Improving scalability with loop transformations and message aggregation in parallel object-oriented frameworks for scientific computing
Application codes reliably achieve performance far less than the advertised capabilities of existing architectures, and this problem is worsening with increasingly-parallel machines. For large-scale numerical applications, stencil operations often impose the great part of the computational cost, and the primary sources of inefficiency are the costs of message passing and poor cache utilization. This paper proposes and demonstrates optimizations for stencil and stencil-like computations for both serial and parallel environments that ameliorate these sources of inefficiency. Achieving scalability, they believe, requires both algorithm design and compile-time support. The optimizations they present are automatable because the stencil-like computations are implemented at a high level of abstraction using object-oriented parallel array class libraries. These optimizations, which are beyond the capabilities of today compilers, may be performed automatically by a preprocessor such as the one they are currently developing.
- Research Organization:
- Los Alamos National Lab., Computing, Information, and Communications Div., NM (United States)
- Sponsoring Organization:
- USDOE Assistant Secretary for Management and Administration, Washington, DC (United States)
- DOE Contract Number:
- W-7405-ENG-36
- OSTI ID:
- 677091
- Report Number(s):
- LA-UR-98-1175; CONF-980464-; ON: DE98006294; TRN: AHC29821%%226
- Resource Relation:
- Conference: 4. USENIX conference on object-oriented technologies and systems (COOTS), Santa Fe, NM (United States), 27-30 Apr 1998; Other Information: PBD: [1998]
- Country of Publication:
- United States
- Language:
- English
Similar Records
Improving Cache Utilization of Linear Relaxation Methods: Theory and Practice
The Specification of Source-to-source Transformations for the Compile-time Optimization of Parallel Object-oriented Scientific Applications