We are building up our testing infrastructure and as it currently stands it may still be a little too catered to our specific build environment. If freesurfer builds for you than I would skip the 'make check' stage. The best sanity check would be to make a copy of the bert subject folder, re-recon that subject, and compare that with the original bert shipped with freesurfer

On 02/28/2017 01:10 AM, Francis Tyson Thomas wrote:

Hi,

I have been trying to build the freesurfer dev version with GPU support as per the information I had from my previous try a couple of months ago that Richard help me to successfully complete and I have not been successful in that.

I followed the instructions from https://surfer.nmr.mgh.harvard.edu/fswiki/freesurfer_linux_developers_page

with slight changes,

After running, cd ~/freesurfer ./setup_configure

I edited the configure file that it generated and then commented out, # with_cuda="" and I changed from CUDA_LIBS="-L$CUDA_DIR/lib $LIB_CUDA -lcudart" to CUDA_LIBS="-L$CUDA_DIR/lib64 $LIB_CUDA -lcudart"

then I executed, ./configure --disable-Werror --with-pkgs-dir=/home/username/centos6-x86_64-packages --prefix=/home/username/fs_install --disable-xawplus-apps --disable-tcltk-apps --enable-fermi-gpu --with-cuda="/usr/local/cuda"

Subsequently when I ran make -j4 it worked without any errors. But after that when I run make check I get errors and it fails.

I had sent a previous email regarding the same in another thread but it looks like it has not shown up. So I'm re-sending this now. Infact I have recloned the freesurfer dev version to try it once again and all I get is new errors. Following are the errors now,

Testsuite summary for Freesurfer dev

# TOTAL: 1 # PASS: 0 # SKIP: 0 # XFAIL: 0 # FAIL: 1 # XPASS: 0

# ERROR: 0

See utils/test/mriBuildVoronoiDiagramFloat/test-suite.log Please report to freesurfer@nmr.mgh.harvard.edu

mailto:freesurfer@nmr.mgh.harvard.edu

make[5]: *** [test-suite.log] Error 1 make[5]: Leaving directory `/media/data1/tyson/freesurfer/utils/test/mriBuildVoronoiDiagramFloat' make[4]: *** [check-TESTS] Error 2 make[4]: Leaving directory `/media/data1/tyson/freesurfer/utils/test/mriBuildVoronoiDiagramFloat' make[3]: *** [check-am] Error 2 make[3]: Leaving directory `/media/data1/tyson/freesurfer/utils/test/mriBuildVoronoiDiagramFloat' make[2]: *** [check-recursive] Error 1 make[2]: Leaving directory `/media/data1/tyson/freesurfer/utils/test' make[1]: *** [check-recursive] Error 1 make[1]: Leaving directory `/media/data1/tyson/freesurfer/utils' make: *** [check-recursive] Error 1

If you want the log files, let me know and I'll send it as well.

The idea behind trying to build this was see how freesurfer v6 dev version works with GPU. Also would it be possible to use the GPU binaries of mri_em_register and the other register binary and combine it with the current opnmp implemenation in freesurfer v6?

My objective all along has been to reduce it from the current best performance of 4hrs and 38mins to something even lower like 2hrs or lesser on a per subject basis since the subjects are pushed to the server dynamically.

Thanks, Tyson

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer