Hq. Ding et Rd. Ferraro, AN 18 GFLOPS PARALLEL CLIMATE DATA ASSIMILATION PSAS PACKAGE, Computers & mathematics with applications, 35(7), 1998, pp. 55-63
We have designed and implemented a set of highly efficient and highly
scalable algorithms for an unstructured computational package, the PSA
S data simulation package, as demonstrated by detailed performance ana
lysis of systematic runs up to 512 nodes of an Intel Paragon. The prec
onditioned Conjugate Gradient solver achieves a sustained 18 Gflops pe
rformance. Consequently, we achieve an unprecedented 100-fold reductio
n in time to solution on the Intel Paragon over a single head of a Gra
y C90. This not only exceeds the daily performance requirement of the
Data Assimilation Office at NASA's Goddard Space Flight Center, but al
so makes it possible to explore much larger and challenging data assim
ilation problems which are unthinkable on a traditional computer platf
orm such as the Gray C90.