[CPMD-list] one typo and one problem

[Christian Tuma]

Dear CPMD developers,

there is a typo in "prowfn.F", line 526:

  ...
  IF(ABC(IAT4).GT.CUT4O) THEN
  ...

It should probably be
  
  ...
  IF(ABCD(IAT4).GT.CUT4O) THEN
  ...

otherwise problems occur during runtime.


Furthermore, I encountered a general problem on platforms where CPMD runs
in 64 bit mode: The routine MEMORY has the argument LENGTH (second argument).
This argument is of the type INTEGER, and the Variable LEN (calculated from
LENGTH) is an INTEGER*8 which allows for memory allocations beyond the 32 bit
limit later on. So all calls to MEMORY have to have a second argument of the
type INTEGER which limits the memory allocation to 8 times the 32 bit limit
(in most cases probably 8 x 2 GB = 16 GB). Is that correct?
I encountered a situation where this is not sufficient: In the routine PROWFN
(prowfn.F, line 77) there is a call to MEMORY to allocate memory for N4 numbers
of the type REAL*8. Unfortunately, for my systems N4 does not fit into an
INTEGER variable, and this makes CPMD stop in the subroutine MEMORY (negative
value for LENGTH).
I don't know whether this type of problem already occured to other users at
maybe different points of the code or whether it is not worthwhile to mention.
Since simply declaring LENGTH in the subroutine MEMORY as INTEGER*8 won't work,
I suggest to introduce a second subroutine similar to MEMORY where LENGTH is
of the type INTEGER*8 and which is called at all points in the CPMD code
(with an INTEGER*8 argument) where the 32 bit limit might be easily
exceeded, e.g. in non-parallel parts of the code (like my problem, there
shouldn't be too many of them I guess).

What do the developers think about that?

cheers,
Christian.


-- 
Christian Tuma             Humboldt-Universitaet Berlin
ct at chemie.hu-berlin.de     Arbeitsgruppe Quantenchemie (Prof. Sauer)
phone: +49-30-20937140     Brook-Taylor-Str. 2, 12489 Berlin, GERMANY
fax:   +49-30-20937136     http://www.chemie.hu-berlin.de/ag_sauer