Computing Environment

We are using 3 different platforms in the course

• 1. The CSME program's Bang cluster.
  • Instructions for setting up your environment on Bang.
  • Running non-interactive batch jobs on Bang.

• 2. Two Graphical processing unit platforms
  • Dirac, a GPU cluster located at NERSC.
    Like the Bang cluster, Dirac provides batch queues that support exclusive access to hardware resources. Dirac also provides interactive access, but you will need to offload some of your code development onto other platforms.

  • Development platforms. Though there are interactive nodes on Dirac, we're a large class and will also have access to local (UCSD) interactive platforms. These machines should be used for development.
    • 28 Mac Minis, each with a low cost NVIDIA GeForce 320M device (1.2 capability). The Mac Minis are located in the Math Department in APM 2402, but may be accessed remotely. There are constraints on how these machines may be accessed, so be sure to read Getting started with the Roomful of Mac Minis to learn more.
    • Interactive GPU stations, located in the CSE Department. Most of the devices use Fermi though configurations are different from Dirac's.

Obtaining Device Characteristics

You can obtain information about the device (GPU) with the deviceQuery app, which lives in the SDK directory structure.

$SDK_HOME/C/bin/$OS/release/deviceQuery

• On Macs: /Developer/GPU\ Computing, darwin
• On Lilliput/CseclassNN: /opt/nvidia/class/NVIDIA_GPU_Computing_SDK, linux
• On Dirac: /usr/common/usg/cuda-sdk/4.1, linux

Obtaining Multicore Processor Characteristics

Here is a description of the memory hierarchy details of the multicore processors used in the course. This list will continue to grow

Here is how to Explore your processor's attributes.

Maintained by [Tue Oct 16 19:36:03 PDT 2012]