[ HPF Home | Versions | Compilers | Projects | Publications | Applications | Benchmarks | Events | Contact ]

---

Princeton Ocean Model

Back to Applications

Steve A. Piacsek, 
Naval Research Laboratory,
piascek@nrlssc.navy.mil

Ocean modeling, 3D; hydrostatic and nonhydrostatic (3D Helmboltz solver required), Implicit time marching (2D Helmboltz solver), Explicit time marching (no solver needed)

Number of Lines of Code: 600-1000

Target Platforms and HPF Compilers Used:

T3E- F77 and F90, SGI O2K- F77 and F90, IBM SP2- F77 and F90

Coding Styles (data decompositions, computational methods):

F77 - DO LOOPS, 5 pt. Algorithms moderate bounding updating. F90- WHERE, 5 pt. Algorithms, moderate bounding updating.

Extrinsic Interfaces Used (and reasons):

MAXVAL MINVAL SUM - too lazy to write my own parallel version

Performance Information, if Available (including any possible comparisons to MPI and/or OpenMP):

HPF SFI O2K 1024x512 (2D)              (3D benchmark not finished)
Proc            Speedup
  4                1.0
  8                1.87/2
 16                3.43/4
 32                6.35/8
 64                11.2/16

Please comment on any aspects of the application that might be interesting, including any problems using HPF effectively:

Boundary updating will slow down overall performance; my CMS code suffered from this. Number of operations performed in one of the directions (I,2,K) - too many K+1, K-1, etc - outweighs operations in other directions by a factor of 3 (i.e. vertical mixing radiation, etc in atmosphere ocean.)