The LINPACK benchmark on the AP1000
130. R. P. Brent,
The LINPACK benchmark on the AP1000,
Proceedings of Frontiers '92
(McLean, Virginia, October 1992), IEEE Press, 1992, 128-135.
Abstract:
dvi (2K),
pdf (62K),
ps (26K).
Paper:
dvi (26K),
pdf (179K),
ps (72K).
Transparencies (1 per page):
pdf (51K),
ps (52K).
Transparencies (4 per page):
pdf (68K),
ps (47K),
Abstract
We describe an implementation of the Linpack Benchmark
on the Fujitsu AP 1000.
Design considerations include communication primitives, data distribution,
use of blocking to reduce memory references,
and effective use of the cache.
Some details of our blocking strategy appear to be new.
The results show that the Linpack Benchmark can be implemented
efficiently on the AP 1000,
and it is possible to attain about 80 percent
of the theoretical peak performance for large problems.
Comments
For related work see [136].
Go to next publication
Return to Richard Brent's index page