[gmx-developers] stuff in double precision

hessb at mpip-mainz.mpg.de hessb at mpip-mainz.mpg.de
Mon Feb 5 19:41:27 CET 2007


> Hi,
>
> On Feb 5, 2007, at 7:20 PM, David van der Spoel wrote:
>
>>
>> If you do an operation in C in which a double is involved, then as
>> far as I know everything is converted to double, possibly unless
>> you use brackets.
>>
>
> Formally yes, but modern compilers relax this unless you enforce IEEE-
> conformance at the cost of performance.
>
>> The problem David experienced is just due to an in a sense ill-
>> posed problem, since looking for 1 in 10^6 energy differences (i.e.
>> subtracting large numbers) is asking for trouble (think diverging
>> trajectories due to round off with different average energies).
>> This is why people generally shy away from absolute free energies.
>>
>> A very simple workaround would be to always use the free energy
>> loops, also for "reference no-free energy simulations", then you
>> are running the same code twice.
>
> That might actually be a smarter option. I'll see if I can find my
> once-upon-a-time-smart-optimization to skip the free energy loop when
> lambda==0 and disable it.

These optimizations are long gone.

Berk.






More information about the gromacs.org_gmx-developers mailing list