The fastest way to GPGPU results
The perfect marriage, NVIDIA Tesla GPU workstation with CAPS DevDeck Linux. The suite includes CAPS HMPP, a simple directive-based compiler to build parallel GPU accelerated applications. Also included is HMPP Performance analyser, and HMPP Wizard, to guide your development choices. Based on a C and FORTRAN directives, HMPP source-to-source compiler preserves your original code investment. HMPP simplifies code parallelisation, memory management, to fully leverage the computing power of GPU processors, while integrating with the massively data-parallel backend for NVIDIA CUDA. The HMPP runtime also ensures application deployment on multi-GPU systems.
Work smarter! Use a 3 step methodology to reduce the cost:
This preliminary code study will highlight the potential of optimisation of your code on hybrid architecture.
Functional migration to GPU
CPU code optimisation, exhibition of SIMT parallelism, hotspot push on GPU and validation of CPU-GPU execution.
GPU Code Tuning for improving performance
By using this methodology you can implement the migration in a fast, easy and incremental way.
For HPC solutions and further information, please use the following contact details.
01727 201860 01727 201888 firstname.lastname@example.org
© 2011 NVIDIA Corporation. NVIDIA, the NVIDIA logo, GeForce, 3D Vision, and The way it’s meant to be played are trademarks and/or registered trademarks of NVIDIA Corporation. All rights reserved.