Google Git
Sign in
llvm / llvm-project / pstl / 1e1f0ea0d6597290b298c9e3e737d73e9e154ef3
commit1e1f0ea0d6597290b298c9e3e737d73e9e154ef3[log] [tgz]
authorLouis Dionne <ldionne@apple.com>Tue Aug 13 11:50:26 2019 +0000
committerCopybara-Service <copybara-worker@google.com>Tue Sep 01 14:44:30 2020 -0700
treefe2a854a8612a815a935e415cc3c3f09516859f7
parent4de88b585bc57a84135cb1fd41a89b75e2dad9d0 [diff]
[pstl] Rename PARALLELSTL_BACKEND to PSTL_PARALLEL_BACKEND

It makes more sense to name configuration options as PSTL_XXX.
Also, I'm naming it PSTL_PARALLEL_BACKEND because we might introduce
the ability to customize the vectorization backend, in which case
PSTL_BACKEND would become ambiguous.

llvm-svn: 368672
GitOrigin-RevId: 5065e7817393d9acf2689d3f666a1c375238a7e2
2 files changed
tree: fe2a854a8612a815a935e415cc3c3f09516859f7
  1. cmake/
  2. docs/
  3. include/
  4. test/
  5. .arcconfig
  6. .clang-format
  7. CMakeLists.txt
  8. CREDITS.txt
  9. LICENSE.txt
  10. README.md
README.md

Parallel STL

Parallel STL is an implementation of the C++ standard library algorithms with support for execution policies, as specified in ISO/IEC 14882:2017 standard, commonly called C++17. The implementation also supports the unsequenced execution policy specified in Parallelism TS version 2 and proposed for the next version of the C++ standard in the C++ working group paper P1001. Parallel STL offers efficient support for both parallel and vectorized execution of algorithms. For sequential execution, it relies on an available implementation of the C++ standard library.

Prerequisites

To use Parallel STL, you must have the following software installed:

  • C++ compiler with:
    • Support for C++11
    • Support for OpenMP* 4.0 SIMD constructs
  • Threading Building Blocks (TBB) which is available for download at https://github.com/01org/tbb/

Known issues and limitations

  • unseq and par_unseq policies only have effect with compilers that support #pragma omp simd or #pragma simd.
  • Parallel and vector execution is only supported for the algorithms if random access iterators are provided, while for other iterator types the execution will remain serial.
  • The following algorithms do not allow efficient SIMD execution: includes, inplace_merge, merge, nth_element, partial_sort, partial_sort_copy, set_difference, set_intersection, set_symmetric_difference, set_union, sort, stable_partition, stable_sort, unique.
  • The initial value type for exclusive_scan, inclusive_scan, transform_exclusive_scan, transform_inclusive_scan shall be DefaultConstructible. A default constructed-instance of the initial value type shall be the identity element for the specified binary operation.
  • For max_element, min_element, minmax_element, partial_sort, partial_sort_copy, sort, stable_sort the dereferenced value type of the provided iterators shall be DefaultConstructible.
  • For remove, remove_if, unique the dereferenced value type of the provided iterators shall be MoveConstructible.
  • The following algorithms require additional O(n) memory space for parallel execution: copy_if, inplace_merge, partial_sort, partial_sort_copy, partition_copy, remove, remove_if, rotate, sort, stable_sort, unique, unique_copy.