@sand15 

It being impossible to do these calculations, except for very small examples, in an interpreted language, the matrices were constructed using a custom program written in C++. They were then output to a text file for reading by Maple. I used Maple for this purpose because I have not built an interface to allow for out an interactive analysis of the matrices.

Constructing these matrices with entries in GMP rationals, along with the many data structures necessary to produce the matrices, consumed about 370 GB of RAM.

All this said, it seems that I may have to overcome my natural laziness and use a tool better suited to the task at hand.

@sand15 

Thank you for creating the examples. For the following reasons, I am not sure what to make of your examples:

1. My matrices are relatively sparse and I specified "storage=sparse" when creating them.
2. In the past have successfully openned larger matrices of a similar type to the 120,000 x 120,000 matrices that I am considering at the moment.

Perhaps, by some numerical quirk, the matrices produced by this particular example may be less sparse than other examples that I have considered.

@Rouben Rostamian  

Thank you for the suggestion.

I tried setting "ulimit unlimited" in bash before starting maple. After this, setting kernelopts(stacklimit=130560), the error occurred at approximately 7 GB of RAM usage. Restarting maple and setting kernelopts(stacklimit=261120), the error occurred at approximately 6.5 GB of RAM usage, so "ulimit unlimited" does not appear to have changed the behaviour on this system.