To build and run GCHP your compute environment needs the following software:
Make (or GNUMake)
CMake version ≥ 3.13
Compilers (C, C++, and Fortran):
Intel compilers version ≥ 18.0.5, or
GNU compilers version ≥ 8.3
MPI (Message Passing Interface)
OpenMPI ≥ 3.0, or
other MPI libraries might work too
NetCDF (with C, C++, and Fortran support)
ESMF version ≥ 8.0.0
Your system administrator should be able to tell you if this software is already available on your cluster, and if so, how to activate it. If it is not already available, they might be able to build it for you.
If you need to build GCHP’s dependencies yourself, see Building GCHP’s Dependencies.
These are GCHP’s hardware requirements. Note that high-end HPC infrastructure is not required to use GCHP effectively. Gigabit Ethernet and 2 nodes is enough for returns on performance compared to GEOS-Chem Classic.
Recommended Minimum Requirements¶
These recommended minimums are adequate to effectively use GCHP in scientific applications:
2 nodes, preferably ≥24 cores per node
Gigabit Ethernet (GbE) interconnect or better
100 GB memory per node
1 TB of storage
Bare Minimum Requirements¶
These bare minimum requirements are sufficient for running GCHP at C24. The are adequate for trying GCHP out, and for learning purposes.
32 GB of memory
100 GB of storage for input and output data
Big Compute Recommendations¶
These hardware recommendations are for users that are interested in tackling large bleeding-edge computational problems:
A high-performance-computing cluster (or a cloud-HPC service like AWS)
>24 cores per node (the more the better), preferably Intel Xeon
High throughput and low-latency interconnect, preferably InfiniBand if using ≥500 cores
Lots of storage. Several TB is sufficient, but tens or hundreds of TB is better.
General Hardware and Software Recommendations¶
Hyper-threading may improve simulation throughput, particularly at low core counts
MPI process should be bound sequentially across cores and nodes (e.g., a simulation with 48-processes with 24 processes per node should bind rank 0 to CPU L#0, rank 1 to CPU L#1, etc. on the first node, and rank 24 to CPU L#0, rank 1 to CPU L#1, etc. on the second node). This should be the default, but it’s worth checking if your performance is lower than expected. With OpenMPI the –report-bindings argument will show you how processes are ranked and binded.
If using IntelMPI include the following your environment setup to avoid a run-time error:
export I_MPI_ADJUST_GATHERV=3 export I_MPI_ADJUST_ALLREDUCE=12
If using OpenMPI and a large number of cores (>1000) we recommend setting
WRITE_RESTART_BY_OSERVER: YESin config file
GCHP.rc. This enables the MAPL o-server functionality for writing restart files, thereby speeding up the mdoel. This is set automatically when executing