High-performance implementation of the level-3 BLAS

By Kazushige Goto, R. V. D. Geijn. TOMS, 2008

372 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/62fdc203262e62af885c757c2549444d5aa2c620

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

1. Yunquan Zhang
2. Qian Wang
3. Xianyi Zhang
4. Qing Yi
5. Christos Boutsidis
6. Petros Drineas
7. Anand Madhavan
8. Rajat Raina

In Hardware and Architecture

1. Yunquan Zhang
2. Qian Wang
3. Xianyi Zhang
4. Qing Yi
5. Anand Madhavan
6. Rajat Raina
7. Andrew Y. Ng
8. R. V. D. Geijn

Citations in the corpus

• Caffe con Troll: Shallow Ideas to Speed Up Deep Learning (2015, background): “Although our contributions in this paper leverage decades of work in high-performance computing (specifically, the advancements in optimizing matrix multiplications [7,14]), we omit discussion of this due to space constraints.”
• AUGEM: Automatically generate high performance Dense Linear Algebra kernels on x86 CPUs (2013, methodology): “Here the TRSM computation has been transformed into two steps: 1) B1 = L −1 11 ∗ B1; 2) B2 = B2 − L21 ∗ B1 [13].”
• Model-driven Level 3 BLAS Performance Optimization on Loongson 3A Processor (2012, methodology): “Because the general matrix multiplication (GEMM) operation is representative of every level 3 BLAS function [6], we focus on optimizing GEMM performance with a single thread and multiple threads on the Loongson 3A CPU.”
• Anatomy of high-performance matrix multiplication (2008, methodology): “How the techniques described in this paper impact the implementation of level-3 BLAS is discussed in [Goto and van de Geijn 2006].”
• A Class of Parallel Tiled Linear Algebra Algorithms for Multicore Architectures (2007, background): “…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 this algorithm are…”
• Large-scale deep unsupervised learning using graphics processors (2009, methodology): “The CPU-based method was implemented using two highly optimized multithreaded linear algebra packages: ATLAS BLAS (Whaley et al., 2001) and Goto BLAS ( Goto & Van De Geijn, 2008 ).”
• Bundle Adjustment in the Large (2010, methodology): “We use GotoBLAS2 [24] for dense linear algebra and CHOLMOD [17] for sparse Cholesky factorization.”
• A Randomized Algorithm for Approximating the Log Determinant of a Symmetric Positive Definite Matrix (2015, positive): “For experiments with Elemental, we used OpenBlas, which is an extension of GotoBlas [10], for its parallel prowess; Eigen provides built-in BLAS and LAPACK.”