HMPP——Benefit from the performance of manycore systems
Benefit from the performance of manycore systemswhile reducing your development efforts
A complete hybrid compiler
with powerful data-parallel code generators
"By insulating hardware specific versions of functions from the native application source, HMPP™ preserves the legacy code, keeps the application independent from fast changing hardware platforms and offers interoperability."
Leverage the computing power of GPUs and FPGAs co-processors
Copyright © 2008 CAPS entreprise. All rights reserved.
Home
Based on OpenMP like directives:
a standard way of developing manycore applications
Dynamic application scaling
The HMPP™ directives allow to define hardware-specific implementations of functions, i.e. codelets, to be offloaded in accelerating coprocessors (NVIDIA© Tesla™, AMD FireStream™, SSE...). Their call site, data transfers and synchronization barriers are also indicated with the HMPP directives in the main application source.
The use of HMPP™ directives preserves application source from hardware-specific coding thus protecting software investments.
Complementary to OpenMP™ and MPI, HMPP™ enables the development of parallel applications that mix the best of the today's available parallel tools.
While preserving the application source from hardware-specific coding, HMPP™ applications dynamically adapts its execution to the system workload and platform configuration. This guarantees the scalability and interoperability of manycore applications.
HMPP™ Workbench includes a C and Fortran compiler, hardware-specific code generators and a runtime that seamlessly integrate in your environment and make use of the hardware vendor development tools and drivers.
The code generators are specifically designed to extract the most of data parallelism from C and Fortran kernels and translate them into the language of the targets such as NVIDIA® CUDA™ or SSE
http://www.caps-entreprise.com/hmpp.html
页:
[1]