Hi Wei,
A quick test and it seems that sander.PIMD is not being built by default
either in serial or parallel but the default test suite is setup to run the
PIMD test cases.
Can you check this please.
On another note what is the status of NEB under the new system? Does it need
to be run with sander.PIMD and I assume sander.PIMD.MPI? Also are things
linear in memory for this as well?
On one other note ;-) does this mean a lot the ifdef PIMD stuff will be
going away? I notice for example that the qm_mm call in force is much much
simpler which is nice. However, there are still a lot of PIMD ifdefs in
force.f. Although a number of these are around variable declarations like
spring_energy or alloc_error. I assume these can go away now?
All the best
Ross
/\
\/
|\oss Walker
| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk | PGP Key available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
> -----Original Message-----
> From: owner-amber-developers.scripps.edu
> [mailto:owner-amber-developers.scripps.edu] On Behalf Of Wei Zhang
> Sent: Tuesday, October 17, 2006 09:26
> To: amber-developers.scripps.edu
> Subject: amber-developers: Major change on CVS (for new PIMD
> implementation)
>
> Dear All,
>
> I just checked in a new implementation of PIMD which is based on
> multi-sander framework.
> Meanwhile, the following executables has been remove from Makefile:
> sander.PIMD, sander.CMD
> sander.PIMD.A1ST, sander.CMD.A1ST. In the new implementation, all PIMD
> simulations are
> run through "sander.MPI". A new control parameter "ipimd" has been
> introduced to indicate
> simulation type, "ipimd=0" means normal PIMD simulation,
> "ipimd=1" means
> primitive PIMD,
> "ipimd=2" means normal-mode PIMD or CMD (depend on
> "adiab_param"). Test
> case for this
> new implementation can be found under directory
> "amber10/test/full_pimd".
>
> The new implementation has the following advantages. Firstly,
> it is more
> efficient because the
> linear memory layout. Secondly, it is much easier to make it work with
> other new features of
> sander, for example: we can run PIMD with AMOEBA force field now.
> Moreover, allthough AMOEBA
> itself has not been parallelized yet, we can run AMEOBA-PIMD
> in parallel
> as long as number of CPUs
> is less than number of beads.
>
> The shortcoming of the new implementation is also obvious, we will not
> be able to run partial-PIMD
> anymore. To overcome this, the old implementation will not be deleted
> but will be merged into
> sander.LES and we will be able to run partial-PIMD through sander.LES.
> It will also be fast if the
> quantum portion is reasonably small.
>
> As you know, this is a major change, although I have tested the
> code(with ifort and g95), there are still
> some chances that some of your staff have been broken. Thus, if you
> experienced some
> unusual behavior of sander after cvs update, please contact me, and we
> can figure it out together.
>
> Thank you very much !
>
> Sincerely,
>
> Wei
>
>
Received on Wed Oct 18 2006 - 06:07:29 PDT
