Skip to content

Portable and vendor neutral framework for parallel programming on heterogeneous platforms.(clang based transpiler)

License

Notifications You must be signed in to change notification settings

IuriiKobein/occa-pre-okl2.0

 
 

Repository files navigation

occa

 

license discussions slack github-ci codecov twitter

 

Performance, Portability, Transparency

OCCA is an open source, portable, and vendor neutral framework for parallel programming on heterogeneous platforms. The OCCA API provides unified models for heterogeneous programming concepts—such as a device, memory, or kernel—while the OCCA Kernel Language (OKL) enables the creation of portable device kernels using a directive-based extension to the C-language.

Mission critical computational science and engineering applications from the public and private sectors rely on OCCA. Notable users include the U.S. Department of Energy and Shell.

Key Features

  • Multiple backends—including CUDA, HIP, Data Parallel C++, OpenCL, OpenMP (CPU), and Metal
  • JIT compilation and caching of kernels
  • C, C++, and Fortran language support
  • Interoperability with backend API and kernels
  • Transparency—easy to understand how your code is mapped to each platform

Requirements

Minimum

  • CMake v3.17 or newer
  • C++17 compiler
  • C11 compiler

Optional

  • Fortan 90 compiler
  • CUDA 9 or later
  • HIP 3.5 or later
  • SYCL 2020 or later
  • OpenCL 2.0 or later
  • OpenMP 4.0 or later

Build, Test, Install

OCCA uses the CMake build system. Checkout the installation guide for a comprehensive overview of all build settings and instructions for building on Windows or Mac OS.

Linux

For convenience, the shell script configure-cmake.sh has been provided to drive the CMake build. Compilers, flags, and other build parameters can be adjusted there. By default, this script uses ./build and ./install for the build and install directories.

The following demonstrates a typical sequence of shell commands to build, test, and install occa:

$ ./configure.sh
$ cmake --build build --parallel <number-of-threads>
$ ctest --test-dir build --output-on-failure
$ cmake --install build --prefix install

If dependencies are installed in a non-standard location, set the corresponding environment variable to this path.

Use

Environment

During installation, the Env Modules file <install-prefix>/modulefiles/occa is generated. When this module is loaded, paths to the installed bin, lib, and include directories are appended to environment variables such as PATH and LD_LIBRARY_PATH.

Building an OCCA application

For convenience, OCCA provides CMake package files which are configured during installation. These package files define an imported target, OCCA::libocca, and look for all required dependencies.

For example, the CMakeLists.txt of downstream projects using OCCA would include

find_package(OCCA REQUIRED)

add_executable(downstream-app ...)
target_link_libraries(downstream-app PRIVATE OCCA::libocca)

add_library(downstream-lib ...)
target_link_libraries(downstream-lib PUBLIC OCCA::libocca)

Command-line Interface

The OCCA command-line interface can be found in <install-prefix>/bin/occa. This tool can be used to query information about hardware and the configuration of OCCA on a given platform.

For example, calling occa info will available OCCA backends and related hardware specs, while occa env display the values of OCCA related environment variables. To see the list of all available options, call occa --help.

$ occa info
========+======================+=================================
 CPU(s) | Processor Name       | AMD EPYC 7532 32-Core Processor 
        | Memory               | 251.6 GB                        
        | Clock Frequency      | 2.4 MHz                         
        | SIMD Instruction Set | SSE2                            
        | SIMD Width           | 128 bits                        
        | L1d Cache Size       |   1 MB                          
        | L1i Cache Size       |   1 MB                          
        | L2 Cache Size        |  16 MB                          
        | L3 Cache Size        | 256 MB                          
========+======================+=================================
 OpenCL | Platform 0           | NVIDIA CUDA                     
        |----------------------+---------------------------------
        | Device 0             | NVIDIA A100-PCIE-40GB           
        | Device Type          | gpu                             
        | Compute Cores        | 108                             
        | Global Memory        | 39.40 GB                        
========+======================+=================================
 CUDA   | Device Name          | NVIDIA A100-PCIE-40GB           
        | Device ID            | 0                               
        | Memory               | 39.40 GB                        
========+======================+=================================

Community

Support

Need help? Checkout the repository wiki or ask a question in the Q&A discussions category.

Feedback

To provide feedback, start a conversation in the general or ideas discussion categories.

Acknowledgements

This work was supported in part by

  • Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC02-06CH11357
  • The Exascale Computing Project (17-SC-20-SC), a joint project of the U.S. Department of Energy’s Office of Science and National Nuclear Security Administration, responsible for delivering a capable exascale ecosystem, including software, applications, and hardware technology, to support the nation’s exascale computing imperative
  • The Center for Efficient Exascale Discretizations (CEED), a co-design center within the U.S. Department of Energy Exascale Computing Project.
  • Intel
  • AMD
  • Shell

License

OCCA is available under a MIT license

About

Portable and vendor neutral framework for parallel programming on heterogeneous platforms.(clang based transpiler)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 87.7%
  • C 3.9%
  • Fortran 3.2%
  • Python 1.8%
  • CMake 1.0%
  • Makefile 0.8%
  • Other 1.6%