Re: AW: update: segmentation faults with AMBER topology and RAMD tcl scripts

From: yudaqi (dqyu_at_ctb.pku.edu.cn)
Date: Wed Feb 24 2010 - 06:37:25 CST

Hi all,

I have had similar problems for several months. In fact, I have
submitted a related question in the maillist about 3 months ago. As far
as I know, the memory problem only happens when the system is large
enough. For some versions, the memory problem also could be solved by
tune the maximum locked memory in the system or the parameter
pairlistMinProcs if multinodes are used.

I really expect the developers could help us.

Yours sincerely
Daqi Yu
Peking University

Vlad Cojocaru дµÀ:
> Balazs,
>
> Great ! Thanks for the effort to compile .
> To use the AMBER force field, it looks as if you need to compile older
> CVS code (8 August 2009 works for sure) of namd2.7 or namd 2.6 .
>
> Francesco,
>
> So it seems that if the new CVS code is compiled against tcl8.4.19,
> the RAMD scripts run with the CHARMM force field with no error
> message. Could you confirm that as well ?
> I also have a new version of the RAMD that is compatible with tcl8.3
> ... It seems that tcl8.3 does not like the "eq" operand and some if
> statements with multiple tests ... Don't ask me why ...
>
> other NAMD users,
>
> Is there anybody that is able to run the new CVS code with AMBER
> topology files ?
>
> Greetings
> Vlad
>
> Balazs Jojart wrote:
>> Dear Vlad,
>> The namd-CVS version was compiled against Tcl8.4.19.
>> The RAMD calculations runs with charmm psf files, without any error
>> message, but with this compilation I obtain again a segmentation
>> fault error with the amber topology files.
>> Don't worry about the troubles, we thank for your work to implement
>> the code into namd.
>> Cheers,
>> Balazs
>>
>> Vlad Cojocaru wrote:
>>> Dear Balazs,
>>>
>>> Did you try to run the tcl forces tutorials from the NAMD website?
>>> Do they run correctly ? If you did not, I suggest you do that first.
>>> Like this we will see if you have a general tcl problem or its only
>>> the RAMD scripts. Let me know about it ... I am saying this because
>>> I had similar problems with tcl 8.5 with other tcl scripts I have
>>> been using ( like the former ABF module )...
>>>
>>> If the tcl forces tutorials exit with the same type of errors, then
>>> there is a general problem with your compilation. If these errors
>>> only appear while running RAMD, then the next test would be if you
>>> could compile against tcl 8.4.19 ...
>>>
>>> Actually maybe to put a correct diagnostic, you could compile namd
>>> 2.6 both against tcl 8.4.19 and 8.5 and try the RAMD scripts with
>>> this NAMD code ....
>>>
>>> What is actually interesting that you don't get the seg faults
>>> anymore although you are using AMBER topology. ....
>>>
>>> I am also trying to test against different tcl versions to see what
>>> is actually going wrong . ....
>>>
>>> Sorry for the troubles .... but we are running these scripts so
>>> successfully here that I didn't even imagine that all these errors
>>> could appear ...
>>>
>>> Vlad
>>>
>>> Balazs Jojart wrote:
>>>> Dear Vlad,
>>>> I think the topology is OK, because it was used in a former
>>>> simulation only using standard MD. In the output file (RAMD), the
>>>> the structure information looks ok. I tested the loadcoords part as
>>>> well, by adding "puts coordinates(1)" lines in the tcl code, and
>>>> the coordinates are also missing - same error "no such element in
>>>> array".
>>>> Our system administrator compiled the newest CVS of namd against
>>>> tcl8.5, fortunately I don't obtain any segfault error, but
>>>> unfortunately, the same error was occurred with the example files
>>>> as well, as described below. The outputs are attached.
>>>> Do we have to compile with another version of tcl (for example
>>>> Tcl8.4.19)?
>>>> Cheers,
>>>> Balazs
>>>>
>>>> Cojocaru,Vlad wrote:
>>>>> Dear Balazs,
>>>>>
>>>>> This is a more friendly error ...
>>>>> It looks as if your definition of the ramd and/or protein atoms is
>>>>> somehow wrong .. You have an atom that somehow does not have a
>>>>> mass .. Can you double check ?
>>>>>
>>>>> Otherwise, if you could send me a zipped file with the files
>>>>> required to reproduce your problem (topology, coordinates, box
>>>>> info files) I can try to see what is wrong ..
>>>>>
>>>>> vlad
>>>>>
>>>>>
>>>>> -----Urspr¨¹ngliche Nachricht-----
>>>>> Von: Balazs Jojart [mailto:jojartb_at_gmail.com]
>>>>> Gesendet: Di 23.02.2010 17:05
>>>>> An: Cojocaru,Vlad
>>>>> Betreff: Re: namd-l: update: segmentation faults with AMBER
>>>>> topology and RAMD tcl scripts
>>>>>
>>>>> Dear Vlad,
>>>>> Using the RAMD method with the compiled namd2.7beta2 against
>>>>> Tcl8.5 and charmm topologies I obtain the following error message:
>>>>>
>>>>> TCL: can't read "masses(5962)": no such element in array
>>>>> FATAL ERROR: can't read "masses(5962)": no such element in array
>>>>> while executing
>>>>> "expr $ligMass + $masses($ramdAtom)"
>>>>> ("foreach" body line 2)
>>>>>
>>>>> the whole error message is attached.
>>>>> I'm going to test the charmm topologies, with the namd2.6 + tcl8.5
>>>>> version compilation.
>>>>> Cheers,
>>>>> Balazs
>>>>>
>>>>>
>>>>> Vlad Cojocaru wrote:
>>>>>> Dear all,
>>>>>>
>>>>>> To test the errors I've been commenting on recently I compiled
>>>>>> today's CVS code of NAMD (23 February 2010) using exactly the
>>>>>> same protocol as for my last running compilation (13 August
>>>>>> 2009). Tcl version is tcl8.4.19. Architecture: x86_64 + infniband
>>>>>> + intel 11.0 compiler + mvapich1.4rc1. The compilation worked
>>>>>> absolutely fine.
>>>>>>
>>>>>> Then I ran the RAMD tests which resulted in "Segmentation
>>>>>> violation" . Then I tested the new executable for one of my
>>>>>> standard MD simulation using AMBER topology. The same undefined
>>>>>> error occurred.
>>>>>>
>>>>>> To conclude, the latest CVS code produces segmentation faults on
>>>>>> AMBER topologies regardless whether RAMD is switched on or not.
>>>>>> Therefore, this error apparently has nothing to do with the RAMD
>>>>>> scripts. It looks as if the amber topology is read correctly but
>>>>>> the program exists with segmentation violation immediately after
>>>>>> reading the topology (below last lines of the output)
>>>>>>
>>>>>> I could not test against CHARMM topologies as I don't have one
>>>>>> but another NAMD user mentioned recently that the code runs fine
>>>>>> on CHARMM topologies while observing the same errors with AMBER
>>>>>> topologies.
>>>>>>
>>>>>> To test the RAMD scripts until the problem with the AMBER
>>>>>> topologies will be fixed, I suggest to compile namd 2.6 against
>>>>>> tcl8.4.19 or above.
>>>>>> If you still see errors, please let me know.
>>>>>>
>>>>>> Cheers
>>>>>> Vlad
>>>>>>
>>>>>> -----------------output sample------------------------------
>>>>>> Reading parm file (1r9o_1.top) ...
>>>>>> PARM file in AMBER 7 format
>>>>>> Warning: Encounter 10-12 H-bond term
>>>>>> Warning: Found 18886 H-H bonds.
>>>>>> Info: SUMMARY OF PARAMETERS:
>>>>>> Info: 62 BONDS
>>>>>> Info: 131 ANGLES
>>>>>> Info: 49 DIHEDRAL
>>>>>> Info: 0 IMPROPER
>>>>>> Info: 0 CROSSTERM
>>>>>> Info: 0 VDW
>>>>>> Info: 210 VDW_PAIRS
>>>>>> Info: TIME FOR READING PDB FILE: 0.158216
>>>>>> Info:
>>>>>> Info: Reading from binary file 1r9o_1_eq10_rst.500000.coor
>>>>>> Info: ****************************
>>>>>> Info: STRUCTURE SUMMARY:
>>>>>> Info: 64265 ATOMS
>>>>>> Info: 64360 BONDS
>>>>>> Info: 13994 ANGLES
>>>>>> Info: 29662 DIHEDRALS
>>>>>> Info: 0 IMPROPERS
>>>>>> Info: 0 CROSSTERMS
>>>>>> Info: 98534 EXCLUSIONS
>>>>>> Info: 60475 RIGID BONDS
>>>>>> Info: 132320 DEGREES OF FREEDOM
>>>>>> Info: 22676 HYDROGEN GROUPS
>>>>>> Info: 4 ATOMS IN LARGEST HYDROGEN GROUP
>>>>>> Info: 22676 MIGRATION GROUPS
>>>>>> Info: 4 ATOMS IN LARGEST MIGRATION GROUP
>>>>>> Info: TOTAL MASS = 394234 amu
>>>>>> Info: TOTAL CHARGE = -1.48128e-05 e
>>>>>> Info: MASS DENSITY = 1.00973 g/cm^3
>>>>>> Info: ATOM DENSITY = 0.0991208 atoms/A^3
>>>>>> Info: *****************************
>>>>>> Info:
>>>>>> Info: Entering startup at 30.2467 s, 156.332 MB of memory in use
>>>>>> Info: Startup phase 0 took 0.00019908 s, 156.332 MB of memory in use
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:55:29 CST