English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Basic Linear Algebra Subprograms (BLAS) is a core library in scientificcomputing and machine learning. This paper presents FT-BLAS, a newimplementation of BLAS routines that not only tolerates soft errors on the fly,but also provides comparable performance to modern state-of-the-art BLASlibraries on widely-used processors such as Intel Skylake and Cascade Lake. Toaccommodate the features of BLAS, which contains both memory-bound andcomputing-bound routines, we propose a hybrid strategy to incorporate faulttolerance into our brand-new BLAS implementation: duplicating computinginstructions for memory-bound Level-1 and Level-2 BLAS routines andincorporating an Algorithm-Based Fault Tolerance mechanism for computing-boundLevel-3 BLAS routines. Our high performance and low overhead are obtained fromdelicate assembly-level optimization and a kernel-fusion approach to thecomputing kernels. Experimental results demonstrate that FT-BLAS offers highreliability and high performance -- faster than Intel MKL, OpenBLAS, and BLISby up to 3.50%, 22.14% and 21.70%, respectively, for routines spanning allthree levels of BLAS we benchmarked, even under hundreds of errors injected perminute.

FT-BLAS: A High Performance BLAS Implementation With Online Fault Tolerance