Solving the Table Maker's Dilemma on Current SIMD Architectures

Christophe Avenel 1, 2 Pierre Fortin 2 Mourad Gouicem 2, 3 Zaidi Samia 2
2 PEQUAN - Performance et Qualité des Algorithmes Numériques
LIP6 - Laboratoire d'Informatique de Paris 6
3 ECO - Exact Computing
LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier
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.
Document type :
Journal articles
Complete list of metadatas

https://hal.sorbonne-universite.fr/hal-01356850
Contributor : Pierre Fortin <>
Submitted on : Friday, August 26, 2016 - 4:43:44 PM
Last modification on : Tuesday, November 5, 2019 - 4:18:03 PM

Links full text

Identifiers

Citation

Christophe Avenel, Pierre Fortin, Mourad Gouicem, Zaidi Samia. Solving the Table Maker's Dilemma on Current SIMD Architectures. Scalable Computing : Practice and Experience, West University of Timisoara, 2016, 17 (3), ⟨10.12694/scpe.v17i3.1183⟩. ⟨hal-01356850⟩

Share

Metrics

Record views

559