[gmx-users] eneconv and timestamps

[Matthias Ernst]

Hi,

I don't understand the behaviour of eneconv(_d) in Gromacs 4.6.7
I have a long trajectory saved every .5ps in single precision and want
to save it every ps.
Unfortunately, there is no -skip option. So, I figured with your help
some time ago to first save the edr file in double precision using
eneconv_d (as using single precision, timestamps get wrong after
>16.7something mus but double precision works) and then use use -dt 1
when the timestamps are correct to omit the wrong frames.

Strangely, a gmxcheck on either the single or the double precision file
with .5ps timestamp show no errors. But, when I use the -dt 1 option, I
get errors like
Reading energy frame 308000 time 154000.000         Writing frame time
154000
WARNING: missing energy sums at time 154065.000000
Reading energy frame 2067000 time 1033500.000         Writing frame time
1.033e+06
WARNING: missing energy sums at time 1033800.000000
Reading energy frame 2675000 time 1337500.000         Writing frame time
1.337e+06
WARNING: missing energy sums at time 1337947.000000
Reading energy frame 3899000 time 1949500.000         Writing frame time
1.949e+06
WARNING: missing energy sums at time 1949972.000000

When I do a gmxcheck on the resulting energy file, again no errors are
reported. Also, when I just run eneconv_d on the resulting energy file
and save it as new file without doing anything else or using again -dt
1, it works.

Why are there errors using -dt 1 when before, just converting from
single to double precision at .5ps timestamp, there were none?
If it was corrupt data, shouldn't the problem arise at the first step?
Why are there no more errors when I further process the frame? Are the
energy sums somehow re-calculated?

Thanks,
Matthias


-- 
Progress... http://www.phdcomics.com/comics/archive.php?comicid=1611