BayeSmith

Setup

./build.sh

Run

Learn

$ bingo/learn -reuse -analysis_type [ interval | taint ] -debug $BENCH_NAME

e.g. bingo/learn -reuse -analysis_type interval -debug sort

Logs (learn.log) and output (.dl file) will be generated under learn-out directory. One may change the name of the directory with option -out_dir $DIRNAME.

Test

$ bingo/learn -test -analysis_type [ interval | taint ] -out_dir test-out $BENCH_NAME

e.g. bingo/learn -test -analysis_type interval -out_dir test-out -dl_from $PATH_TO_DL_FILE sort

Logs (learn.log) and output (.dl file) will be generated under test-out directory. One may run test with existing datalog rule file with option -dl_from $PATH_TO_DATALOG_FILE. One may run test with custom rule weights with option -rule_prob_from $PATH_TO_RULE_PROB_TXT_FILE. One may change the name of the directory with option -out_dir $DIRNAME.

Report

$ script/pldi19/report.sh $TIMESTAMP

e.g. script/pldi19/report.sh baseline will show statistics for baseline performace. Similarly, script/pldi19/report.sh pldi will show statistics for pldi21 submitted version performace.

Visualize Bayesian Network

$ bingo/visualizer $BNET_DIR

Above command shall prompt a shell starting with visualizer> . Then, one may run further commands to get the visuals. Here, we illustrate commonly used commands. For more information, just enter 'help' in the prompted shell as visualizer> help.

Then, the following command will output an svg file showing partial BNet that shows common ancestor between two alarms:

visualizer> common $ALARM_1 $ALARM_2 $OUT_FILE_NAME

The following command will output an svg file showing partial BNet that shows derivation of a single alarm. :

visualizer> single $ALARM $OUT_FILE_NAME

By default, it will visualize up to 100 close nodes used for alarm derivation. As one may want to track more nodes, will use the following command:

visualizer> single $ALARM $SIZE $OUT_FILE_NAME

$SIZE describes the maximum number of nodes to visualize.