PolyBench/Python v0.0

Copyright (c) 2020 Universidade da Coruña.

Contact:
    Miguel Ángel Abella González <[email protected]>

PolyBench/Python is the reimplementation of PolyBench in the Python
programming language. It is a benchmark suite of 30 numerical computations
with static control flow, extracted from operations in various application
domains (linear algebra computations, image processing, physics simulation,
dynamic programming, statistics, etc.).


System requirements:
------------------------------------------------------------------------------
Runtime:
    - Linux operating system running on x86 or AMD64 (the later recommended)
    - Python 3.6 or newer (CPython 3.8 and PyPy 7.3.1 (Python 3.6.9) tested)

Dependencies:
    - Python development packages for the selected interpreter(s)
    - GNU C compiler
    - NASM

Optional:
    - Python virtualenv


Installation
------------------------------------------------------------------------------
This section asumes a Linux distribution. Other Unix-like operating systems
may work up to some extent, but PolyBench/Python relies on some Linux-specific
system calls for some of its options which probably will not be available in
other platforms.

Installing system dependencies using package managers

    On Debian and derivatives (Ubuntu, etc.)
        $ sudo apt install python3 python3-dev
        $ sudo apt install pypy3 pypy3-dev
        $ sudo apt install libpapi-dev
        $ sudo apt install gcc nasm
        $ sudo apt install python3-virtualenv

    On Red-Hat based systems (Fedora, CentOS, etc.)
        $ sudo dnf install python3 python3-devel
        $ sudo dnf install pypy3 pypy3-devel
        $ sudo dnf install papi-devel
        $ sudo dnf install gcc nasm
        $ sudo dnf install python3-virtualenv


Installing PolyBench/Python dependencies
    At the root level of PolyBench/Python (where this README file is):
        $ pip install -r requirements.txt --user

    This will install all required packages for the active user. It may take
    some time for compiling dependencies.

    Alternatively, it is recommended to issue the previous command on a
    Python virtual environment. At the root level of PolyBench/Python issue
    the following set of commands for creating a virtual environment for a
    given interpreter...
        For CPython3:
            $ virtualenv -p `which python3` venv-cpython
        For PyPy3:
            $ virtualenv -p `which pypy3` venv-pypy
    Once the virtual environment is created, activate it and issue the
    instalation commands on it. For instance, using the previously created
    virtual environment for PyPy3:
        $ source venv-pypy/bin/activate
        (venv-pypy) $ pip install -r requirements.txt

    Note: once inside a virtual environment, the command "python" is aliased
        to the selected interpreted of the virtual environment.


Usage:
------------------------------------------------------------------------------
There are two main tools in PolyBench/Python, both implemented in Python. Both
are command line tools accepting the help option (both -h and --help) which
prints some useful usage information and available options.

create-benchmark.py
    Allows to create a new benchmark given a benchmark name and a category
    where to locate it. For instance, for creating the "correlation" benchmark
    on the "datamining" category:
        $ python create-benchmark.py -C datamining/correlation -N Correlation

run-benchmark.py
    Allows to run one or all benchmarks by issuing a single command, accepting
    a set of options which may alter the execution behavior.

    Some usage examples of this tool:
    - List existing benchmarks:
        $ python run-benchmark.py

    - Running an existing benchmark:
        $ python run-benchmark.py benchmarks/datamining/correlation/correlation.py

    - Running an existing benchmark and printing the execution time:
        $ python run-benchmark.py benchmarks/datamining/correlation/correlation.py --polybench-options POLYBENCH_TIME

    - Verifying a benchmark's implementation against a compatible output file
      named "bench.out":
        $ python run-benchmark.py benchmarks/datamining/correlation/correlation.py --verify-file bench.out

    Available PolyBench options (--polybench-options):
        POLYBENCH_TIME: (default off) prints the time elapsed during the
            execution of the kernel's code in seconds.
        POLYBENCH_DUMP_ARRAYS: (default false) outputs the returning arrays of
            the benchmark into stderr.
        POLYBENCH_PADDING_FACTOR: (default 0) adds N elements at the end of
            every array's dimension.
        POLYBENCH_PAPI: (default off) enables PAPI counters. This will not
            work when POLYBENCH_TIME is enabled
        POLYBENCH_PAPI_VERBOSE: (default false) print the PAPI counter name
            next to its value.
        POLYBENCH_CACHE_SIZE_KB: (default 32770) the size, in KiloBytes, of
            the data structure used for flushing the cache.
        POLYBENCH_NO_FLUSH_CACHE: (default flush) disable cache flushing.
        POLYBENCH_CYCLE_ACCURATE_TIMER: (default false) use the processor's
            timestamp counter (TSC) on compatible systems.
        POLYBENCH_LINUX_FIFO_SCHEDULER: (default false) use the FIFO scheduler
            for this process. This requires superuser privilege.

    Examples using multiple polybench options:
    - Printing verbose PAPI counters:
        $ python run-benchmark.py benchmarks/datamining/correlation/correlation.py --polybench-options POLYBENCH_PAPI,POLYBENCH_PAPI_VERBOSE
    - Print the execution time and use a larger cache flush size:
        $ python run-benchmark.py benchmarks/datamining/correlation/correlation.py --polybench-options POLYBENCH_TIME,POLYBENCH_CACHE_SIZE_KB=65538