Orio is an open-source extensible framework for the definition of domain-specific languages and generation of optimized code for multiple architecture targets, including support for empirical autotuning of the generated code.

For more detailed documentation, refer to the Orio website, https://brnorris03.github.io/Orio/.

Orio is implemented in Python 3. Some search methods (e.g., Mlsearch) require the pandas and sklearn packages. The simplest way to install Orio is to run

pip install orio

This will install the most recent release of Orio and the packages it uses in your current Python environment. You can also add the --user option if the above command requires superuser privileges.

If you want to build Orio from a git clone, you can use pip install -e . in the top-level directory. Note that you can simply run orcc (and the other top-level command-line scripts) directly from the git clone without installing anything. Testing is provided through pydev, to run all available tests, runpytest or pytest -v in the top-level Orio directory.

To test whether Orio has been properly installed in your system, try to execute orcc command as given below as an example. If you used the --user option, you can find orcc under your home directory, e.g., in ~/.local/bin on Unix.

  $ orcc --help

  description: compile shell for Orio

  usage: orcc [options] <ifile>
    <ifile>   input file containing the annotated code

  options:
    -h, --help                     display this message
    -o <file>, --output=<file>     place the output to <file>
    -v, --verbose                  verbosely show details of the results of the running program

After making sure that the orcc executable is in your path, you can try some of the examples included in the testsuite subdirectory, e.g.:

 $ cd examples
 $ orcc -v axpy5.c

The same directory contains two more examples of Orio input -- one with a separate tuning specification file (orcc -v -s axpy5.spec axpy5-nospec.c) and another with two transformations specified using a Composite annotation (orcc -v axpy5a.c). To see a list of options, orcc -h. To keep all intermediate code versions, use the -k option. You can also enable various levels of debugging output by using the -d <NUM> option, setting <NUM> to an integer between 1 and 6, e.g., for the most verbose output -d 6. This is the recommended setting when submitting sample output for bug reports.

To use machine learning-based search (Mlsearch), install numpy, pandas, and scikit-learn modules. Alternatively, if using conda, simply run conda install pandas to obtain all prerequisites if needed.

If Orio reports problems building the code, adjust the compiler settings in the tuning spec included in the axpy5.c example.

Please report bugs at https://github.com/brnorris03/Orio/issues and include complete examples that can be used to reproduce the errors. Send all other questions and comments to: Boyana Norris, brnorris03@gmail.com .

Principal Authors:

• Boyana Norris, University of Oregon
• Albert Hartono, Intel
• Azamat Mametjanov, Argonne National Laboratory
• Prasanna Balaprakash, Argonne National Laboratory
• Nick Chaimov, University of Oregon

• B. Norris, A. Hartono, and W. Gropp. Annotations for productivity and performance portability. Petascale Computing: Algorithms and Applications, pp. 443–462. Chapman & Hall / CRC Press, Taylor and Francis Group, Computational Science, 2007, http://www.mcs.anl.gov/uploads/cels/ papers/P1392.pdf.

• Azamat Mametjanov, Daniel Lowell, Ching-Chen Ma, and Boyana Norris. 2012. Autotuning Stencil-Based Computations on GPUs. In Proceedings of the 2012 IEEE International Conference on Cluster Computing (CLUSTER '12). IEEE Computer Society, USA, 266–274. DOI:https://doi.org/10.1109/CLUSTER.2012.46

• Prasanna Balaprakash, Stefan M. Wild, Boyana Norris, SPAPT: Search Problems in Automatic Performance Tuning, Procedia Computer Science, Volume 9, 2012, Pages 1959-1968, ISSN 1877-0509, https://doi.org/10.1016/j.procs.2012.04.214.

• N. Chaimov, B. Norris, and A. Malony. Toward multi-target autotuning for accelerators. Proceedings of the 20th IEEE International Conference on Parallel and Distributed Systems, December 16-19, 2014, Hsinchu, Taiwan, 2014, http://ix.cs.uoregon.edu/~norris/icpads14.pdf.

• Lim, Robert V., B. Norris and A. Malony. “Autotuning GPU Kernels via Static and Predictive Analysis.” 2017 46th International Conference on Parallel Processing (ICPP) (2017): 523-532. https://arxiv.org/pdf/1701.08547

• Orio's old webpage: http://trac.mcs.anl.gov/projects/performance/wiki/Orio

• Some hidden links: http://trac.mcs.anl.gov/projects/performance/wiki/AnnPerformance (Old results for tuning Pluto)