| blob | d478ad1ed09eef66bc2ae206fc503789c3a7bb91 |
2 <HTML>
3 <HEAD>
5 </HEAD>
6 <BODY>
10 Since this FAQ is written for version 2.0, not all of the solutions
11 may apply to previous GROMACS versions.
13 <DL>
19 is supported.</A>
29 <DD><A HREF="#convert">I want to convert my structure from a .gro, .tpr or trajectory file to a .pdb file.</A>
31 missing atoms.</A>
46 to analyze.</A>
48 same number/sequence of atoms.</A>
50 over and over again.</A>
52 for.</A>
56 <DD><A HREF="#none">My problem is none of the above or none of the solutions below seem to work.</A>
57 </DL>
58 <P>
60 <DL>
63 <P>
66 agreement</A> and send it to us. For academic institutions the
67 license is free. We will give you a username and password so you can
68 download GROMACS.
70 <P> (More info can be found in the
73 <P>
77 my system is supported. Now what?</FONT></B>
79 <P>
89 problems should exist. Especially SUN (SunOS), SGI (IRIX) and LINUX
90 are known to be virtually problem-free. It is even possible to use
93 The basic requirements are a version of GNU make (usually called gmake)
94 and a C/C++ compiler. For better performance a good
95 FORTRAN compiler should be available too.
97 <P>
98 Systems for which a default Makefile is present are listed below.
99 For all these systems GROMACS normally compiles without any problems,
100 and can also run in parallel (providing a good version of MPI or PVM is
101 present).
122 </DL>
123 </multicol>
125 <P> (More info can be found in the
128 <P>
132 <P>
134 instructions on
137 for when things go wrong.
138 If all else fails, please don't hesitate to
140 a problem and found a solution, please
142 do the work you already did all over again.
144 <P>
149 <P>
150 <DD>
151 You will need an up to date version of MPI or PVM.
154 <PRE>
155 cd $GMXHOME/src
156 gmake clean
157 gmake
158 </PRE>
159 On most architectures you will now have new binaries in the
160 $GMXHOME/bin/$GMXCPU directory. The mdrun in that directory will run
161 both in parallel and non-parallel.
162 On some architectures (e.g. IBM SP) you will have a parallel mdrun binary
163 in $GMXHOME/bin/$GMXCPU_mpi (or _pvm).
164 <P>
165 The extra things needed to start a run on multiple processors are:
169 with the appropriate options.
170 For instance on an SGI with the newest verison of MPI:
171 <PRE>
174 </PRE>
175 NOTE: you have to specify the full path of the MPI mdrun binary.
178 <P>
179 Newer MPI versions on SGI's can officially not run in the background.
180 To prevent mdrun from hanging when you log out, use the following csh script:
181 <PRE>
182 #!/bin/csh -f
184 exit
185 </PRE>
186 <P>
191 <P>
192 <DD>
194 <P> First rest assured that GROMACS is amongst the fastest MD programs
195 currently available. We are often a factor of 2 faster than some of
196 the other leading packages.</P>
198 <P> Ultimately, one would want to simulate seconds of whole ensembles
199 of proteins. Since that is not nearly possible (we only come 9 orders
200 of magnitude short in time and 18 orders of magnitude short in number
201 of particles), we suggest the following things to at least get the
202 maximum performance currently available: </P>
204 <UL>
206 <LI>Turn off double precision by selecting a non-double precision CPU
208 effect on the outcome of your simulations, the accuracy is normally
209 not restricted by the precision of the floating point notation, but by
210 the numerical integration scheme itself. <em>This is a free-for-all
212 Challenge, the double precision version runs some 30% slower than
213 single precision.)
215 <LI>Use Fortran inner loops (providing you have a good Fortran
216 compiler available for your system). If you have a DEC alpha running
217 Linux, get the Compaq compilers
220 <LI>If your system consists partly of water, use water specific
221 non-bonded optimizations (set
225 depending on your system and your machine type.
227 <LI>Run GROMACS parallel on multiple processors (see
229 Although this increases overall performance, effiency may be lost due
230 to sub-linear scaling of the parallel system, e.g. 2 processors
231 perform at 85% of twice the performance of one processor. This gets
232 progressively worse for more processors, for more than 8 processors
233 efficiency drops to about 60%. Because of this, it is often better to
234 run multiple runs on one or a few processors. Check out or <A
236 both single- and multi-processor systems.
238 <LI>Preliminary tests show that PME and PPPM are
240 nm on some systems. Unfortunately PME and PPPM do not scale as well as
241 cut-off electrostatics when running in parallel.
245 which controls the constraining of hydrogen atoms. This eliminates the
246 highest freqency motions in your system, enabling you to increase the
248 with negligible loss of accuracy!
250 <LI>If your system has relatively slow disk-IO, and/or you write
251 frames and energies out very often, and/or you have a very large
252 system the performance might be limited by disk access. In that case,
253 you might consider writing less frames to your trajectories
257 <LI>And last, if you do not yet have the fastest computer available,
259 another one or two.
261 </UL>
264 details on GROMACS performance. Also be sure to check it out in case
265 you are still considering what type of machine you should buy to
266 perform your simulations.</P>
268 <P>
271 <DT><HR><B><FONT COLOR="#FF0000">I want to simulate a very large system using GROMACS. Now what?</FONT></B>
273 <P>
275 <DD> Just do it. The limit on the number of atoms in the calculation
276 is only determined by the amount of memory available in your computer
280 your system (your mileage may vary). Due to the fact that we initially
281 developed GROMACS to run on our home-built parallel machine, with
282 only 8Mb of memory per processor, the code is quite well optimized for
283 memory use. To get an indication of scaling of GROMACS performance as
284 a function of system size, have a look at the
287 <P>
292 <P>
295 Start where it says "eiwit.pdb" (this is somewhere at the top).
296 More detailed info can be found in the
298 section, you probably can start where it says "Ribonuclease S-Peptide".
299 <P>
304 <P>
307 but won't be able to make inter-subunit cystine bridges.
309 have different chain identifiers.
311 <P>
314 <DT><HR><B><FONT COLOR="#FF0000">I want to convert my structure from a .gro, .tpr or trajectory file to a .pdb file. Now what?</FONT></B>
316 <P>
323 structure files can be fitted with
332 view them.
334 <P>
338 long bonds and/or missing atoms. Now what?</FONT></B>
340 <P>
341 <DD> There are probably atoms missing earlier in the
345 will tell you which one is missing. Then add the atoms in your pdb file,
346 energy minimization will put them in the right place,
347 or fix the side chain with e.g. the
349 <P>
355 <P>
356 <DD> This could mean that you have started a very long simulation, or
357 you are using a very slow computer. You might want to check if your
358 simulations really should be as long as it is, or get a faster
359 computer. Either way, there is nothing to worry about because GROMACS
361 also mean that your simulation somehow caused Not-a-Numbers to be
362 generated (check the energies in your logfile). In that case look at
364 produced</A>''.
365 <P>
371 <P>
373 actually indicates that some atoms have gotten a very large velocity,
374 which usually means that (part of) your molecule(s) is (are)
375 exploding. If you are using LINCS for constraints, you probably also
376 already got a number of LINCS warnings. When using SHAKE this will
377 give rise to a SHAKE error, which halts your simulation before the
378 "1-4 not within cutoff" error can appear.
379 <P>
380 There can be a number of reasons for the large velocities in
381 your system. If it happens at the
382 beginning of the simulation, your system might be not equilibrated
383 well enough (e.g. it contains some bad contacts). Try a(nother) round
384 of energy minimization to fix this. Otherwise you might have a very
385 high temperature (which might be caused by incorrect use of
387 timestep. Experiment with these parameters till the error stops
388 occurring.
390 <P>
393 <DT><HR><B><FONT COLOR="#FF0000">My run seems to be running, but no output is produced. Now what?</FONT></B>
395 <P>
397 <UL>
399 <LI>Your simulation might simply be (very) slow, and since output is
400 buffered, it can take quite some time for output to appear in the
401 respective files. If you are trying to fix some problems and you want
402 to get output as fast as possible, you can set the environment
405 buffering on again.
407 <LI>Something might be going wrong in your simulation, causing
409 the result of e.g. division by zero). Subsequent calculations with
411 everything down by orders of magnitude. On a SGI system this will
412 usually result in a large percentage of CPU time being devoted to
416 <LI>You might have all
419 this will suppress most output.
421 <LI>Your disk might be full. Eventually this will lead to
425 <LI>If you are running with MPI on an SGI you need a script,
428 </UL>
429 <P>
434 <P> <DD> If you are using simulated annealing (see
437 simulated annealing run</A>''. If you are not using simulated
438 annealing, you might have very close contacts or a too large time
439 step. This causes inaccurate integration which will usually result in
440 a large positive temperature drift. Try some more energy minimization
441 to get rid of the close contacts, or if that still doesn't help, try a
442 short equilibration run with a smal(ler) time step. <P>
445 <DT><HR><B><FONT COLOR="#FF0000">I get very strange temperatures during my simulated annealing run. Now what?</FONT></B>
447 <P>
449 <DD> This probably means that the initial time in your simulation is
450 not zero ps. The temperature during simulated annealing is controlled
451 by two points on a linear curve: temperature will be zero K at
454 ps. This means that if you do not start your simulation at zero ps,
457 simulation is larger than that, the temperature in your simulation
459 will rise indefenitely with increasing time. Below two schematic
460 graphs are included to clarify things a bit.
462 <P>
463 <TABLE>
465 <TR>
467 <TD>
469 <TR>
471 <TD>
473 <TR>
477 </TABLE>
478 <P>
483 <P>
484 <DD>
486 The following answer is only valid for crashes which are not due to
487 mdrun, like a system crash, a full disk, or a kill by the queuing
489 <P>
491 To really continue a simulation as if nothing had happened, you will
492 need coordinates and velocities in full precision (i.e.
495 reduced precision (only 3 decimal places after the decimal point) and
496 do not contain velocity information at all. Feed this trajectory and
500 to specify the one-but-last frame from your
502 last frame is likely to be corrupted due to the crash. With the
505 restart your simulation.
507 <P>
509 After the continuation run is finished, you will have your simulation
510 split up in separate files, which you will probably want to combine.
511 This can be done as follows (the same command works for xtc-files):
512 <PRE>
514 </PRE>
515 The energy files can be concatenated in a similar manner:
516 <PRE>
517 <TT><a href="online2.0/eneconv.html">eneconv</a> -o whole.edr part1.edr part2.edr part3.edr</TT>
518 </PRE>
519 Since tpbconv sets the time in the continuation runs the files are
520 automatically sorted and overlapping frames removed. If you have a mix of
521 runs continued with tpbconv and grompp you might have to set the times yourself
522 (see the manual pages for details).
523 <P>
525 It is ofcourse possible to start a simulation from the coordinates in
527 velocities will have to be generated resulting in a 'kink' in the
528 simulation. To prevent this you should write coordinates and
530 during your simulations. Do this by setting
534 these frames very often (every 10 ps or so), but remember that when
536 calculated after the last frame in the <TT><a
538 recalculated for a proper continuation.
540 <P>
544 which I want to analyze. Now what?</FONT></B>
546 <P>
547 <DD>
550 proteins</A>.
552 <P>
554 called "<TT>eiwit.pdb</TT>". Then this is what you would do:
556 <PRE>
557 pdb2gmx -f eiwit.pdb -reth -ter -n
558 </PRE>
562 hydrogens which are present in your input file. It will also <B>not
563 add</B> any missing hydrogens, so your molecules should be
565 types for which you must select 'none' for both C- and N-terminus.
570 which exactly corresponds with the molecule(s) in your input file.
571 It also writes a coordinate file
574 <P>
575 The next step is:
576 <PRE>
578 </PRE>
589 sequentially.
598 each of your structures to the center of the box (the one you specify
602 <P>
603 A not very exiting but obligatory step is:
604 <PRE>
605 grompp -f grompp.mdp -c conf.gro -p topol.top
606 </PRE>
607 This will generate a run input file
612 available. You can probably use it 'as is', but you might want or need
613 to modify some thing. In any case you are encouraged to
617 <P>
618 Now, suppose you want to calculate all cross-rmsd values for all
619 structures. You will do:
620 <PRE>
621 g_rms -f eiwit.xtc -s topol.tpr -m
622 </PRE>
625 to output an RMSD matrix in
629 <P>
630 Of course there are many more analysis tools available. For example
633 manual</A>.
635 <P>
639 the same number/sequence of atoms. Now what?</FONT></B>
640 <P>
641 <DD>Just type:
642 <PRE>
644 </PRE>
645 g_confrms accepts
648 The program will ask you to select subgroups of both structures for the
649 (non mass weighted) LSQ fit. These subgroups must have the same number of atoms, however the two
651 with the same number of residues but not the same type of residues can be
652 fitted on c-alpha's. You will be warned when the atomnames in the fit groups
653 do not match, but the program will go on.
655 one the first.
657 on top of each other.
658 </P>
662 index group over and over again. Now what?</FONT></B>
663 <P>
664 <DD>
671 favorite GROMACS analysis tool. Now this one group will get selected
672 automatically every time an index group is needed.
673 <P>
677 have a program for. Now what?</FONT></B>
678 <P>
679 <DD>
680 <P> Try the following minimal C program:
681 <HR>
682 <SMALL>
683 <PRE>
684 #include "sysstuff.h"
685 #include "macros.h"
686 #include "statutil.h"
687 #include "smalloc.h"
688 #include "copyrite.h"
689 #include "gstat.h"
691 void main(int argc,char *argv[])
692 {
693 t_topology *top;
694 int status;
695 int natoms;
696 real t;
697 rvec *x;
698 matrix box;
700 static char *desc[] = {
701 "In the future, this program might perform your analysis."
702 };
704 static char *bugs[] = {
705 "This program does not do anything."
706 };
708 static bool bPBC=TRUE;
709 static int nr=1;
710 static real frac=0.5;
711 t_pargs pa[] = {
715 };
717 t_filenm fnm[] = {
718 { efTPX, NULL, NULL, ffREAD },
719 { efTRX, NULL, NULL, ffREAD }
720 };
721 #define NFILE asize(fnm)
723 CopyRight(stderr,argv[0]);
724 parse_common_args(&argc,argv,PCA_CAN_TIME,TRUE,
725 NFILE,fnm,asize(pa),pa,asize(desc),desc,asize(bugs),bugs);
727 /* read topology (atomnames, bonds etc.): */
728 /* (if you don't need those you can omit this step) */
729 top=read_top(ftp2fn(efTPX,NFILE,fnm));
731 /* initialize reading trajectory: */
732 natoms=read_first_x(&status,ftp2fn(efTRX,NFILE,fnm),&t,&x,box);
734 /* check topology against trajectory: */
735 /* your trajectory can be smaller than your topology, */
736 /* if you omitted molecules at the end of your topology, */
737 /* if you did not read a topology, omit this also */
740 top->atoms.nr,natoms);
742 /* start analysis of trajectory */
743 do {
744 if (bPBC)
745 /* make molecules whole again */
746 rm_pbc(top->idef,natoms,box,x,x);
748 /**************************/
749 /* */
750 /* PUT YOUR ANALYSIS HERE */
751 /* */
752 /**************************/
753 </PRE></FONT><PRE>
754 } while (read_next_x(status,&t,natoms,x,box));
756 /* clean up */
757 sfree(x);
758 close_trj(status);
760 }
761 </PRE>
762 </SMALL>
763 <HR>
765 Note that this has to be edited before it will do anything usefull.
766 You have to be sure all the GROMACS include files and library are
767 available when compiling this. The easiest way to do this is to make a
770 (look how other programs are in there and copy that). After that typing
772 <PRE>
773 % gmake g_myanal
774 </PRE>
776 should do the trick. If you don't want to clutter up the
780 fancy options in your program, try modifying one of the smaller
782 although you might want to take one which already does something
783 similar to what you want to do.
785 <P>Finally, if you wrote an analysis tool which, to your opinion, adds
786 something that is really missing in GROMACS, please
788 users, can also benefit from it.
789 </P>
792 <DT><HR><B><FONT COLOR="#FF0000">My problem is none of the above or none of the solutions seem to work. Now what?</FONT></B>
793 <P>
794 <DD>
796 see if your problem (and/or a possible solution) is mentioned
797 there. Be sure to check the
799 download page. Also try
801 where a guided tour of GROMACS is provided.
803 will tell you if you missed any essential steps in setting up a run.
807 list</A> will solve a number of potential problems. In general it
809 all the GROMACS programs you (tried to) use. If all this still leaves
811 don't hesitate to contact us</A>.
812 </P>
814 <P>For summary, the pages to check:
815 <UL>
816 <LI>The <A HREF="readme1.6.html">download page</A> (esp. the <A HREF="readme1.6.html#problems">problems section</A>).
818 guided tour
824 </UL>
825 </P>
826 </DL>
828 <HR>
829 <BR>
830 </BODY>
831 </HTML>