> The profiling results you presented
> at https://mail.gna.org/public/relax-devel/2006-10/msg00133.html
> (Message-id: <1161682551.7703.107.camel@mrspell>) is probably because
> the python profiler is notoriously inconsistent with its timings! I
> get similar inconsistencies when I profiled model-free optimisation on
> test data that I'll soon add to the 'test_suite' branch (synthetic
> data where S2f = 0.952, S2s = 0.582, and ts = 32 ns).
>
I was using the time utility for those timings, not the Python profiler.
This simply measures the the amount of time the relax process runs. I've
been meaning to look into this issue in more detail, but haven't found
time where my computer is sure to be idle for long enough. Whatever the
results mean, they certainly reflect real variation in the time relax
takes to perform the calculation on my machine, they are not just timing
inconsistencies. For instance the variations are supported by the
time-stamps on the various output files from those processes.
The CPU time should really be measured by counting the CPU clock
ticks. The amount of variability in the time utility and the standard
Python profiler would be due to activity of the operating system and
any other running processes. I did a lot of profiling a while back
(2004 I think) to optimise the speed of the calculation of the
model-free equations, relaxation equations, chi-squared equation,
etc., and from my experience timing or using the standard profiler
will drive you insane :P. Especially if you make a change that really
speeds things up (on average) but the first profile timing result says
it took longer!?! I wouldn't get too hung up on the timings.
> Would you also know the status of the 'auto_select',
> 'auto_select_merged', and 'macraild' branches of the repository
> (http://svn.gna.org/viewcvs/relax/branches/)? They haven't been
> touched for over half a year, do you think it would be ok to delete
> them?
Yes, anything in those branches that hasn't made it to the trunk by now
is almost certainly a non-starter. Feel free to delete them.
I'll do it tomorrow (if they're still there :). They can always be
recovered if necessary, they will always exist within the repository
at the revision of the deletion (minus one).
Edward
Re: apply