Auto-tuning techniques for linear algebra routines on hybrid platforms

Bernabé, Gregorio; Cuenca, Javier; García, Luis-Pedro; Giménez, Domingo

doi:10.1016/j.jocs.2015.04.032

Cited by 7 publications

(8 citation statements)

References 24 publications

(23 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The optimization of linear algebra computations in multicore + GPU heterogeneous systems is analyzed in Bernabé et al For us to balance the work assigned to each computational component and so improve performance, an adequate partition of the matrices is decided. The matrices are split statically (the dynamic selection of the split sizes is discarded because it would produce a high overhead when the linear algebra routine is used in codes of higher level).…”

Section: Autotuning Techniquesmentioning

confidence: 99%

“…Two autotuning techniques are traditionally applied to balance matrix distribution between the CPU and the coprocessors. The methodology, called empirical modeling, is based on a theoretical model of the execution time of the routine . With the more complex and heterogeneous systems we are working with now, it is much more difficult to obtain a simple, effective model.…”

Section: Autotuning Techniquesmentioning

confidence: 99%

“…In the empirical modeling technique, a simple (but effective) model of the execution time is obtained, and it is used to decide the values for some algorithmic parameters to minimize the total execution time . The work distribution between the different computing units (the GPUs and the CPU) is decided.…”

Section: Work In Processmentioning

confidence: 99%

“…This paper considers the adaptation to clusters of multicore CPUs plus multiGPU or multiMIC of techniques previously applied for autotuning basic linear algebra routines for simpler computational systems: multicore, multicore + GPU, and multicore + multicoprocessors …”

Section: Introductionmentioning

confidence: 99%

“…applied for autotuning basic linear algebra routines for simpler computational systems: multicore, 2, 16 multicore + GPU, 17 and multicore + multicoprocessors. 18 The basic computational components (multicore, GPU, and MIC) have different architectures, and their programming paradigms are also different.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Guided installation of basic linear algebra routines in a cluster with manycore components

Cuenca

García

Giménez

et al. 2017

Concurrency and Computation

View full text Add to dashboard Cite

Summary Computational systems are nowadays composed of basic computational components that share multiprocessors and coprocessors of different types, typically several graphics processing units (GPUs) or many integrated cores (MICs), and those computational components are combined in heterogeneous clusters of nodes with different characteristics, including coprocessors of different types, with varying numbers of nodes at different speeds. The software previously developed and optimized for simpler system needs to be redesigned and reoptimized for these new, more complex systems. The adaptation to hybrid multicore + multiGPU and multicore + multiMIC of autotuning techniques for basic linear algebra routines is analyzed. The matrix‐matrix multiplication kernel, which is optimized for different computational system components through guided experimentation, is studied. The routine is installed for each node in the cluster, and the information generated from individual installations may be used for a hierarchical installation in a cluster. The basic matrix‐matrix multiplication may, in turn, be used inside higher level routines, which delegate their efficient execution to the optimization of the lower level routine. Experimental results are satisfactory in different multicore + multiGPU and multicore + multiMIC systems. So the guided search of execution configurations for satisfactory execution times proves to be a useful tool for heterogeneous systems, where the complexity of the system means a correct use of highly efficient routines and libraries is difficult.

show abstract

Section: Autotuning Techniquesmentioning

confidence: 99%

Section: Autotuning Techniquesmentioning

confidence: 99%

Section: Work In Processmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations