Dyndatarace

This repo includes the precompiled executables and source codes of Dyndatarace, a binary instrumentation based hybrid dynamic data race detection tool. All the required data for executing the experiments described in below publications are included as well.

Onder Kalaci, DYNAMIC DATA RACE DETECTION IN CONCURRENT PROGRAMS, MS Thesis, Bogazici University, 2014. Supervisor: Alper Sen

Onder Kalaci, Alper Sen, "Hybrid Data Race Detection for Multithreaded Programs", Technical Report, Bogazici University, 2014.

The distribution includes four different pintools are under the folder "pintools". Our algorithm is in SegmentBasedHybrid directory.

In our experiments, we used the following vesions of tools:

PIN version: "pin-2.1256759-gcc-4.4.7". Apache httpd web server version: 2-2.22 pbzip2 version: Parallel BZIP2 v1.1.6 Parsec version: PARSEC 3.0 Firefox: Firefox 28.0 (used in only in the technical report)

=========== How to compile from sources

First, set the "PIN_ROOT" variable as the root directory of your PIN directory. Then, for each of the detector (sources/PureHappensBefore, sources/PureLockset, sources/HybridDetector), change the directory, and execute a single "make" command. For each detector, a folder named "obj-intel64" will be created. The file with the ".so" extention (ie. PureLockset.so) is the pintool which is required for insturmentation.

=========== Inputs for httpd and pbzip2 is under "inputs" folder. We do not include the inputs for PARSEC 3.0 benchmark applications. They are already freely avalible on the web "parsec.cs.princeton.edu/download.htm".

inputs/httpd: a) index.html is the 8.3k static web page that is used in the experiments. b) generator.py is the script that generates the clients which connect to the httpd web server, and GETs the 8.3k web page. c) executeHTTPDTest.py is the script that executes httpd web server with insturmentation, and writes the results to a file.
inputs/pbzip a) executePBZIPTest.py is the script that executes pbzip compressor with insturmentation, and writes the results to a file. b) example.txt is the file that is used as the input to the pbzip2 compression tool. c) generator.py is the script that is used to generate some other input files for different tests.
inputs/parsec a) As already mentioned, we do not include the inputs for parsec 3.0, which are already freely avalible on the web "parsec.cs.princeton.edu/download.htm". b) How to run parsec experiments: i) Firstly, set the directories for parsec and pin executable from the first two lines of the script. (PARSEC_DIR and PIN_EXECUTABLE variables) ii) Add the following lines to the for loop on the 296th line of the script:

For lockset implementataion: example.executeOurImpelemtation("purelocset", "PureLocksetImp.so",{} , "purelockset_") For hb implementataion: example.executeOurImpelemtation("purehb", "PureHappensBeforeImp.so",{}, "purehb_") For hybrid implementataion with no optimizations enabled: example.executeOurImpelemtation("hybrid", "HybridDetector.so",{} , "tsan_1_0") iii) How to execute segment based hybird implementation with optimizations: Add optimizations with values to the "executeOurImpelemtation" call as the 3rd input.

Optimization 1 :maxVCHistoryCount Optimization 2: enable_signle_access Optimization 3: segmentCountForThisExecution (this must be given relative to total segment count in the original execution. Refer to the paper.) Optimization 4: sample_rate
```
 	example.executeOurImpelemtation("hybrid", "HybridDetector.so",{ "enable_signle_access" : ""}, "tsan_131072_1")
     example.executeOurImpelemtation("hybrid", "HybridDetector.so",{"maxVCHistoryCount" :100] } , "tsan_1_0")
 	example.executeOurImpelemtation("hybrid", "HybridDetector.so",{"sample_rage" :0.1] } , "tsan_1_0")
 	executeOurImpelemtation("hybrid", "HybridDetector.so",{"max_single_segment_set_size":12000 } , "tsan_1_0")
 	or combinations:
 	example.executeOurImpelemtation("hybrid", "HybridDetector.so",{"max_single_segment_set_size":segmentCountForThisExecution, "enable_signle_access" : "", "sample_rate" :str(SAMPLE_RATES[counter]) } , "tsan_1_0")
```

c) How to get output: Fourth parameter given to "executeOurImpelemtation" is the file where the outputs are written for each application. Moreover, in the end, a file with name format "HH:MM:SS_DD_MM_YYYY_execution_history.txt" is written, which includes all the test results executed with "executeOurImpelemtation" function call. This file is an XML file, which is easy to read and understand. ie: xml <OWN> <RACE_COUNT>0</RACE_COUNT> <ALGORITHM>hybrid</ALGORITHM> <maxVCHistoryCount>10</maxVCHistoryCount> <APP_NAME>swaptions</APP_NAME> <segment_history_table_size>63.6363636364</segment_history_table_size> <TIME>70.6933810711</TIME> <INPUT_TYPE>simsmall</INPUT_TYPE> <enable_signle_access/> </OWN>

===========

onderkalaci/dyndatarace

Dyndatarace