GPU Devel Nodes

From oldwiki.scinet.utoronto.ca
Revision as of 14:28, 13 April 2010 by Northrup (talk | contribs) (Created page with '{{Infobox Computer |image=center|300px|thumb |name=Cell Development Cluster |installed=June 2010 |operatingsystem= Linux |logi…')
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
Cell Development Cluster
300px-Cell Broadband Engine Processor.jpg
Installed June 2010
Operating System Linux
Interconnect Infiniband
Ram/Node 32 Gb
Cores/Node 2 PPU + 16 SPU
Login/Devel Node cell-srv01 (from login.scinet)
Vendor Compilers ppu-gcc, spu-gcc

The Cell Development Cluster is a technology evaluation cluster with a combination of 14 IBM PowerXCell 8i "Cell" nodes and 8 Intel x86_64 "Nehalem" nodes. The QS22's each have two 3.2GHz "IBM PowerXCell 8i CPU's, where each CPU has 1 Power Processing Unit (PPU) and 8 Synergistic Processing Units (SPU), and 32GB of RAM per node. The Intel nodes have two 2.53GHz 4core Xeon X5550 CPU's with 48GB of RAM per node.

Login

First login via ssh with your scinet account at login.scinet.utoronto.ca, and from there you can proceed to cell-srv01 which is currently the gateway machine.

Compile/Devel/Compute Nodes

Cell

You can log into any of 12 nodes blade[03-14] directly to compile/test/run Cell specific or OpenCL codes.

Nehalem (x86_64)

You can log into any of 8 nodes cell-srv[01-08] directly, however they are not configured to work with the Cell blades yet.

Local Disk

This test cluster currently cannot see the global /home and /scratch space so you will have to copy (scp,etc..) your code to a separate local/home dedicated for this cluster. Also initially you will probably want to copy/modify your scinet .bashrc, .bashrc_profile, and .ssh directory onto this system.

Programming Frameworks

Currently there are two programming frameworks to use. The Cell SDK is the most full featured and "supported" but is architecture specific, whereas the OpenCL should allow code portability to other accelerators (like GP-GPU's) however is from IBM alphaworks so should be considered beta.

OpenCL

Examples are in

/usr/share/doc/OpenCL-0.1-ibm

Cell SDK

Demos, examples, and build details available in

/opt/cell/sdk/

Compilers

NOTE: gcc has been replaced, use ppu-gcc for most cases where gcc whould be used

  • ppu-gcc -- GNU GCC PPC compiler
  • spu-gcc -- GNU GCC SPU compiler
  • xlcl -- IBM XL OpenCL compiler

MPI

Still a work in progress.


Documentation

Further Info

IBM Cell Broadband engine tutorials http://publib.boulder.ibm.com/infocenter/ieduasst/stgv1r0/index.jsp


User Codes

Please discuss put any relevant information/problems/best practices you have encountered when using/developing for CELL and/or OpenCL

MHD

CFD