[gmx-developers] npme broken in latest cvs?

[Berk Hess]

Hi,

It is a bug I put in recently.

A fast fix is:
diff -r1.119 pme.c
608c608,609
<       rvec_inc(f[i],atc->f[local_pos]);
---
 >       //rvec_inc(f[i],atc->f[local_pos]);
 >         copy_rvec(atc->f[local_pos],f[i]);
612c613,614
<       rvec_inc(f[i],pme->bufv[buf_index[node]]);
---
 >       //rvec_inc(f[i],pme->bufv[buf_index[node]]);
 >         copy_rvec(pme->bufv[buf_index[node]],f[i]);

I have to think a bit to understand if this is a proper solution (also 
working
single processor) or not.

Berk.


Atte Sillanpää wrote:
> Hi,
>
> I've tried compiling the latest cvs-versions (mdrun version 55) on two 
> separate Opteron clusters using gnu and pgi-compilers but I get 
> crashing code. The cluster that I refer to below has two dual core 
> cpu:s on each node. I've managed to reproduce this with any input 
> files I've tried. See below, for a gmxcheck between a two step dd and 
> dd:npme run.
>
> 1) If I request only domain decomposition, the job runs ok.
> 2) If I specifically ask for npme with the mdrun flag the jobs either 
> stall or crashe (see below).
> 3) If I let mdrun to decide division to dd/npme (no dd/npme flags), 
> the code does not stall, but crashes within some tens of steps as 
> waters can't be settled etc.
>
> With the same configure options (nothing fancy, fftw 3.1.2) I get 
> working code (the source was checked out on 11th April). Any ideas 
> what could be the problem?
>
> Cheers,
> Atte
>
> gmxcheck output between two runs with "-dd 2 2 1" and "-dd 2 1 1 -npme 
> 2":
> ...
> comparing energy file e_pme.edr and e_dd.edr
>
> Opened e_pme.edr as single precision energy file
> Opened e_dd.edr as single precision energy file
> There are 162 terms in the energy files
>
> There are 162 terms to compare in the energy files
>
> Reading energy frame      1 time    0.000           Vir-ZY           
> step   0:       1058.05,  step   0:      1056.45
> Pres-ZY (bar)    step   0:      -95.3608,  step   0:     -95.2099
> Reading energy frame      3 time    0.004           Kinetic En.      
> step   2:       58262.7,  step   2:      57779.1
> Total Energy     step   2:       -362651,  step   2:      -363088
> Temperature      step   2:       322.793,  step   2:      320.114
> Pressure (bar)   step   2:       489.052,  step   2:      148.929
> pV               step   2:       10406.1,  step   2:      3168.91
> Vir-XX           step   2:       14573.3,  step   2:      17932.7
> Vir-XY           step   2:      -2524.57,  step   2:     -2370.18
> Vir-XZ           step   2:       2051.88,  step   2:      1733.56
> Vir-YX           step   2:      -2632.26,  step   2:     -2475.74
> Vir-YY           step   2:       16702.3,  step   2:      19691.3
> Vir-YZ           step   2:       1213.29,  step   2:      675.571
> Vir-ZX           step   2:       1978.46,  step   2:      1733.89
> Vir-ZY           step   2:       1226.37,  step   2:      612.046
> Vir-ZZ           step   2:         11378,  step   2:      15401.7
> Pres-XX (bar)    step   2:       497.122,  step   2:      167.411
> Pres-XY (bar)    step   2:        246.06,  step   2:      232.524
> Pres-XZ (bar)    step   2:      -179.978,  step   2:     -150.756
> Pres-YX (bar)    step   2:       256.182,  step   2:      242.446
> Pres-YY (bar)    step   2:       263.995,  step   2:     -31.6074
> Pres-YZ (bar)    step   2:       -112.03,  step   2:     -61.5908
> Pres-ZX (bar)    step   2:      -173.078,  step   2:     -150.787
> Pres-ZY (bar)    step   2:      -113.259,  step   2:     -55.6198
> Pres-ZZ (bar)    step   2:       706.039,  step   2:      310.982
> #Surf*SurfTen    step   2:       2056.54,  step   2:      1535.89
> Coul-SR:CHOL-CHOL  step   2:   -0.00815666,  step   2:  -0.00819171
> Coul-SR:CHOL-SOL  step   2:      -361.949,  step   2:     -362.405
> Coul-SR:CHOL-Na  step   2:       3.20663,  step   2:      3.20127
> Coul-SR:CHOL-Cl  step   2:    -0.0473165,  step   2:   -0.0468598
> T-DPPC_POPC_POPG_CHOL  step   2:       314.493,  step   2:      313.836
> T-SOL_Na_Cl      step   2:        329.66,  step   2:      325.308
> Last energy frame read 3 time    0.004          Files read succesfully
>
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the www 
> interface or send it to gmx-developers-request at gromacs.org.
>
> This email was Anti Virus checked by Astaro Security Gateway. 
> http://www.astaro.com
>