Trouble with Intel compilers after recent commits

[wcwitt]

I wasn't able to build with Intel today, failing at cutil.c. I don't recall the full error (sorry), but it worked again after I reverted to aadb4a0a10b90601735c9a1439e8f04bd8abb70b.

[albapa]

This must be because of my md5sum commit. Which intel compiler did you use?

[albapa]

Can you please submit some more details? With the 2019 intel suite it compiles fine for me.

[wcwitt]

Sure - just ran through it again:

git clone --recursive https://github.com/libAtoms/QUIP.git
cd QUIP
export QUIP_ARCH=linux_x86_64_ifort_icc
make config
[lapack/blas: -L${MKLROOT}/lib/intel64 -lmkl_rt -lpthread -lm -ldl]
[n to preconditioned minimisation]
[y to GAP]
make

with

$ ifort -v
ifort version 19.1.2.254
$ icc -v
icc version 19.1.2.254 (gcc version 4.8.5 compatibility)

The trouble for me does begin in the neighborhood of the md5 stuff

icc -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -isystem /home/wcw28/.conda/envs/calphy2/include -fPIC -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/fox/objs.linux_x86_64_ifort_icc/finclude -O3 -g -traceback -DDUMP_CORE_ON_ABORT -DGETARG_F2003 -DFORTRAN_UNDERSCORE -D'GIT_VERSION="https://github.com/libAtoms/QUIP.git,6f2ad16d1"' -D'GAP_VERSION=1651824457' -D'QUIP_ARCH="linux_x86_64_ifort_icc"' -D'SIZEOF_FORTRAN_T=2' -DHAVE_GAP -DHAVE_QR  -fPIC -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/fox/objs.linux_x86_64_ifort_icc/finclude -O3 -g -traceback -DDUMP_CORE_ON_ABORT -DGETARG_F2003 -DFORTRAN_UNDERSCORE -D'GIT_VERSION="https://github.com/libAtoms/QUIP.git,6f2ad16d1-dirty"' -D'GAP_VERSION=1651824457' -D'QUIP_ARCH="linux_x86_64_ifort_icc"' -D'SIZEOF_FORTRAN_T=2' -DHAVE_GAP -DHAVE_QR  -c /home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/md5.c
icc: command line warning #10006: ignoring unknown option '-fno-plt'
icc: warning #10193: -vec is default; use -x and -ax to configure vectorization
icc -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -isystem /home/wcw28/.conda/envs/calphy2/include -fPIC -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/fox/objs.linux_x86_64_ifort_icc/finclude -O3 -g -traceback -DDUMP_CORE_ON_ABORT -DGETARG_F2003 -DFORTRAN_UNDERSCORE -D'GIT_VERSION="https://github.com/libAtoms/QUIP.git,6f2ad16d1"' -D'GAP_VERSION=1651824457' -D'QUIP_ARCH="linux_x86_64_ifort_icc"' -D'SIZEOF_FORTRAN_T=2' -DHAVE_GAP -DHAVE_QR  -fPIC -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms -I/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/fox/objs.linux_x86_64_ifort_icc/finclude -O3 -g -traceback -DDUMP_CORE_ON_ABORT -DGETARG_F2003 -DFORTRAN_UNDERSCORE -D'GIT_VERSION="https://github.com/libAtoms/QUIP.git,6f2ad16d1-dirty"' -D'GAP_VERSION=1651824457' -D'QUIP_ARCH="linux_x86_64_ifort_icc"' -D'SIZEOF_FORTRAN_T=2' -DHAVE_GAP -DHAVE_QR  -c /home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/cutil.c
icc: command line warning #10006: ignoring unknown option '-fno-plt'
icc: warning #10193: -vec is default; use -x and -ax to configure vectorization
/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/cutil.c(170): error: expected an expression
        for(int i = 0; i < 16; i++)
            ^

/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/cutil.c(170): error: identifier "i" is undefined
        for(int i = 0; i < 16; i++)
                       ^

/home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/cutil.c(273): warning #810: conversion from "void *" to "int" may lose significant bits
    return ((int) p);
           ^

compilation aborted for /home/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/src/libAtoms/cutil.c (code 2)
make[1]: *** [cutil.o] Error 2
make[1]: Leaving directory `/rds/user/wcw28/hpc-work/nonhydrostatic/codes/cclake/test/QUIP/build/linux_x86_64_ifort_icc'
make: *** [libAtoms] Error 2

[jameskermode]

Looks like moving the definition of int i outside the scope of the for loop and cleaning up the case should fix this.

This is why it would be good to have Intel compilers in the CI, hence #329, but unfortunately we didn't get it working.

[albapa]

Yes - the intel compiler should really interpret the code according to the 11 (or even 99) standard, which allows this syntax. But I agree with James that it's best moving the declaration outside, hence the pull request. @wcwitt if you can confirm that it compiles for you, I'll close this issue.

[wcwitt]

https://github.com/libAtoms/QUIP/pull/442 compiles for me. Thanks. I confess I thought it must be something else.