Download Benchmarks - Revision history

Tcarlson at 13:22, 12 October 2012

2012-10-12T13:22:46Z

2012-08-10T04:04:02Z

2012-04-07T13:51:15Z

2012-04-07T13:50:34Z

2012-02-19T17:02:11Z

2012-02-19T17:01:56Z

2012-02-19T17:01:16Z

2012-02-19T16:59:51Z

2012-01-16T14:33:58Z

2012-01-16T14:32:00Z

@@ Line 22: / Line 22: @@
 * tar xjf sniper-benchmarks.tbz
 * # or
-* git clone http://snipersim.com/git/benchmarks.git
+* git clone http://snipersim.org/git/benchmarks.git
 <br>
 * cd benchmarks

@@ Line 10: / Line 10: @@
 * <code>-n numcores</code> (required): number of threads to run. Note that this is the total number of threads that will run (equal to the number of available cores in the simulated machine). For Parsec, this means that the <code>-n</code> passed to the benchmark is usually lower since many Parsec programs start additional helper threads or use multiple threads per requested core.
 * <code>-i inputsize</code> (required): input set name. Usually <code>test</code>, <code>small</code> and <code>large</code> are available, as defined in our [[Paper:Iiswc2011Heirman|IISWC 2011]] paper.
 * <code>-c</code>, <code>-g</code>, <code>-s</code>: passed to <code>sniper/run-sniper</code> unmodified
@@ Line 28: / Line 29: @@
 * make
 * ./run-sniper -p splash2-fft -i test -n 4 -c gainestown
 </code>

@@ Line 7: / Line 7: @@
 * Make sure to define the GRAPHITE_ROOT (pointing to Sniper) and BENCHMARKS_ROOT (pointing to the location where sniper-benchmarks.tbz is installed)
 * Run <code>run-sniper</code> without any commandline options to see a list of supported benchmarks
-* <code>-p suitename-benchmarkname</code> (required): suite (splash2, parsec) and name (fft, barnes, ...) of the benchmark
+* <code>-p suitename-benchmarkname</code> (required): suite (splash2, parsec) and name (fft, blackscholes, ...) of the benchmark
 * <code>-n numcores</code> (required): number of threads to run. Note that this is the total number of threads that will run (equal to the number of available cores in the simulated machine). For Parsec, this means that the <code>-n</code> passed to the benchmark is usually lower since many Parsec programs start additional helper threads or use multiple threads per requested core.
 * <code>-i inputsize</code> (required): input set name. Usually <code>test</code>, <code>small</code> and <code>large</code> are available, as defined in our [[Paper:Iiswc2011Heirman|IISWC 2011]] paper.

@@ Line 1: / Line 1: @@
-A package with benchmarks is available at [http://snipersim.org/packages/sniper-benchmarks.tbz sniper-benchmarks.tbz]. These have been slightly adapted to:
+A package with benchmarks is available at [http://snipersim.org/packages/sniper-benchmarks.tbz sniper-benchmarks.tbz] (compressed archive) or http://snipersim.com/git/benchmarks.git (Git repository). These have been slightly adapted to:
 * Build using the compilation switches that Sniper requires (e.g. -mno-sse4), see [[Porting applications to Sniper]].
 * Mark the start and end of the parallel section (region of interest, ROI), see [[Simulator hooks]].
@@ Line 20: / Line 20: @@
 * wget http://snipersim.org/packages/sniper-benchmarks.tbz
 * tar xjf sniper-benchmarks.tbz
 * cd benchmarks
 * export GRAPHITE_ROOT=/path/to/sniper

@@ Line 14: / Line 14: @@
 This distribution currently contains the following benchmark suites:
 * SPLASH-2
-* [http://parsec.cs.princeton.edu/ Parsec 2.1]
+* [http://parsec.cs.princeton.edu/ PARSEC 2.1]
 Usage example: