In article <[log in to unmask]>, Dan Hollis
<[log in to unmask]> says:
>
>Is it possible to use mapped files in CM mode? HPFOpen is NM callable only,
>but if someone were to write a switch for it, would the pointer work in CM
mode?
>
>-Dan
>.----------------------------------------------.
>|Dan Hollis -- Pharmacy Computer Services, Inc.|
>[log in to unmask]      -     (503)476-3139|
>`----------------------------------------------'
 
While it is possible to get a long or short pointer using a CM/NM
switch scheme, making use of it is not so easy. The CM emulator actually
has several new "MOVE" group instructions to move data to/from a LONG or
SHORT "NM" address to/from the CM stack; so if you get the NM pointer and
store it in a Double(Short Pointer) or Long(Long Pointer) data type
using SPL, then assemble the appropriate "move-to-virtual" or
"move-from-virtual" instructions after pushing the source/target
addresses onto the stack, it works just fine,  The performance is about
the same as MTDS/MFDS, etc.;so unless you're moving fairly big chunks, the
overhead of the move sequence might a bit much. Also, the SPL compiler
won't recognize the "new" instruction mnemonics, so you'll have to define
constants with the right bit pattern for these in order to assemble them.
Without an SPL compiler, there's no way to generate these instruction that
I know of (except for modifying dummy code with debug);
 
If you get the pointer and succeed in moving the data to from the stack,
you may still have a problem with posting changes to the file.  It would
probably be safer and more practical to write a native procedure to do
whatever functions are needed with mapped access.....
 
Winston