Performance of a subroutine library on vector-processing machines

Daly, Chris; Croz, J. J. Du

doi:10.1016/0010-4655(85)90151-1

Cited by 9 publications

(6 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Approaches (2) and (3) should be considered as extensions of (l), not as alternatives. Implementors should not consider translating the model implementation into extended FORTRAN or assembly language without first considering whether the structure of the model implementation is well adapted to their machine.…”

Section: Notes On Implementationmentioning

confidence: 99%

Algorithm 656: an extended set of basic linear algebra subprograms: model implementation and test programs

Dongarra

Croz

Hammarling

et al. 1988

ACM Trans. Math. Softw.

146

View full text Add to dashboard Cite

This paper describes a model implementation and test software for the Level 2 Basic Linear Algebra Subprograms (Level 2 BLAS). Level 2 BLAS are targeted at matrix-vector operations with the aim of providing more efficient, but portable, implementations of algorithms on high-performance computers. The model implementation provides a portable set of FORTRAN 77 Level 2 BLAS for machines where specialized implementations do not exist or are not required. The test software aims to verify that specialized implementations meet the specification of Level 2 BLAS and that implementations are correctly installed. SCOPE OF THE ALGORITHMIn [7] we have defined the specification of a set of Basic Linear Subprograms for selected matrix-vector operations usually referred to as "Level 2 BLAS" or "Extended BLAS." They provide a standard framework to develop modular, portable, and efficient FORTRAN 77 code for many computational problems in J. J.

show abstract

Section: Notes On Implementationmentioning

confidence: 99%

Algorithm 656: an extended set of basic linear algebra subprograms: model implementation and test programs

Dongarra

Croz

Hammarling

et al. 1988

ACM Trans. Math. Softw.

146

View full text Add to dashboard Cite

show abstract

“…Then, in Section 6, we describe the tools that perform these transformations and how they are integrated into Toolpack [2]. In Section 7 we report the results of timing experiments in which we compare machine-transformed with human-transformed versions of the codes described in [4]. Finally, in Section 8, we suggest future directions for this work.…”

Section: Continuementioning

confidence: 99%

“…The seven special paradigms that the tools currently recognize were derived from the Fortran structures that occur in several generic linear algebra routines, for example, the routines of [4]. The tools also transform more general structures using general algorithms; they issue warnings when this occurs because data dependencies may render the code incorrect and, in any case, its performance is apt to be poor.…”

Section: A Typical Paradigmmentioning

confidence: 99%

“…Since ISTUD, ISTCD, and ISTSB are parse tree flattening tools, the parser, ISTYP, must be invoked after ISTUD to produce a parse tree/symbol table for input to ISTCD, and after ISTCD to produce a parse tree/symbol table for input to ISTSB. 4 Moreover, whenever ISTCD or ISTSB carries out a transformation, it sets the termination status to one of the flags termflag_0 or termflag_1 as explained in the comments. The termination status determines which further tool sequences are invoked.…”

Section: The Transformation Toolsmentioning

confidence: 99%

“…Fortran' programs that include matrix and vector calculations usually call highly optimized linear algebra routines from a library; see, for example, [7,4]. The nested DO loops in such linear algebra routines have properties that make the code amenable to the application of certain transformations, described in [5], aimed at causing a vectorizing Fortran compiler to produce code that is better adapted to vector architecture.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Transforming FORTRAN DO loops to improve performance on vector architectures

Cowell

Thompson²

1986

ACM Trans. Math. Softw.

View full text Add to dashboard Cite

The performance of programs executing on vector computers is significantly improved when the number of accesses to memory can be reduced. Unrolling Fortran DO loops, followed by substitutions and eliminations in the unrolled code, can reduce the number of loads and stores. In this paper we characterize the unrolling transformation and associated transformations of Fortran DO loops and describe a set of software tools to carry out these transformations. The tools use the machinery available in Toolpack and have been integrated into that environment. We describe the results of applying these tools to a collection of linear algebra subroutines.

show abstract

Basic linear algebra computations on the Sperry ISP

Croz¹,

Wasniewski

1988

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Performance of a subroutine library on vector-processing machines

Cited by 9 publications

References 2 publications

Algorithm 656: an extended set of basic linear algebra subprograms: model implementation and test programs

Algorithm 656: an extended set of basic linear algebra subprograms: model implementation and test programs

Transforming FORTRAN DO loops to improve performance on vector architectures

Basic linear algebra computations on the Sperry ISP

Contact Info

Product

Resources

About