Investigate numerical test failures on Hamilton 8

[danielhollas]

We tried to compile ABIN on the new Hamilton 8 cluster in Durham on node cn7g03, and we're seeing numerical differences in restart.xyz files for the following tests: PIMD, ABINITIO, SPLINE, MORSE, HARMON All these tests are running PIMD with NHC thermostat. We tried GCC-8 and GCC-11.2.

We have verified that the tests pass on Hamitlon7 head node and also on hamilton7:cn7g02, with gcc-8.2. Given that we have the same compiler version and same CPUs, I don't know what the heck is happening.

Here's the kernel version and ldd output on hatmilton8:g03

$ uname -a
Linux [cn7g03.ham8.dur.ac.uk](http://cn7g03.ham8.dur.ac.uk/) 4.18.0-348.20.1.el8_5.x86_64 #1 SMP Thu Mar 10 20:59:28 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

$ ldd bin/abin
	libgfortran.so.5 => /lib64/libgfortran.so.5 (0x00007fc26fd48000)
	libm.so.6 => /lib64/libm.so.6 (0x00007fc26f9c6000)
	libstdc++.so.6 => /lib64/libstdc++.so.6 (0x00007fc26f631000)
	libgomp.so.1 => /lib64/libgomp.so.1 (0x00007fc26f3f9000)
	libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fc26f1e1000)
	libquadmath.so.0 => /lib64/libquadmath.so.0 (0x00007fc26efa0000)
	libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fc26ed80000)
	libc.so.6 => /lib64/libc.so.6 (0x00007fc26e9bb000)
	libz.so.1 => /lib64/libz.so.1 (0x00007fc26e7a3000)
	/lib64/ld-linux-x86-64.so.2 (0x00007fc2701c7000)
	libdl.so.2 => /lib64/libdl.so.2 (0x00007fc26e59f000)

[jack-t-taylor]

Further to the comment above, we are still seeing numerical differences in restart.xyz files for the following tests when trying to compile ABIN on the Hamilton8 cluster (node cn7g03) in Durham: PIMD, ABINITIO, H2O_SCHWENKE_PIMD, HARMON and MORSE. The same is true also for force.xyz files for MORSE.