From: Marcelo C. R. Melo (melomcr_at_gmail.com)
Date: Thu Oct 10 2019 - 17:58:12 CDT
I came across a curious behavior of GCC when compiling NAMD.
After pulling from the gerrit repository and configuring a NAMD version
(tested with multicore, +/-CUDA, +/-MPI, all with the same result), the
compilation process stops with a segmentation fault.
I am using GCC version 8.3.1 from February 23rd, 2019. This seems to me to
be the only real difference between my case and other machines I have used
to successfully compile NAMD, and others who I have contacted.
I just checked with today's version of the source code in the master branch
and the behavior is the same. Previous versions from months ago have the
same issue (I was comparing variations of the QM/MM code).
I am using Charm++ version 6.8.2 also compiled from gerrit.
My compilation consistently breaks at line "src/GoMolecule.C:1456:9",
though I wonder how reproducible this will be.
Has anyone had any issues with this compiler, or can think of something I
am missing here?
I copied the final full error message below.
###### - error message
g++ -m64 -std=c++0x -I/.../gerrit/charm//multicore-linux-x86_64/include
-DCMK_OPTIMIZE=1 -Isrc -Iinc -Iplugins/include -Icolvars/src
-DSTATIC_PLUGIN -I.rootdir/tcl/include -DNAMD_TCL -I.rootdir/fftw/include
FTW -DNAMD_VERSION=\"2.13\" -DNAMD_PLATFORM=\"Linux-x86_64-multicore\"
-DUSE_NODEPATCHMGR -DBONDED_CUDA -DUSE_HOMETUPLES -DUSE_CKLOOP=1 -O3
s -ffast-math -o obj/GoMolecule.o -c src/GoMolecule.C
during GIMPLE pass: recip
src/GoMolecule.C: In member function ‘BigReal
Molecule::get_go_force2(BigReal, BigReal, BigReal, int, int, BigReal*,
src/GoMolecule.C:1456:9: internal compiler error: Segmentation fault
BigReal Molecule::get_go_force2(BigReal x,
This archive was generated by hypermail 2.1.6 : Sat Dec 07 2019 - 23:20:52 CST