DocumentCode :
3228653
Title :
Adaptive Strassen and ATLAS´s DGEMM: a fast square-matrix multiply for modern high-performance systems
Author :
D´Alberto, Paolo ; Nicolau, Alexandru
Author_Institution :
Dept. of Electr. & Comput. Eng., Carnegie Mellon Univ., Pittsburgh, PA
fYear :
2005
fDate :
1-1 July 2005
Lastpage :
52
Abstract :
Strassen´s algorithm has practical performance benefits for architectures with simple memory hierarchies, because it trades computationally expensive matrix multiplications (MM) with cheaper matrix additions (MA). However, it presents no advantages for high-performance architectures with deep memory hierarchies, because MAs exploit limited data reuse. We present an easy-to-use adaptive algorithm combining Strassen´s recursion and high-tuned version of ATLAS MM. In fact, we introduce a last step in the ATLAS-installation process that determines whether Strassen´s may achieve any speedup. We present a recursive algorithm achieving up to 30% speed-up versus ATLAS alone. We show experimental results for 14 different systems
Keywords :
adaptive codes; mathematics computing; matrix multiplication; memory architecture; ATLAS DGEMM; adaptive Strassen algorithm; data reuse; deep memory hierarchy; high-performance systems; matrix additions; matrix multiplications; recursive algorithm; Adaptive coding; Computer architecture; Computer science; Equations; High performance computing; IEEE members; Kernel; Memory architecture; Personal digital assistants; Software packages;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
High-Performance Computing in Asia-Pacific Region, 2005. Proceedings. Eighth International Conference on
Conference_Location :
Beijing
Print_ISBN :
0-7695-2486-9
Type :
conf
DOI :
10.1109/HPCASIA.2005.18
Filename :
1592249
Link To Document :
بازگشت