Padd (Prolicleve/Adaptive Domain Decomposition )
Back to Applications
Principal Contact Person and Organization (including e-mail address):
Cecile Germaine
LRI University Paris Suol
cecile.germaine@lri.fr
Brief Description of Application:
Turbulence simulation with adaptive domain decomposition coupling with a remeshing program through Java.
Number of Lines of Code: not usable yet (40)
Target Platforms and HPF Compilers Used:
SP2, PGI Adaptor, and Linux PC’s networked (Myrinet), PGI ADAPTOR
Coding Styles (data decompositions, computational methods):
Blocks (because of efficiency) - cycicle would be better. Iterative sparse compilation (dual Sehur).
Extrinsic Interfaces Used (and reasons):
HPF local/F77 local standing for C-local, C wrapper for calls/call backers to Java U7
Performance Information, if Available (including any possible comparisons to MPI and/or OpenMP):
Not yet - fall 98
Please comment on any aspects of the application that might be interesting, including any problems using HPF effectively:
Full F90 not implemented - needs a lot of code rewriting from furiously written code.
Coupling HPF with Java is easy. Future work: a Java native interface(and corresponding ***** API) for HPF
Trend HPF code environment is high performance server code *****.