↢

programming

Table of Contents

C

Language server

For C programming in emacs, clangd is available on debian and can be installed via: 

apt install clangd

Or with a specific version number as: 

apt install clangd-19

If a library header is not found by clangd, one can define it in the .clangd file placed in the root directory of the project. For example, to include glib.h on debian: 

CompileFlags:
  Add:
  - "-I/usr/include/glib-2.0"
  - "-I/usr/lib/x86_64-linux-gnu/glib-2.0/include"

Tricks

From Klemens (2014): 

go_libs="-lm"
go_flags="-g -Wall -include monster.h -O3"
alias go_c="cc -xc - $go_libs $go_flags"

with monster.h being, for example: 

#include <math.h>
#include <stdio.h>
#include <stdlib.h>

Example usage: 

go_c << "XXXX"
         int main(){ printf("Hello, world.\n"); }
         XXXX

HDF5

h5ls prints information about a file or dataset. 

h5ls <fname.h5>
h5dump

h5dump displays HDF5 file contents.

It’s helpful to inspect the headers of a HDF5 file first. This can be done using the --header flag: 

h5dump --header <fname.h5>

Suppose this gives you something like follows: 

HDF5 "<fname.h5>" {
   GROUP "/" {
   ATTRIBUTE "time" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SCALAR
   }
   DATASET "darcy_velocity.cell.0" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SIMPLE { ( 894740, 1 ) / ( 894740, 1 ) }
   }
   DATASET "darcy_velocity.cell.1" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SIMPLE { ( 894740, 1 ) / ( 894740, 1 ) }
   }
   DATASET "darcy_velocity.cell.2" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SIMPLE { ( 894740, 1 ) / ( 894740, 1 ) }
   }
   }

To display an attribute (such as time), we use the flag -a: 

h5dump -a /time <fname.h5>

This gives: 

HDF5 "<fname.h5>" {
   ATTRIBUTE "time" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SCALAR
   DATA {
   (0): 4.7952e+07
   }
   }
   }

Datasets can be displayed with the -d flag: 

h5dump -d /darcy_velocity.cell.0 <fname.h5>

This will return something like: 

DATASET "/darcy_velocity.cell.0" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SIMPLE { ( 894740, 1 ) / ( 894740, 1 ) }
   DATA {
   (0,0): 7.05594e-08,
   (1,0): 4.07452e-09,
   (2,0): 1.41819e-09,
   (3,0): 1.10755e-09,
   (4,0): 1.01312e-09,
   (5,0): 9.20153e-10,
   ...

We can use the --start and --count flags to look at a slice of the data. For example, to print the first 3 entries of a dataset: 

h5dump -d /darcy_velocity.cell.0 --start="1,0" --count="3,1" <fname.h5>

Gives: 

HDF5 "<fname.h5>" {
   DATASET "/darcy_velocity.cell.0" {
   DATATYPE  H5T_IEEE_F64LE
   DATASPACE  SIMPLE { ( 894740, 1 ) / ( 894740, 1 ) }
   SUBSET {
   START ( 1, 0 );
   STRIDE ( 1, 1 );
   COUNT ( 3, 1 );
   BLOCK ( 1, 1 );
   DATA {
   (1,0): 4.07452e-09,
   (2,0): 1.41819e-09,
   (3,0): 1.10755e-09
   }
   }
   }
   }

See also the man page of h5dump.

Creating HDF5 files with C++

A problem when working with multidimensional arrays is that ISO C++ does not allow runtime initialization of the length. The obvious solution is to dynamically allocate memory for these arrays. However, HDF5 requires continuous chunks of memory that can be directly dumped into the file. Using pointers to pointers does not guarantee this.

Hence, we are stuck with one-dimensional arrays. When flattening a three-dimensional array, we need to do indexing work. This can be done as follows:

Suppose the dimensions of the array A are given by nx, ny, nz. Then, the entry at i, j, k can be accessed by: 

A(i, j, k) = A[i + nx * (j + ny * k)]

Note that this is just computing offsets. Given an entry at i, j, we skip to the entry at i, j+1 by skipping nx * j entries. Now, to skip in a three-dimensional matrix from i, j, k to i, j + 1, k + 1, the offset is nx * j + nx * ny * k. This can be manipulated to give the expression above.

Calling gnuplot from C

Define GNUPLOT as 

#define GNUPLOT "gnuplot -persist"

Plot with gnuplot as 

FILE *fp;
fp = popen(GNUPLOT, "w");
fprintf(fp, "plot ';-' w p lt -1 pt 6 title 'e and z'\n");

for (int i = 0; i &lt; N * (ngp + 2); i ++) {
  fprintf(fp, "%f %f\n", x[i], e[i]);
  fprintf(fp, "\033");

  fflush(fp);
  fclose(fp);
 }

gnuplot specifics

Specify the delimiter character: 

set datafile separator "<char>"

Specify comment character: 

set datafile commentschar "<char>"

The terminal determines how the plot is displayed. I like the terminals dumb, qt, and png. Specify the terminal: 

set term dumb # plots to shell
set term qt   # graphical output
set term png  # plot to image file

The font can be specified through the term. For example 

set term qt font "Times, 12"

or 

set term pngcairo font "monospace"

Allows to plot several graphs in one figure, in a vertically or horizontally stacked way: 

set multiplot layout 2,1 rowsfirst

set title 'surface flow'
plot '1.0e-06/surface_outlet_flux.dat'; u 1:2 w l title 'i = 1.0e-06 m/s'

set title 'subsurface flow'
plot '1.0e-06/subsurface_outlet_flux.dat' u 1:2 w l title 'i = 1.0e-06 m/s'

unset multiplot

Use this for plotting three-dimensional data: 

splot "<file>" u 1:2:3 w l

SERGHEI

Compiling SERGHEI on Debian Trixie

I have pnetcdf installed via package manager. Verify through apt search pnetcdf. It should return something like: 

libnetcdf-pnetcdf-22/stable,now 1:4.9.3-2 amd64  [Installiert,automatisch]
  Schnittstelle für den Zugriff auf wissenschaftliche Daten

libnetcdf-pnetcdf-dev/stable,now 1:4.9.3-2 amd64  [installiert]
  creation, access, and sharing of array-oriented scientific data

libpnetcdf-dev/stable,now 1.14.0-2 amd64  [installiert]
  Development files for the parallel netCDF library

libpnetcdf6/stable,now 1.14.0-2 amd64  [Installiert,automatisch]
  Libraryfor reading and writing parallel NetCDF files

pnetcdf-bin/stable 1.14.0-2 amd64
  Programs for reading and writing parallel NetCDF files

Follow the README and compile Kokkos: 

bash buildKokkos [GPU_ARCHITECTURE]

A good way to discover cmake configuration flags is as follows: 

mkdir build
cd build
cmake ..
cmake -LAH

In the build folder, the integrated surface–subsurface solver in SERGHEI can be compiled using: 

cmake -S ../ -DPnetCDF_C_LIBRARY=/usr/lib/x86_64-linux-gnu/libpnetcdf.so -DPnetCDF_C_INCLUDE_DIR=/usr/include -DSERGHEI_SWE_RE=ON -DSERGHEI_ENABLE_TESTS=ON

Note that if you want to run the test cases, you must enable centred friction discretisation. This requires the flag -DSERGHEI_FRICTION_CENTERED=ON. If no errors are thrown, complete the install: 

cmake --build . -j$(nproc)

$(nproc) uses all available CPUs for parallel compilation.

R

Language server

It's possible to get autocomplete and other IDE-like features in emacs, using lsp-mode. For this, R needs to install its own language server: 

install.packages("languageserver")

impressum

Author: ilhan özgen xian

Created: 2026-03-19 Do 14:59

Validate