MPI gibbs

From: Jarrod Smith <jsmith_at_chazin.structbio.vanderbilt.edu>
Date: Thu 28 Sep 2000 10:26:45 -0500

We are trying to get MPI gibbs to work in AMBER6. The calculation we're
trying to run does not use dynamically modified windows, so I commented
out the mexit statement on line 316 of gib.f and compiled on Intel/Linux
with PGF77 and MPIPRO. When np==1, the binary runs fine, and reproduces
the correct result within rounding/precision errors. When I use more
than one CPU, the energies begin to deviate from each other
significantly even after 10 steps, and within the first 100 steps, the
calculation stops with the message:

     Routine TORCON (internal constraints) failed.
     Required change is too large.

And then at the bottom of the output file:

          ***** ERROR TERMINATION DUE TO SHAKE OR TORCON *****
     AMBER/MPI: ERRORS ENCOUNTERED

The exact energy deviation and step at which the calculation exits seem
to be a function of the number of processors used.

Basically, we'd really like to get this working and are looking for
suggestions. The control input appears below my signature if you want
to have a look.

Thanks
 
Jarrod A. Smith
Research Asst. Professor, Biochemistry
Asst. Director, Center for Structural Biology
Computation and Molecular Graphics
Vanderbilt University
jsmith_at_structbio.vanderbilt.edu



PMF via FEP for virtual bond cons P
 &cntrl
        timlim = 9999999.0, irest = 1, ibelly = 0, ichdna = 0,
        ntx = 7, ntxo = 1, ig = 2347613,
        tempi = 0.0, heat = 0.0, ntb = 2,
        iftres = 1, ibxrd = 0, nrun = 1,
        ntt = 5, temp0 = 298.0, dtemp = 5.0,
        tautp = 0.20, tauts = 0.20, isolvp = 6932,
        nsel = 0, dtuse = 1.0, ntp = 2,
        npscal = 1, pres0 = 1.0, comp = 44.6,
        taup = 1.0, ndfmin = 6, ntcm = 0,
        nscm = 0, istay = 0, nstay = 999999,
        natrcm = 0, isvat = 1, nstlim = 1000,
        init = 4, t = 0.0,
        dt = 0.0010, ntc = 3,
        nconp = 0, tol = 0.00050, tolr2 = 0.00010,
        ncorc = 1, ishkfl = 1, ntf = 3 ,
        ntid = 0, ntnb = 1,
        nsnb = 25, idiel = 1,
        ielper = 0, imgslt = 0, idsx0 = 0,
        itrslu = 1, ioleps = 0, intprt = 5,
        itip = 0, cut = 9.0, scnb = 2.0,
        scee = 1.2, dielc = 1.0, ntpr = 500,
        ntwx = 1000, ntwv = -1, ntwe = -1,
        ntwxm = 999999, ntwvm = 999999, ntwem = 999999,
        ioutfm = 0, isande = 0,
        iperat = 0, iatcmp = 0, ntatdp = 0,
        ntr = 0, ntrx = 1,
        taur = 1.0, intr = 1, ibigm = 0,
        isftrp = 0, rwell = 5.0, iftime = 0,
        ctimt = 0.0, almda = 0.0, almdel = 0.002,
        isldyn = 3, idifrg = 0, nstmeq = 2000,
        nstmul = 3000, ndmpmc = 0, idum2 = 0,
        idwide = 0, ibndlm = 0, iavslp = 8,
        iavslm = 2, islp = 0, corrsl = 0.80,
        amxmov = 0.10, iavdel = -1, iavdem = 2,
        amxdel = 1.0, almdl0 = 0.00010, dlmin = 1.0D-6,
        dlmax = 0.10, amxrst = 0.50, norsts = 0, ntsd = 0,
        dte = 0.0010, dta = 0.0010, ivcap = 0,
        natcap = 0, fcap = 0.0
 &end
 1218 1706 1 0 2 0.0 1.0
      40.0 20.83 40.0 45.83 0 0
Received on Thu Sep 28 2000 - 08:26:45 PDT
Custom Search