ExBLAS: Reproducible and Accurate BLAS Library - Archive ouverte HAL Access content directly
Conference Papers Year : 2015

ExBLAS: Reproducible and Accurate BLAS Library


Due to non-associativity of floating-point operations and dynamic scheduling on parallel architectures, getting a bit-wise reproducible floating-point result for multiple executions of the same code on different or even similar parallel architectures is challenging. We address the problem of reproducibility in the context of fundamental linear algebra operations – like the ones included in the BLAS library – and propose algorithms that yield both reproducible and accurate results (correct rounding, except for triangular solver). We present implementations of these algorithms for the BLAS routines along with the performance results in parallel environments such as Intel desktop and server CPUs, Intel Xeon Phi, and both NVIDIA and AMD GPUs.
Fichier principal
Vignette du fichier
exblas.pdf (222.78 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01202396 , version 1 (20-09-2015)
hal-01202396 , version 2 (09-10-2015)
hal-01202396 , version 3 (21-12-2015)


  • HAL Id : hal-01202396 , version 3


Roman Iakymchuk, Caroline Collange, David Defour, Stef Graillat. ExBLAS: Reproducible and Accurate BLAS Library. NRE: Numerical Reproducibility at Exascale, Nov 2015, Austin, TX, United States. ⟨hal-01202396v3⟩
742 View
694 Download


Gmail Facebook X LinkedIn More