Run CP2K Docker Application

This guide describes how to run the CP2K Docker application. Go to the Application Configuration screen.

For MI250 queue

Paste the following into the Run Script text box to run the H2O-DFT-LS-NREPS2 benchmark on an 8-GPU node.

CPU_COUNT=96; RANKS=12; GPU_TYPE=MI250;
benchmark H2O-DFT-LS-NREPS2 --arch VEGA90A --gpu-type $GPU_TYPE --rank-stride 8 --omp-thread-count 8 --ranks $RANKS --gpu-count 8 --cpu-count $CPU_COUNT --output /home/aac/H2O-DFT-LS-NREP2-8GPU
For the 32-H2O-RPA benchmark on an 8-GPU node, use:

CPU_COUNT=96; RANKS=12; GPU_TYPE=MI250; benchmark 32-H2O-RPA --arch VEGA90A --gpu-type $GPU_TYPE --rank-stride 8 --omp-thread-count 8 --ranks $RANKS --gpu-count 8 --cpu-count $CPU_COUNT --output /home/aac/32-H2O-RPA-8GPU

## For MI210 queue

Paste the following into the **Run Script** text box to run H2O-DFT-LS-NREPS2 on an 8-GPU node.

CPU_COUNT=128; RANKS=16; GPU_TYPE=MI210; benchmark H2O-DFT-LS-NREPS2 --arch VEGA90A --gpu-type $GPU_TYPE --rank-stride 8 --omp-thread-count 8 --ranks $RANKS --gpu-count 8 --cpu-count $CPU_COUNT --output /home/aac/H2O-DFT-LS-NREP2-8GPU For 32-H2O-RPA on an 8-GPU node:

CPU_COUNT=128; RANKS=16; GPU_TYPE=MI210;
benchmark 32-H2O-RPA --arch VEGA90A --gpu-type $GPU_TYPE --rank-stride 8 --omp-thread-count 8 --ranks $RANKS --gpu-count 8 --cpu-count $CPU_COUNT --output /home/aac/32-H2O-RPA-8GPU

Extract the figure of merit (FOM)

Output files are under /home/aac. To find the FOM for H2O-DFT-LS-NREPS2:

grep "CP2K " /home/aac/H2O-DFT-LS-NREP2-*

Here’s an example output that shows 10.940 as the FOM.

CP2K 1 1.0 0.012 0.029 10.940 10.940

To find the FOM for the 32-H2O-RPA sub-benchmark:

grep -E 'Input.*H2O.*inp|CP2K ' /home/aac/32-H2O-RPA-*