[gmx-developers] mdrun not working with VERSION 3.3_beta_20050202

Florian Haberl Florian.Haberl at chemie.uni-erlangen.de
Mon Feb 14 18:58:41 CET 2005


Hi,

i post the following to developers mailing list cause it^s a problem with a 
cvs version.
I^ve done some benchmarking on an Itanium 2 system,
specs for them are: 

HP zx6000 with 
2 Itanium 2 (McKinley)-CPUs with 900MHz and 1.5MB L3-Cache
10GB Ram
and some awsome fast scsi raid discs ( > 100 MB/s writing speed)
coupled with gbit lan.

Gromacs  3.2.1 (actuall version) 
vs
VERSION 3.3_beta_20050202 (from cvs)

(Benchmarks are after my signature).

both compiled with
mpich-1.2.6-intel8
intel_cc_81-026
intel_fc_81-023

Single cpu runs without problems mpi with gmx 3.2.1 too, but with more than 
one cpu some strange errors occur.

With more than 1cpu benchmarks always died with an error msg:

[quote]
-------------------------------------------------------
Program mdrun, VERSION 3.3_beta_20050202
Source code file: ns.c, line: 314

Fatal error:
ns.c, 315: Negative number of short range atoms.
Call your Gromacs dealer for assistance.
-------------------------------------------------------

"It's Not Dark Yet, But It's Getting There" (Bob Dylan)

Error on node 0, will try to stop all the nodes
Halting parallel program mdrun on CPU 0 out of 2
-------------------------------------------------------
Program mdrun, VERSION 3.3_beta_20050202
Source code file: ns.c, line: 314

Fatal error:
ns.c, 315: Negative number of short range atoms.
Call your Gromacs dealer for assistance.
-------------------------------------------------------

"It's Not Dark Yet, But It's Getting There" (Bob Dylan)

Error on node 1, will try to stop all the nodes
Halting parallel program mdrun on CPU 1 out of 2

gcq#281: "It's Not Dark Yet, But It's Getting There" (Bob Dylan)

[0] MPI Abort by user Aborting program !
[0] Aborting program!
p0_31546:  p4_error: : -1
    p4_error: latest msg from perror: No such file or directory
[/quote]

also i tried with 4 cpus 2cpus 2cpus round robin always same error.

Md0.log file say:

[quote]
Initializing LINear Constraint Solver
  number of constraints is 94
  average number of constraints coupled to one constraint is 0.5

   Rel. Constraint Deviation:  Max    between atoms     RMS
       Before LINCS         0.010152     87     88   0.004417
        After LINCS         0.000002    352    354   0.000001

Going to use C-settle (653 waters)
wo = 0.888096, wh =0.0559521, wohh = 18.0154, rc = 0.081665, ra = 0.00645837
rb = 0.051255, rc2 = 0.16333, rone = 1, dHH = 0.16333, dOH = 0.1

Constraining the coordinates at t0-dt (step -1)
   Rel. Constraint Deviation:  Max    between atoms     RMS
       Before LINCS         0.000484    195    196   0.000165
        After LINCS         0.000023      1      2   0.000004

Started mdrun on node 0 Mon Feb 14 18:42:01 2005
Initial temperature: 308.342 K
-------------------------------------------------------
Program mdrun, VERSION 3.3_beta_20050202
Source code file: ns.c, line: 314

Fatal error:
ns.c, 315: Negative number of short range atoms.
Call your Gromacs dealer for assistance.
-------------------------------------------------------
[/quote]

Greetings,

Florian

-- 
-------------------------------------------------------------------------------
 Florian Haberl                            Universitaet Erlangen/
 Computer-Chemie-Centrum    Nuernberg
                                                      Naegelsbachstr. 25
                                                      D-91052 Erlangen
  Mailto: florian.haberl AT chemie.uni-erlangen.de
 -------------------------------------------------------------------------------

GFlops

			cpu	d.dppc	d.lzm	d.villin
mc		icc	1	0.48		0.74		1.02
mc-r	r	icc	2	0.71		1.16		1.62
mc		icc	2	0.77		1.31		1.81
mc		icc	4	1.14		1.58		1.91
mc cvs	icc	1	0.60		0.97		1.56

time in min:ss

mc			1	146:27	26:45	2:35
mc-r	r		2	98:19	17:19	1:32
mc			2	96:26	15:18	1:22
mc			4	61:40	13:15	1:18
mc-cvs		1	119:08	20:48	1:48

ps/NODE hour

mc			1	4.09		44.8		248.2
mc-rr		2	6.10		69.3		391.3
mc			2	6.22		78.4		439.0
mc			4	9.73		90.6		461.5
mc-cvs		1	5.04		57.7		331.5

mc-rr means round robin
mc means machine name here
icc Intel compiler as mentioned above




More information about the gromacs.org_gmx-developers mailing list