Compiling CASINO on Bluegene/Cetus

General discussion of the Cambridge quantum Monte Carlo code CASINO; how to install and setup; how to use it; what it does; applications.
Post Reply
Kayahan
Posts: 22
Joined: Fri Jun 07, 2013 5:56 pm

Compiling CASINO on Bluegene/Cetus

Post by Kayahan »

Dear CASINO users,

I have been trying to compile CASINO on cetus.alcf.anl.gov with the arch file (bluegene-xlf-cobalt-parallel.cetus) provided in CASINO v2.13.569. However at the end of the compilation I get the following error, as you can see in the attached document.

Nevertheless, I tried using the executable using the runqmc command with various options:

i=4 8 16 32 64 and j=1 2 4 8

/home/kayahan/SOFTWARE/CASINO/bin_qmc/utils/bluegene-xlf-cobalt-parallel.cetus/runqmc -n 128 --ppn=$i --tpp=$j --user.shmemsize=2000 -N test -T 1h

This is a LiCoO2 2x1x1 supercell and blip file is around 0.8 gb. I set --user.shmemsize=2000 as I would eventually work with larger supercells. Even the first step in vmc_opt estimates 1-2 days for completion while increasing $i makes it even worse. Same calculation I do on our old local cluster with the CASINO I compiled, takes 7 hours using 48 processors only.

I am having a hard time understanding why this is happening, any recommendation would be greatly helpful.

Best regards,
Kayahan
Attachments
output.sh.tar
(30 KiB) Downloaded 881 times
Kayahan
Posts: 22
Joined: Fri Jun 07, 2013 5:56 pm

Re: Compiling CASINO on Bluegene/Cetus

Post by Kayahan »

Ok, let me answer to myself... I think decreasing the optimization level of the compilation can give healthy compilation, but I am not sure how that is going to affect overall performance of the code. Nevertheless, decreasing from "-O4 -qipa" to "-O3 -g" in the cetus arch file made things work. Besides, I found out that CASINO's internal script runs parallel, which causes problems in getting the code running. When compiled in parallel, the compilation works fine, but when a job is submitted, it causes problems in reading input files such as correlation.data. So after modifying the arch file as above and setting CASINO_ARCH by hand:

export CASINO_ARCH=bluegene-xlf-cobalt-parallel.cetus

then running make -j 1 OpenmpShm in the CASINO home directory gets the code compiled and running without a problem.

Kayahan
Post Reply