-source software released under liberal licensing terms.
• What open-source license does Thrust use?
  • Thrust is licensed under the Apache License v2.0.

Functionality

• Can I create a thrust::device_vector from memory I've allocated myself?
  • No. Instead, wrap your externally allocated raw pointer with thrust::device_ptr and pass it to Thrust algorithms.
• How do I find the array index of the element with the maximum/minimum value?
  • Use thrust::max_element or thrust::min_element, which are found int he file <thrust/extrema.h>
• Can I call Thrust algorithms inside a CUDA kernel?
  • No, it is not currently possible to call Thrust algorithms inside a __global__ or __device__ function.
• Can I call Thrust algorithms from CUDA Fortran?
  • Yes! This example shows how to call Thrust's sort algorithm from Fortran.

Troubleshooting

• If you're targeting the CUDA backend:
  • Make sure you are using CUDA 4.0 or greater:
    • run nvcc --version
  • Make sure you're compiling files that #include Thrust with nvcc.
  • Make sure that files that #include Thrust have a .cu extension. Other extensions (e.g., .cpp) will cause nvcc to treat the file incorrectly and produce an error message.
• If you're targeting the OpenMP backend:
  • Make sure you've enabled OpenMP code generation:
    • -fopenmp with g++
    • /openmp with cl.exe
• If all else fails, send a message to thrust-users and we'll do our best to assist you.