We are happy to announce the new version 4.0 of rCUDA, cooked for more than one year. As for previous versions, it has been developed as a joint collaboration between the Parallel Architectures Group from the Universitat Politecnica de Valencia and the High Performance Computing and Architectures Group from the Universitat Jaume I. This new rCUDA version has been succesfully presented in the Mellanox booth at SC12, held at Salt Lake City, Utah, USA.
The rCUDA Framework enables the concurrent usage of CUDA-compatible devices remotely.
It can be useful in three different environments:
- Clusters. To reduce the number of GPUs installed in High Performance Clusters. This leads to increase GPUs use and to energy savings, as well as other related savings like acquisition costs, maintenance, space, cooling, etc. It also allows providing a large amount of GPUs to a single application.
- Academia. In commodity networks, to offer access to a few high performance GPUs concurrently to many students.
- Virtual Machines. To enable the access to the CUDA facilities on the physical machine.
The new rCUDA v4.0 version implements most of the functions in the CUDA Runtime API version 5.0, excluding only those related with graphics interoperability. It additionally includes highly optimized TCP and low-level InfiniBand pipelined communications as well as full multi-thread and multi-node capabilities. rCUDA targets the Linux and Windows OSs (for 32- and 64-bit architectures) on both client and server sides.
In opposition to previous versions of rCUDA, which provided the CU2rCU code conversion utility, this new version provides full compatibility support with CUDA 5 (libraries included), so that your running programs do not need to be recompiled. Furthermore, this new rCUDA version also includes an initial integration with the SLURM scheduler, thus allowing rCUDA to dynamically use all the CPUs in your cluster.
If you are interested in using rCUDA, please proceed to the software request form page. The rCUDA team will be glad to send you a copy of the software at no cost. It is distributed for free.
For further information, please refer to the papers listed in the developer's personal webpage.