Error when offloading to Intel Xeon Phi coprocessors

LAMMPS LAMMPS Mailing List Mirror
1

Hello,

I ran the ‘in.intel.rhodo’ test in USER-INTEL/TEST directory with and without offloading to the Intel MICs. The test runs just fine without offloading, but, when offloading, we get the following error:

……

---------------- Step 0 ----- CPU = 0.0000 (sec) ----------------

TotEng = -407822.9223 KinEng = 343117.3005 Temp = 299.0280

PotEng = -750940.2229 E_bond = 38475.9361 E_angle = 174741.9873

E_dihed = 83388.5840 E_impro = 3416.1854 E_vdwl = -36925.8120

E_coul = 3281502.6969 E_long = -4295539.8006 Press = 26.2604

Volume = 4927920.5360

ERROR on proc 4: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 5: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 7: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 0: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 2: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 3: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 6: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 1: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 14: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 10: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 12: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 9: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 13: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 11: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 8: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

ERROR on proc 15: Out of range atoms - cannot compute PPPM (…/pppm.cpp:1920)

We built the LAMMPS 5Nov2016 version with the Intel 16.1.150 compiler and MKL and IntelMPI 5.1.2.150. We followed the /MAKE/OPTIONS/Makefile.intel_coprocessor make file. Is there something we are missing in the compilation or a potential bug? Please let us know if more information is required.

Thank you,

Amity

2

Can you send the commandline you are using for launch and the coprocessor SKU? Thanks. - Mike

3

Dear Michael,

The command line I use to launch the job over two nodes.

mpirun -n 8 ./lmp_cascade2 -sf intel -pk intel 1 -in in.intel.rhodo

Our system has 1440 nodes (Intel Xeon E5-2670 processor, 16 core/node) with each node having two Intel Xeon Phi 5110P coprocessors.

Thanks,

Amity