Re: [AMBER] unspecified launch failure launching kernel kCalculateMolecularVirial

From: Ross Walker <ross.rosswalker.co.uk>
Date: Fri, 12 Aug 2011 09:02:24 -0700

Please try searching in the archive before posting.
http://archive.ambermd.org/

This is a known bug (ntb=2,ntr=1 is broken right now) and will be fixed in
the upcoming patch...

Damn, I sound like a broken record... (or maybe it should be a corrupted MP3
these days ;-) )...

> -----Original Message-----
> From: Sangeetha B [mailto:sangeetha.bicpu.edu.in]
> Sent: Thursday, August 11, 2011 9:24 PM
> To: AMBER Mailing List
> Subject: Re: [AMBER] unspecified launch failure launching kernel
> kCalculateMolecularVirial
>
> I am running a simulation of a dimeric protein with a total of 576
> residues, solvated
> using TIP3P water model in a cubic box of 10 angstroms dimensions. The
> whole system contains approximately 40000 atoms. I run the simulations
> on NVIDIA GPU using PMEMD implementation of Amber 11.
>
> Here I perform NVT and NPT simulations with restraints on some parts
> of the protein.
> The NVT simulations run properly, whereas the NPT simulations fail
> with the following error.
>
> When the whole protein is restrained the simulations were running
> properly and when a part of the system is constrained I get the
> following error in the terminal whereas no errors are printed in the
> mdout file.
>
>
> cudaMalloc GpuBuffer::Allocate failed out of memory
> At line 111 of file inpcrd_dat.f90
> Fortran runtime error: End of file
>
> or
>
> cudaMemcpyToSymbol: SetSim copy to cSim failed all CUDA-capable devices
> are busy
> or unavailable
> At line 111 of file inpcrd_dat.f90
> Fortran runtime error: End of file
>
>
>
> The following is the input file that i used..
>
>
> # DYnamics with constant pressure with restraints at 300K for 500ps
> &cntrl
> imin=0,irest=1,ntx=5,
> nstlim=250000,dt=0.002,
> ntc=2,ntf=2,
> cut=10, ntb=2, ntp=1, taup=1.0,
> ntpr=500, ntwx=500,
> ntt=3, gamma_ln=2.0,
> temp0=300.0,
> ntr=1,
> /
> Restraints
> 2.0
> RES 1 45 59 134 153 270 289 333 347 422 441 558
> END
> END
>
>
>
> I did some trials by removing and adding the restraints one at a
> time...
> with the first three restraint groups, the process runs whereas when
> the last three groups are used individually or in combination with the
> first three, the process fails with the errors...
>
> Is this problem related only to memory..? NPT+restraints will always
> cause this error?
>
> Because i tried another simulation with only a monomer (288 amino
> acids and 20000 atoms) which runs fine even with NPT + restraints..
>
> It would be grateful if anyone can tell me why this error occurs only
> for dimer while it works fine for a monomer...
>
>
> --
> Sangeetha
>
>
>
>
> On Fri, Aug 12, 2011 at 5:59 AM, Ross Walker <ross.rosswalker.co.uk>
> wrote:
> > Oh in that case just run the equilibration step using the CPU code
> and then
> > set ntr=0 and run the production calculation on the GPU and you
> should be
> > good.
> >
> > Let me know if you see this error when ntr=0.
> >
> > All the best
> > Ross
> >
> >> -----Original Message-----
> >> From: arikcohen [mailto:arikcohen.ucla.edu]
> >> Sent: Thursday, August 11, 2011 5:18 PM
> >> To: AMBER Mailing List
> >> Subject: Re: [AMBER] unspecified launch failure launching kernel
> >> kCalculateMolecularVirial
> >>
> >> The restraints are during the eq. step and are being removed in the
> >> production run.
> >>
> >> Thanks a lot
> >>
> >> Arik
> >> On 8/11/2011 4:06 PM, Ross Walker wrote:
> >> > Yeap... This is the NTB=2, ntr=1 bug. The fix really is coming
> soon.
> >> >
> >> > As I say for now either do not use restraints (why do you want
> >> restraints
> >> > after the equilibration anyway?) or set ntb=1 and run with NVT.
> >> >
> >> > All the best
> >> > Ross
> >> >
> >> >> -----Original Message-----
> >> >> From: arikcohen [mailto:arikcohen.ucla.edu]
> >> >> Sent: Thursday, August 11, 2011 3:54 PM
> >> >> To: AMBER Mailing List
> >> >> Subject: Re: [AMBER] unspecified launch failure launching kernel
> >> >> kCalculateMolecularVirial
> >> >>
> >> >> Sorry, here is the eq.in
> >> >>
> >> >> equilibration run (0 - 2 ns)
> >> >> &cntrl
> >> >> imin=0,                   ! No minimization, plain MD(default)
> >> >> irest=1,                  ! restart calculation
> >> >> ntx=5,                    ! Read X and V formatted, if ntb>  0
> read
> >> >> box,
> >> >> use V if irest = 1
> >> >> ntb=2,                    ! Use pbc Pressure = const
> >> >> pres0=1.0,                ! Reference pressure
> >> >> ntp=1,                    ! constant pressure dynamics
> >> >> taup=2.0,                 ! Pressure relaxation time
> >> >> cut=10,                   ! Non-bonded interaction cutoff
> >> >> ntc=2,                    ! SHAKE: Constrain bonds involving H
> atoms
> >> >> ntf=2,                    ! Typically ntf=ntc enable TIP3P
> constrain
> >> >> tempi=300.0, temp0=300.0, ! Initial and final(reference)
> temperature
> >> if
> >> >> ntt>0
> >> >> ntt=3,                    ! Switch temperature scaling (3 Use
> >> Langevin
> >> >> dynamics)
> >> >> gamma_ln=1.0,             ! Collision frequency for Langevin
> >> dynamics
> >> >> nstlim=1000000, dt=.002,  ! Number of MD steps and step size(in
> ps)
> >> >> ntpr=100,                 ! Print energy every
> >> >> ntwx=100,                 ! Print coor every
> >> >> ntwr=500,                 ! Print for restart option every
> >> >> ntr=1,                    ! Restrain atoms(cartesian space) using
> >> >> harmonic potentail
> >> >> &end
> >> >> Group input for restrained atoms
> >> >> 30
> >> >> RES   1 8
> >> >> RES  10
> >> >> RES  12 47
> >> >> RES  49
> >> >> RES  51 99
> >> >> RES 101 126
> >> >> RES 128 199
> >> >> RES 201 220
> >> >> RES 222 248
> >> >> END
> >> >> END
> >> >>
> >> >> Thanks a lot again,
> >> >>
> >> >> Arik
> >> >>
> >> >> On 8/11/2011 3:12 PM, Ross Walker wrote:
> >> >>> Hi Arik,
> >> >>>
> >> >>> Please send you input files. I suspect the bug is NPT +
> restraints
> >> >> which is
> >> >>> fixed in our development tree. Just waiting some final testing
> >> before
> >> >>> release.
> >> >>>
> >> >>> For now you can run NVT and you should be fine.
> >> >>>
> >> >>> All the best
> >> >>> Ross
> >> >>>
> >> >>>> -----Original Message-----
> >> >>>> From: arikcohen [mailto:arikcohen.ucla.edu]
> >> >>>> Sent: Thursday, August 11, 2011 3:05 PM
> >> >>>> To: AMBER Mailing List
> >> >>>> Subject: Re: [AMBER] unspecified launch failure launching
> kernel
> >> >>>> kCalculateMolecularVirial
> >> >>>>
> >> >>>> No. just a regular enzyme-substrate calculation.
> >> >>>>
> >> >>>> Arik
> >> >>>>
> >> >>>> On 8/11/2011 3:03 PM, Scott Le Grand wrote:
> >> >>>>> Organic solvent or lipid bilayer simulation by any chance?
> >> >>>>> On Aug 11, 2011 2:59 PM, "arikcohen"<arikcohen.ucla.edu>
> >> wrote:
> >> >>>>>> Dear Users,
> >> >>>>>>
> >> >>>>>> While running on a GPU cluster I encountering from time time
> the
> >> >>>>>> following persistent error:
> >> >>>>>>
> >> >>>>>> *unspecified launch failure launching kernel
> >> >>>> kCalculateMolecularVirial*
> >> >>>>>> I would be most thankful for any help/suggestion you might
> have.
> >> >>>>>>
> >> >>>>>> Thanks,
> >> >>>>>>
> >> >>>>>> Arik
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>
> >> >>>>>> _______________________________________________
> >> >>>>>> AMBER mailing list
> >> >>>>>> AMBER.ambermd.org
> >> >>>>>> http://lists.ambermd.org/mailman/listinfo/amber
> >> >>>>> _______________________________________________
> >> >>>>> AMBER mailing list
> >> >>>>> AMBER.ambermd.org
> >> >>>>> http://lists.ambermd.org/mailman/listinfo/amber
> >> >>>> _______________________________________________
> >> >>>> AMBER mailing list
> >> >>>> AMBER.ambermd.org
> >> >>>> http://lists.ambermd.org/mailman/listinfo/amber
> >> >>> _______________________________________________
> >> >>> AMBER mailing list
> >> >>> AMBER.ambermd.org
> >> >>> http://lists.ambermd.org/mailman/listinfo/amber
> >> >>
> >> >> _______________________________________________
> >> >> AMBER mailing list
> >> >> AMBER.ambermd.org
> >> >> http://lists.ambermd.org/mailman/listinfo/amber
> >> >
> >> > _______________________________________________
> >> > AMBER mailing list
> >> > AMBER.ambermd.org
> >> > http://lists.ambermd.org/mailman/listinfo/amber
> >>
> >>
> >> _______________________________________________
> >> AMBER mailing list
> >> AMBER.ambermd.org
> >> http://lists.ambermd.org/mailman/listinfo/amber
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 12 2011 - 09:30:03 PDT
Custom Search