Top Page > Browsing
Instability of a function Band_DFT_NonCol in the case of GCC
Date: 2021/08/02 20:56
Name: Naoya Yamaguchi

Dear all,

I often encountered several noncollinear calculation cases with segmentation faults in Ver. 3.9.2 in using GCC, and I was able to avoid them by suppressing the optimization level for a function `Band_DFT_NonCol`.

The way is wrapping the function with pragma statements, as you may know.
#pragma GCC optimize("O2")
double Band_DFT_NonCol(
...
#pragma GCC reset_options

My makefile included the following:
CC = mpicc -Dkcomp -O3 -ffast-math -march=znver2 -mfma -fomit-frame-pointer -fopenmp -I${MKLROOT}/include/fftw
FC = mpif90 -Dkcomp -O3 -ffast-math -march=znver2 -mfma -fomit-frame-pointer -fopenmp
LIB= -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -lmkl_gf_lp64 -lmkl_gnu_thread -lmkl_core -lgfortran -lpthread

You can see my modification in
https://gist.github.com/Ncmexp2717/6355db544fecd711725fa5958b2959cf
The above also includes some treatment to avoid issues reported in the following.
http://www.openmx-square.org/forum/patio.cgi?mode=view&no=2728
http://www.openmx-square.org/forum/patio.cgi?mode=view&no=2747

Note: pragma statements of the Intel style (#pragma optimization_level 1) doesn't seem to work properly with GCC.

Regards,
Naoya Yamaguchi
メンテ
Page: [1]

Re: Instability of a function Band_DFT_NonCol in the case of GCC ( No.1 )
Date: 2021/08/07 12:07
Name: T. Ozaki  <t-ozaki@issp.u-tokyo.ac.jp>

Hi,

Thank you very much for your report.
I will take account of the pragma statement in an upcoming patch.

Regards,

TO
メンテ

Page: [1]

Thread Title (must) Move the thread to the top
Your Name (must)
E-Mail (must)
URL
Password (used in modification of the submitted text)
Comment (must)

   Save Cookie