[Swift-commit] r6066 - SwiftApps/SciColSim/benchmarks

[davidk at ci.uchicago.edu]

Author: davidk
Date: 2012-11-20 09:47:29 -0600 (Tue, 20 Nov 2012)
New Revision: 6066

Added:
   SwiftApps/SciColSim/benchmarks/README
Log:
README for benchmark scripts


Added: SwiftApps/SciColSim/benchmarks/README
===================================================================
--- SwiftApps/SciColSim/benchmarks/README	                        (rev 0)
+++ SwiftApps/SciColSim/benchmarks/README	2012-11-20 15:47:29 UTC (rev 6066)
@@ -0,0 +1,57 @@
+SciColSim Benchmark Scripts
+===========================
+
+Prerequisites
+-------------
+Please ensure that Swift is in your $PATH. This script
+has been tested on the following machines:
+
+beagle
+midway
+pads
+raven
+uc3 
+
+How to Run
+----------
+The start-benchmark.sh script is used to start a run. Each successful
+invocation of this script will create a new run directory with the name
+of run<NNN>.
+
+The options you pass to this script will determine how the script is run.
+
+Usage: start-benchmark.sh -site sitename -start startval -stop stopval -step stepval [-limit limitval]
+
+Valid options for sitename are beagle, midway, pads, raven, and uc3.
+The start value represents the minimum value for target innovation.
+The stop value represents the maximum value for target innovation.
+The step value determines the frequency of tasks to run between start and stop values
+Limit represents a maximum time limit, in seconds. If the task exceeds this limit, the program will end and return a result of -1
+
+Example:
+
+-----
+./start-benchmark.sh -site beagle -start 1000 -stop 2000 -step 100 -limit 3600
+-----
+
+This will create 10 tasks on Beagle with target innovation values of 1000, 1100, 
+1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900 and 2000. The tasks will run for a 
+maximum of 1 hour. 
+
+Time logs
+---------
+As tasks complete, a log file will be created for each task. The file name for these
+time logs have the format of times/times-<tivalue>.out. The format of these files are:
+
+<tivalue> <seconds>
+
+Errors
+------
+The errors directory contains the standard error output of the benchmark.sh script. This
+can be useful while debugging.
+
+Plot
+----
+Plots get generated by a gnuplot script called plot.gp. The resulting plot file called be called
+results/plot.png.
+