http:^^www.cs.cornell.edu^info^projects^sp-2^
来自「This data set contains WWW-pages collect」· EDU^INFO^PROJECTS^SP-2^ 代码 · 共 160 行
EDU^INFO^PROJECTS^SP-2^
160 行
MIME-Version: 1.0
Server: CERN/3.0
Date: Wednesday, 20-Nov-96 19:52:18 GMT
Content-Type: text/html
Content-Length: 5924
Last-Modified: Saturday, 02-Mar-96 21:48:01 GMT
<html><head><title>CUCS IBM SP-2</title></head><body><center><h1>Computer Science IBM SP-2</h1></center><hr><h2>Using the CUCS SP-2</h2><ul><li>The machine is called <i>granita</i>. The eight nodes are <i>granita1</i>through <i>granita8</i>.<li>If you have a CUCS login, you can use the SP-2.<li>Log into <i>granita1</i> or <i>granita2</i> which we've designated asinteractive nodes. Shells installed: <code>sh, bsh, csh, ksh, tcsh, bash, tsh</code>. If you experience problemsduring your first login, try to remove operating-system specific stufffrom your shell configuration file (for example, AIX does not havethe <code>arch</code> command; you can use <code>uname</code> instead). The file <code>/usr/lpp/bos/README</code> contains informationabout the release of AIX used on our SP-2. In addition to <code>man</code>you can use <i>InfoExplorer</i> to get more information about commandsand usage of the machine. To use this program, set up your remote display properly and type <code>info</code>.<li>Use <code>poe</code> to run parallel jobs that use neither ActiveMassages nor Split-C (and <code>info -l pe</code>or <code>man poe</code> to read more about <code>poe</code>).<li>Read below about how to run parallel programs that use Active Messagesor Split-C.</ul>More information about:<ul><li><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><!WA0><a href="http://www.tc.cornell.edu/UserDoc/Hardware/">IBM SP-2hardware</a>.<li><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><!WA1><a href="http://www.tc.cornell.edu">Cornell Theory Center SP-2</a>.</ul><hr><h2>Homegrown software</h2>In general, local software is installed in <code>/usr/u/sww</code>. Besure that <code>/usr/u/sww/sp2/bin</code> and <code>/usr/u/sww/sp2/gnu/bin<code> are in your path.<ul><li><h3>Split-C</h3><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><!WA2><a href="http://www.cs.cornell.edu/Info/Projects/Split-C/">Split-C</a> is a simple extension to C forparallel computing. It provides a global address space though globalpointers which can be dereferenced just like regular pointers.Split-phase assignment statements allow programmers to hide the latencyof remote accesses by overlapping computation and communication.Examples and makefiles can be found in <code>~sww/sp2/split-c-bench/cu-bench</code>.<p>Before working with Split-C, source <code>~sww/sp2/etc/sp2-setenv</code>. Users of non-csh shellsshould execute commands in <code>~sww/sp2/etc/sp2-setenv-non-csh</code>.To compile Split-C programs, create a <code>Makefile</code>(look at samples in various directories in <code>~sww/sp2/split-c-bench/cu-bench</code>) and type <code>gmake</code>.You must include <code>Make.split-c</code> in your <code>Makefile</code>!Split-C programs are run in the same way asActive Messages programs, i.e. using <code>amr</code> scriptslocated in <code>/usr/u/sww/sp-2/bin</code>. For example, to run a program<code>foo</code> on 3 processors type <code>amr3 ./foo</code><p><li><h3>Debugging Split-C</h3>To debug a Split-C program, the following steps need to be done:<ul><li> include split-c/debug.h<li> insert splitc_debug() as the first statement to be executed aftersplitc_main()<li> compile and run your program as described in the previous section<li> you will see the following message in node 0 (most commonly run ongranita1): Debugging Split-C -- hit enter to continue:"<li> before hitting return, log onto the node you want to debug (if youwant to debug the master node, open a new shell)<li> go to the directory where your program source is located<li> run gdb<li> inside gdb, do: <code>file am_run</code>, and then<code>attach pid</code>, where pid thethe proc id of the am_run process on the node being debugged<li> hit return on node 0 to let computation proceed<li> once you've attached gdb to am_run, am_run is stopped by gdb, andyou can set breakpoints, look at stack frames, etc.</ul><p><li><h3>Active Messages</h3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><!WA3><a href="http://www.cs.cornell.edu/Info/Projects/CAM">Active Messages</a> is a low-overhead communication layerthat offers high-performance communication on many parallel machines.A native Active Messages layer (SP2AM) is now available for the SP-2.The main performance characteristics of SP2 AM are a one-word round-triplatency of 51 us and an asymptotic network bandwidth of 34.3 MB/s.<p>The SP2AM library is found in <code>/usr/u/sww/sp-2/lib/libsp2gam.a</code>and the header file is in <code>/usr/u/sww/sp-2/include</code>. Beforerunning programs that use Active Messages, source <code>~sww/sp2/etc/sp2-setenv</code> and read<code>/usr/u/sww/sp-2/gam-1.0/doc/RunningPrgms</code>. The <code>amr</code> scripts are located also in <code>/usr/u/sww/sp-2/bin</code>.<p><li><h3>MPI</h3><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><!WA4><a href="http://www.mcs.anl.gov/mpi/index.html">MPI</a> is a popularmessage passing interface for portable parallel programs. We have animplementation of MPI (based on the MPICH library) running overActive Messages on the SP-2.The header files are located in <code>/usr/u/sww/sp-2/include</code>.The library file is located in <code>/usr/u/sww/sp-2/lib</code>.The easiest way to compile and link is with the script file "ampicc" (whichis built on top of xlC):<p><code>ampicc -O3 foo.c -o foo<p>You can also compile MPI programs with xlC, gcc, and split-cc (please lookat the examplesin the directory ~sww/sp2/ampi/examples for information about this).MPI programs are run exactly likeordinary Active Messages programs (i.e. "<code>amr4 foo</code>").Be sure to source <code>~sww/sp2/etc/sp2-setenv</code>.</ul><p><p></ul><hr><h2>Other software</h2>Software available on granita1 and granita2 also includes tcsh, bash, C Set ++ (xlC), Fortran (xlf), xpdbx, X11, matlab. GNU software installed in<code>~sww/sp2/gnu</code> includes emacs, gmake, gcc, g++, gdb, bison. Some of it is replicated locally in /usr/local/gnu/bin.<p><hr><h2>Problems</h2>If you experience difficulties with the SP-2, please contact the SP-2czar <!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><!WA5><a href="http://www.cs.cornell.edu/Info/People/grzes/grzes.html">Grzegorz Czajkowski</a>.</html>⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?