BlackScholes benchmark

Benchmark computing Black Scholes formula using different technologies.

Prerequisites

• icc, if compiling native benchmarks. Intel Distribution for Python* 2019 Gold benchmarks used icc 17.0.1.
• mkl, if compiling native benchmarks with MKL.

Setup

Linux & Mac

• Run . activate-conda.sh to install miniconda on Linux and Mac.
• Run make to build and run native benchmarks.
  • Run make mkl to build and run MKL version.
  • Run make nomkl to build and run non-MKL version.
  • Run make black_scholes_mkl to only build MKL version.
  • Run make black_scholes to only build non-MKL version.

Windows

• Download & install Miniconda3 and MSYS2.
• Run bash from MSYS2 and activate miniconda environment.
• Run ./install-conda-envs.sh to install Python environments.

Usage

Native benchmarks

• Non-MKL version: Run the compiled binary ./black_scholes.
• MKL version: Run the compiled binary ./black_scholes_mkl.

Python benchmarks

usage: {bs_erf_*.py|run.sh} [-h]
                       [--steps STEPS] [--step STEP] [--chunk CHUNK]
                       [--size SIZE] [--repeat REPEAT] [--dask DASK]
                       [--text TEXT]


optional arguments:
  -h, --help       show this help message and exit
  --steps STEPS    Number of steps
  --step STEP      Factor for each step
  --chunk CHUNK    Chunk size for Dask
  --size SIZE      Initial data size
  --repeat REPEAT  Iterations inside measured region
  --dask DASK      Dask scheduler: sq, mt, mp
  --text TEXT      Print with each result

See also

"Accelerating Scientific Python with Intel Optimizations" by Oleksandr Pavlyk, Denis Nagorny, Andres Guzman-Ballen, Anton Malakhov, Hai Liu, Ehsan Totoni, Todd A. Anderson, Sergey Maidanov. Proceedings of the 16th Python in Science Conference (SciPy 2017), July 10 - July 16, Austin, Texas