Releases: LLNL/dataracebench
Releases · LLNL/dataracebench
DataRaceBench v1.4.1
DataRaceBench v1.4.0
- Added data race ground truth encoding into all microbenchmarks.
- Added 4 new C/C++ microbenchmarks and 2 Fortran microbenchmarks
- Added LLOV into test harness
- Provided high-precision evaluation scripts, log parsers for Intel Inspector, ThreadSanitizer, ROMP, Coderrect and LLOV
- Added DRB-metainfo.json for metadata information
DataRaceBench v1.3.2
- Various bug fixes for some Fortran tests and scripts based on user inputs
- Added the support for a static analysis tool.
DataRaceBench v1.3.1
- Added equivalent Fortran versions for existing C/C++ benchmark programs.
- Additional benchmark programs added to cover some of the latest OpenMP 5.0 language features.
- 222 new benchmarks added in total: 56 are in C and 166 are in Fortran.
- Collected new kernels representing new patterns from literature and other benchmarks (e.g., NAS Parallel Benchmark).
- Improved the test process to allow users to select customized subsets of benchmarks to evaluate their tools.
- Dockerized tools to streamline the entire benchmarking process to quickly generate a dashboard showing the state-of-the-art.
- Re-evaluated several available data race detection tools and generate more comprehensive results showing their strengths and limitations.
- Updated documentation.
DataRaceBench v1.2.0
- Increased the test coverage of OpenMP directives and clauses(up to OpenMP 4.5 ), including threadprivate, collapse, taskloop, teams, distribute, atomic, flush, depend, critical, barrier, taskwait, taskgroup, copyin and copyprivate, linear, if, etc.
- Increased the coverage of data-sharing attribute rules defined in OpenMP 4.5, including rules related to function arguments, static and dynamic storage duration, variables which are not referenced in a construct, mapped array sections, and clauses (collapse and order) associating multiple for loops, etc.
- Renamed all microbenchmarks to be DRBxxx-namewhere xxx is the 3-digit benchmark ID number.
- Added a few tests related to taskwait, taskgroup and ordered. They are contributed by the Rice University.
- Improved the test scripts to support C++ code, better report testing errors
- Miscellaneous fixes to typos
DataRaceBench v1.1.1
V1.1.1 2017-10-04
- refined source code comments to be consistent: using C-style /* ... */
- specified data race pairs in comments when applicable: using a format of var1@line_no:col_no vs. var2@line_no:col_no
DataRaceBench v1.1.0
- added 7 additional micro-benchmarks using omp flush, task depend, single, etc., making it 79 microbenchmarks in total.
- merged user-suggested fixes for private(j) to enforce only a single-pair dependence per test when possible (Thanks to jprotze ).
- refined the outofbounds benchmarks to only have valid wrap around memory accesses
- swapped file names of two benchmarks: antidep1-var-yes.c and antidep2-orig-yes.c
DataRaceBench v1.0.2
- fixed Ubuntu 16 linker issue for option --run
- added status message while evaluation is performed
- improved time measurement output precision in test script
DataRaceBench v1.0.1
Updated 2 microbenchmarks (added initialization for variables) and extended the evaluation script with an additional error log.
This release contains 72 C benchmarks and an evaluation scrpit for the tools Helgrind, Thread Sanitizer, Archer, and Intel Inspector.
DataRaceBench v1.0.0
Initial release with 72 C99 data race benchmarks (microbenchmarks) and a script for evaluating the tools Helgrind, Thread Sanitizer, Archer, and Intel Inspector.