Consider Cython as a means to accelerate exponential format conversion from the strings

[bskinn]

A well constructed bit of Cython code might enable very rapid intake of the exponential values from the CUBE files, exploiting the knowledge of the format of the values.

It might end up being necessary to inspect the first value imported, in order to know how many digits are expected after the decimal and in the exponent, but the time taken for that inspection should be DWARFED by the time saved by optimized parsing of the values.

Of course, I should profile the intake operation so that I can be sure the parsing is actually where the bottleneck is.