In a message dated 1/10/2000 9:05:48 AM Pacific Standard Time,
[log in to unmask] writes:
> How do you know you have a "memory leak"? What are the symptoms?
If you find that you run out of memory after a large number of repetitions of
the same sequence of operations, or your memory use meter keeps creeping
upwards until a crash occurs, or you have to split your case into separate
runs to get it to work reliably, it becomes tempting to blame "memory leak."
In order to back up such an assumption, it may be necessary to examine the
generated code to see where temporary arrays are created, and determine that
the crash occurs at one of these points.
Regrettably, it seems that one way to make this happen is to use an f90
compiler which has a C back end with a "better" version of the C compiler
than the one which the vendor has tested most thoroughly, or a slightly
different version than the one used to compile the run-time.
Considering that some of the compiler test suites have not been able to run
under Windows versions prior to Windows 2000 RC2, I believe that current
production Windows versions tend to have memory leaks associated with
run-time exceptions -- if a section is aborted, its allocated memory is not
reclaimed.
Tim
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|