Automated empirical optimizations of software and the ATLAS project

By R. C. Whaley, Antoine Petitet, Jack J. Dongarra. Parallel Computing, 2001

1,072 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/8dc2184214ee39b31e2c0d623842b66c0141984b

How researchers see this paper

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

In Computer Networks and Communications

In Hardware and Architecture

Citations in the corpus

Automated Solution of Differential Equations by the Finite Element Method: The FEniCS Book (2012, methodology): “There exist a number of optimized implementations of BLAS, including hand-optimized vendor implementations, empirically and automatically tuned libraries (Whaley et al., 2001) and formal methods for automatic derivation of algorithms (Bientinesi et al.”
A column pre-ordering strategy for the unsymmetric-pattern multifrontal method (2004, methodology): “This BLAS increases the performance of UMFPACK4 by about 50% for large sparse matrices on this computer, as compared with the ATLAS 3.4.1 BLAS [Whaley et al. 2001].”
Model-driven Level 3 BLAS Performance Optimization on Loongson 3A Processor (2012, methodology): “Additonally , OpenBLAS had higher speedups on the functions of single precision real and complex, which showed that ATLAS couldn't utilize single precision SIMD instructions on Loongson 3A. V. RELATED WORK Goto [3] analyzed many matrix multiplication algorithms and proposed an optimized method.”
Finite Discrete Gabor Analysis (2007, methodology): “Fast implementations exists for both algorithms, see [108, 39].”
Scaling Distributed Machine Learning with the Parameter Server (2014, positive): “Beyond convenience, this interface design leads to efficient code by leveraging CPUefficient multithreaded self-tuning linear algebra libraries such as BLAS [16], LAPACK [3], and ATLAS [49] .”
Automatic performance tuning of sparse matrix kernels (2003, methodology): “In particular, ATLAS improves on PHiPAC ideas, extending their applicability to the entire BLAS standard [325, 324].”
Nodal discontinuous Galerkin methods on graphics processors (2009, neutral): “The calculations used ATLAS 3.8.2 [ 25] for element-local operations if such use proved advantageous.”
FreeContact: fast and free software for protein contact prediction from residue co-evolution (2014, methodology): “ATLAS provides a highly efficient machine-specific implementation of BLAS [27] and LAPACK [22]; it automatically adapts itself during the build process to the host architecture in order to optimize performance.”
Sparsity: Optimization Framework for Sparse Matrix Kernels (2004, methodology): “Search has been effectively used in automatic optimization frameworks for dense matrix kernels (Bilmes et al., 1997; Whaley et al., 2000).”
Automatic Differentiation of Algorithms: From Simulation to Optimization (2002, methodology): “Approaches such as Atlas [501] can potentially be applied.”
Lightweight modular staging: a pragmatic approach to runtime code generation and compiled DSLs (2010, neutral): “Sym(25) Plus(Sym(11),Sym(14)) Sym(8) Plus(Sym(0),Sym(4))”
A Class of Parallel Tiled Linear Algebra Algorithms for Multicore Architectures (2007, background): “…linear algebra libraries due to their scalability limits in the context of multicore computing, Section 3 describes fine granularity, tiled algorithms for the Cholesky, LU and QR factorizations and presents a programming model for their asynchronous and dynamic execution; performance results for…”