Execution time of multiple-precision BLAS Level 1 operations on CPU and GPU platforms

This dataset contains the execution time of four BLAS Level 1 operations - ASUM, DOT, SCAL and AXPY - implemented using multiple-precision software for central processing units (CPUs) and CUDA compatible graphics processing units (GPUs). Each log file provided contains the results of three test runs, and also details about the experimental setup.

The following software packages are considered:
1. For CPUs:
• MPFR: A C library for multiple-precision floating-point computations with correct rounding (https://www.mpfr.org)
• ARPREC: An arbitrary precision package for Fortran and C++ (https://www.davidhbailey.com/dhbsoftware)
• MPDECIMAL: A package for correctly-rounded arbitrary precision decimal floating point (https://www.bytereef.org/mpdecimal)
• MPACK: Multiple-precision versions of BLAS and LAPACK (http://mplapack.sourceforge.net)
• XBLAS: A reference implementation of extended and mixed precision BLAS routines (https://www.netlib.org/xblas)
2. For GPUs:
• GARPREC: A port of the ARPREC package for CUDA-enabled GPUs (https://code.google.com/archive/p/gpuprec/downloads)
• CAMPARY: A multiple-precision library that uses floating-point expansions to represent extended precision numbers (http://homepages.laas.fr/mmjoldes/campary)
• CUMP: A library for arbitrary precision arithmetic on CUDA based on the GNU MP Bignum Library (https://github.com/skystar0227/CUMP)
• MPRES-BLAS: Multiple-precision GPU accelerated BLAS functions based on Residue number system (https://github.com/kisupov/mpres-blas)

The complete source code for the benchmarks can be found at https://github.com/kisupov/mpres-blas