Minimizing development and maintenance costs in supporting persistently optimized BLAS

293 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Quantized Convolutional Neural Networks for Mobile Devices (2015, methodology): “We compare the theoretical and realistic speed-up in Section 5.4, and discuss the effect of adopting the BLAS library for acceleration.”
Paleo: A Performance Model for Deep Neural Networks (2016, background): “For instance, achieving peak FLOPS is a difficult proposition, usually requiring customized libraries developed by organizations with intimate knowledge of the underlying hardware, e.g., Intel’s MKL (int, 2009), ATLAS (Whaley & Petitet, 2005), and cuDNN.”
Matrix eQTL: ultra fast eQTL analysis via large matrix operations (2011, methodology): “There are faster versions of BLAS available for R called ATLAS (32 bit only) by (Whaley and Petitet, 2005) and GOTO (64 bit only) by (Goto and Geijn, 2008) adapted for R by Ei-Ji Nakama [http://prs.ism.ac.jp/∼nakama/SurviveGotoBLAS2/binary/windows/x64/].”
Multicore bundle adjustment (2011, background): “An interesting direction for future work is a framework along the lines of FFTW [7] or ATLAS [14] for automatically tuning the parameters of the system either at compile time or at runtime.”
Fast R Functions for Robust Correlations and Hierarchical Clustering (2012, methodology): “On an 8-core system using a faster BLAS matrix multiplication (Whaley and Petitet 2005), the speedup compared to the standard R correlation calculation ranges from about 5 when the fraction of missing data is relatively large (here 0.”