Difference between revisions of "GPU Devel Nodes"

From oldwiki.scinet.utoronto.ca
Jump to navigation Jump to search
m
 
(88 intermediate revisions by 6 users not shown)
Line 1: Line 1:
{{Infobox Computer
+
{| style="border-spacing: 8px; width:100%"
|image=[[Image:GeForce_9800_GT_3qtr_low.png|center|300px|thumb]]
+
| valign="top" style="cellpadding:1em; padding:1em; border:2px solid; background-color:#f6f674; border-radius:5px"|
|name=GPU Development Cluster
+
'''WARNING: SciNet is in the process of replacing this wiki with a new documentation site. For current information, please go to [https://docs.scinet.utoronto.ca https://docs.scinet.utoronto.ca]'''
|installed=June 2010
+
|}
|operatingsystem= Linux
 
|loginnode= cell-srv01 (from <tt>login.scinet</tt>)
 
|numberofnodes=8
 
|rampernode=48 Gb
 
|corespernode=8
 
|interconnect=Infiniband,GigE
 
|vendorcompilers=gcc,nvcc
 
}}
 
  
The Intel nodes have two 2.53GHz 4core Xeon X5550 CPU's with 48GB of RAM per node with 3 containing NVIDIA 9800GT GPUs. 
+
<span style="color:#772222">The ARC GPU have been decommisioned. The head node, arc01, is still up, however for GPU computations, users are encouraged to move to the [[Gravity]] cluster.  For visualization, new [[Visualization Nodes]] are being setup.</span>
 
 
===Login===
 
 
 
First login via ssh with your scinet account at <tt>login.scinet.utoronto.ca</tt>, and from there you can proceed to <tt>cell-srv01</tt> which
 
is currently the gateway machine.
 
 
 
==Compile/Devel/Compute Nodes==
 
 
 
=== Nehalem (x86_64) ===
 
You can log into any of 8 nodes '''<tt>cell-srv[01-08]</tt>''' directly however each
 
of the nodes have differing configurations.
 
 
 
* '''<tt>cell-srv01</tt>''' - login node & nfs server, GigE connected
 
* '''<tt>cell-srv[02-05]</tt>''' - no GPU, GigE connected
 
* '''<tt>cell-srv[06-07]</tt>''' - 1x NVIDIA 9800GT GPU, Infiniband connected
 
* '''<tt>cell-srv08</tt>''' - 2x NVIDIA 9800GT GPU, GigE connected
 
 
 
==Local Disk==
 
 
 
This test cluster currently cannot see the global <tt>/home</tt> and <tt>/scratch</tt> space so you will have to copy (scp,etc..) your code to a separate local<tt>/home</tt> 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, NVIDIA's CUDA framework or OpenCL.
 
 
 
=== CUDA ===
 
 
 
<pre>
 
/usr/local/cuda/
 
</pre>
 
 
 
=== OpenCL ===
 
Demos, examples, and build details available in
 
 
 
===Compilers===
 
 
 
* '''nvcc''' -- Nvidia GCC compiler
 
 
===MPI===
 
 
 
Still a work in progress.
 
 
 
 
 
== Documentation ==
 
* CUDA
 
** http...
 
 
 
* OpenCL
 
** http...
 
 
 
== Further Info ==
 
 
 
 
 
 
 
== User Codes ==
 
 
 
Please discuss put any relevant information/problems/best practices you have encountered when using/developing for CUDA and/or OpenCL
 

Latest revision as of 19:26, 31 August 2018

WARNING: SciNet is in the process of replacing this wiki with a new documentation site. For current information, please go to https://docs.scinet.utoronto.ca

The ARC GPU have been decommisioned. The head node, arc01, is still up, however for GPU computations, users are encouraged to move to the Gravity cluster. For visualization, new Visualization Nodes are being setup.