Contact Us
Brown Deer Technology Home
COPRTHR

COPRTHR-1

COPRTHR SDK. The CO-PRocessing THReads (COPRTHR) SDK provides libraries and tools that simplify the use of heterogeneous computing platforms for developers targeting accelerators and co-processors.

Features. COPRTHR (a.k.a. "copper threads") provides the following libraries and tools:

  • STDCL® - A portable API for targeting compute offload accelerators and co-processors. (learn more)
  • Tools - The clcc tools support a standard workflow for accelerators and co-processors, generating pre-compiled executables and libraries. (learn more)
  • CLETE - An expression template engine used to accelerate the processing of C++ container classes. (learn more)
  • libcoprthr_opencl - An open-source OpenCL™ implementation for x86_64, ARM, Epiphany, and MIC multicore processors. (learn more)

Latest Release. Version 1.6.2 (Freewill) is available for download.

COPRTHR (CO-PRocessing THReads) Support. Support is available for end-user developers and includes application and algorithm porting, design, and optimization for co-processor architectures using the COPRTHR SDK. Additional technical services are also available including support for porting the COPRTHR SDK to additional processor architectures.

COPRTHR: Tools

Tools. The COPRTHR SDK provides tools for developers targeting heterogeneous computing platforms with accelerators and co-processors.

clcc The clcc tools support a standard workflow for accelerators and co-processors, generating pre-compiled executables and libraries.

OpenCL is a trademark of Apple Inc. used by permission by The Khronos Group
The Khronos Group develops and maintains the OpenCL standard.

STDCL

STDCL® is a portable API for targeting compute offload accelertors and co-processors. STDCL is designed in a style inspired by familiar and traditional UNIX APIs for C programming. The implementation of STDCL leverages OpenCL™ to provide broad portability across modern accelerator and co-processor architectures.

Features. STDCL provides the following features and support:

  • - Default compute contexts
  • - Dynamic loader for compute-offload programs
  • - Distributed memory management
  • - Compute offload kernel execution
  • - Synchronization

Documentation. A detailed description of STDCL can be found in the STDCL Reference Manual.

Example. The following simple example shows how SDTCL can be used to simplify offloading computations to an accelerator:

libcoprthr_opencl

libcoprthr. The libcoprthr_opencl library is an open-source OpenCL™ run-time implementation supporting x86_64, ARM, Epiphany and MIC multicore processors.  Although the current release has not yet passed conformance testing, the libary has been tested as a drop-in replacement supporting several complex OpenCL applications and exhibited good performance.

CLETE

CLETE. The Compute Layer Expression Template Engine is a software technology that enables the automatic acceleration of C++ container classes using a co-processor device such as a GPU or MIC accelerator. The use of CLETE requires no special tool or compiler and works with any conformant C++ compiler. All that is required is the inclusion of the appropriate header for the STDCL-extended container class and a single #define setting the desired behavior.

Features. CLETE provides the following features and support:

  • - Automatic kernel generation
  • - Automatic host-code instrumentation
  • - Programmer control over automatic, semi-automatic and manual operation

Example. The following simple example shows the use of the clvector container class with fully automatic acceleration enabled with CLETE.  The simple vector expression will be executed on a co-processor device such as a GPU or MIC accelerator:

Copyright © 2017 Brown Deer Technology, LLC.  |  Trademarks