How to write data file for function of two outputs

[Adamszk]

Greetings,
How to write data file (for testing purpose in gold folder) for a function that generates two outputs? It is easy to implement but the real concern is that will the automake file recognize it? Say data file with five columns instead four etc.

[mateunho]

Currently there's only workaround approach, for example see @kfitch's p_sincos test and data file.

PS I'm working on a more flexible test framework, including multivariate/multitype input and output, with minimal changes to existing code. But it's still in development, so if you have any suggestions feel free to share :)

[Adamszk]

Thanks for clues, I will look into that.
I am not familiar with automake programming, but after few months of using it, I am impressed of what it does.  It is amazing tool and whoever did this programming for pal testing, did a good job.   I have some suggestions for improvements or to make it easier for everyone....
1)  In status report - I think code comparison should be based on code, not hardware.  For example, let take a function sin(x). It is a standard function in most c libraries.  Therefore, memory and speed sin(x) comparison should be between standard sin(x) c code and new developed p_sin(x) c code. etc.  No need to compare function that is not standard in c code such ODE functions etc.
2)  if still want to show speed performance based on hardware then at least show some specs of hardware used (description of epiphany core, ARM and x86 cpu used.)  But to me, it is hardly relevant for software developer because it is telling me about hardware performance not software.
3)  I am not sure if that is possible or that I am asking too much from you, the tables of functions listed on readme.md are great.  Its neat. Coloring them by font or highlighting them will help the development team to focus and prioritize.  For example, function p_abs(x) marked green could mean that it is generally accepted as final and there is no need for improvement.  Other developers can look into it as an example to follow.  Yellow could mean that it is good but more improvements can be made.  Red means that needs development.  Blue means someone is working on it so that other developer does have to waste time doing the same thing and to avoid collision between two developers. Colors will help to see a big picture of pal library .
4) I had a case when I modified data gold file, run the test and the function was still using old data file hidden somewhere.  I ended up thinking that I am doing something wrong with the code.  Everytime a make check is made, flush out old object and other files so that make will start with clean library code only.  It is a bit slower check but more thorough, easier and faster for developers in the long run.
5)  Some clues how to do or new functions to add that will measure new code performance or benchmark tool (see 1) every time make check is made.  For example, I made p_sin(x) and did make check... and it pass nicely but it tells me that the new code is slower than standard sin(x) function is c compiler.  It will give developer a better guide when to commit a code and Ola or anyone else does not have to check every new code when it arrives.  Also, benchmark should bypass codes that are not standard.

Other than that, it is a good test framework.  Since I am new in this, I do not want to  unwittingly add the unrealistic expectations from you.  If I will come up with something, I will let you know.