Assumed size arrays: Colon vs. asterisk - DIMENSION(:) arr vs. arr(*)

The form

real, dimension(:) :: arr

declares an assumed-shape array, while the form

real :: arr(*)

declares an assumed-size array.

And, yes, there are differences between their use. The differences arise because, approximately, the compiler 'knows' the shape of the assumed-shape array but not of the assumed-size array. The extra information available to the compiler means that, among other things, assumed-shape arrays can be used in whole-array expressions. An assumed-size array can only be used in whole array expressions when it's an actual argument in a procedure reference that does not require the array's shape. Oh, and also in a call to the intrinsic lbound -- but not in a call to the intrinsic ubound. There are other subtle, and not-so-subtle, differences which your close reading of the standard or of a good Fortran book will reveal.

Some advice for new Fortran programmers is to use assumed-shape arrays when possible. They were not available before Fortran 90, so you will see lots of assumed-size arrays in old code. Assumed-shape arrays are better in new code, because the shape and size functions can be used to query their sizes to avoid out-of-bounds error and to allocate arrays whose dimensions depend on the dimensions of input arrays.

High Performance Mark's answer explains the difference between the two statements - in short: yes, there's a difference; only one declares an assumed-size array - and the implications.

However, as dimension(:) but is mentioned, seemingly against not dimension(*), I'll add one thing.

real, dimension(:) :: arr1real, dimension(*) :: arr2

is equivalent to

real :: arr1(:)real :: arr2(*)

or even using dimension statements. [I don't want to encourage that, so I won't write out the example.]

The important difference in the question is the use of * and :, not dimension.

Perhaps there was some conflation of assumed-size with dummy argument? It is as a dummy argument where this choice is most common.