PSTSWM Benchmark Results

Evaluation of Early Systems -- recent results


PSTSWM Benchmark Results

As part of the Evaluation of Early Systems project and the CHAMMP climate modeling program, benchmarks drawn from the Parallel Community Climate Model PCCM2 are being used to evaluate parallel machines of interest to the U.S. High Performance Computing and Communications (HPCC) Program. One stage of this ongoing study was completed in April 1995, with benchmark results obtained on an Intel Paragon MP system, on an IBM SP2 system, and on a Cray Research Inc. T3D using the Parallel Spectral Transforms Shallow Water Model (PSTSWM).

PSTSWM is a message-passing benchmark code and parallel algorithm testbed that solves the nonlinear shallow water equations on a rotating sphere using the spectral transform method. PSTSWM was developed to evaluate parallel algorithms for the spectral transform method as it is used in global atmospheric circulation models. Multiple parallel algorithms are embedded in the code and can be selected at run-time, as can the problem size, number of processors, and data decomposition. PSTSWM is also used in the fair evaluation of parallel computers for spectral transform-based weather codes, fair in the sense that the code can be easily optimized for each platform before timing measurements are taken.

PSTSWM is a communication-intensive code, requiring global communication of the prognostic fields during each timestep, and is a very demanding (and realistic) benchmark. When using the optimal parallel algorithm and communication protocols, all three platforms show good performance and reasonable scaling, as illustrated in the following two figures. Note that the choice of parallel algorithm and its implementation must be made carefully to achieve this performance.

Computational rate and computational rate per processor for PSTSWM on an Intel Paragon (P), on an IBM SP2 (S), and on a Cray Research Inc. T3D (T) for a range of numbers of vertical levels (L) and numbers of processors when using a problem resolution of T170 and 64 bit precision. Foster, Toonen, and Worley, Performance of parallel computers for spectral atmospheric models, The SP2 has the highest computational rate for a given number of processors, followed by the T3D and the Paragon, respectively.

Evaluation of Early Systems Project


Patrick H. Worley / (worleyph@ornl.gov)
Last Modified Tuesday, 12-Sep-2000 13:01:40 EDT.