rCUDA 4: GPGPU as a Service in HPC Clusters

Print Friendly, PDF & Email

httpv://www.youtube.com/watch?v=fzRMXoyaRLQ

In this video, Antonio J. Peña from the Technical University of Valencia and Rafael Mayo from Jaume University present: rCUDA 4: GPGPU as a Service in HPC Clusters.

With rCUDA, GPU-accelerated applications are enabled to seamlessly interact with any GPU of a cluster independently of its exact physical location. This provides the possibility of sharing accelerators among different nodes, thus reducing power requirements. Decoupling GPUs from nodes, creating pools of accelerators, brings additional flexibility to cluster deployments.”

Recorded at the HPC Advisory Council Spain Workshop 2012 in Malaga. Download the slides (PDF).

Comments

  1. Sir i’m doing project in rCUDA ,i need to know how the task is allocated to GPU ,can i provide a task in hyperlink to connect to remote server???kindly provide me the details sir..
    thanks in advance…

  2. hi, iam doing project in rCUDA ,i need to connect the rCUDA server with rCUDA client .but i donot know how to set the LD_LIBRARY path and RCUDA environment variable..if anyone knows kindly reply me
    thanks in advance…