k-Wave

1. ZJIANG
   

   Member
   Joined: May '18
   Posts: 3

   Hi,

   I've put kspaceFirstOrder-OMP file into k-wave binaries floder and 'dir(getkWavePath('binaries'))' confirms it (on Linux). When I run the code, 'SaveToDisk' generate input h5 files succefully but I get the error'chmod: cannot access ‘./kspaceFirstOrder-OMP’: No such file or directory' all the time. Could you please help to find out what is going wrong?

   here is my code:

   module load matlab

   cd $PBS_O_WORKDIR
   chmod +x ./kspaceFirstOrder-OMP
   ./kspaceFirstOrder-OMP -i test3.h5 -o testout.h5

   thanks!

   Zheng

   Posted 4 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Zheng,

   Does $PBS_O_WORKDIR point to /k-Wave/binaries/ ? It sounds like you're in the wrong folder.

   Brad.

   Posted 4 years ago #

3. ZJIANG
   

   Member
   Joined: May '18
   Posts: 3

   Hi Bradley,

   Thanks for replying. You're right, I was looking at the wrong folder. After fixing it, I got these:

   ./kspaceFirstOrder-OMP: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.20' not found (required by ./kspaceFirstOrder-OMP)
   ./kspaceFirstOrder-OMP: /lib64/libstdc++.so.6: version `CXXABI_1.3.8' not found (required by ./kspaceFirstOrder-OMP)
   ./kspaceFirstOrder-OMP: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by ./kspaceFirstOrder-OMP)

   What does it mean? Is it because of the matlab script?

   bests,

   Zheng

   Posted 4 years ago #

4. vich
   

   Member
   Joined: Jul '19
   Posts: 19

   What do you get if you run the following?

   strings /lib64/libstdc++.so.6 | grep CXX

   Depending on your Linux distribution, you may have to update a package called gcc, libstdc++, or something like that.

   Posted 4 years ago #

5. ZJIANG
   

   Member
   Joined: May '18
   Posts: 3

   Hi Vich,

   Thanks for helping. It appered to be a problem caused by our server.

   I have another question. While runing a job that requires large memory, I keep getting errors saying exceed cpu numbers. I used these lines:
   export OMP_PLACES={0}:200:1
   export OMP_PROC_BIND=true

   and added '-t 200' after OMP. Do I have to increase the cpu numbers or is there a way to limit the use of cpus?

   thanks,

   Zheng

   Posted 4 years ago #

6. Jiri Jaros
   

   Developer
   Joined: Feb '12
   Posts: 115

   Hi ZJIANG,

   As I understood, you try to execute k-Wave on a cluster.

   - How many cores does a node have? 200 seems to be a huge number. If you don't have a massive NUMA machine, you'd very likely to be limited to 16-64 threads.

   - If you submit a job using PBS you can specify everything using the qsub command

   qsub -l select=1:ncpus=16:ompthreads=16

   This, e.g., asks for one node with 16 cores, one thread per core. The -t parameter is not necessary. I'd also recommend you to display the OMP environment by
   export OMP_DISPLAY_ENV=true

   - Could you post here the node specification and the job script?

   - Answering the previous question, you have to load a GCC module 7.3 or newer.

   Best
   Jiri

   Posted 4 years ago #

RSS feed for this topic

Reply

You must log in to post.