Level 3 BLAS for distributed memory concurrent computers

Conference · Wed Jan 01 00:00:00 EST 1992

OSTI ID:7133964

Choi, Jaeyoung; Walker, D W ^[1]; Dongarra, J J ^[2]

---

1. Oak Ridge National Lab., TN (United States)
2. Oak Ridge National Lab., TN (United States) Tennessee Univ., Knoxville, TN (United States). Dept. of Computer Science

This paper discusses issues in developing a version of the Level 3 BLAS for distributed memory concurrent computers. The Level 3 BLAS are particularly important when using machines with hierarchical memory as they maximize reuse of data in the upper levels of the memory hierarchy. In implementing the distributed Level 3 BLAS a block scattered decomposition is employed. Details of the parallel implementation of DGEMM, for performing matrix-matrix multiplication, and DTRSM, for solving triangular systems, are given, and results are presented for runs on the Intel Touchstone Delta computer.

• Research Organization: Oak Ridge National Lab., TN (United States)
• Sponsoring Organization: USDOE; USDOD; USDOE, Washington, DC (United States); Department of Defense, Washington, DC (United States)
• DOE Contract Number: AC05-84OR21400
• OSTI ID: 7133964
• Report Number(s): CONF-9209274-2; ON: DE93003613
• Resource Relation: Conference: CNRS-NSF collaboration workshop on environments and tools for parallel scientific computing, St. Hilaire du Touvet (France), 7-8 Sep 1992
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
DISTRIBUTED DATA PROCESSING
MEMORY MANAGEMENT
L CODES
ALGEBRA
ALGORITHMS
ARRAY PROCESSORS
FACTORIZATION
ITERATIVE METHODS
MATRICES
MEMORY DEVICES
PARALLEL PROCESSING
PERFORMANCE
CALCULATION METHODS
COMPUTER CODES
DATA PROCESSING
MATHEMATICAL LOGIC
MATHEMATICS
PROCESSING
PROGRAMMING
990200* - Mathematics & Computers