Dear developers,
I encountered a Segmentation fault error (as shown in the picture) when executing the post-processing of exciton distribution using lumen-2.1.0-sp-omp and lumen-2.0.1-sp-omp:
~/codes/lumen-2.0.1-sp-omp/bin/ypp -e w -J GW-final -v all
However, it can be executed successfully when using yambo-5.3.0-sp-omp. What is the reason for this, and how should I solve it?
Looking forward to your reply! Thank you!
xiaomeng
Exciton distribution
Exciton distribution
- Attachments
-
- ab9046b57660450c1fff52361061c200.zip
- (91.94 KiB) Downloaded 276 times
Re: Exciton distribution
Dear Xiaomeng,
the code for plotting the exciton wave-fnuctions has been strongly optimized in the lumen fork.
1) It might be that, by doing so, an OpenMP bug has been introduced.
Can you try to run on a single thread? Just run this in your terminal, before executing ypp
It might be faster than 5.3 in multi-thread mode anyway.
Please, let us know if this works.
2) The other option I see, is that in lumen ypp is using a bit more memory, and the seg-fault is due to memory.
If this is the case, please try the bug-fixes of the lumen fork. I've just introduced a memory optimization commit.
Best,
D.
the code for plotting the exciton wave-fnuctions has been strongly optimized in the lumen fork.
1) It might be that, by doing so, an OpenMP bug has been introduced.
Can you try to run on a single thread? Just run this in your terminal, before executing ypp
Code: Select all
export OMP_NUM_THREADS=1Please, let us know if this works.
2) The other option I see, is that in lumen ypp is using a bit more memory, and the seg-fault is due to memory.
If this is the case, please try the bug-fixes of the lumen fork. I've just introduced a memory optimization commit.
Best,
D.
Davide Sangalli, PhD
Piazza Leonardo Da Vinci, 32, 20133 – Milano
CNR, Istituto di Struttura della Materia (ISM)
https://sites.google.com/view/davidesangalli
Piazza Leonardo Da Vinci, 32, 20133 – Milano
CNR, Istituto di Struttura della Materia (ISM)
https://sites.google.com/view/davidesangalli
Re: Exciton distribution
Dear davide
I followed your instructions: first entered export OMP_NUM_THREADS=1, then ran the command ~/codes/lumen-2.1.0-sp-omp/bin/ypp -F ypp.in -J GW-final, and the exciton distribution plot was generated successfully.
Thank you very much for your guidance.
Xiao meng
I followed your instructions: first entered export OMP_NUM_THREADS=1, then ran the command ~/codes/lumen-2.1.0-sp-omp/bin/ypp -F ypp.in -J GW-final, and the exciton distribution plot was generated successfully.
Thank you very much for your guidance.
Xiao meng
Re: Exciton distribution
Nice. Could you also try the bug-fixes with 96 threads?
Best,
D.
Best,
D.
Davide Sangalli, PhD
Piazza Leonardo Da Vinci, 32, 20133 – Milano
CNR, Istituto di Struttura della Materia (ISM)
https://sites.google.com/view/davidesangalli
Piazza Leonardo Da Vinci, 32, 20133 – Milano
CNR, Istituto di Struttura della Materia (ISM)
https://sites.google.com/view/davidesangalli