Sarah
I am a little confused. Are you saying that recon-all has a bug, or that
super-recon has a bug?
If you are saying recon-all has a bug, you should start a new thread so
others who don't care about "faster and parallel..." are aware.
If super-recon is screwed up, please tell me what you think needs to be
added/changed, or send a code snippet.
Thanks
Daniel
[EMAIL PROTECTED] wrote:
hi there,
i'm actually working on parallelizing recon-all on a grid...i
*think* the script won't work all the way thru bcs calling
recon-all with the -normalization2 flag is supposed to call
both mri_normalize AND mri_mask (according to the DevTable);
but it looks as if the recon-all script only calls
mri_normalize with that flag...this means mri_mask is never
called and brain.finalsurfs.mgz is not produced...so when you
get to -finalsurfs it can't find the file...alas
i just changed the recon-all script to call mri_mask when it
gets the -norm2 flag...this seems to have worked.
sarah
---- Original message ----
Date: Wed, 22 Mar 2006 18:09:11 -0500
From: Daniel Goldenholz <[EMAIL PROTECTED]>
Subject: Re: [Freesurfer] Faster and Parallel recon-all
To: "Goradia, Dhruman D" <[EMAIL PROTECTED]>
Cc: freesurfer@nmr.mgh.harvard.edu
Hi again,
I have totally revamped the script in light of the fact that
recon-all
-notify did not do what I thought it did at all!
The new and improved (version 1.02) super-recon is quite a
bit fancier.
Now, I leave all notifcations and waiting and emailing to
super-recon
itself, so that recon-all assumptions are reduced.
If you want to run a complete recon-all -all but have the
hemisphere
parts run in parallel, you run a command like this one:
super_recon all $SUBJECTS_DIR $SUBJECT /space/dicom/file1
/space/dicom/file2 [EMAIL PROTECTED]
Using this new and improved command, if you have a cluster
with daul
processors, then you should be able to run multiple
subjects faster than a standard recon-all command.
The script in "all" mode will run the same stuff as recon-
all -all.
If you wish to run a subset of the full set of commands, you
can simply
comment out a specific line you don't want.
There are also several other ways to call the script now,
since in
reality it calls itself recursively with different options.
Hopefully this one will do what the original one intended to
do.
Good luck,
Daniel
Goradia, Dhruman D wrote:
Hi Daniel,
The script is ok but it still does not make it to the end.
As the script
is designed to send me an email at the end of the last
step, I got an
email within 15 min after I executed the script saying
that "recon-all
14777 finished with errors at Tue Mar 21 18:46:15 EST
2006". However it
went on the complete the "gcareg" step.
I have attached the status and log files for your
reference. If you get
a chance have a look at it. I believe this problem is not
due to
super-recon script but due to recon-all script. Please let
me know.
I am very interested in the super-recon script, as it takes
about 30 hrs
to process 1 scan currently.
-Dhruman
-----Original Message-----
From: Daniel Goldenholz [mailto:[EMAIL PROTECTED]
Sent: Tuesday 21 March 2006 17:09
To: Goradia, Dhruman D
Cc: freesurfer@nmr.mgh.harvard.edu
Subject: Re: [Freesurfer] Faster and Parallel recon-all
Dhruman
Thanks for catching that. Spelling was never my strong
point. I have
corrected the script on my website.
As for you second problem, did you properly remove the
waitfor
step1-00.done from the motion correction step, since you
say you skipped step1-00, i.e. the invol step? I mention
this in a
comment near the top of the script.
D
Goradia, Dhruman D wrote:
Hi Daniel,
I was having a second look at the script and apparently
the script is
fine with a small spelling error. The STAT variable which
is set at 3
different instances has spelling error in "scripts"
where "t" is
missing.
Once this was corrected the script is fine but it's still
not working.
It initiates all the processes and then just sits idle.
After waiting
for half an hour it still does not make step1-00.done (I
start directly
with motion cor) keeping all the other processes waiting.
I am still
probing into this. If you can think of something that
might be of help
then let me know.
-Dhruman
-----Original Message-----
From: Daniel Goldenholz [mailto:[EMAIL PROTECTED]
Sent: Tuesday 21 March 2006 11:25
To: Goradia, Dhruman D
Subject: Re: [Freesurfer] Faster and Parallel recon-all
Dhruman
That's confusing. I think you should take one line of
executable code
from your script that produces that behavior and then
email it to the
freesurfer list.
It sounds to me like the recon-all script ASSUMES that the
log file
exists if you specify a name. I thought it actually
creates one first,
but I could have been mistaken there, and if so I will
need to add a
few
extra lines of code.
PLEASE follow up on this. It will help others.
The other thing to try is to create a text file first,
such as like
this:
echo "#Starting recon-all manually..." >
/Applications/freesurfer/subjects/ddg6/scripts/recon-all-
status.log
and then see if my script is ok.
Keep me posted.
D
Goradia, Dhruman D wrote:
Hi Daniel,
I tried to use the script super-recon on mac OX S using
bash shell. I
made the necessary changes to make it operable in bash.
However when I
run the script its gives me the following error:
Tcsh:/Applications/freesurfer/subjects/ddg6/scripts/recon-
all-status.l
o
g
:not a file or directory.
I tried to run the first 3 steps of the script (motioncor,
nuintensitycor, talairach) and it gives the above error.
However if I
remove the log and status flags, the first 3 steps
executes properly.
Do
you have any idea why? Do you see any problem if I delete
the log and
status flags from the whole script and use it?
-Dhruman
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf
Of Daniel
Goldenholz
Sent: Thursday 16 March 2006 15:17
Cc: Freesurfer Mailing List; Nick Schmansky
Subject: Re: [Freesurfer] Faster and Parallel recon-all
Ah. Silly me, I didn't notice the -log and -status
options. Thanks
Nick.
I have now included those into my super-recon script as
well. Problem
solved.
(if you have seen my script already, you may need to hit
RELOAD on
your
browser to see the new version)
Daniel
Nick Schmansky wrote:
The script looks good. The one thing it needs though
is for the
right
and left hemi steps to log to separate files, otherwise
output to
recon-
all.log will be interleaved, making debugging
impossible.
On Thu, 2006-03-16 at 12:15 -0500, Daniel Goldenholz
wrote:
Ooops. I sent the below with the wrong subject tag.
Here it is
again:
Daniel Goldenholz wrote:
Freesurfers
I was hoping someone else would do this, but well, it
seems that
no
one did. So I went ahead and made this myself.
Amusingly, I don't NEED to reconstruct any brains
right now, so I
don't have anything to test this out with.
Please someone try it, and let me know if it is
helpful.
What I did was write a script that runs each and
every part of the
recon-all -all stream, one command at a time. (based
on the NEW
version 3 stream).
Each one is loaded with -waitfor and -notify so that
the
dependency
structure is intact.
However, ALL of the commands start running in the
background! If
you have a 2 more processor system, then recon-all
should complete
in about HALF the time as it usually does, since all
the
hemisphere
specific stuff can be done in parallel.
At the completion of each hemipshere, it will email
you.
I included a command for each minute step so that
people can
easily
fiddle with options and commands and so forth.
Obviously, if you run recon-all -autoreon1, then you
can then run
my script (after deleting the step1-??.done lines).
If you run recon-all -autorecon2, then you can run my
script
(after
deleting the step1-??.done and step2-??.done lines)
And so forth.
To see the script, go here:
http://www.nmr.mgh.harvard.edu/~daniel
click on fMRI scripts
click on super-recon
Enjoy, and PLEASE email me if you use this. If there
are
corrections I can easily post them.
Daniel
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfe
r
--
Daniel Goldenholz
-----------------------------------------------------
Cell: 617-935-9421 http://people.bu.edu/danielg/
"Mavet v'chaim b'yad halashon"
"Life and death are in the hands of language"
- Proverbs 18:21
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
--
Daniel Goldenholz
-----------------------------------------------------
Cell: 617-935-9421 http://people.bu.edu/danielg/
"Mavet v'chaim b'yad halashon"
"Life and death are in the hands of language"
- Proverbs 18:21
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
