Hi all,
I'm attempting to build FreeSurfer with OpenMP enabled.
After checking out a copy of the main dev trunk from the CVS repository, I did the configuration with "--enable-openmp" option and run 'make'. Compilation successfully finished and all of 'make check' tests passed.
Now I have two questions.
(1) Although 'recon-all' finished without error, the results are slightly different from those obtained by the binary distribution (stable v5.1.0). I also find that the message format in a log file (recon-all.log) is not the same as that of the stable version, so I wonder if the source code in the main dev trunk is modified/updated from the stable version... Is it okay to ignore the slight difference of the results?
(2) Despite the openmp option being enabled, it looks like the program runs as a single-thread mode. I actually find the message "NUMBER OF OPENMP THREADS = 1" in a log file. Do I need to specify some option when running 'recon-all' script?
Any suggestion and/or help would be appreciated.
Best, Akio Yamamoto
Akio,
the results of our dev tree will be different from the stable branch (as we are continually incorporating improvements in the dev tree which eventually make their way into the public stable release). The new OpenMP code will also change the results, but we have found not significant differences.
To enable usage of more than one thread (currently just in mri_ca_register), you must add the flag -openmp <numthreads> to the recon-all command line (where numthreads is, say, 8, to use 8 cores).
Nick
On Tue, 2012-06-12 at 18:25 +0900, Akio Yamamoto wrote:
Hi all,
I'm attempting to build FreeSurfer with OpenMP enabled.
After checking out a copy of the main dev trunk from the CVS repository, I did the configuration with "--enable-openmp" option and run 'make'. Compilation successfully finished and all of 'make check' tests passed.
Now I have two questions.
(1) Although 'recon-all' finished without error, the results are slightly different from those obtained by the binary distribution (stable v5.1.0). I also find that the message format in a log file (recon-all.log) is not the same as that of the stable version, so I wonder if the source code in the main dev trunk is modified/updated from the stable version... Is it okay to ignore the slight difference of the results?
(2) Despite the openmp option being enabled, it looks like the program runs as a single-thread mode. I actually find the message "NUMBER OF OPENMP THREADS = 1" in a log file. Do I need to specify some option when running 'recon-all' script?
Any suggestion and/or help would be appreciated.
Best, Akio Yamamoto
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Nick,
I confirmed that the program run in multi-thread mode by adding the "-openmp" flag. Thank you so much for your help.
May I ask a few more questions?
(1) Again, I find minor difference between the two results obtained by running in single-thread and multi-thread mode (say, numthreads is 4 or 8). Is it also possible to ignore this minor difference? Or, this means the program does not function correctly?
(2) Is there any special reason why we can use more than one thread only in 'mri_ca_register'? At this point in time, I would like to speed up 'mri_em_register' as well (and some others programs too, in the future). Any comment or suggestion?
Best regards, Akio
(2012/06/13 2:05), Nick Schmansky wrote:
Akio,
the results of our dev tree will be different from the stable branch (as we are continually incorporating improvements in the dev tree which eventually make their way into the public stable release). The new OpenMP code will also change the results, but we have found not significant differences.
To enable usage of more than one thread (currently just in mri_ca_register), you must add the flag -openmp <numthreads> to the recon-all command line (where numthreads is, say, 8, to use 8 cores).
Nick
On Tue, 2012-06-12 at 18:25 +0900, Akio Yamamoto wrote:
Hi all,
I'm attempting to build FreeSurfer with OpenMP enabled.
After checking out a copy of the main dev trunk from the CVS repository, I did the configuration with "--enable-openmp" option and run 'make'. Compilation successfully finished and all of 'make check' tests passed.
Now I have two questions.
(1) Although 'recon-all' finished without error, the results are slightly different from those obtained by the binary distribution (stable v5.1.0). I also find that the message format in a log file (recon-all.log) is not the same as that of the stable version, so I wonder if the source code in the main dev trunk is modified/updated from the stable version... Is it okay to ignore the slight difference of the results?
(2) Despite the openmp option being enabled, it looks like the program runs as a single-thread mode. I actually find the message "NUMBER OF OPENMP THREADS = 1" in a log file. Do I need to specify some option when running 'recon-all' script?
Any suggestion and/or help would be appreciated.
Best, Akio Yamamoto
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Hi Akio
we are going to work our way through the time-intensive binaries and parallelize them, but it will take us time to get to them.
cheers Bruce On Thu, 14 Jun 2012, Akio Yamamoto wrote:
Hi Nick,
I confirmed that the program run in multi-thread mode by adding the "-openmp" flag. Thank you so much for your help.
May I ask a few more questions?
(1) Again, I find minor difference between the two results obtained by running in single-thread and multi-thread mode (say, numthreads is 4 or 8). Is it also possible to ignore this minor difference? Or, this means the program does not function correctly?
(2) Is there any special reason why we can use more than one thread only in 'mri_ca_register'? At this point in time, I would like to speed up 'mri_em_register' as well (and some others programs too, in the future). Any comment or suggestion?
Best regards, Akio
(2012/06/13 2:05), Nick Schmansky wrote:
Akio,
the results of our dev tree will be different from the stable branch (as we are continually incorporating improvements in the dev tree which eventually make their way into the public stable release). The new OpenMP code will also change the results, but we have found not significant differences.
To enable usage of more than one thread (currently just in mri_ca_register), you must add the flag -openmp <numthreads> to the recon-all command line (where numthreads is, say, 8, to use 8 cores).
Nick
On Tue, 2012-06-12 at 18:25 +0900, Akio Yamamoto wrote:
Hi all,
I'm attempting to build FreeSurfer with OpenMP enabled.
After checking out a copy of the main dev trunk from the CVS repository, I did the configuration with "--enable-openmp" option and run 'make'. Compilation successfully finished and all of 'make check' tests passed.
Now I have two questions.
(1) Although 'recon-all' finished without error, the results are slightly different from those obtained by the binary distribution (stable v5.1.0). I also find that the message format in a log file (recon-all.log) is not the same as that of the stable version, so I wonder if the source code in the main dev trunk is modified/updated from the stable version... Is it okay to ignore the slight difference of the results?
(2) Despite the openmp option being enabled, it looks like the program runs as a single-thread mode. I actually find the message "NUMBER OF OPENMP THREADS = 1" in a log file. Do I need to specify some option when running 'recon-all' script?
Any suggestion and/or help would be appreciated.
Best, Akio Yamamoto
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Aiko,
Per question (1), there will always be slight differences due to the nature of our algorithms, but the thing to do to convince yourself that the differences are not significant is to conduct a group analysis (following the procedures we describe in our tutorials) comparing say 4threads to 8threads. You shouldnt find statistically different results in thickness or structure volumes. Let us know if you do and where.
Nick
Hi Nick,
I confirmed that the program run in multi-thread mode by adding the "-openmp" flag. Thank you so much for your help.
May I ask a few more questions?
(1) Again, I find minor difference between the two results obtained by running in single-thread and multi-thread mode (say, numthreads is 4 or 8). Is it also possible to ignore this minor difference? Or, this means the program does not function correctly?
(2) Is there any special reason why we can use more than one thread only in 'mri_ca_register'? At this point in time, I would like to speed up 'mri_em_register' as well (and some others programs too, in the future). Any comment or suggestion?
Best regards, Akio
(2012/06/13 2:05), Nick Schmansky wrote:
Akio,
the results of our dev tree will be different from the stable branch (as we are continually incorporating improvements in the dev tree which eventually make their way into the public stable release). The new OpenMP code will also change the results, but we have found not significant differences.
To enable usage of more than one thread (currently just in mri_ca_register), you must add the flag -openmp <numthreads> to the recon-all command line (where numthreads is, say, 8, to use 8 cores).
Nick
On Tue, 2012-06-12 at 18:25 +0900, Akio Yamamoto wrote:
Hi all,
I'm attempting to build FreeSurfer with OpenMP enabled.
After checking out a copy of the main dev trunk from the CVS repository, I did the configuration with "--enable-openmp" option and run 'make'. Compilation successfully finished and all of 'make check' tests passed.
Now I have two questions.
(1) Although 'recon-all' finished without error, the results are slightly different from those obtained by the binary distribution (stable v5.1.0). I also find that the message format in a log file (recon-all.log) is not the same as that of the stable version, so I wonder if the source code in the main dev trunk is modified/updated from the stable version... Is it okay to ignore the slight difference of the results?
(2) Despite the openmp option being enabled, it looks like the program runs as a single-thread mode. I actually find the message "NUMBER OF OPENMP THREADS = 1" in a log file. Do I need to specify some option when running 'recon-all' script?
Any suggestion and/or help would be appreciated.
Best, Akio Yamamoto
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
freesurfer@nmr.mgh.harvard.edu
-
Akio Yamamoto -
Bruce Fischl -
Nick Schmansky