icpc: error #10106: Fatal error

[Luke_D_]

Encountered on the devcloud...

icpc: error #10106: Fatal error in /glob/development-tools/versions/intel-parallel-studio/compilers_and_libraries_2019.3.199/linux/bin/intel64/mcpcom, terminated by kill signal

Command argmuments

-Wall -Wextra  -wd1011 -O3   -qopenmp -std=gnu++17 

Attached preprocessed source.

[Luke_D_]

Preprocessed source.

[Viet_H_Intel]

What compiler version is it? 

I can't untar the tar ball.

$ ls -lt Filters.cpp.bz2 
-rw-r--r-- 1 vahoang ome 204825 Dec 13 09:11 Filters.cpp.bz2
$ tar -xvjf Filters.cpp.bz2 
tar: This does not look like a tar archive
tar: Skipping to next header
tar: Exiting with failure status due to previous errors
$

[Luke_D_]

It's not a tarball... it's just a bzip-ed preprocessed source. Try bunzip?

******@login-1:~/cswarm/mrwt/build$ icpc --version
icpc (ICC) 19.0.3.199 20190206
Copyright (C) 1985-2019 Intel Corporation.  All rights reserved.

[Luke_D_]

******@login-1:~/temp$ bunzip2 Filters.cpp.bz2 
******@login-1:~/temp$ icpc -Wall -Wextra  -wd1011 -O3   -qopenmp -std=gnu++17 -c Filters.cpp -o Filters.o
icpc: error #10106: Fatal error in /glob/development-tools/versions/intel-parallel-studio/compilers_and_libraries_2019.3.199/linux/bin/intel64/mcpcom, terminated by kill signal
compilation aborted for Filters.cpp (code 1)
******@login-1:~/temp$

It runs for a long time... maybe it's just running OOM? It's not giving me any information about what kind of signal I'm getting...

[Viet_H_Intel · Accepted solution]

It compiled OK for me.

$ icpc -Wall -Wextra  -wd1011 -O3   -qopenmp -std=gnu++17 -c Filters.cpp -o Filters.o

$ icpc -V
Intel(R) C++ Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 19.0.3.199 Build 20190206
Copyright (C) 1985-2019 Intel Corporation.  All rights reserved.

$ ls -lt Filters.o
-rw-r--r-- 1 vahoang ome 299832 Dec 13 11:48 Filters.o
 

[Luke_D_]

Okay. I guess that it's system specific to intel's devcloud environment... maybe something ulimit based. I'll try and figure out who to contact about that. Thanks

[Viet_H_Intel]

Ok, it took about 2mins to compile, btw

[Luke_D_]

Viet Hoang (Intel) wrote:

Ok, it took about 2mins to compile, btw

Yeah, that's consistent with what I see. We've always had very long compiles (and ICE) with our use of "modern" C++ and icc. Generally I use clang and gcc during development cycles and then go back to icc at the end of a cycle for performance analysis. I'm looking forward to the Next version to see if icc can play a more direct role during development. I can compile on one of the compute nodes that has a newer icc and different ulimits, so I can compile Filter... now I'm just getting a normal ICE elsewhere from EDG.