evalhf/fortran - and its relation to Fortran
<Text-field style="Heading 2" layout="Heading 2" bookmark="info">Description</Text-field>
Typically a Fortran program or a set of functions can be easily converted into a Maple function which can be executed by evalhf. The speed of a function executed by evalhf compared to a function compiled in optimized Fortran is on a ratio anywhere between 1:5 to 1:50. Converting Fortran into Maple-evalhf is still one or two orders of magnitude faster than running the equivalent code under standard Maple.
The Whetstone benchmark gives a ratio of 1:35 in favor of compiled Fortran (under a VAX running Unix BSD 4.3).
The following differences and problems should be observed when converting Fortran into Maple-evalhf:
The only type handled by evalhf is floating point (double precision). Integers and Booleans are treated as floats also.
There is no equivalent of the common or equivalence statements.
Any Fortran expression which will evaluate over the integers, in particular expressions assigned to integer variables, should be surrounded by the function trunc().
Array declarations are dynamic with the array() function, and not static.
All variables should be declared as local variables.
Fortran may return values through assigned arguments. This will not work under evalhf. Arrays with a single element may solve this problem.
Fortran is very liberal with the array dimensions and will allow a subroutine to work with an array that has a declaration different from the caller. This is not allowed in Maple-evalhf; furthermore, arrays can only be passed as a whole, not just by the mention of a single element.
Returned values in Maple are the last value computed. In Fortran these values are assigned to a variable with the same name as the function.
There is neither a read nor format statement.