Choosing the optimal distribution for Code_aster, Octave, Scilab?

Question

linux

So, I'm starting to work in a new place where there are no licenses for FEM ( ANSYS , Abaqus , etc.) programs and Matlab , which are the main weapon of an engineer (after Excel and a calculator, of course). Therefore, I will use free analogues: Code_aster , Octave and Scilab .
There is a special CAElinux distribution, which is “sharpened” for these tasks, and these programs are already preinstalled. But, judging by the rumors, there will be no more distribution updates (but it will be relevant for another year or two). On the other hand, as I understand it, using a ports system like Gentoo gives a performance boost, which is extremely important for finite element analysis, where for large calculations, a performance increase of 10-15% can be worth several hours of calculation time savings.
Which begs the question, is it worth the hassle (installing Gentoo and compiling programs from source) for a person with basic knowledge of Linux, and roughly how much performance gain should I expect?
Expected working computer configuration: i7-3740QM, 16Gb RAM, 256Gb SSD.

Answer 1

With a basic knowledge of Linux, when compiling Gentoo, not only can you not get a noticeable gain, but vice versa. 10-15% is actually very optimistic even for experienced people. A more realistic figure is not higher than 10%, depending on the tasks.
Better install CAElinux now and start learning Gentoo. Until CAElinux becomes irrelevant, you will already have far from basic knowledge and will do everything you need, the way you need it.

Answer 2

I'm too lazy to give an exact link (and yes Google will help you), but a couple of years ago, comparative tests of free distributions for highly loaded scientific computing were carried out. CentOs was recognized as the fastest (by the way, your CAElinix is ​​based on it). Since I myself am engaged in the calculation of technological processes, I note such things. I think it is much more profitable for you to study the technologies of parallel computing organizations as a priority, judging by the computer, for example, CUDA NVidia is just for you. This will give a much greater increase in productivity, and this knowledge is less prone to obsolescence and is much closer to your main profile.
As for compilation, you can assemble any distribution kit "for yourself" from source codes, it's just that Gentoo elevated this principle to the rank of a central idea. Z.Y. if you're interested, I myself use debian-based distributions, but this is just a habit, and "catching flies" in speed percentages is not so important for me.