Several C/C++ and Fortran compilers are available on all NCAR HPC systems. The information on this page applies to all of those systems except where noted.
Compiler commandsAll supported compilers are available via the module utility. After loading the compiler module you want to use, refer to the Compilers available on NCAR systems table below to identify and run the appropriate compilation wrapper command. If your script already includes one of the following generic MPI commands, there is no need to change it:
Build any libraries that you need to support an application with the same compiler, compiler version, and compatible flags used to compile the other parts of the application, including the main executable(s). Also, before you run the applications, be sure you have loaded the same module/version environment in which you created the applications. This will help you avoid job failures that can result from missing MPI launchers and library routines. Compiler man pages To refer to the man page for a compiler, log in to the system where you intend to use it, load the module, then execute man for the compiler. For example:
You can also use -help flags for a description of the command-line options for each compiler. Follow this example:
Where to compileChoose where to compile your code based on where you intend to run your code. Note the types of processors and operating systems as shown in the following table.
The operating systems provide different versions of some standard libraries, which may be incompatible with each other. If your code will run on both Skylake and Cascade Lake nodes on Casper, specify Skylake nodes.
Compile on Derecho if…
Compile on Casper if...
Compile on Cheyenne if…
Changing compilersTo change from one compiler to another, use module swap. In this example, you are switching from Intel to NVIDIA:
When you load a compiler module or change to a different compiler, the system makes other compatible modules available. This helps you establish a working environment and avoid conflicts. If you need to link your program with a library*, use module load to load the library as in this example:
Then, you can invoke the desired compilation command without adding link options such as -l netcdf. Here's an example:
Compiling CPU codeUsing the Cray compiler collectionDerecho users have access to the Cray Compiling Environment (CCE) using the cce module. The compiler collection includes cc, CC, and ftn for compiling C, C++, and Fortran codes. To see which versions of the compiler are available, use the module avail command:
CCE base compilers are available by default in the ncarcompilers module. Loading the ncarcompilers module simplifies building code with dependencies such as netCDF. For example, compiling a simple Fortran code using netCDF is as follows with the compiler wrappers:
Meanwhile, if you did not have the ncarcompilers module loaded, you would need to run the following command instead, with the linker flags and include-paths:
Using Cray MPICH MPIUnlike other MPI libraries, Cray MPICH does NOT provide MPI wrapper commands like mpicc, mpicxx, and mpif90. Rather, use the same cc, CC, and ftn commands you use to compile a serial code. The Cray Programming Environment (CPE) will add MPI build flags to your commands whenever you have the cray-mpich module loaded. As many application build systems expect the MPI wrappers, our ncarcompilers module will translate a call to “mpicc” to “cc” (and likewise for the other languages) as a convenience, typically eliminating the need to alter pre-existing build scripts. Cray MPICH also supports GPU devices. If you are using an MPI application compiled with GPU support, enable CUDA functionality by loading a cuda module and setting or exporting this environment variable before calling the MPI launcher in your job by including this in your script:
Also, if your GPU-enabled MPI application makes use of managed memory, you also need to set this environment variable:
At runtime, you will also need to pass information about job parallelism to the mpiexec (or mpirun / aprun) launcher because this information is not automatically taken from the PBS job script. You can pass this information by setting environment variables or by using mpiexec options. Full details of runtime settings for launching parallel programs can be found by running man mpiexec. The primary settings you will need are:
Cray MPICH has many tunable parameters you can set through environment variables. Run man mpi for a complete listing of these environment variables. Example PBS select statements and corresponding MPI launch options are shown below for binding a hybrid MPI + OpenMP application (144 MPI ranks, and 4 OpenMP threads per MPI rank, which requires 5 nodes but does not fully subscribe the last node). Examples of both methods – setting environment variables and passing options to mpiexec – are provided. Environment variable example
mpiexec options example
Other compilersThese additional compilers are available on Derecho.
When using non-Cray compilers, you can use either the compiler collection’s own commands (e.g., ifort, nvfortran) or the equivalent CPE command (e.g., ftn) as long as you have loaded your desired compiler module. If you do not have the ncarcompilers module loaded and you are using the cray-mpich MPI, you will need to use a CPE command. Compilers available on NCAR systems
Using Intel compilersThe Intel compiler suite is available via the intel module. It includes compilers for C, C++, and Fortran codes. It is NOT loaded by default. To see which versions are available, use the module avail command.
To load the default Intel compiler, use module load without specifying a version.
To load a different version, specify the version number when loading the module. Similarly, you can swap your current compiler module to Intel by using the module swap command.
Extensive documentation for using the Intel compilers is available online here. To review the manual page for a compiler, run the man command for it as in this example:
Optimizing your code with Intel compilersIntel compilers provide several different optimization and vectorization options. By default, they use the -O2 option, which includes some optimizations. Using -O3 instead will provide more aggressive optimizations that may not improve the performance of some programs, while -O1 enables minimal optimization. A higher level of optimization might increase your compile time significantly. You can also disable any optimization by using -O0.
ExamplesTo compile and link a single Fortran program and create an executable, follow this example:
To enable multi-threaded parallelization (OpenMP), include the -qopenmp flag as shown here:
Compiling GPU codeOn Derecho, GPU applications should be built with either the Cray compilers or the NVIDIA HPC SDK compilers and libraries. In the following examples, we demonstrate the use of NVIDIA’s tools. Additional compilation flags for GPU code will depend in large part on which GPU-programming paradigm is being used (e.g., OpenACC, OpenMP, CUDA) and which compiler collection you have loaded. The following examples show basic usage, but note that many customizations and optimizations are possible. You are encouraged to read the relevant man page for the compiler you choose. OpenACCTo compile with OpenACC directives, simply add the -acc flag to your invocation of nvc, nvc++, or nvfortan. A Fortran example:
You can gather more insight into GPU acceleration decisions made by the compiler by adding -Minfo=accel to your invocation. Using compiler options, you can also specify which GPU architecture to target. This example will request compilation for both V100 (as on Casper) and A100 GPUs (as on Derecho):
Specifying multiple acceleration targets will increase the size of the binary and the time it takes to compile the code. OpenMPUsing OpenMP to offload code to the GPU is similar to using OpenACC. To compile a code with OpenMP offloading, use the -mp=gpu flag. The aforementioned diagnostic and target flags also apply to OpenMP offloading.
CUDAThe process for compiling CUDA code depends on whether you are using C++ or Fortran. For C++, the process often involves multiple stages in which you first use nvcc, the NVIDIA CUDA compiler, and then your C++ compiler of choice.
Using the nvcc compiler driver with a non-NVIDIA C++ compiler requires loading a cuda environment module in addition to the compiler of choice. The compiler handles CUDA code directly, so the compiler you use must support CUDA. This means you should use nvfortran. If your source code file ends with the .cuf extension, nvfortran will enable CUDA automatically. Otherwise, you can specify the -Mcuda flag to the compiler.
GPU compilers for CasperTo compile CUDA code to run on the Casper data analysis and visualization nodes, use the appropriate NVIDIA compiler command:
For more information on compiling code on Casper nodes, see: Native commandsWe recommend using the module wrapper commands described above. However, if you prefer to invoke the compilers directly, unload the NCAR default compiler wrapper environment by entering this on your command line:
You can still use the environment variables that are set by the modules that remain loaded, as shown in the following examples of invoking compilers directly to compile a Fortran program. Intel compiler
NVIDIA HPC compiler
GNU compiler collection (GCC)
* In addition to multiple compilers, CISL keeps available multiple versions of libraries to accommodate a wide range of users' needs. Rather than rely on the environment variable LD_LIBRARY_PATH to find the correct libraries dynamically, we encode library paths within the binaries when you build Executable and Linkable Format (ELF) executables. To do this, we use RPATH rather than LD_LIBRARY_PATH to set the necessary paths to shared libraries. This enables your executable to work regardless of updates to new default versions of the various libraries; it doesn't have to search dynamically at run time to load them. It also means you don't need to worry about setting the variable or loading another module, greatly reducing the likelihood of runtime errors. |