The
future of high-performance computing relies on the efficient
and scalable use of clusters with symmetric multiprocessor
(SMP) nodes and low-latency, high-bandwidth interconnection
networks. Current examples of such platforms include
Sun Ultra HPC machines, Compaq AlphaServers with Quadrics
switches, SGI Origins, and the IBM SP system with SMP
nodes. Moreover, the future of NASA mission-critical
computing for computational aerosciences relies on the
success of computational clusters (e.g., SMP Linux clusters
at Goddard Space Flight Center, and large SGI Origin
arrays at Ames Research Center). Hardware benchmark results
reveal awesome performance rates for each component;
however, few applications on SMP clusters ever reach
a fraction of these peak speeds. While methodologies
for symmetric multiprocessors (e.g., OpenMP or POSIX
threads) and message-passing primitives for clusters
(e.g., MPI) are well developed, performance dictates
the use of a hybrid solution. We present preliminary
results of our complexity model and programming methodology
that is based hierarchically upon realistic model components
for message-passing and for symmetric multiprocessor
parallel architectures. The current deployment of teraflops
and the future development of petaflops systems will
certainly require the exploitation of similar hybrid
programming models.
Our
goal is to validate and refine a core complexity
model, efficient primitives and algorithmic libraries
needed to support the effective use of SMP clusters
for computational aeroscience (CAS) and earth and
space science (ESS) codes. In addition, we will show
how to design, implement, analyze, and refine algorithms
that take advantage of the hybrid programming model
and contribute to significant speed-ups for real
computational science problems. These will include
some basic combinatorial support tasks used in numerical
aerodynamics simulations (e.g., sorting integers
for particle-in-cell codes) and selected numerical
computations (e.g., fast transforms).
These
algorithmic kernels exhibit promising performance
improvements. Our hybrid programming environment
also must support mission-critical high-performance
computing applications. Thus, we use real computational
aeroscience and Earth and space science codes drive
our algorithmic development. These applications include
segmentation and classification of remotely-sensed
imagery, and navigation, calibration, and registration
of terascale data sets, for remote sensing (AVHRR
and MODIS) and three-dimensional numerical relativity
with the CACTUS computational astrodynamics toolkit.
