rscimark: Wrapper for the SciMark 2.0 benchmark.

In berndbischl/rscimark: SciMark 2.0 Benchmark for Scientific and Numerical Computing

Description

This function is a simple wrapper around the ANSI C version of the SciMark 2.0 benchmark which is a benchmark for numerical and scientific computing. Concicely performance measurements for the computational kernels Fast Fourier Transformation (FFT), Gauss-Seidel relaxation, Sparse matrix-multiply, Monte Carlo integration and dense LU factorization are computed.

In order to isolate effects of memory hierarchy the problem sizes, e.g., the size of the matrix fpr the dense LU matrix factorization, are pretty small. However, addressing the performance of the memory subsystem is possible by setting the large argument to TRUE.

Usage

rscimark(large = FALSE, minimum.time = 2)

Arguments

large	[logical(1)] Run large version of benchmark? Default is FALSE.
minimum.time	[numeric(1)] Minimum time to run each of the benchmarks, in seconds. Default is 2.

Value

[numeric] Named vector of time measurements with the following components:

Composite

Mean value of the remaining components.

FFT

Performance of the Fast Fourier Transformation (FFT).

SOR

Performance of the Jacobi Successive Over-relaxation (SOR).

MC

Performance of a Monte Carlo integration.

SMM

Performance of a spare matrix multiplication.

LU

Performance of a dense LU matrix factorization.

berndbischl/rscimark documentation built on May 12, 2019, 7:24 p.m.