Strictly speaking, if your compiled code "does'nt run", there can be no question of convergence. If you meant to say that it does run but does not produce results similar to what you were used to from the old version of the compiler, then there is something to work with.
There can be any number of reasons why a Fortran program compiled with two different compilers or (two versions of one compiler) give different results. The most common reason is buggy code that happened to give correct results with one compiler. Some assumptions about initialization or lifetimes of variables may have been made that are not in agreement with the Fortran standard. If your program makes much use of floating point calculations, the chances for variations in results are increased.
The Intel compiler has many options. You need to study the meanings of those options, and evaluate the possibility that changing some of them from their default settings may solve the problem for you.
If you can reproduce the problem with a small example, you may try to obtain help here. Otherwise, you have a difficult debugging task facing you.
