Dear community, this is a very simple question that can have a great impact on the performance on parallel computation. My question is threefold.
1. - Where does allocatable arrays and static reside? (cache, register, RAM). I've read that static arrays deliver faster run time because they access a different part of the memory layout.
2.- If static arrays are faster, then for the performance we should avoid allocatable arrays. How can we pass arguments to Fortran at a compiled time using bash scripting or a makefile in order to enforce static arrays?
3.- in MPI we usually use domain decomposition, thus the only way I have found to accommodate a different number of nodes in each rank is by using allocatable arrays. Any suggestion to enforce static arrays?
Thank you very much dear community!!