Compiling - Atos HPCF and ECS Introduction Tutorial

Reference Documentation

HPC2020: Building your programs

The following exercises will help you understand how to build your own software on the ATOS HPCF or ECS.

Building a simple program

Ensure your environment is clean by running:
```
module reset
```
Create a directory for this tutorial and cd into it:
```
mkdir -p compiling_tutorial
cd compiling_tutorial
```

We are going to use a simple program that will display versions of different libraries linked to it. Create a file called versions.c using your favourite editor with the following contents:

versions.c

#include <stdio.h>
#include <hdf5.h>
#include <netcdf.h>
#include <eccodes.h>

int main() {
    #if defined(__INTEL_LLVM_COMPILER)
        printf("Compiler: Intel LLVM %d\n", __INTEL_LLVM_COMPILER);
    #elif defined(__INTEL_COMPILER)
        printf("Compiler: Intel Classic %d\n", __INTEL_COMPILER);
    #elif defined(__clang_version__)
        printf("Compiler: Clang %s\n",  __clang_version__);
    #elif defined(__GNUC__)
        printf("Compiler: GCC %d.%d.%d\n", __GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__);
    #else
        printf("Compiler information not available\n");
    #endif
    
    // HDF5 version
    unsigned majnum, minnum, relnum;
    H5get_libversion(&majnum, &minnum, &relnum);
    printf("HDF5 version: %u.%u.%u\n", majnum, minnum, relnum); 

    // NetCDF version
    printf("NetCDF version: %s\n", nc_inq_libvers());

    // ECCODES version
    printf("ECCODES version: ");
    codes_print_api_version(stdout);
    printf("\n");

    return 0;
}

Try to naively compile this program with:
```
$CC -o versions versions.c
```
The compilation above fails as it does not know where to find the different libraries. We need to add some additional flags so the compiler can find both the include headers and link to the actual libraries.
1. Let's use the existing software installed on the system with modules, and benefit from the corresponding environment variables *_DIR which are defined in them to manually construct the include and library flags:
```
$CC -o versions versions.c -I$HDF5_DIR/include -I$NETCDF4_DIR/include -I/$ECCODES_DIR/include -L$HDF5_DIR/lib -lhdf5 -L$NETCDF4_DIR/lib -lnetcdf -L$ECCODES_DIR/lib -leccodes
```
  Load the appropriate modules so that the line above completes successfully and generates the versions executable:
  You will need to load the following modules to. have those variables defined.:
  module load hdf5 netcdf ecmwf-toolbox $CC -o versions versions.c -I$HDF5_DIR/include -I$NETCDF4_DIR/include -I/$ECCODES_DIR/include -L$HDF5_DIR/lib -lhdf5 -L$NETCDF4_DIR/lib -lnetcdf -L$ECCODES_DIR/lib -leccodes
  The versions executable should now be in your current directory:
  ls versions
2. Run ./versions. You will get an error such as the one below:
```
$ ./versions
./versions: error while loading shared libraries: libhdf5.so.200: cannot open shared object file: No such file or directory
```
  While you passed the location of the libraries at compile time, the program cannot not find the libraries at runtime. Inspect the executable with ldd to see what libraries are missing
  ldd is a utility that prints the shared libraries required by each program or shared library specified on the command line:
  $ ldd versions linux-vdso.so.1 (0x00007ffffada9000) libhdf5.so.200 => not found libnetcdf.so.19 => not found libeccodes.so => not found libc.so.6 => /lib64/libc.so.6 (0x000014932ff36000) /lib64/ld-linux-x86-64.so.2 (0x00001493302fb000)
3. Can you make that program run successfully?
  While you passed the location of the libraries at compile time, the program cannot not find the libraries at runtime. There are two solutions:
  Use the environment variable LD_LIBRARY_PATH- not recommended for long term
  Use the environment variable LD_LIBRARY_PATH. Check that ldd with the environment variable defined reports all libraries found:
  LD_LIBRARY_PATH=$HDF5_DIR/lib:$NETCDF4_DIR/lib:$ECCODES_DIR/lib ldd ./versions
  Use rpath - robust solution
  Use the rpath strategy to engrave the library paths into the actual executable at link time, so it always knows where to find them at runtime. Rebuild your program with:
  $CC -o versions versions.c -I$HDF5_DIR/include -I$NETCDF4_DIR/include -I/$ECCODES_DIR/include -L$HDF5_DIR/lib -Wl,-rpath,$HDF5_LIB -lhdf5 -L$NETCDF4_DIR/lib -Wl,-rpath,$NETCDF4_DIR/lib -lnetcdf -L$ECCODES_DIR/lib -Wl,-rpath,$ECCODES_DIR/lib -leccodes
  Check that ldd with the environment variable defined reports all libraries found:
  unset LD_LIBRARY_PATH ldd ./versions
  Final version
  For convenience, all those software modules define the *_INCLUDE and *_LIB variables:
  module show hdf5 netcdf4 ecmwf-toolbox | grep -e _INCLUDE -e _LIB
  You can use those in for your compilation directly, with the following simplified compilation line:
  $CC -o versions versions.c $HDF5_INCLUDE $NETCDF4_INCLUDE $ECCODES_INCLUDE $HDF5_LIB $NETCDF4_LIB $ECCODES_LIB
  Now you can run your program without any additional settings:
  $ ./versions Compiler: GCC 8.5.0 HDF5 version: <HDF5 version> NetCDF version: <NetCDF version> of <date> $ ECCODES version: <ecCodes version>
Can you rebuild the program so it uses the "old" versions of all those libraries in modules? Ensure the output of the program matches the versions loaded in modules? Do the same with the latest.
You need to load the desired versions or the modules:
module load hdf5/old netcdf4/old ecmwf-toolbox/old
And then rebuild and run the program:
$CC -o versions versions.c $HDF5_INCLUDE $NETCDF4_INCLUDE $ECCODES_INCLUDE $HDF5_LIB $NETCDF4_LIB $ECCODES_LIB ./versions
The output should match the versions loaded by the modules:
echo $HDF5_VERSION $NETCDF4_VERSION $ECCODES_VERSION
Repeat the operation with
module load --latest hdf5 netcdf4 ecmwf-toolbox

To simplify the build process, let's create a simple Makefile for this program. With your favourite editor, create a file called Makefile in the same directory with the following contents:

Makefile

#
# Makefile
#
# Make sure all the relevant modules are loaded before running make

EXEC = versions

# TODO: Add the necessary variables into CFLAGS and LDFLAGS definition
CFLAGS = 
LDFLAGS = 

all: $(EXEC)

%: %.c 
	$(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS)

test: $(EXEC)
	./$(EXEC)

clean:
	rm -f $(EXEC)

You can test it works by running:

make clean ldd test

Edit the Makefile and add the *_INCLUDE and *_LIB variables which are defined by the modules:

Makefile

#
# Makefile
#
# Make sure all the relevant modules are loaded before running make

EXEC = versions

CFLAGS = $(HDF5_INCLUDE) $(NETCDF4_INCLUDE) $(ECCODES_INCLUDE)
LDFLAGS = $(HDF5_LIB) $(NETCDF4_LIB) $(ECCODES_LIB)

all: $(EXEC)

%: %.c 
	$(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS)

test: $(EXEC)
	./$(EXEC)

ldd: $(EXEC)
	@ldd $(EXEC) | grep -e netcdf.so -e eccodes.so -e hdf5.so 

clean:
	rm -f $(EXEC)

Then run it with:

make clean ldd test

Using different toolchains: prgenv

So far we have used the default compiler toolchain to build this program. Because of the installation paths of the library, it is easy to see both the version of the library used as well as the compiler flavour with ldd:

$ make ldd
        libhdf5.so.200 => /usr/local/apps/hdf5/<HDF5 version>/GNU/8.5/lib/libhdf5.so.200 (0x000014f612b7d000)
        libnetcdf.so.19 => /usr/local/apps/netcdf4/<NetCDF version>/GNU/8.5/lib/libnetcdf.so.19 (0x000014f611f2a000)
        libeccodes.so => /usr/local/apps/ecmwf-toolbox/<ecCodes version>/GNU/8.5/lib/libeccodes.so (0x000014f611836000)

Rebuild the program with:
1. The default GNU GCC compiler.
2. The default Classic Intel compiler.
3. The default LLVM-based Intel compiler.
4. The default AMD AOCC.
Use the following command to test and show what versions of the libraries are being used at any point:
```
make clean ldd test
```
You can perform this test with the following one-liner, exploiting the prgenv module:
for pe in gnu intel intel-llvm amd; do ml prgenv/$pe; make clean ldd test; echo "******************"; done
Pay attention to the following aspects:
- The Lmod module command informs you that it has reloaded the corresponding modules when changing the prgenv. This ensures the libraries used in your program are built with the same compiler for maximum compatibility.
- The compiler command changes automatically, since we are using the environment variable $CC in the Makefile.
- The include and library flags in the compilation lines are adapted automatically based on the libraries loaded.
- The final binary is linked with the corresponding libraries for the version of the compiler as shown by ldd output.

Real-world example: CDO

To put into practice what we have learned so far, let's try to build and install CDO. You would typically not need to build this particular application, since it is already available as part of the standard software stack via modules or easily installable with conda. However, it is a good illustration of how to build a real-world software with dependencies to other software packages and libraries.

The goal of this exercise is for you to be able to build CDO and install it under one of your storage spaces (HOME or PERM), and then successfully run:

<PREFIX>/bin/cdo -V

You will need to:

Familiarise yourself with the installation instructions of this package in the official documentation.
Decide your installation path and your build path.
Download the source code from the CDO website.
Set up your build environment (i.e. modules, environment variables) for a successful build
Build and install the software
Test that works with the command above.

Make sure that CDO is built at least with support to:

NetCDF
HDF5
SZLIB (hint: use AEC)
ecCodes
PROJ
CMOR
UDUNITS

Building in Batch

It is strongly recommended you bundle all your build process in a job script that you can submit in batch. That way you can request additional cpus and speed up your compilation exploiting build parallelism with make -j

If you would like a starting point for such a job, you can start from the following example, adding and amending the necessary bits as needed:

build_cdo.sh

#!/bin/bash
#SBATCH -J build_cdo
#SBATCH -o %x.out
#SBATCH -n 8

set -x
set -e
set -u
set -o pipefail

# Get the URL and VERSION for the latest CDO available
URL=$(curl -s https://code.mpimet.mpg.de/projects/cdo/files | grep attachments/download  | sed -e "s:.*href=\"\(.*\)\".*:https\://code.mpimet.mpg.de\1:" | head -n 1)
VERSION=$(echo $URL | sed -e "s:.*/cdo-\(.*\).tar.gz:\1:")

# TODO: Define installation prefix and build directory
# Hint: Use somewhere in your PERM for installation.
PREFIX=
BUILDDIR=

# Move to our BUILD DIRECTORY
mkdir -p $BUILDDIR
cd $BUILDDIR

# Download source
[ -f cdo-$VERSION.tar.gz ] || wget $URL
[ -d cdo-$VERSION ] || tar xvf cdo-$VERSION.tar.gz
cd cdo-$VERSION

# TODO: Define the environment for the build 


# TODO: Configure the build


# Make sure we start a clean build
make clean

# Build
make -j $SLURM_NTASKS

# Install
make install

# Check installed binary
$PREFIX/bin/cdo -V

We will take a step-by-step approach, taking the build_cdo.sh template above as the starting point.

First thing we need to do is to decide where to install your personal CDO. A good choice would be using your PERM space, and we may use the same structure as the production installation. Because in the script we already have a variable called $VERSION containing the CDO version to install, we can also use that. This way we could have multiple versions of the same package installed alongside should we require it in the future:

# Define installation prefix and build directory
PREFIX=$PERM/apps/cdo/$VERSION

Next comes the decision on what to use for the build itself. Since it is a relatively small build, for performance you might use $TMPDIR which is local to the node.

BUILDDIR=$TMPDIR

However, if you are going to submit it as a batch job, the directory will be wiped at the end. While you are putting the build script together, it may be more practical to have the build directory somewhere that is not deleted after a failed build so you can inspect output files and troubleshoot. As an example, we could pick the following directory in PERM:

BUILDDIR=$PERM/apps/cdo/build

Let's look at the environment for the build. All of dependencies listed above are already available, so we may leverage that by loading all the corresponding modules:

# Define the environment for the build 
module load hdf5 netcdf4 aec ecmwf-toolbox proj cmor udunits

At this point, we need to refer to the installation instructions of this package in the official documentation. We can see it is a classic autotools package, which is typically built with the configure - make - make install sequence. We should then look at the configure --help to see how to enable all the extra options to configure the build:

# Configure the build
./configure --help && exit

Since we are just getting some short help, we can just run the script locally to get the configure output.

bash ./build_cdo.sh

We should inspect the output of the configure help command, and identify what options are to be used:

  --with-szlib=<yes|no|directory> (default=no)  
  --with-hdf5=<yes|no|directory> (default=no)
  --with-netcdf=<yes|no|directory> (default=no)
  --with-udunits2=<directory>
  --with-cmor=<directory> Specify location of CMOR library.
  --with-eccodes=<yes|no|directory> (default=no)
  --with-proj=<directory> Specify location of PROJ library for cartographic
                          projections.

Since all those dependencies are not installed on system paths, we will need to specify the installation directory for each one of them. We may then use the *_DIR environment variables defined by the corresponding modules we load just before.

We will also define where to install the package with the --prefix option. Let's amend the configure line with:

# Configure the build
./configure --prefix=$PREFIX --with-hdf5=$HDF5_DIR --with-netcdf=$NETCDF4_DIR --with-szlib=$AEC_DIR  --with-eccodes=$ECCODES_DIR  --with-proj=$proj_DIR --with-cmor=$CMOR_DIR --with-udunits2=$UDUNITS_DIR

Note that for PROJ, since the variable $PROJ_DIR has a special meaning in the package itself, we must use the lowercase version $proj_DIR.

We are now ready to attempt our first build. Submit it the build script to the batch system with:

sbatch build_cdo.sh

While it builds, we may keep an eye on the progress with:

tail -f build_cdo.out

At this point CDO build and installation should complete successfully, but the execution of the newly installed CDO at the end fails with:

$ grep -v ECMWF-INFO build_cdo.out | tail
make[1]: Leaving directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0/test/pytest'
make[1]: Entering directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
make[2]: Entering directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
make[2]: Nothing to be done for 'install-exec-am'.
make[2]: Nothing to be done for 'install-data-am'.
make[2]: Leaving directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
make[1]: Leaving directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
+ /perm/user/apps/cdo/2.3.0/bin/cdo -V
/perm/user/apps/cdo/2.3.0/bin/cdo: error while loading shared libraries: libproj.so.25: cannot open shared object file: No such file or directory

If we inspect the resulting binary with ldd, we will notice there are a few libraries that cannot be found at runtime:

$ ldd $PERM/apps/cdo/2.3.0/bin/cdo
        linux-vdso.so.1 (0x00007ffc3fbb6000)
        libproj.so.25 => not found
        libeccodes.so => not found
        libcmor.so => /usr/local/apps/cmor/3.7.1/lib/libcmor.so (0x00001484b970d000)
        ...

We are missing the PROJ and ecCodes libraries. We will need to explicitly set RPATHs when we build CDO to make sure the libraries are found. In Autotools packages, and as shown in the configure help we ran earlier, we may pass any extra link flags through the LDFLAGS environment variable. We can amend our build script setting that variable just after loading the dependant modules:

# Define the environment for the build 
module load hdf5 netcdf4 aec ecmwf-toolbox proj cmor udunits

# We will need to explicitly set rpath for proj and eccodes since CDO build system will not
export LDFLAGS="-Wl,-rpath,$proj_DIR/lib -Wl,-rpath,$ECCODES_DIR/lib"

If we submit the the build job again and wait for it to complete, we should see something like:

$ grep -v ECMWF-INFO build_cdo.out | tail -n 25
make[2]: Nothing to be done for 'install-exec-am'.
make[2]: Nothing to be done for 'install-data-am'.
make[2]: Leaving directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
make[1]: Leaving directory '/etc/ecmwf/nfs/dh2_perm_a/user/apps/cdo/build/cdo-2.3.0'
+ /perm/user/apps/cdo/2.3.0/bin/cdo -V
Climate Data Operators version 2.3.0 (https://mpimet.mpg.de/cdo)
System: x86_64-pc-linux-gnu
CXX Compiler: g++ -std=gnu++17 -g -O2 -fopenmp -pthread
CXX version : g++ (GCC) 8.5.0 20210514 (Red Hat 8.5.0-15)
C Compiler: gcc -g -O2 -fopenmp -pthread -pthread
C version : gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-15)
F77 Compiler: gfortran -g -O2
F77 version : GNU Fortran (GCC) 8.5.0 20210514 (Red Hat 8.5.0-15)
Features: 7/503GB 16/256threads c++17 OpenMP45 Fortran pthreads HDF5 NC4/HDF5/threadsafe OPeNDAP sz udunits2 proj cmor sse2                                                                
Libraries: yac/3.0.1 NetCDF/4.9.1 HDF5/1.12.2 proj/9.1.1 cmor/3.7.1
CDI data types: SizeType=size_t
CDI file types: srv ext ieg grb1 grb2 nc1 nc2 nc4 nc4c nc5 nczarr
     CDI library version : 2.3.0
 cgribex library version : 2.1.1
 ecCodes library version : 2.30.2
  NetCDF library version : 4.9.1 of Feb  9 2023 13:54:09 $
    exse library version : 1.5.0
    FILE library version : 1.9.1

For reference, this is the complete job script:

build_cdo.sh

#!/bin/bash
#SBATCH -J build_cdo
#SBATCH -o %x.out
#SBATCH -n 8

set -x
set -e
set -u
set -o pipefail

# Get the URL and VERSION for the latest CDO available
URL=$(curl -s https://code.mpimet.mpg.de/projects/cdo/files | grep attachments/download  | sed -e "s:.*href=\"\(.*\)\".*:https\://code.mpimet.mpg.de\1:" | head -n 1)
VERSION=$(echo $URL | sed -e "s:.*/cdo-\(.*\).tar.gz:\1:")

# Define installation prefix and build directory
PREFIX=$PERM/apps/cdo/$VERSION
#BUILDDIR=$TMPDIR
BUILDDIR=$PERM/apps/cdo/build

# Move to our BUILD DIRECTORY
mkdir -p $BUILDDIR
cd $BUILDDIR

# Download source
[ -f cdo-$VERSION.tar.gz ] || wget $URL
[ -d cdo-$VERSION ] || tar xvf cdo-$VERSION.tar.gz
cd cdo-$VERSION

# Define the environment for the build 
module load hdf5 netcdf4 aec ecmwf-toolbox proj cmor udunits

# We will need to explicitly set rpath for proj and eccodes since CDO build system will not
export LDFLAGS="-Wl,-rpath,$proj_DIR/lib -Wl,-rpath,$ECCODES_DIR/lib"

# Configure the build
./configure --prefix=$PREFIX --with-hdf5=$HDF5_DIR --with-netcdf=$NETCDF4_DIR --with-szlib=$AEC_DIR  --with-eccodes=$ECCODES_DIR  --with-proj=$proj_DIR --with-cmor=$CMOR_DIR --with-udunits2=$UDUNITS_DIR

# Make sure we start a clean build
make clean

# Build
make -j $SLURM_NTASKS

# Install
make install

# Check installed binary
$PREFIX/bin/cdo -V

Space shortcuts

Page tree

Building a simple program

Using different toolchains: prgenv

Real-world example: CDO