[gmx-developers] May be bug in 64-bit !

[Mark Abraham]

Utkal Ranjan Pradhan wrote:
>  >For reference, the errors actually never occur is nsgrid.c. However, 
> 
>  >when your coordinates become invalid (usually NaN) for some reason 
> 
>  >like horribly bad structure, compiler bug, file IO error, etc., the 
> 
>  >first place we notice is when we try to put them on the grid to do 
> 
>  >neighborsearching.
> 
>  
> 
> Hi
> 
>  
> 
> Thank you all for your valuable inputs, I could manage to fix the rename 
> issue of "temp.top" to "cpeptide.top".
> 
>  
> 
> Can anyone guide me in finding the clue for the *invalid "ci" range* 
> value issue. Probably I am missing out on some important step. I would 
> require your help in getting this issue resolved.

This is just the place where other problems manifest themselves. There's 
very likely nothing wrong with the grid searching code.

> *An Observation:* While testing with the “demo” script, If I’ll pass 
> “10” as runtime parameter (Force Field) to pdb2gmx (it asks for input), 
> at least the FIRST PASS of “mdrun” do not show the range check error. 
> But it raises the same range boundary check error in the 2^nd pass of 
> “mdrun”
> 
>  
> 
>  >>This error comes up all the time on all OS when people haven't 
> followed a sufficiently gentle minimization and equilibration regime<<
> 
> Sorry to say but I am unaware of terms in molecular dynamics so any 
> guidance would be appreciated.

That's at least a six month undergraduate lecture course... :-)

Something's broken, most likely it's with your compilers or libraries. 
Unfortunately nobody seems to have the knowledge or incentive to help 
you with your problem. So far you've made no case whatsoever for the 
need for a working native Win64 port.

Mark