GNU Astronomy Utilities



2.1.7 Building custom programs with the library

In Cosmological coverage and visualizing tables, we repeated a certain calculation/output of a program multiple times using the shell’s for loop. This simple way of repeating a calculation is great when it is only necessary once. However, if you commonly need this calculation and possibly for a larger number of redshifts at higher precision, the command above can be slow. Please try it out by changing the sequence command in the previous section to ‘seq 0.1 0.01 10’. It will take about 11 seconds32! This can be improved by hundreds of times! This section will show you how.

Generally, repeated calls to a generic program (like CosmicCalculator) are slow, because a generic program can have a lot of overhead on each call. To be generic and easy to operate, CosmicCalculator has to parse the command-line and all configuration files (see Option management and configuration files) which contain human-readable characters and need a lot of pre-processing to be ready for processing by the computer. Afterwards, CosmicCalculator has to check the sanity of its inputs and check which of its many options you have asked for. All the this pre-processing takes as much time as the high-level calculation you are requesting, and it has to re-do all of these for every redshift in your loop.

To greatly speed up the processing, you can directly access the core work-horse of CosmicCalculator without all that overhead by designing your custom program for this job. Using Gnuastro’s library, you can write your own tiny program particularly designed for this exact calculation (and nothing else!). To do that, copy and paste the following C program in a file called myprogram.c.

#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <gnuastro/cosmology.h>

int
main(void)
{
  double area=4.03817;          /* Area of field (arcmin^2). */
  double z, adist, tandist;     /* Temporary variables.      */

  /* Constants from Plank 2018 (arXiv:1807.06209, Table 2) */
  double H0=67.66, olambda=0.6889, omatter=0.3111, oradiation=0;

  /* Do the same thing for all redshifts (z) between 0.1 and 5. */
  for(z=0.1; z<10; z+=0.01)
    {
      /* Calculate the angular diameter distance. */
      adist=gal_cosmology_angular_distance(z, H0, olambda,
                                           omatter, oradiation);

      /* Calculate the tangential distance of one arcsecond. */
      tandist = adist * 1000 * M_PI / 3600 / 180;

      /* Print the redshift and area. */
      printf("%-5.2f %g\n", z, pow(tandist * 60,2) * area / 1e6);
    }

  /* Tell the system that everything finished successfully. */
  return EXIT_SUCCESS;
}

Then run the following command to compile your program and run it.

$ astbuildprog myprogram.c

In the command above, you used Gnuastro’s BuildProgram program. Its job is to simplify the compilation, linking and running of simple C programs that use Gnuastro’s library (like this one). BuildProgram is designed to manage Gnuastro’s dependencies, compile and link your custom program and then run it.

Did you notice how your custom program created the table almost instantaneously? Technically, it only took about 0.03 seconds! Recall that the for loop of Cosmological coverage and visualizing tables took more than 11 seconds (or \(\sim367\) times slower!).

Please run the ls command to see a listing of the files in the current directory. You will notice that a new file called myprogram has been created. This is the compiled program that was created and run by the command above (its in binary machine code format, not human-readable any more). You can run it again to get the same results by executing it:

$ ./myprogram

The efficiency of your custom myprogram compared to repeated calls to CosmicCalculator is because in the latter, the requested processing is comparable to the necessary overheads. For other programs that take large input datasets and do complicated processing on them, the overhead is usually negligible compared to the processing. In such cases, the libraries are only useful if you want a different/new processing compared to the functionalities in Gnuastro’s existing programs.

Gnuastro has a large library which is used extensively by all the programs. In other words, the library is like the skeleton of Gnuastro. For the full list of available functions classified by context, please see Gnuastro library. Gnuastro’s library and BuildProgram are created to make it easy for you to use these powerful features as you like. This gives you a high level of creativity, while also providing efficiency and robustness. Several other complete working examples (involving images and tables) of Gnuastro’s libraries can be see in Library demo programs.

But for this tutorial, let’s stop discussing the libraries here and get back to Gnuastro’s already built programs (which do not need C programming). But before continuing, let’s clean up the files we do not need any more:

$ rm myprogram* z-vs-tandist*

Footnotes

(32)

To measure how much time the loop of Cosmological coverage and visualizing tables takes on your system, you can use the time command. First put the whole loop (and pipe) into a plain-text file (to be loaded as a shell script) called z-vs-tandist.sh. Then run this command: time -p bash z-vs-tandist.sh. The relevant time (in seconds) is shown after real.