[gmx-developers] Hands-on Tutorial - Brazil - 04/11 - For David

Erik Marklund erikm at xray.bmc.uu.se
Sun Nov 9 15:38:28 CET 2008

David van der Spoel skrev:
> Erik Marklund wrote:
>> David van der Spoel skrev:
>>> Elton Carvalho wrote:
>>>> Hey there!
>>>> David, what's the probability that gcc4.1 causes Gromacs (or some
>>>> analysis tools) do 2+2=5 or something like that?
>>>> Would this be the reason for the non-reproduction of the results at
>>>> the tutorial sessions in São Carlos earlier this week, since those
>>>> machines had gcc4.1 running?
>>> Good point. I will try it on other machines first, but usually gcc 4 
>>> causes SEGVs. If it is broken on other machines as well we have a 
>>> problem, but probably localized in g_hbond.
>> Hi,
>> Intresting. There WAS a memory issue in g_hbond that involved writing 
>> out of array bounds that cause SEGVs and outright erroneous results. 
>> The fix did go into the development branch before the release of 4.0 
>> and also into the 3.3-patches as far as I know. Which version 
>> generated irreproducible results at the tutorial?
> The current release branch (4.0.1 soon te be 4.0.2) can not reproduce 
> the numbers from my 2006 paper on alcohols. It seems the error is in 
> the ACF rather than in the fitting procedure, as one can do the 
> fitting manually with xmgrace which gives the same result as g_hbond. 
> Hence the underlying data, the ACF must be wrong.
What about the hbond existence matrix? One could use that raw data with 
a bit of coding/scripting to produce one's own ACF, just to double 
check. There has been a fix for a bug where the last occurence of each 
hbond was lost. When hbonds are rare that error could be significant (i 

What about 4.0.0? Can it reproduce the numbers?

Erik Marklund, PhD student
Laboratory of Molecular Biophysics,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,    75124 Uppsala, Sweden
phone:    +46 18 471 4537        fax: +46 18 511 755
erikm at xray.bmc.uu.se    http://xray.bmc.uu.se/molbiophys

More information about the gromacs.org_gmx-developers mailing list