지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학술대회자료
- 저자정보
- 발행연도
- 2011.5
- 수록면
- 277 - 282 (6page)
이용수
초록· 키워드
오류제보하기
This paper reports how to improve the parallel multifrontal solution method that is used widely in finite element analysis. To solve the large scale problems with the limited in-core memory, many direct solvers employ an out-of-core memory. The conventional multifrontal solver was used only with the out-of-core memory. To improve its performance, factored data blocks should be cached in physical memory rather than the secondary storage. To optimize the solver performance, novel ways of managing the in-core memory and caching data is proposed; frontal stack compaction, inverse stack data structure, and selective data caching and recovery were developed. The packing frontal stack and inverse stack removes the additional physical memory for intermediate front data to increase available in-core memory. Selective data caching is an improved technique for increasing the memory space to load the factored data from the out-of-core to physical memory during the substitution phase. The enhanced solver is more efficient than other direct solvers, which can handle the out-of-core storage. The parallel efficiency of the improved method is much higher than that of the original method.
목차
ABSTRACT
OPTIMIZING PARALLEL MULTIFRONTAL SOLVER
PERFORMANCE OF IMPROVED SOLVER
CONCLUSION
REFERENCES
참고문헌 (0)
참고문헌 신청이 논문의 저자 정보
최근 본 자료
댓글(0)
0
UCI(KEPA) : I410-ECN-0101-2013-410-000695520