Solving the Table Maker's Dilemma on Current SIMD Architectures
Journal Title: Scalable Computing: Practice and Experience - Year 2016, Vol 17, Issue 3
Abstract
Correctly-rounded implementations of some elementary functions are recommended by the IEEE 754-2008 standard, which aims at ensuring portable and predictable floating-point computations. Such implementations require the solving of the Table Maker's Dilemma which implies a huge amount of computation time. These computations are embarrassingly and massively parallel, but present control flow divergence which limits performance at the SIMD parallelism level, whose share in the overall performance of current and forthcoming HPC architectures is increasing. In this paper, we show that efficiently solving the Table Maker's Dilemma on various multi-core and many-core SIMD architectures (CPUs, GPUs, Intel Xeon Phi) requires to jointly handle divergence at the algorithmic, programming and hardware levels in order to scale with the number of SIMD lanes. Depending on the architecture, the performance gains can reach 10.5x over divergent code, or be constrained by different limits that we detail.
Authors and Affiliations
CHRISTOPHE AVENEL, PIERRE FORTIN, MOURAD GOUICEM, SAMIA ZAIDI
Keywords
Related Articles
- EP ID EP204682
- DOI 10.12694/scpe.v17i3.1183
- Views 25
- Downloads 0
Cloud based Dynamic Course Selection Framework using Network Graphs with Term Difficulty Estimation
The system developed in this paper uses a cloud based technology to implement and design a software as a service (SAAS) application for adaptive course selection and term difficulty estimation for a networked curriculum....
An Analysis for Parallel Wind Simulation Speedup using OpenFOAM
An analysis of speedup for parallel execution of OpenFOAM software for wind simulation over rugged terrain is presented in the paper. Runtime speedup is analyzed using small and medium resolution DEM models for icoFoam a...
Semiempirical Atom-centered Density Matrix Propagation Approach to Temperature-dependent Vibrational Spectroscopy of Irinotecan
In the present study, a molecular dynamics study of irinotecan molecule with the atom-centered density matrix propagation scheme was carried out at AM1 semiempirical level of theory, at series of different temperatures,...
Analysis and Verification of XACML Policies in a Medical Cloud Environment
The connectivity of devices, machines and people via Cloud infrastructure can support collaborations among doctors and specialists from different medical organisations. Such collaborations may lead to data sharing and jo...
Impact of Single Parameter Changes on Ceph Cloud Storage Performance
In a general purpose cloud system efficiencies are yet to be had from supporting diverse applications and their requirements within a storage system used for a private cloud. Supporting such diverse requirements poses a...