new version of mddnmr available

387 views
Skip to first unread message

Vladislav Orekhov

unread,
Oct 1, 2010, 8:52:38 AM10/1/10
to mddnmr
new version of mddnmr program is released and can be downloaded from
http://pc8.nmr.gu.se/~mdd/Downloads/
or
http://dl.dropbox.com/u/12390190/MDD/mddnmr1.7_01Oct2010.tgz

Vladislav
---------------------------------------------------------------------
Vladislav Y. Orekhov
Associate Professor
Swedish NMR Centre at Gothenburg University
Box 465, Goteborg
SE 405 30, Sweden

Vladislav Orekhov

unread,
Jan 30, 2012, 8:27:05 AM1/30/12
to mdd...@googlegroups.com
new version 2.1 of mddnmr program is released and can be downloaded from
http://pc8.nmr.gu.se/~mdd/Downloads/

The new version contains the following improvements:
1. Introduction of threading both for MDD and CS parts. This allows
dramatic speedup of calculations for comuptes with multiple CPU and
Cores. Use parameter MDD_THREADS.
2. For MDD, possibility to introduce residuals in to the reconstructed
spectrum. Parameter MDD_NOISE
3. Several improvements in qMDD
4. Fixing of multiple bugs.

Vladislav
---------------------------------------------------------------------
Vladislav Y. Orekhov

Douglas Kojetin

unread,
Feb 6, 2012, 9:21:32 AM2/6/12
to Vladislav Orekhov, mdd...@googlegroups.com
Hi Vladislav,

Do you know if it is possible to replace the 'mddnmr' binary that comes with Bruker's Topspin 3 software with this newer version? There is no 'mddnmr' binary distributed in the mddnmr2.1.tgz archive, so it doesn't appear to be as easy as replacing the older version distributed with Topspin 3 with your new one.

Thanks,
Doug

> --
> You received this message because you are subscribed to the Google Groups "mddnmr" group.
> To post to this group, send email to mdd...@googlegroups.com rom your registered email address.
> To unsubscribe from this group, send email to mddnmr-un...@googlegroups.com
> For more options, visit this group at http://groups.google.com/group/mddnmr

rharris

unread,
Feb 7, 2012, 12:11:41 PM2/7/12
to mddnmr
Hi Vladislav

I am trying out the new version and whilst the mddsolver works, I get
a segmentation fault with the cssolver (works under 2.0)
I am running on a 64bit Ubuntu computer and the Install gets the
correct bin directory and 'running' just cssolver gives the list of
parameters.
The command that gives the Segmentation fault is:
cssolver ./MDD/region01 CS_alg=IRLS CS_norm=0 CS_niter=10
CS_lambda=1.0 CS_ZF=2 > ./MDD/region01.log

I checked ldd cssolver and all of the required library files are
available.

I also noticed that when running with MDD_THREADS 4, it only appears
to be using 2 (time taken is exactly twice that if I split the
regions.runs file into 4 and run them separately)

any ideas?
thanks
Richard

On Jan 30, 8:27 am, Vladislav Orekhov <vl.orek...@gmail.com> wrote:
> new version 2.1 of mddnmr program is released and can be downloaded fromhttp://pc8.nmr.gu.se/~mdd/Downloads/

Vladislav Orekhov

unread,
Feb 7, 2012, 12:51:11 PM2/7/12
to mdd...@googlegroups.com
Dear Douglas,
Mddnmr updates for Topspin 3 are distributed by Bruker. The updates
typically follow all the developments in the standalone version. None of
the executable distributed from our web site can be used with TopSpin.

Regards,
Vladislav

Vladislav Orekhov

unread,
Feb 8, 2012, 6:37:06 AM2/8/12
to mdd...@googlegroups.com
Hi Richard,
thank you for the bug report. We will try to fix it ASAP. In the
meantime you can use cssolver from the previous version of the program.

Regards,
Vladislav

Vladislav Orekhov

unread,
Feb 17, 2012, 10:33:24 AM2/17/12
to mdd...@googlegroups.com
Hi Rharris,
we fixed the bug with standalone cssolver in 2.1. Please update the
program using qMDD or download and install new version from
http://pc8.nmr.gu.se/~mdd/Downloads/mddnmr2.1.tgz

Regards,
Vladislav

On 2012-02-07 18.11, rharris wrote:

rharris

unread,
Feb 18, 2012, 8:15:07 AM2/18/12
to mddnmr
Hi Vladislav

Thanks for the fix of the cssolver.

Can I ask how to use the MDD_THREADS value? Does this only work within
qMDD? I am running the proc.sh script in a terminal (rather than
through using qMDD) on an i7 (8 core) computer and it doesn't matter
what value I put in for MDD_THREADS it always uses only two of the
cpus (even with MDD_THREADS 1).

thanks
Richard

On Feb 17, 10:33 am, Vladislav Orekhov <vl.orek...@gmail.com> wrote:
> Hi Rharris,
> we fixed the bug with standalone cssolver in 2.1. Please update the
> program using qMDD or download and install new version fromhttp://pc8.nmr.gu.se/~mdd/Downloads/mddnmr2.1.tgz

Vladislav Orekhov

unread,
Feb 18, 2012, 9:55:08 AM2/18/12
to mdd...@googlegroups.com
Hi Richard,
mddnmr always works from script proc.sh . qMDD only prepares the script
and start it.
Threading is organized over sub-regions, so that maximal number of
threads cannot exceed number of sub regions. Number of sub regions is
defined by size of processed region (ROISW) and size of sub region
(SRSIZE). Both parameters are in ppm. Threading works only when
mddnmr4pipeN.sh is used. It may be that you use too small ROISW or too
large SRSIZE, so that you have only 1 or 2 sub-regions. Note that number
of threads shown by command top or in Activity Monitor is 1+n, there n
is number of CS threads.

Current version of standalone cssolver does not thread, but one can run
the solver for individual sub-regions in parallel. In the future we plan
to implement threading in cssolver.


Regards,
Vladislav

Reply all
Reply to author
Forward
0 new messages