hello friends,
i was just trying to see the performance of different compiler
(gfortran and ifort) and there is a problem inthis part of the code:
"===============Creating Initial Configuration=============="
58 write(*,&
59 '(1x,"Minimum distance between any two particle is",
1x,f6.4)'),rcut !rcut=1.5
60 !-----------------------------------------------
61 ! INITIALISE POSITION
62 !-----------------------------------------------
63 10 do i=1,ntot
64 x(i)=boxl*(ran(seed)-0.0)
65 y(i)=boxl*(ran(seed)-0.0)
66 z(i)=boxl*(ran(seed)-0.0)
67 call pbc(x,y,z,boxl)
68 end do
69 !- - - - - - - - - - - - - - - - - - - - - - - -
70 ! KEEPING MINIMUM DIST. BETWEEN ANY TWO
71 ! ATOM GREATER THEN RCUT
72 !- - - - - - - - - - - - - - - - - - - - - - - - ...

In article p25g2000pri.googlegroups.com>,
rudra gmail.com> writes:

Why subtract 0.0?

this is a subroutine and seed, boxl is supplied`

rudra wrote:

Are you certain that it necessarily will converge? Are you getting the
same random sequence from the two compilers?

What have you done to try to characterize the nonconvergence? Do you
even have the slightest idea where the two compilers are diverging? I
would strongly recommend that you study that, using write statements or
a debugger.

In article <91157a9b-7293-4334-b813-5b611fb805d7@p25g2000pri.googlegroups.com>,
rudra gmail.com> writes:

Nothing like deleting all context from your follow up.
When I asked 'what is seed?', I literally meant give
us the value. Did you actually RTFM, which tells you
how to use RAN() in gfortran? When I asked 'What is
boxl?' Either show us the code or at least tell us
what boxl() does.

Craig Powers wrote:

I agree. A program could be very sensitive to roundoff such that
it doesn't converge or doesn't seem to converge. In some cases,
the result can alternate between values differing in the low order
but, and so will 'fail to converge' if compared for equality.

If it diverges (goes to infinity) then that is very different.

-- glen

"glen herrmannsfeldt" ugcs.caltech.edu> wrote in message
news:mKednd9g_6UrCrXVnZ2dnUVZ_srinZ2d@comcast.com...

Some form of fuzzy logic?

Regards,

Mike Metcalf

Michael Metcalf wrote:

http://en.wikipedia.org/wiki/Periodic_point

http://en.wikipedia.org/wiki/Limit_set

When using something like Newton-Raphson, it is non
unusual that the limit, even for a scalar, has a
period greater than one. For multi-dimensional systems
the result can be much more complicated.

Since the OP didn't give many details, it is hard to know,
but it might be that on one system it does converge to
a fixed point (period one cycle), and on others it doesn't.

-- glen

glen herrmannsfeldt wrote:

"glen herrmannsfeldt" ugcs.caltech.edu> wrote in message
news:5fudneccWLcaL7XVnZ2dnUVZ_gudnZ2d@comcast.com...

Well, I looked as those references, but 'low order buts' were nowhere
mentioned :-).

Regards,

Mike Metcalf