Have you tried it on another GPU?
Can you show the steps you took in leap to build your model? If they
might have included trying things manually at the leap cmd line, I would
put together a script and rebuild the prmtop 'clean', in case leap
propagated some bad state that only appears in cuda.
Bill
On 2/23/17 6:23 PM, Alicia Merlino wrote:
> Hi, after the heating I succesfully switched back to 2 fs but I still have
> troubles running with cuda. The same SHAKE error appears.
>
> 2017-02-23 2:01 GMT-03:00 Alicia Merlino <amerlino.fcien.edu.uy>:
>
>> I'll let you know!
>>
>>
>> 2017-02-23 1:38 GMT-03:00 Bill Ross <ross.cgl.ucsf.edu>:
>>
>>> It would be interesting to see if you could switch back to 2fs once
>>> things equilibrate.
>>>
>>> Bill
>>>
>>>
>>> On 2/22/17 8:32 PM, Alicia Merlino wrote:
>>>> That was with a time step of 1 fs if I change to 2 fs the calculation
>>>> crashes after 400 steps or so
>>>>
>>>> 2017-02-23 1:00 GMT-03:00 Alicia Merlino <amerlino.fcien.edu.uy>:
>>>>
>>>>> I've run the heating with pmemd for 5000 steps and it ended ok. I don´t
>>>>> understand what's happening. I'll try a longer heating with ntpr=1 to
>>> see
>>>>> what happens.
>>>>>
>>>>> Thank you!
>>>>>
>>>>> 2017-02-23 0:30 GMT-03:00 Alicia Merlino <amerlino.fcien.edu.uy>:
>>>>>
>>>>>> Hi, with pmemd the simulation gets started and after 500 steps
>>> everything
>>>>>> seems to be ok.
>>>>>>
>>>>>>
>>>>>> 2017-02-22 23:26 GMT-03:00 David Case <david.case.rutgers.edu>:
>>>>>>
>>>>>>> On Wed, Feb 22, 2017, Alicia Merlino wrote:
>>>>>>>
>>>>>>>> Dear David, the calculation crashes while reading atomic coordinates
>>>>>>> and
>>>>>>>> velocities with the error message that I reported before.
>>>>>>>>
>>>>>>>> Executable path: pmemd.cuda
>>>>>>> OK...now I see that you are running pmemd.cuda. Can you try a short
>>>>>>> simulation using pmemd itself? Doesn't need to be very many steps:
>>> just
>>>>>>> set ntpr=1 so you can see if the simulation gets started OK.
>>>>>>>
>>>>>>>
>>>>>>>>>>> 80: 177 - 178 and 81: 177 - 179
>>>>>>>>>>> Hydrogen atom 177 appears to have multiple bonds to atoms 178 and
>>>>>>> 179
>>>>>>>>>> which
>>>>>>>>>>> is illegal for SHAKEH.
>>>>>>>>>>> Exiting due to the presence of inconsistent SHAKEH hydrogen
>>>>>>> clusters.
>>>>>>> Amber developers: does anyone know if the atom numbers printed here
>>> (from
>>>>>>> gpu.cpp) start from zero or from 1?
>>>>>>>
>>>>>>> ....dac
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> AMBER mailing list
>>>>>>> AMBER.ambermd.org
>>>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>>>>
>>>>>>
>>>>>> --
>>>>>> Dra. Alicia Merlino
>>>>>> Profesora Adjunta del Laboratorio de Qca. Teórica y Computacional
>>>>>> Instituto de Química Biológica
>>>>>> Facultad de Ciencias, Universidad de la República
>>>>>> Iguá 4225, 11400 Montevideo, Uruguay
>>>>>> Tel: (+598) 2525 2186 <+598%202525%202186>
>>>>>> Fax: (+598) 2525 0749 <+598%202525%200749>
>>>>>>
>>>>>
>>>>> --
>>>>> Dra. Alicia Merlino
>>>>> Profesora Adjunta del Laboratorio de Qca. Teórica y Computacional
>>>>> Instituto de Química Biológica
>>>>> Facultad de Ciencias, Universidad de la República
>>>>> Iguá 4225, 11400 Montevideo, Uruguay
>>>>> Tel: (+598) 2525 2186 <+598%202525%202186>
>>>>> Fax: (+598) 2525 0749 <+598%202525%200749>
>>>>>
>>>>
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>
>>
>> --
>> Dra. Alicia Merlino
>> Profesora Adjunta del Laboratorio de Qca. Teórica y Computacional
>> Instituto de Química Biológica
>> Facultad de Ciencias, Universidad de la República
>> Iguá 4225, 11400 Montevideo, Uruguay
>> Tel: (+598) 2525 2186 <+598%202525%202186>
>> Fax: (+598) 2525 0749 <+598%202525%200749>
>>
>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 23 2017 - 19:30:02 PST
