Russell, Martin and Andreas: Rene is optimistic that NEC will be soon be releasing a compiler that includes support for Cray pointers on assumed-size arrays.
Meanwhile, we are working on reducing our use of Cray pointers, and finding workarounds that do not degrade performance. We hope to have something out by the end of summer. Russell Fiedler writes: > Martin, > > I would be interested in seeing the changes that you made to the MOM3 > code. I'm about to make some modifications to get MOM4 running on a SX-5 > and was going to base my fixes on Rene's work although I was rather loath > to do all the copyin-copyout changes that he had to do. > > As a note to others, the problem is that the SX Fortran 90 compiler > doesn't provide support for Cray pointers on assumed sized arrays. > > Cheers, > Russ Fiedler > > > On Thu, 13 Jun 2002, Martin Schmidt wrote: > > > Dear MOM4 users, > > > > I was working some full weeks to have the MOM3-code running on a > > NEC SX-6. There have been two mojor problems: > > > > - pointers which are not initialized in the mpp-code but are > > considered as assigned in the further program flow, > > - Cray - pointers. I had made also an official request to NEC > > concerning CRAY-pointers, - no answer. > > Cray- pointers are used in MOM to reorganize the inner structure > > of fields without copying the data., e.g. to condense two > > dimensions into one. To my knowledge, CRAY-pointers are not F90 > > standard and should be avoided in future code. My work around > > was recoding the mpp-modules. There was an earlier fix by Rene > > Redler and I based my changes on his code. However, he really > > copied large fields at some places. So I introduced F90-pointers > > and included the field dimensions in the procedure interface. > > Now MOM3 runs with MPI on a SX-6, however, I am not sure what > > FORTRAN90 really does, may be some copy process is hidden in the > > mpp-procedure calls. This needs tests and I do not have the F90 > > knowledge to do this. > > > > If requested, I will e-mail the changed mpp-files. However, it > > is old code and not the last version as used in MOM4. > > > > Greetings, > > Martin Schmidt > > > > _______________________________________________________________ > > > > Don't miss the 2002 Sprint PCS Application Developer's Conference > > August 25-28 in Las Vegas - >http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink > > > > _______________________________________________ > > FMS-mom4 mailing list > > [EMAIL PROTECTED] > > https://lists.sourceforge.net/lists/listinfo/fms-mom4 > > > > > _______________________________________________________________ > > Don't miss the 2002 Sprint PCS Application Developer's Conference > August 25-28 in Las Vegas - >http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink > > _______________________________________________ > FMS-mom4 mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/fms-mom4 > -- Balaji 1 609.452.6516 SGI/GFDL Princeton University _______________________________________________________________ Don't miss the 2002 Sprint PCS Application Developer's Conference August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink _______________________________________________ FMS-mom4 mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fms-mom4
